NASA-CR-193013
I
NASA
third International Symposium on
Space terahertz technology
JPL
Th« University
of Michigan
Symposium Proceedings
so in
r\i I CO
r- i r»
s 3 s w
M Ot N t
I X I —
m ►- ro U
o i o* c
Z I Z Z5
o
Integrated on Si section
Machined gain and phasing section
March 24-26, 1992
University of Michigan
Ann Arbor, Michigan
UJ
u r
< o
Q. >-
to \n
Q O
OC Z Q.
•-•OX
I >■
t-rw
3
I V) >
»- o o
a. o
z -J
*- >- o
-H X
o JU
m < uj
^ZK
•H O
I •-" M
<X t- *-
u < a.
I Z UJ
in UJ <
< t- e*
Z Z ui
c
3
c
9
r
u
o
z
a
UJ
uj a.
u
o o
oc >*
a r»
Sponsored by:
NASA Office of Aeronautics and Space Technology (OAST), University Space Engineering Research Centers
Program, with cooperative sponsorship by the Microwave Theory and Techniques Society of IEEE.
Organized Jointly by:
The University of Michigan's NASA Center for Space Terahertz Technology and JPL's Center for Space
Microelectronics Technology.
Proceedings of the
Third International Symposium on
Space Terahertz Technology
March 24-26, 1992
University of Michigan
Ann Arbor, Michigan
Symposium Co-chairs:
Technical Co-chairs:
Organizing Committee
Fawwaz T. Ulaby, University of Michigan
Carl A. Kukkonen, Jet Propulsion Laboratory
Gabriel M. Rebeiz, University of Michigan
Margaret A. Frerking, Jet Propulsion Laboratory
Local Arrangements: Valerie Kabat, University of Michigan
Symposium Proceedings: Valerie Kabat, University of Michigan
Group Photo
A group photo of some of the 1992 Symposium participants.
ORIGINAL PAGE
BLACK AND WHITE PHOTOGRAPH
Preface
The Third International Symposium on Space Terahertz Technology was held at the
University of Michigan in Ann Arbor, Michigan, on March 24-26, 1992. The Symposium,
which was attended by approximately 165 scientists and engineers from the U.S., Europe,
and Japan, featured papers relevant to the generation, detection, and use of the terahertz
spectral region for space astronomy and remote sensing of the Earth's upper atmosphere.
The program included thirteen sessions covering a wide variety of topics including solid-
state oscillators, power-combining techniques, mixers, harmonic multipliers, antennas and
antenna arrays, submillimeter receivers, and measurement techniques.
The Symposium was sponsored by the University Space Engineering Research Centers
Program of NASA's Office of Aeronautics and Space Technology (OAST), and organized
jointly by the University of Michigan's NASA Center for Space Terahertz Technology and
JPL's Center for Space Microelectronics Technology. The Microwave Theory and
Techniques Society of IEEE served as a cooperative sponsor of the Symposium, as well as a
medium for publication of some of the papers that were presented at the Symposium in
the form of a mini-special issue (April 1993) of the IEEE-MIT Transactions.
The Fourth International Symposium on Space Terahertz Technology will be held at the
University of California, Los Angeles, on March 30-April 1, 1993.
Fawwaz T. Ulaby
Carl A. Kukkonen
Contents
Opening Session Chain Fawwaz Ulaby
NASA's OAET Sensors Program *
M. Sokoloski
NASA's Astrophysics Program in High-Resoluton THz Spectroscopy *
L. Caroff
Applications of Terahertz Technology to Astronomy *
T. G. Phillips
Coherent Systems in the Terahertz Range: Elements, Operation & Examples 1 — /
P. F. Goldsmith
Session L Sources I Chain CarlKukkonen
Broadband Millimeter-Wave GaAs Transmitters and Receivers Using
Planar Bow-Tie Antennas 24~2^
Y. Konishi, M. Kamegawa, M. Case, R. Yu, M. J.W. Rodwell,
R. A. York, D. B. Rutledge
Active CPW-Fed Slot Antennas for Power Combining Applications 32— >2>
B. K. Kormanyos, G. M. Rebeiz
2x2 Quasi-Optical Power Combiner Array at 20 GHz 37-^/
S. Kawasaki, T. Itoh
Monolithic Millimeter-Wave Diode Array Beam Controllers:
Theory and Experiment 45 ~~>
L. B. Sjogren; H-X. L. Liu; F. Wang; T. Liu; W. Wu; X-H. Qin; E. Chung;
C. W. Domier; N. C. Luhmann, Jr.; J. Maserjian; M. Kim; J. Hacker;
D. B. Rutledge; L. Florez; J. Harbison
A Study of Subterahertz HEMT Monolithic Oscillators 58""'^
Y. Kwon and D. Pavlidis
Session 2: Varactors I Chair: Tatsuo Itoh
Varactor Diodes for Millimeter and Submillimeter Wavelengths 73 ""V
B. J. Rizzi, J. L. Hesler, H. Dossal, T. W. Crowe
A Schottky/2-DEG Varactor Diode for Millimeter and _
Submillimeter Wave Multiplier Applications 93 ^ O
W. C. B. Peatman, T. W. Crowe, M. Shur, B. Gelmont
Thermionic Emission Current in a Single Barrier Varactor 110 ~" /
H. Hjelmgren, J. R. East, E. Kollberg
*Presentation only.
Progress on Single Barrier Varactors for Submillimeter Wave
Power Generation 115 "10
S. M. Nilsen, H. Gronqvist, H. Hjelmgren, A. Rydberg, E. Kollberg
Effect of Cooling on the Efficiency of Schottky Varactor
Frequency Multipliers at Millimeter Waves 134-7/
J. Louhi, A. Raisanen, N. Erickson
Session 3: Varactors n Chair RayBlundell
Superlattice Barrier Varactors 146 "/ *■ —
C. Raman, J. P. Sun, W. L. Chen, G. Munns, J. East, G. Haddad
A New Fabrication Technique for Back-to-Back Varactor Diodes 158 W "^
R. P. Smith, D. Choudhury, S. Martin, M. A. Frerking, J. K. Liu,
F. A. Grunthaner
A 200 GHz Tripler Using Single Barrier Varactor 164 -7 7
D. Choudhury, M. A. Frerking, P. D. Batelaan
A Submillimeter Tripler Using a Quasi-Waveguide Structure 181 " *
N. Erickson, G. Cortes-Medellin
Session 4: SIS Receivers I Chain Neville Luhmann
A 380 GHz SIS Receiver Using Nb/A10 x /Nb Junctions for a
RadioAstronomical Balloon-Borne Experiment: PRONAOS 189 -/&
P. Febvre, P. Feautrier, C. Robert, J. C. Pernot, A. Germont, M. Hanus,
R. Maoli, M. Gheudin, G. Beaudin, P. Encrenaz
A Low Noise 410-495 Heterodyne Two Tuner Mixer, Using Submicron
Nb/Al203/Nb Tunneljunctions 210 "pi
G. de Lange, C. E. Honingh, M. M. T. M. Dierichs, R. A. Panhuyzen,
H. H. A. Schaeffer, T. M. Klapwijk, H. van de Stadt, M. W. M. de Graauw
Double Dipole Antenna SIS Receivers at 100 and 400 GHz 222 — / %
A. Skalare, H. van de Stadt, Th. de Graauw, R. A. Panhuyzen,
M. M. T. M. Dierichs
Slot Antenna SIS Mixers for Submillimeter Wavelengths** 234 ~ * I
J. Zmuidzinas, H. G. LeDuc, J. A. Stern
Session 5: SIS Receivers II Chain David Rutledge
A Planar Quasi-Optical SIS Receiver for Array Applications 235^
P. A. Stimson, R. J. Dengler, P. H. Siegel, H. G. LeDuc
A Fixed Tuned Broadband Matching Structure for Submillimeter .
SIS Receivers** 243 ~-0.\
T. H. Buttgenbach, H. G. LeDuc, P. D. Maker, T. G. Phillips
* Abstract only.
Modeling and Performance of Nb SIS Mixers in the 1.3 mm
and 0.8 mm Bands 244 ~v?2_^
A. Karpov, M. Carter, B. Lazareff, D. Billon-Pierron, K.H. Gundlach
Comparison of Measured and Predicted Performance
of a SIS Waveguide Mixer at 345 GHz 251-^3
C. E. Honingh, G. deLange, M. M. T. M. Dierichs, H. H. A. Schaeffer,
J. Wezelman, J. v. d. Kurr, Th. de Graauw, T. M. Klapwijk
A Low Noise 492 GHz SIS Waveguide Receiver 266 ^~^¥
C. K. Walker, J. W. Kooi, M. Chan, H. G. LeDuc, P. L. Schaffer, /
J. E. Carlstrom, T. G. Phillips
Session 6: Antennas & Networks I Chair: Gabriel Rebeiz
Slot-Line End-Fire Antennas for THz Frequencies 280~^-D
H. Ekstrom, S. Gearhart, P. R. Acharya, H. Dave, G. Rebeiz, S. Jacobsson,
E. Kollberg, G. Chin
Quasi-Optical Antenna-Mixer-Array Design for Terahertz Frequencies 291 "W &
Y. Guo, K. A. Potter, D. B. Rutledge
Analysis of a Novel Non-Contacting Waveguide Backshort 298 ~~i? /
T. M. Weller, L. P. B. Katehi, W. R. McGrath
Silicon Micromachined Waveguides for Millimeter
and Submillimeter Wavelengths 316 ~b2^P
M. Yap, Y.C. Tai, W.R. McGrath, C. Walker
Session 7: Antennas & Networks II Chair: Linda Katehi
Progress in Integrated-Circuit Horn Antennas for Receiver Applications:
Parts I & II 324 -^ ^3^
G. V. Eleftheriades, W. Y. Ali-Ahmad, G. M. Rebeiz
Zone Plate Lens Antennas for Millimeter and Submillimeter Wavelengths 345
P. F. Goldsmith
-31
Onset of Dispersion in Nb Microstrip Transmission Lines
at Submillimeter Wave Frequencies 362 ^j5 2-»
H. H. S. Javadi, W. R. McGrath, B. Bumble, H. G. LeDuc
Double-Slot Antennas on Extended Hemispherical Dielectric Lenses 382 "~ * ^
D. F. Filipovic, S. J. Gearhart, B. K. Kormanyos, G. M. Rebeiz
Session 8: SIS Theory & Fabrication I Chain MarcFeldman
Embedding Impedance Approximations in the Analysis of SIS Mixers 394 ""^2 j/
A. R. Kerr, S. K. Pan, S. Withington /
Submicron Area Nb/ AlOx/Nb Tunnel Junctions / ~
for Submillimeter Mixer Applications 408 ^ ' pb
H. G. LeDuc, B. Bumble, S. R. Cypher, A. J. Judas, J. A. Stern
Noise in Josephson Effect Mixers and the RSJ Model** .419 - 3"?
R. Schoelkopf, T. Phillips, J. Zmuidzinas
Fabrication and Characterization of High Current-Density, ^
Submicron NbN/MgO/NbN Tunnel Junctions 420 "3 L
J. A. Stern, H. G. LeDuc, A. J. Judas
Session 9: Sources II Chain James Mink
A Quasioptical Resonant-Tunneling-Diode />
Oscillator Operating Above 200 GHz** 439 ^3^
E. R. Brown, C D. Parker, K. M. Molvar, A. R. Calawa, M. J. Manfra
Transit-Time Devices as Local Oscillators for Frequencies -
above 100 GHz 440-3/
H. Eisele, C. Kidner, G. I. Haddad
Negative Differential Resistance (NDR) Frequency , n
Conversion with Gain 457^ H
R. J. Hwu, R. W. Aim, S. C. Lee
Modeling, Design, Fabrication and Testing of
InP Gunn Devices in the D-Band (110 GHz-170 GHz) 477 ^i\ /
R. Kamoua, H. Eisele, J. R. East, G. I. Haddad, G. Munns, M. Sherwin
Session 10: SIS Theory & Fabrication II Chair: Anthony Kerr
Recent Advances in Superconducting-Mixer Simulations 494 **h ^
S. Withington, P. R. Kennedy
Submillimeter Wave Detection with Superconducting . . . «
Tunnel Diodes 502 ~*j£
M. J. Wengler
Evaluation of Integrated Tuning Elements with SIS Devices 522
M. M. T. M. Dierichs, C. E. Honingh, R. A. Panhuyzen, B. J. Feenstra,
A. Skalare, J. J. Wijnbergen, H. v. d. Stadt, Th. de Graauw
Source Conductance Scaling for High Frequency ^\Jh
Superconducting Quasiparticle Receivers 538 "-'
Q. Ke, M. J. Feldman
** Abstract only.
Session 11: Sources m Chair: George Haddad
Resonant Tunneling Diodes as Sources for
Millimeter and Submillimeter Wavelengths 548— y^
O. Vanbesien, R. Bouregba, P. Mounaix, D. Lippens, L. Palmateer, J. C. Pernot,
G. Beaudin, P. Encrenaz, E. Bockenhoff, J. Nagle, P. Bois, F. Chevoir, B. Vinter
Simulation of Electron Transport in Quantum Well Devices 560 — V '
D. R. Miller, K. K. Gullapalli, V. R. Reddy, D. P. Neikirk
Parallel Arrays of Josephson Junctions for Submillimeter Local Oscillators 575 ~ V O
A. Pance, M. J. Wengler
Monolithic Millimeter- Wave Diode Grid Frequency Multipler Arrays 595 "77
H. X. L. Liu; X. H. Qin; L. B. Sjogren; W. Wu, E. Chung;
C. W. Domier; N. C. Luhmann, Jr.
Session 12: Mixers and Detectors I Chairs: Margaret Frerking
Planar GaAs Diodes for THz Frequency Mixing Applications 600 "**i?Z5
W. L. Bishop, T. W. Crowe, R. J. Mattauch, H. Dossal
Planar Doped Barrier Subharmonic Mixers 616 "^ '
T. H. Lee, J. R. East, G. I. Haddad.
New Approach to the Design of Schottky Barrier Diodes for THz Mixers 631 , ~*<3 ^*»
A. Jelenski, A. Grub, V. Krozer, H. L. Hartnagel
Electrical and Infrared Properties of Thin Niobium Microbolometers Near T c .643 ~>S3
E. N. Grossman, J. E. Sauvageau, D. G. McDonald
Session 13: Mixers & Detectors II Chair: Thomas Crowe
Measurements of the Single Sideband Suppression for a
650 GHz Heterodyne Receiver 654 — <£5^/
S. Crewell, H. Nett /
InGaAs/InP Heteroepitaxial Schottky Barrier Diodes for Terahertz _
Applications 661 "--0-3
U. V. Bhapkar, Y. Li, R. J. Mattauch
A Broadband THz Receiver for Low Background Space Applications 678 -~ Q1&
C. Hagmann, D. J. Benford, A. C. Clapp, P. L. Richards, P. Timbie
AlGaAs/GaAs Quasi-Bulk Effect Mixers: Analysis and Experiments 688 ^£? /
K. S. Yngvesson, J.-X. Yang, F. Agahi, D. Dai, C. Musante, W. Grammer, K. M. Lau
All-Solid-State Radiometers for Environmental Studies to 700 GHz 706 —S3
R. Riidiger, P. Zimmermann
A 492 GHz Cooled Schottky Receiver for Radio Astronomy 724
J. Hernichel, R. Schieder, J. Stutzki, B. Vowinkel, G. Winnewisser,
P. Zimmermann
Third International Symposium on Space Terahertz Technology Page 1
N93-27727
COHERENT SYSTEMS IN THE TERAHERTZ
FREQUENCY RANGE : ^ oSJ £
ELEMENTS C. -
OPERATION
& EXAMPLES
PAUL F. GOLDSMITH
MILLITECH CORPORATION
South Deerfield MA, 01373
and
FIVE COLLEGE RADIO ASTRONOMY OBSERVATORY
Department of Physics and Astronomy
University of Massachusetts, Amherst MA 01003
8 e 2 Third International Symposium on Space Terahertz Technology
TERAHERTZ COHERENT SYSTEMS APPLICATIONS
RADIOMETRY / SPECTROSCOPY
ASTRONOMY
ATMOSPHERIC REMOTE SENSING
ALL-WEATHER SYNTHETIC VISION SYSTEMS
CONTRABAND DETECTION
HIGH POWER
PLASMA HEATING
HIGH ENERGY ACCELERATORS
PLASMA DIAGNOSTICS
THERMAL IMAGING
DENSITY PROBING
BACKSCATTER MEASUREMENTS
COMMUNICATIONS
PERSONAL & VEHICULAR
DIGITAL DATA LINKS
TV REMOTE / STUDIO LINKS
MATERIALS MEASUREMENT AND COMMERCIAL PROCESS CONTROL
PAPER MAKING
HV CABLE MANUFACTURING
RADAR SYSTEMS
MILITARY - SEEKERS, INSTRUMENTATION, AND MODELING
AUTOMOTIVE COLLISION AVOIDANCE
ATMOSPHERE, METEOROLOGY, GROUND, ICE, AND FOLIAGE
Third International Symposium on Space Terahertz Technology
Page 3
COMPONENTS OF COHERENT SYSTEMS
AT MILLIMETER & SUBMILUMETER WAVELENGTHS
/
/ /
INPUT OPTICS
SIGNAL
PROCESSING
ELEMENTS
COLUMAT1NG MIRRORS
AND LENSES
POLARIZING GRIDS
WAVEPLATES
@— XN
COHERENT SOURCE
LOCAL OSCILLATOR;
TRANSMITTER
<
DIPLEXER
%-rv— 1
COMBINATION OF
LOCAL OSCILLATOR
AND SIGNAL
1L
ANTENNA/FEED
ELEMENTS
MIXER
IF SYSTEM
EFFICIENCY, BEAMWIDTH,
BANDWIDTH, CONSTRUCTION
INTEGRABIUTY
CONVERSION LOSS; NOISE;
LO. POWER; BANDWIDTH
NOISE; BANDWIDTH
•S-KHD
^o
S-KHD
DETECTION/SIGNAL
PROCESSING
SPECTROMETERS:
FREQUENCY COVERAGE;
RESOLUTION; FLEXIBILITY;
POWER CONSUMPTION
Page 4 Third International Symposium on Space Terahertz Technology
BRIEF OVERVIEW OF SELECTED COMPONENTS
EMPHASIZE AREAS THAT I FEEL DESERVE MORE ATTENTION
THAN THEY ARE RECEIVING AT PRESENT
[1] MATERIALS MEASUREMENT
FUNDAMENTAL FOR MANY ASPECTS OF SYSTEMS DESIGN
NEED MORE DATA, BETTER DATA, AND BETTER ACCESS
REXOLITE DATA FROM G. J. SIMONIS, J. P. SATTLER, T. L.
WORCHESKY, AND R. P. LEAVITT, INT. J. INFRARED
AND MILLIMETER WAVES, VOL. 5, 57 - 72, 1984.
BORON DATA FROM. A. J. GATESMAN, R. H. GILES, AND J.
NITRIDE WALDMAN PROC. MATERIALS RESEARCH SOCIETY
SYMPOSIUM ON WIDE BANDGAP SEMICONDUCTORS, 1991
FALL MEETING, BOSTON
INTERCOMPARISON OF TECHNIQUES FOR DETERMINATION OF
NEAR MILLIMETER DLELECTRIC PROPERTIES
JAMES BIRCH ET AL. - NATIONAL PHYSICAL LABORATORY
TEDDINGTON, MIDDLESEX
U.K. TW110LW
REPORT DES 115, OCTOBER 1991
Third International Symposium on Space Terahertz Technology p a g e 5
[2] QUASIOPTICAL COMPONENTS
HOW CAN THEY BE FABRICATED IN SUBMILLIMETER REGION?
TRADITIONAL MACHINING METHODS BECOME VERY DIFFICULT
AND EXPENSIVE-
NEED TO FIND CONSTRUCTIVE COMBINATIONS OF METAL-
WORKING AND SEMICONDUCTOR PROCESSING APPROACHES SUCH
AS SELECTIVE ETCHING
EXAMPLES:
PROCESSING SILICON TO FABRICATE TWO DIMENSIONAL
IMAGING HORN ANTENNA ARRAYS (REBEIZ ET AL. IEEE
MTT 38, 1473 (1990))
ETCHING AND PLATING SILICON TO MAKE DICHROIC PLATE
HIGH PASS FILTERS IN 1000 GHZ RANGE
(SIEGEL AND LICHTENBERGER 1990 MTT-S SYMP. DIGEST,
1341)
Pige 6 Third International Symposium on Space Terahertz Technology
RADIOMETRY AND SPECTROSCOPY : ASTRONOMY
OBSERVING LOCATION DEPENDS PRIMARILY ON FREQUENCY:
GROUND - BASED
AIRPLANE AND BALLOONS: KAO; SOFIA
SPACE: SWAS; SMIM; FIRST
[1] SENSITIVITY
HIGHEST SENSITIVITY ALWAYS REQUIRED
CRYOGENIC COOLING IS ACCEPTABLE
BROADBAND SYSTEMS WILL BE REQUIRED FOR FUTURE SYSTEMS
[2] IMAGING SYSTEMS
FOCAL PLANE ARRAYS DEVELOPED FOR MILLIMETER RANGE:
FCRAO 15 -ELEMENT QUARRY ARRAY 85-115 GHZ
NRAO 8-ELEMENT ARRAY IN 230 GHZ RANGE
CANNOT SACRIFICE FEED EFFICIENCY SIGNIFICANTLY JUST TO
OBTAIN LARGER NUMBER OF ELEMENTS DUE TO COST AND
COMPLEXITY OF ASSOCIATED SIGNAL PROCESSING.
[3] OTHER COMPONENT DEVELOPMENT
RAPID PROGRESS IN FREQUENCY MULTIPLIER SOURCES, BUT
FURTHER DEVELOPMENT REQUIRED FOR GREATER BANDWIDTH
AND REACHING HIGHER FREQUENCIES
Third International Symposium on Space Terahertz Technology
Page 7
PLANAR HETERODYNE ARRAY USING A DIELECTRIC-FILLED PARABOLA
Receive
Element
(Schottky
Dlod« or SIS
Tunnel Junction)
Focal Point of
Dielectric— Riled
Parabola
Coplanar
Lines for
IF it DC
Signal
Input
EXPLODED VIEW (SIDE)
1.0 INCH
I I | I I I I |
TOP VIEW
(COVER REMOVED)
P.H. Siegel
California Institute of Technology Jet Propulsion Laboratory
p <*ge 8 Third International Symposium on Space Terahertz Technology
RADIOMETRY:
AIRCRAFT ALL WEATHER LANDING SYSTEM
APPROACH
FOCAL PLANE IMAGING SYSTEM AT 94 GHZ TO PROVIDE
SYNTHETIC VISION CAPABILITY FOR AIRCRAFT LANDING IN
ALMOST ALL WEATHER CONDITIONS
MILLIMETER - WAVE IMAGING ALLOWS GOOD VISIBILITY OF
RUNWAY BOUNDARIES AND POSSIBLY DANGEROUS OBSTACLES
FROM APPROPRIATE DISTANCE
FOCAL PLANE RADIOMETRIC IMAGING PERMITS REAL-TIME (30/
SECOND) UPDATE RATE
IMAGES READILY INTERPRETABLE WITHOUT EXTENSIVE
PROCESSING
HEADS-UP DISPLAY STRAIGHTFORWARD TO IMPLEMENT
TECHNOLOGY:
FOCAL PLANE ARRAY OF 256 (TO DATE) PIXELS UTILIZING
CONSTANT - WIDTH SLOT ANTENNAS
SINGLE -ENDED HARMONIC MIXERS WITH QUASIOPTICAL LOCAL
OSCILLATOR INJECTION
DICKE-TYPE LOAD COMPARISON ESSENTIAL
MECHANICAL OR ELECTRONIC (QUASIOPTICAL HYBRID OR
MONOLITHIC ) REALIZATIONS POSSIBLE
COMPACT OPTICS
Third International Symposium on Space Terahertz Technology
Page 9
Page 10 Third International Symposium on Space Terahertz Technology
DETECTION OF CONCEALED
WEAPONS AND CONTRABAND MATERIAL
E RQ BI ^M :
• DETECTION OF PLASTIC WEAPONS AND
EXPLOSIVES CONCEALED BENEATH CLOTHING OF
AIRLINE PASSENGERS.
CONSTRAINTS :
• EFFECTIVE PERFORMANCE
• NON-INVASIVE OPERATION
• RAPID PROCESSING
TECHNICAL APPROACH :
• ACTIVE (REFLECTING) AND PASSIVE (RADIOMETRIC)
MILLIMETER-WAVELENGTH IMAGING SYSTEMS
• RADIOMETRIC SYSTEM LEAST INVASIVE AND
OFFERS GOOD FIDELITY
• CLOSE FOCUSED OPTICS AND FOCAL PLANE ARRAY
millitech
Third International Symposium on Space Terahertz Technology
Page 11
Scan 86: 10/81x91 14:39
rr-rt vla»
la talst
x (crt) : -38.88 38.88 8.25 t 241 1
*, (err) : -28.88 VS. 88 8.25 < 381 i
Passive Line Scan 94 GHz Millimeter Wave Image
ORIGINAL PAQE IS
OF POOR QUALITY
/\ millitacfr
Page 12
Third International Symposium on Space Terahertz Technology
RADIOMETRY: ATMOSPHERIC REMOTE SENSING
[1] ISSUES:
MEASUREMENT OF TRACE CONSTITUENTS INCLUDING: H 2
3
CIO
N 2
PHYSICAL CONDITION (TEMPERATURE) PROFILING
DELAY MEASUREMENTS FOR RADAR ALTIMETERS
MESOSPHERIC WIND VELOCITY DETERMINATIONS
TRACE EMISSIONS FROM LOCALIZED SOURCES
[2] OBSERVING LOCATIONS
GROUND - BASED: 3 AND CIO MONITORING NETWORK
ANTARCTIC AND POLAR REGIONS
AIRPLANE:
USEFUL AS TEST PLATFORM AND FOR
STUDY OF LOCALIZED PHENOMENA
SPACE
UARS - SUCCESSFULLY OPERATING !
MAS (SHUTTLE LIMB - SOUNDER)
EOS (EARTH OBSERVING SYSTEM)
AMSU-B / METEOSAT
Third International Symposium on Space Terahertz Technology
Page 13
. LOW ELEVATION
• BEAM
A. \ ~ A
■JJ ' \
HIGH ELEVATION
BEAM
A . *A *
BEAM SWITCHING
CHOPPER
PARTIAL ABSORBER FOR TOTAL POWEH BALANCE
SPECTROMETER
CONTROL
AND
ANALYSIS
COMPUTER.
CONFIGURATION FOR GROUND-BASED RADIOMETER TO STUDY ATMOSPHERIC TRACE GASES
094 ca COLLIMATING
LENS
SIGNAL
DOUBLE DIELECTRIC
SLAB FILTER
TRANSMISSION
MILLIMETER
FEEOHORN
LOCAL
OSCILLATOR
SIGNAL
BANO
IMAGE
BANO
FREQUENCY -—
ABSORBING
LOAO
DIELECTRIC SLAB SINGLE-SIDEBAND FILTER FOR 279 GHZ CIO RADIOMETER
Page 14
Third International Symposium on Space Terahertz Technology
Third International Symposium on Space Terahertz Technology
Page 15
r SCANNING
ANTENNA
SYSTEM
THERMAL RADIATION
PRIMARY
FROM ATMOSPHERIC LIMB /
SPACE VIEW
FOR CALIBRATION
SWITCHING
MIRROR
OICHROIC
PLATE
H
63 GHz
RAOIOMETER
FILTER BANK
POLARIZATION
GRIO
CAL
TARGET
205 GHz
RAOIOMETER
183 GHz
RAOIOMETER
CIO
FILTER BANK
H 2 Oj
FILTER BANK
FILTER BANK
H 2
FILTER BANK
I FILTER BANK
H
COMMAND
AND
DATA
HANOLING
TO
UARS
UARS MICROWAVE LIMB SOUNDER INSTRUMENT SIGNAL FLOW PATH
h/lcm
80
70
O s
H,0
60
50
40
30
20
10 -
,L
CIO
UARS- MLS TARGETS AND ALTITUDE RANGES
Page 16 Third International Symposium on Space Terahertz Technology
PLASMA DIAGNOSTICS
THERMAL IMAGING - RADIO METRY WITH HIGH TIME RESOLUTION
EXTREMELY BROADBAND AND/OR SWEPT - FREQUENCY
DENSITY PROFILING - MEASUREMENT OF ELECTRON COLUMN
DENSITY THROUGH PLASMA
INTERFEROMETERS - EITHER RADIO OR OPTICAL TYPES
DEPENDING ON WAVELENGTH
SCATTERING EXPERIMENTS- PROBE TURBULENCE AND SCALE
OF FLUCTUATIONS IN PLASMA
EXAMPLE OF PLASMA DIAGNOSTIC SYSTEM
2 - MM WAVELENGTH 180 DEGREE BACKSCATTER IMAGING
SYSTEM DEVELOPED BY DR. P. EFTHIMION (PRINCETON
PLASMA LABORATORY) AND E.L. MOORE ET. AL. (MILLITECH
CORPORATION)
INCLUDES PHASELOCKED TRANSMITTER AND 64 ELEMENT
FOCAL PLANE IMAGING ARRAY
Third International Symposium on Space Terahertz Technology Page 17
COMMUNICATIONS
APPLICATIONS:
PERSONAL
VEHICULAR - CAR TRAIN AND PLANE
DIGITAL DATA LINKS - SATELLITE AND GROUND
MILITARY COMMUNICATIONS (MILSTAR)
TV REMOTE - STUDIO LINKS
DEVELOPMENTS IN FIELD HAVE BEEN REVIEWED BY H. MEINEL IN
PROC. 18*k EUROPEAN MICROWAVE CONFERENCE, STOCKHOLM,
pp. 1203-1216, 1988
Page 18 Third International Symposium on Space Terahertz Technology
RADAR SYSTEMS
■ MILITARY RADAR SYSTEMS
INSTRUMENTATION RADARS
SEARCH RADARS
SEEKERS
HELICOPTER OBSTACLE AVOIDANCE SYSTEMS
■ AUTOMOTIVE RADAR
PRESENTLY VERY ACTIVE FIELD
GOALS ARE COLLISION AVOIDANCE AND ULTIMATELY
AUTOMATIC CONTROL OF VEHICLE
ATMOSPHERE CLOUD STRUCTURE (ICE & WATER)
METEOROLOGY
REMOTE SENSING OCEANS
VEGETATION
ICE
MODELING
MILLIMETER / SUBMILLIMETER MODELING OF LOWER
FREQUENCY RADAR SYSTEMS AND TARGETS
Third International Symposium on Space Terahertz Technology
Page 19
GUNN OSCILLATOR
CZD-
POLARIZATION 2
IF OUTPUTS
POWER
DIVIDER
ODD
POLARIZATION-
DIPLEXING GRID
POLARIZATION 1
IF OUTPUTS
LENS-
T
LINEAR/CIRCULAR
TRANSFORMER
DUAL POLARIZATION MONOPULSE LENS ANTENNA
GS07
SPECIFICATIONS SUMMARY
TYPE: PULSE
TRANSMITTER FREQUENCY: 77GHz
PULSE WIDTH: 40ns
RISE/FALL TIME: 4ns
ANTENNA: THREE BEAM SCANNING
BEAM WIDTH: 2' ELEVATION AND AZIMUTH
IF BANDWIDTH: lKHz TO 200MHz
145mm
DIA
OSCILLATOR
PIN SWITCH
MATRIX
FEED
NETWORK
INTERFACE -
CONNECTOR
MILLITECH AUTOMOBILE RADAR FRONT END
_ / \ millitech
I
o
3
S
in
£'
3
a
?
a-
3
3
n
a-
a
Q
88
Third International Symposium on Space Terahertz Technology Page 21
MATERIALS MEASUREMENT AND
MANUFACTURING PROCESS CONTROL
MAJOR CONSIDERATIONS
■ DEMANDS EXTREMELY RUGGED SYSTEMS
■ COST IS A CRITICAL FACTOR
■ MOST INDUSTRIES ARE CONSERVATIVE AND NEED TO BE
CONVINCED OF VALUE OF NEW SYSTEM
■ WHAT ARE THE UNIQUE CAPABILITIES OF TERAHERTZ
RANGE?
APPLICATIONS:
HIGH VOLTAGE CABLE INSPECTION
PAPER MAKING
Page 22
Third International Symposium on Space Terahertz Technology
PAPER MEASUREMENTS AT SUBMILLIMETER WAVE LENGTHS
BEAM FROM
FIR U>SER
BEAM SPLITTER
PAPER SAMPLE
D1
FPI
D2
1.0
0.8 -
[A
0.6
U)
2
W
-z.
<
0.4
a:
\—
0.2
\
_ x
X
70 fi
\
96 \i
V 1.18^
-
\ 570^
\
**
\
X
X
\
\\
, X
X.
\
\ N
\
\
\ ^^ ..
1
— I ~i r- i ~ "" i i i i
20
40
60
80
100
% MOISTURE
TRANSMITTANCE OF 80 fiM NEWSPRINT
AS A FUNCTION OF MOISTURE CONTENT
FROM BOULAY; ET AL, IR & MM WAVES, VOL. 5, PP 1221-1234, 1984
Third International Symposium on Space Terahertz Technology Page 23
CONCLUSIONS
APPLICATIONS OF COHERENT SYSTEMS IN TERAHERTZ RANGE ARE
EXTREMELY DIVERSE AND ARE EXPANDING
RAPID TECHNICAL PROGRESS IS TAKING PLACE ON MANY FRONTS
TRANS - MILLIMETER REGION IS NOW SIMILAR TO MILLIMETER RANGE
JUST A FEW YEARS AGO AND A < 3 MM RANGE IS COMPARABLE TO
MICROWAVE REGION IN RECENT PAST
REAL SUBMILLIMETER REGION STILL HAS MANY CHALLENGES
INCLUDING BASIC QUASIOPTICAL COMPONENTS, FREQUENCY SOURCES,
ANTENNAS (INCLUDING ARRAYS) AND HIGH EFFICIENCY AND RUGGED
MIXERS AND DETECTORS
AN IMPORTANT CONSIDERATION: DD7FERENT APPLICATIONS HAVE
ENORMOUSLY DIVERSE REQUIREMENTS
THE SINGLE GREATEST OBSTACLE TO BROADER COMMERCIAL AND
INDUSTRIAL UTILIZATION OF TERAHERTZ REGION IS COST
WE NEED TO MAKE IT CHEAP AS WELL AS GOOD !
I WOULD LIKE TO ACKNOWLEDGE CONSIDERABLE ASSISTANCE FROM
J. BIRCH, P. EFTHIMION, R. GILES, D. KEAVENEY, R. MCINTOSH, E.
MOORE, A. PARRISH, P. SIEGEL, J. WATERS AND OTHER CO-WORKERS
AT MILLITECH AND AT F. C. R. A. 0.
Page 24 Third International Symposium on Space Terahertz Technology
3z-**~ N93-27728
Is
Broadband Millimeter- Wave GaAs Transmitters and Receivers
D Using Planar Bow-Tie Antennas
Y. Konishi*, M. Kamegawa*, M. Case, R. Yu, M. J. W. Rodwell, R. A. York,
and D. B. Rutledget
Department of Electrical and Computer Engineering.
University of California, Santa Barbara
*On leave from Shimadzu Corp. Kyoto, Japan.
^Division of Engineering and Applied Science. California Institute of Technology
Abstract
We report broadband monolithic transmitters and receivers ICs for mm-wave
electromagnetic measurements. The ICs use non-linear transmission lines (NLTL) and
sampling circuits as picosecond pulse generators and detectors. The pulses are radiated and
received by planar monolithic bow-tie antennas, collimated with silicon substrate lenses and
off-axis parabolic reflectors. Through Fourier transformation of the received pulse, 30-250
GHz free space gain-frequency measurements are demonstrated with = 0.17 dB accuracy,
RMS.
Introduction
For mm-wave and sub-mm wave gain-frequency measurements, convenient,
broadband power sources and detectors have been required for some time. Measurement
systems based upon waveguide components (harmonic mixers, frequency multipliers, and
horn antennas)[l] have played a dominant role, but each component has narrowband
frequency coverage (1.5:1). To measure over a broad bandwidth, many waveguide
systems must be used, which is both inconvenient and very expensive. In addition, above
100 GHz it is difficult and expensive to machine the small waveguides and difficult to
attain efficient device-waveguide coupling. Broadband monolithic mm-wave ICs address
these difficulties.
Several groups have reported superconductor devices such as SIS (Superconductor-
Insulator-Superconductor) detectors[2-4] or oscillators [5] for mm-wave measurement or
Third International Symposium on Space Terahertz Technology Page 25
for radio astronomy. Popular devices based on niobium technology (e.g., Nb/A10x/Nb
junctions) must be cooled to liquid helium temperature, so a large and expensive cooling
system is required. Additionally, due to the very low impedance of superconducting
devices (» 0.1 Q), impedance matching to a 50 Q system is difficult
Antenna-coupled picosecond photoconductors have also been used to generate and
detect picosecond radiated electromagnetic pulses. Though Fourier analysis of the received
signals, several groups have recently demonstrated broadband spectroscopy (~ 50 GHz-
1.5 THz) [6-8]. Such systems require expensive and complex mode-locked lasers (=
$150,000) to excite the photoconductors, and the radiated power is extremely small.
As with the photoconductive systems, our system for mm-wave measurements
radiates and detects picosecond pulses and obtains frequency information through Fourier
transformation. Our system uses solid-state monolithic devices, NLTLs and sampling
circuits for pulse generation and detection[9-ll]. With the NLTLs, we have several
advantages. First, the system has fewer components and is very compact without the laser
or its optics. Second, there is substantially more radiated power than the photoconductive
system. Third, since the NLTL is driven by a microwave synthesizer and the NLTL input
frequency can be varied by as much as one octave, the system can easily be tuned to any
desired mm-wave harmonic frequency. Finally, the transmitters and receivers are
inexpensive components fabricated on GaAs with a 5 mask process at 3 |im device
geometries. No cooling system is required for GaAs ICs as with the superconducting
devices.
Here we will describe the system, especially the broadband bow-tie antenna and its
optics. We have demonstrated the system performance by spectroscopic measurement of a
thin alumina substrate with accuracy of 0.17 dB RMS and reproducibility better than 0.3
dB from 30 to 250 GHz.
NLTLs & sampling circuits
The NLTL is a ladder network of high impedance transmission line sections
periodically loaded with reversed biased monolithic Schottky diodes serving as voltage-
variable capacitors[9]. The resulting voltage-variation in wave propagation velocity results
in the compression of negative-going wavefronts and the formation of picosecond shock-
waves. The NLTL converts an input 7-14 GHz sine wave to a sawtooth waveform. In on-
wafer measurements, ~ 1.5 ps falltime and « 5 V peak to peak voltage swing has been
attained. NLTL-gated sampling circuits attained similar risetime. Such devices allow
Page 26
Third International Symposium on Space Terahertz Technology
generation and detection of transient signal with » 250 GHz bandwidth. The transmitter
NLTL is typically driven by a 10 GHz + 100 Hz sinusoidal wave from a microwave
synthesizer. This NLTL drives an on-wafer bow-tie antenna. The receiver consists of an
NLTL-gated sampling circuit integrated with a bow-tie antenna. The NLTL which
generates the sampler's strobe pulse is typically driven by a 10 GHz sinusoidal waveform
from a second synthesizer. The resulting sampled 100 Hz IF signal is observed on a
standard oscilloscope.
Antenna and quasi-optical system
In the case of a planar antenna on a dielectric substrate, most of the power is radiated
into the substrate, and is trapped. This causes standing waves and resulting resonances
within the GaAs substrate (er= 13). To avoid this, hyper-hemispherical substrate lenses are
used with the bow-tie antennasf 12, 13].
NLTL
hyper-
hemispherical
lens
attenuators
NLTL-gated
sampling
circuit
//
^
/
off-axis
paraboloidal mirror material or array
under test
Figure 1: Measurement system schematic diagram (left-' transmitter, right: receiver)
The output of the transmitter NLTL is connected by a coplanar waveguide (CPW) feed
line to the feedpoint of the bow-tie antenna. This structure also serves as a balun.
Sawtooth waves generated by the NLTL are radiated from the antenna. The bow-tie
antenna is scale-invariant and has frequency-independent radiation impedance and
frequency-independent far-field radiation patterns as long as its linear dimensions are larger
Third International Symposium on Space Terahertz Technology Page 27
than a free space wavelength. The antenna thus acts as a high pass filter, with the 2 mm
length resulting in a ~ 35 GHz low-frequency cut-off[14]. The 55 |im total width of the
CPW feedline defines a ~ 1.3 THz upper frequency limit for the antenna.
The radiation is extracted through a silicon (er= 11.8, 16 mm diameter) hyper-
hemispherical substrate lens on the back side of the IC. Matching of the IC and lens
dielectric constants is very important. For example, a sapphire lens (er= 9.9) causes
standing waves in GaAs substrate due to the discrepancy in er. This results in substantial
resonances at 60 GHz, 120 GHz and 180 GHz. Compared to hemispherical lenses, hyper-
hemispherical lenses improve the poor numerical aperture of the bow-tie antennas, and
provide defocusing of the parasitic reflections arising at the lens-air interface. In contrast,
hemispherical lenses exhibit strong spherical-mode resonances. The radiated beam is
collimated with off-axis parabolic mirrors, and is focused on the receiver through similar
optics. The antenna system loss, including substrate lenses absorption, coupling loss
between the antenna and the lens etc., is = -20 dB as determined by 10 MHz - 40 GHz
network analysis! 14].
Metal surfaces surrounding the experimental apparatus are covered with microwave
absorber (Emerson & Cuming, FGM-40) to suppress reflections. Additionally, imaging the
transmitter antenna onto the receiver produces a resonant cavity because of reflections at the
air-lens and lens-antenna interfaces. To obtain accurate gain-frequency measurements,
these resonances are suppressed by placing =» 5 dB thin-film metal attenuators on both sides
of the sample under test.
Device Fabrication
The circuits were fabricated on GaAs semi-insulating substrates with a five mask
process at 3 jtm design rules. Schottky diodes are formed on GaAs with a 425-nm-thick
exponentially graded N" active layer with a 2x10*7 cm"3 surface doping and 225 nm
exponential grading constant. Beneath the N" layer, a buried 1 Jim-thick N + layer (6x10*8
cm"3) provides the diode cathode connection. Ohmic contacts to the N + later (the diode
cathode connections) are formed by a 0.5 \un recess etch to the N + layer, a self-aligned
AuGe/Ni/Au liftoff, and subsequent alloying. Proton implantation (masked by 1.6 nm gold
on 1.1 ^im polyimide) provides isolation between diodes and defines Schottky contact
areas. The transmission line sections are implemented in CPW, formed with a 1.1 |J.m
Ti/Pt/Au liftoff; Schottky contacts result where this liftoff intersects unimplanted regions.
With two additional mask steps, air-bridge crossovers are formed.
Page 28
Third International Symposium on Space Terahertz Technology
Results
The received signal (Fig. 2) shows that the sawtooth waveform has changed to a pulse
train with initial fast rise and a decay time set by the antenna system's low-frequency cut-
off. The peak-peak amplitude is 167 mV, and the pulse risetime is 2.6 ps as limited by the
speed of sampling circuits, the NLTL, and the antenna system.
Because the far-field radiation pattern is frequency-independent, the antenna effective
aperture size is proportional to A 2 . Consequendy, misalignment selectively attenuates high-
frequency components and limits the system bandwidth. With poor alignment, the pulse
risetime degrades due to the reduced bandwidth.
>
E
3
Q.
O
>
'<D
O
<D
cr
-250 -
-300
2.6 ps risetime, 1 0%-90%
1 67 mV peak-peak
-i — i — i — i — | — i — i — i — i — | i i — i — i — | — i — i — i i
5 10 15
time, ps
Figure 2: Received waveform.
20
To demonstrate the system accuracy, we measured the insertion loss of a 254 p.m
alumina substrate (er= 9.9). From 30 to 250 GHz the measurement values correspond well
to theory. (Fig.3) With three subsequent measurements, the accuracy attained was 0.17 dB
RMS, and the reproducibility was better than 0.3 dB.
Third International Symposium on Space Terahertz Technology
Page 29
I I I I L_l I L.
_1 I I ■ i i
50 100 150 200 250
Frequency.GHz
Figure 3: mm-wave measurement of 254 urn-thick alumina test sample.
S -10
CO
03*
CO
o
c
o
CD
CO
c
-20 -
-30 -
-40 -
-50
J. i i_i i L_i i_
_L
a a
• o •
x
o first measurement
x second measurement
* o
x »
Ox £
x
x *
o -
-i 1 1 1 1 1 1 1 1 j—r—r—i 1 1 1 1 1 r-
50 100 150 200
Frequency.GHz
Figure 4: mm-wave measurement of microwave absorber.
We also measured the insertion loss of a microwave absorbing material (Emerson &
Cuming FGM-40, 1.0 mm thickness ). (Fig.4) A loss minimum is seen at 60 GHz, with
the attenuation improving at higher frequencies.
Page 30 Third International Symposium on Space Terahertz Technology
Above 150 GHz, this measurement is limited by the ~ 35 dB system dynamic range.
This dynamic range can be gready improved by using narrowband signal detection (e.g., a
lock-in amplifier).
Conclusion
We have demonstrated a simple and inexpensive system for broadband mm-wave
electromagnetic measurements. Reproducible, accurate measurements are possible from 30
to 250 GHz. The combination of the bow-tie antennas and the substrate lenses provides
acceptable coupling efficiency over a broad bandwidth, despite the high systems loss («
-20 dB between antennas) and the additional (= 10 dB) attenuation required to suppress
standing waves. The bow-tie antenna is readily integrated with monolithic circuits.
The current system will allow convenient and accurate measurement of materials and
emerging mm-wave quasi-optical amplifier arrays. With attainable improvements in the
diode cut-off frequency, system bandwidth can potentially be extended to 1 THz.
Acknowledgment
This work was supported by the Air Force Office of Scientific Research under grant
number (AFOSR-89-0394)
References
[1] Tektronix, Inc. 1991 Catalog
[2] L. R. D'Addario, "An SIS mixer for 90-120 GHz with gain and wide bandwidth", Int. J. of IR and MM
waves, Vol. 5 , No.ll, pp. 1419-1433, 1984.
[3] T. H. BQttgenbach, R. E. Miller, M. J. Wengler, D. M. Watson, T. G. Phillips, "A Broadband Low
Noise Receiver for Submillimeter Astronomy", IEEE, MTT-S. Digest, pp. 469-472, 1988.
[4] S. Kodaira, J. Inatani, K. Sakai, T. Fukushima, "Phase Locking of SWL Array Junctions in
Submillimeter Mixing", Jpn. J. Appl. Phys. Vol. 29, No. 3, pp. L463-L465, March, 1990.
[5] J. Inatani, Y. Konishi, K. Sakai, and S. Kodaira, "Flux-Flow Oscillator connected with a Bow-Tie
Antenna", ISEC, Tokyo, June, 12-13, 1989.
[6] D. H. Auston and M. C. Nuss, " Electro-optic generation and detection of femtosecond electrical
transients", IEEE, Quantum Electron., Vol. 24, pp.184-197, 1988.
[7] G. Arjavalingam, Y. Pastrol, J. M. Halbut and G. V. Kopcsay, "Broad-band microave measurements
with transiet radiation from optelectronically pulsed antenna", IEEE, Trans. MTT., Vol. 38, No.5, pp. 615-
621, May, 1990.
Third International Symposium on Space Terahertz Technology Page 31
[8] N. Katzenellenbogen and D. R. Grischkowsky, "Efficient generation of 380 fs pulses of THz radiation
by ultrafast laser pulse excitation of a biased metal-semiconductor interface", Appl. Phys. Lett., Vol.58,
No.3, pp. 222-224, January, 1991.
[9] M. J. W. Rodwell, M. Kamegawa, R. Yu, M. Case, E. Carmen, K. S. Giboney, "GaAs Nonlinear
Transmission Lines for Picosecond Pulse Generation and Millimeter-Wave Sampling", IEEE, Trans.
MTT., Vol. 39, No.7, July, 1991.
[10] R. Yu, M. Case, M. Kamegawa, M. Sandram, M J. W. Rodwell and A. Gossard, "275 GHz 3mask
Integrated Sampling Circuit", Elect Lett., Vol. 26, No. 13, pp. 949-951, June, 1990.
[11] R.A. Marsland, C. J. Maden, D. W. Van Der Weide, M. S. Shakouri, and D. M. Bloom, "Monolithic
Integrated Circuits for MM-Wave Instrumentation", in Technical Digest, GaAs IC Symposium, New
Orlens.La. October, 7-10, 1990.
[12] D. B. Rutledge, D. P. Neikirk, and D. P. Kasilingam. "Integrated-Circuit Antenna" in Infrared and
Millimeter Waves, K. J. Button, Ed., Vol. 10, pp. 1-90, New York: Academic Press, 1984.
[13] R. C. Compton, R. C. McPhedran, Z. P. Popovic, G. M. Rebeiz, P. P. Tong and D. B. Rutlegde,
"Bow-Tie Antennas on a Dielectric Half-Space: Theory and Experiment", IEEE, Trans. Antenna Propag.,
AP-35, pp. 622-631, June, 1987.
[14] M. Kamegawa, Y. Konishi, M. Case, R. Yu, and M. J. W. Rodwell, "Coherent, Broadband
Millimeter-Wave Specroscopy Using Monolithic GaAs Circuits", LEOS Summer Topical Meetings,
Newport Beach, July, 24-26, 199 1 .
Page 32 Third International Symposium on Space Terahertz Technology
ACTIVE CPW-FED SLOT ANTENNAS FOR POWER
COMBINING APPLICATIONS
Brian K. Kormanyos and Gabriel M. Rebeiz
NASA/Center for Space Terahertz Technology
Electrical Engineering and Computer Science Department
University of Michigan . . "
Ann Arbor, MI 48109-2122 ^93-27729
ABSTRACT
We have combined integrated circuit antenna technology with microwave oscillator design
to build an active slotroscillator. The design is planar, does not require via holes and
is compatible with monolithic transistor technology. The CPW-fed antenna impedance is
calcualted using a full-wave analysis technique. Slot-oscillators were built at 7, 13, and 22
GHz and the predicted oscillation frequencies agree well with experiments. The design is
easily scaled to millimeter-wave frequencies and can be extended to power combining arrays.
INTRODUCTION
Millimeter-wave systems are becoming increasingly important in many military and com-
mercial applications. Millimeter-wave receivers and transmitters have been traditionally
waveguide-based systems and these are expensive to build at these frequencies. To solve
this problem, several groups researched quasi-optical power combining topologies and active
antennas [1-4]. In this paper, we present a novel active transmitter suitable for low-cost
millimeter- wave applications. The transmitter consists of a cpw-fed slot antenna (or a dual-
slot antenna) on a high-dielectric substrate- lens and a three- terminal device (millimeter- wave
HEMT). The novelty in this approach is that we use the antenna impedance, calculated by
a full-wave analysis method, as a parameter in the design of the oscillator. This results in
a much more compact circuit than the conventional approach which consists of an oscillator
with a 50Q output that is matched to a radiating slot-antenna. In our design, the matching
network is eliminated (or minimized), the circuit is much smaller than a wavelength and this
Third International Symposium on Space Terahertz Technology Page 33
allows the design of a power combining array without trigerring grating lobes. The design
can be easily scaled to millimeter wavelengths when HEMT transistor technology is available
at these frequencies.
OSCILLATOR DESIGN AND MEASUREMENTS
The oscillator design is based on the S-parameters of the transistor used. An indefinite
scattering matrix is employed so that short circuited lengths of CPW may be placed at the
gate and source. Computer optimization is then applied to the lengths of CPW to maximize
the reflection coefficient at the drain of the device. In this way a reflection magnitude greater
than one is obtained without the use of an external feedback network and its associated
complications. A slot antenna is connected to the drain through a length of CPW. In order
for oscillations to build up, the impedance the slot antenna presents to the drain must have a
reflection coefficient magnitude at least as large as the reciprocal of the reflection coefficient
at the drain and the phase must be opposite in sign. The impedance of the CPW fed slot-
antenna on a substrate must be well known and is calculated by a full wave moment method
analysis. The terminals of the FET are DC isolated from eachother to allow bias voltages to
be applied. This is done by integrating metal-insulator-metal capacitors and bypassed slits
in the ground plane.
Slot-oscillators were designed and built at 13GHz and 22GHz (Fig 1.) using commercially
available hetero-junction FETs (NE32100, NE32184). The circuits oscillated near the pre-
dicted frequency when placed at the focus of a one inch diameter elliptical silicon substrate
lens (Fig. 1). The radiation patterns of the oscillators on the substrate lens were measured
(Fig. 2) and are used to estimate the directivity. Total oscillator power is calculated with
the radar equation. The total radiated power measured was 5.4mW at 13.01GHz and 3mW
at 22.45GHz. The DC to RF efficiency is 5.4% at 13.01GHz and 3.8% at 22.45GHz. These
numbers are consistent with the capability of the transistor which is a low noise small signal
devices operated at maximum bias. In the future medium power transistors will be used.
Page 34
Third International Symposium on Space Terahertz Technology
0.312mm
0.25mm
/, mmj^
3.38mm
NEC I „
G I NE32184 I D
S^
3.96mm-
0.073mm
S tarn >
1.192 /
mm •
0.122mm
1.758
ezzzzzzzzzzzzz
IJ-a
ZZZZZZZZZZZZZ21
* E3
1.32
2ZZZ2ZZZZZZZZ
NEC
X
/
NE32100 ^ S
/
802mm — •■ %£
1
Figure 1: 13GHz and 22GHz slot-oscillator designs.
1 1 ii i t ii 1 1 1 1 1 1 ii 1 1 ii i u-y i Hi" m ii 1 1 ii 1 1 ii 1 1 ii i ii 1 1 ii i
CD
a -io
a
o
4)
-15
-20
-25
-30
-Pol (E)-
-pol (H>-
■-- X-Pbl (E) "
x-Poi (H) :
' I l l II I l I I I I I II I 111 I Ill III I Ml I
I I I 1 1 I I I I I M l | I II I IL^U !» ^AH M I I | I I I I I I I I I |l I I I I I I I I
-60 -40 -20 20 40
Angle (degrees)
60
-20 o 20
Angle (degrees)
Figure 2: Radiation patterns of 13GHz and 22GHz slot-oscillators on one inch diameter
silicon substrate lens.
Third International Symposium on Space Terahertz Technology
Page 35
A 7GHz VCO (Fig 3.) was designed using the above method with the incorporation of varac-
tor diodes (Metelics MSV34-60-E28) at the source terminals of the FET. An oscillator tuning
range of 850MHz was achieved from 6.68GHz to 7.35GHz. This shows that electronically
tunable slot-oscillators are possible for phase locked loops or other applications.
0.25mm
T
VA ° 5 P3
// mm £
%
mm
0.776 // //
Z
g g
0.312mm
1
mm /,
2.0
G I
Metelics
MSV34-60-E28
Figure 3: 7 GHz VCO with 850 MHz tuning range
The oscillators are well suited for use in power combining arrays synchronized by the mutual
coupling between antennas. One possible array configuration (Fig. 4) is to place a two
dimensional array on a dielectric block of quarter wavelength thickness. Most of the power
will radiate out the opposite side of the block. A weak substrate mode will exist in the
block and may enhance the mutual coupling. If neccessary a reflector may be used on the
back side of the block to further enhance mutual coupling and improve phase equalization
between elements.
Page 36
Third International Symposium on Space Terahertz Technology
X/4
Reflector
\
\ Slot-Oscillators
TMO
k *
Dielectric Block
T t t Y
adiated Power
Figure 4: Possible slot-oscillator array configuration
ACKNOWLEDGEMENTS
This work is supported by the AF/ Rome-Air Development Center and by the NASA/Center
for Space Terahertz Tehnology at the University of Michigan. We thank Prof. Linda Katehi for
providing us with the full- wave solution of a cpw-fed slot antenna on an infinite dielectric substrate.
REFERENCES
[1] J.W. Mink, "Quasi-Opical Power Combining of Solid State Millimeter wave Sources," IEEE
Trans. Microwave Theory Tech., vol. 34, pp. 273-279, Feb. 1986
[2] Z.B. Popovic, R.M. Weikle, M. Kim, and D.B. Rutledge, "A 100 MESFET Planar Grid oscilla-
tor," IEEE Trans. Microwave Theory Tech., vol. 39, pp. 193-199, Feb. 1991
[3] R.A. York, and R.C. Compton, "Quasi- Optical Power Combining Using Mutually Synchronized
Oscillator Arrays," submitted to IEEE Trans. Microwave Theory Tech., Oct. 1990
[4] N. Camilleri and T. Itoh, "A Quasi-Optical Multiplying Slot-Array," IEEE Trans. Microwave
Theory Tech., vol. 33, pp.1189-1195, Nov.1985
Third International Symposium on Space Terahertz Technology Page 37
2x2 QUASI-OPTICAL
POWER COMBINER ARRAY
AT 20 GHz N9 3-^53^30
/6>oSl3
Shigeo Kawasaki and Tatsuo It oh fL %
Department of Electrical Engineering
University of California
Los Angeles, CA 90024
ABSTRACT
Investigation of a power combiner made of two FET oscillators for an active array
1 are reported. As an approach by a quasi-optical method, a two-dimensional planar array of
strongly coupled oscillators by direct connection through a microstrip line is used. In-
phase condition between the oscillators as well as in-phase condition of each radiation wave
was accomplished by regulating length of feed microstrip lines. The radiation elements of
lX.-slot are embedded in a circuit ground plane. At an operation frequency of 20 GHz, in
both H- and E-planes, reasonable L radiation patterns were obtained which have good
agreement with theoretical patterns.
INTRODUCTION
Quasi-optical circuits have been of growing interest for compact and simple
microwave and millimeter-wave systems. Among many solid-state devices, FETs are
preferred as an active source for applications based on the MMIC technologyfl].
However, due to low efficiency, individual FET has limited capability of the power
generation at higher operating frequencies. Therefore, a power combining technique is
Page 38 Third International Symposium on Space Terahertz Technology
essential for high power and high frequency systems. Further, the need for such systems
has resulted in a variety of power combining configuration using FETs.
As one of such power combiners, the planar grid oscillator using MESFETs has
been developed in a distributed fashion[2]. Also, the weakly coupling oscillator array in a
periodic fashion supported by a reflector element has been reported[3]. Recently, using
external injection locking, the power combiner array with feedback FET oscillators was
made[4]. However, due to a multimoding problem, the power combiner array with strong
coupling has not made great progress to date. Regarding linear spatial power combiner
arrays with strong coupling, we have already reported a few results obtained under the
stratified structure[5],[6]. Through these results, in-phase operation for high combining
efficiency was facilitated.
In this paper, we describe design and experimental results of a quasi-optical spatial
power combiner array made of two negative resistance HEMT oscillators and 4 slots.
Radiation patterns from the slots aligned as 2x2 were compared with theoretical patterns in
both H-plane and E-plane. Further, through these patterns, it was confirmed that the
radiation waves from each slots were in-phase.
DESIGN
In order to demonstrate topology for the MMIC technology, the circuit was made in
a layered structure by adopting slot radiators in a circuit ground plane. Fig. 1 shows the
configuration of 2x2 spatial power combiner array designed at 20 GHz. The circuit
structure made of two negative resistance oscillators and feed lines is etched on the top side
of the substrate, while the slot radiators shown by cross-hatched sections in Fig. 1 are
embedded in the ground plane of the bottom side of the same substrate. Each oscillator
was designed with -50 Q for an input impedance. Since both sides of the substrate can be
effectively used, the slot radiators in the circuit ground plane increase design flexibility.
Third International Symposium on Space Terahertz Technology
Page 39
RF energy generated from one FET oscillator is divided and delivered to two slots
aligned in the E-plane. In addition, a part of this energy is used to lock the other oscillator
through a direct coupling between the two oscillators. The locked and amplified signal
flows into another two slots aligned in the E-plane as well. Each pair of the slots aligned in
the E-plane are connected in series. Therefore, this 2x2 array consists of two slots aligned
in the E-plane and two slots aligned in the H-plane with W separation between the centers
of two slots. The direct connection of two slots in H-plane separated by a thin line
enhances high packaging density.
1
u
I
SSS¥
m
11
jsr
D
FET • CU
a
n
3-
2Zc
B
Zo
J
FET
□
V/.W
.vX;.y
1L
i+4
T
1L
Slot Radiator
(Bottom Side)
*- X/4
Transition
Transformer
/2Zo
Fig. 1 Circuit Configuration
Page 40 Third International Symposium on Space Terahertz Technology
According to the previous results about the direct connection of oscillators[6], when
the distance between two branch-points from the oscillators (shown as A and B in Fig. 1)
is an odd number of half wavelength, the radiation waves from the two branching points A
and B are in-phase. Since the length of a straight line between A and B is IX, a microstrip
line with X/2 should be added. As a result, the radiation waves from the slots in H-plane
can have the same phase.
The distance between two centers of slots is IX. Thus, the divided energy from
one branching point A or B can reach the centers of both slots at the same time. This
results in an anti-phase radiation. For the in-phase condition, one of these pass must have
additional X/2 to invert the phase on one of the slots. By means of this control of pass
length between an oscillator and a slot as well as between two oscillators, the in-phase
condition was obtained in both H-plane and E-plane.
In addition to the consideration of phase condition, input matching condition needs
to be taken into account. A one-stage X/4 transformer is inserted from a center of a slot
toward a branched feed line connected to the branch-point A or B. In order to
electromagnetically couple a feed line with a slot radiator, a X/4 microstrip-to-slotline is
adopted.
On the other hand, the slot radiators were also designed at 20 GHz with a IX, slot
length and a 0.081 X slot width. Since input impedance of the center feed \X slot radiator
provides the circuit with 50 Q load, the slot and the 50 Q feed line are matched in the
steady state oscillation condition.
EXPERIMENTAL RESULTS
The circuit were fabricated on 31 mil thick substrate with £,-=2.33, using NEC
NE32484 package HJFETs. The negative resistance FET oscillator was optimized at 21
GHz with 5 % margin by using small signal S parameters so that the actual circuit can also
Third International Symposium on Space Terahertz Technology
Page 41
be operated at 20 GHz. The maximum operation frequency of 19.5 GHz was observed
from the fabricated circuit as shown in Fig. 2.
It is found that, in this two-
device power combiner case,
a frequency margin of 7.7 %
is required under the design
condition of -50 Q for an
input impedance, in reference
[1], we investigated the
frequency margin of a single
quasi-optical oscillator with a
single slot using the same
v
NE32484. In this case, the
frequency margin of 4.4 %
was obtained.
Compared with this previous results, the difference between the operation frequency and
the design frequency is increased. This implies that, in the case of a large number of FET
combiners, accumulation of a difference of input impedances due to such frequency
difference becomes a cause of load pulling, and, due to this phenomenon, the frequency
difference is increasingly enhanced. The tuning range obtained by changing the applied DC
voltages, V ds (2.6-5.0 V) and V gs (-1.0—1.6 V), was 121 MHz.
Radiation patterns in both the H-plane and the E-plane are shown in Fig. 3. In both
cases, it is easy to find out null points around ±30° which result from array factor. For
comparison, the theoretical analysis was carried out by using the point matching method to
obtain numerical data from the Pocklington type integral equation. In the calculation, the
voltage ratio of two source generators to excite each slot was set to 1, while the phase
difference of these generators was set to 0. Using these data, theoretical radiation patterns
Fig. 2 Operation Speetrum
Page 42
Third International Symposium on Space Terahertz Technology
10 -i
m
•a
a>
3
o
a.
a>
>
Angle (degree)
experiment
theory
100
Experiment
(2x2 H-plane)
Vds=4.0 V
lds=45 mA
Vgs=-1.81 V
fo=19.14 GHz
Theory
Pockilngton I.E.
Voltage Ratio=1.0
Phase Difference=0
Comparision in Radiation Pattern (H Plane)
m
■o
o
a
>
a>
10-1
0"
-10-
-20
100
100
Angle (degree)
experiment
theory
Experiment
(2x2 E-plane)
Vds=4.0 V
lds=40 mA
Vgs=-1.82 V
f0=1 9.15 Ghz
Theory
Pocklington I.E.
Voltage Ratio=1.0
Phase Difference=0
Comparision in Radiation Pattern (E plane)
Fig. 3 Radiation Patterns
Third International Symposium on Space Terahertz Technology Page 43
are obtained as shown in Fig.3 for both the H-plane and the E-plane. Agreement around
the main beams is good.
Sidelobes in the E-plane becomes lager than those in the H-plane because of a
single element factor. This fact is shown in the theoretical patterns in Fig. 3. However, the
sidelobes obtained from the experiment were still large in both H-plane and E-plane. These
discrepancies may result from the inadequate experimental setup.
CONCLUSIONS
Design and experimental results of a 2x2 quasi-optical spatial power combiner was
reported as a prototype of two dimensional quasi-optical power combiner array. Although
the impedance matching conditions as well as the in-phase condition have been taken into
account carefully, the difference of the operation frequency (19.2-19.5 GHz) to the design
frequency (20 GHz) has increased. In the case of a large array, more attention should be
paid to avoid increase of this frequency difference resulting in phenomena such as load
pulling.
In both the H-plane and the E-plane of a 2x2 quasi-optical power combiner array, £
radiation patterns were obtained by controlling the lengths of feed lines. Good agreement
between the experiment and the theory was obtained about mainlobes in both radiation
patterns.
It is believed that, through this prototype circuit, the fundamental data for a
monolithic quasi-optical power combiner array were obtained.
ACKNOWLEDGEMENT
This work was supported by US Army Research Office under contract DAAL 03-
88-K-0005.
Page 44 Third International Symposium on Space Terahertz Technology
REFERENCES
[1] S. Kawasaki and T. Itoh, "24 GHz FET Oscillator with Slot Antenna for Quasi-Optical
Transmitter", 16th In't Conf. on IR & MMW, Switzerland, Aug. 1991, pp 286-287
[2] Z.B. Popovic, R.M. Weikle II, M. Kim and D.B. Rutledge, "A 100-MESFET Planar
Grid Oscillator", IEEE Trans. Microwave Theory Tech., vol. 39, pp 193-200, Feb.
1991
[3] R.A. York and R.C. Compton, "Quasi-Optical Power Combining Using Mutually
Synchronized Oscillator Arrays", IEEE Trans. Microwave Theory Tech., vol. 39, pp
1000-1009, June 1991
[4] J. Birkeland and T. Itoh, "Two-Port FET Oscillators with Applications to Active
Arrays", IEEE Microwave and Guided Wave Lett., vol.1, ppl 12-113, May 1991
[5] S. Kawasaki and T. Itoh, "40 GHz Quasi-Optical Second Harmonic Spatial Power
Combiner Using FETs and Slots", to be published in 1992 IEEE MTT-S Int'l
Microwave Symposium, Albuquerque, NM.
[6] S. Kawasaki and T. Itoh, "6-Element Periodic and Nonperiodic Linear Arrays for
Quasi-Optical Spatial Power Combiner", to be published in URSI Radio Science
Meeting, Chicago IL, July 1992
Third International Symposium on Space Terahertz Technology
Page 45
MONOLITHIC MILLIMETER- WAVE DIODE ARRAY
BEAM CONTROLLERS: THEORY AND EXPERIMENT
N93-27731
L. B. Sjogren, H-X. L. Liu, F. Wang, T. Liu, W. Wu, X-H. Qin, E. Chung, %~~.33
C.W. Domier, N. C. Luhmann, Jr.
Center for High Frequency Electronics /(pO^l^f
Department of Electrical Engineering
University of California p /J
Los Angeles, California 90024-1594 I "
J. Maserjian
Jet Propulsion Laboratory
Pasadena, California 91109
M. Kim, J. Hacker, D.B. Rutledge
California Institute of Technology
Pasadena, California 91125
L. Florez, J. Harbison
Bellcore, Inc.
Red Bank, New Jersey 07701-7030
~\
X
y H
I. Introduction
Power-combining arrays of semiconductor devices offer a promising approach to
the realization of compact, reliable, economical systems for watt-level operation at
millimeter- wave and submillimeter- wave frequencies. Such ("grid") arrays have demon-
strated numerous functions at microwave and millimeter-wave frequencies in recent
experimental efforts. Monolithic diode arrays have demonstrated phase shifting at 93
GHz [1], frequency doubling from 33 to 66 GHz [2] and frequency tripling from 33 to
99 GHz [3]. One dimensional monolithic imaging arrays have been demonstrated at
94 GHz [4]. Additional quasi-optical functions have been demonstrated at microwave
frequencies by arrays employing hybrid technology. These include the oscillator grid
[5-6], mixer grid [7], and amplifier grid [8]. The hybrid grids operate on the same
basic quasi-optical principle as the monolithic arrays, so the functions demonstrated
to-date in hybrid form should be feasible also in a monolithically integrated form at
millimeter- wave frequencies. Additional array design approaches have been suggested
for further development of millimeter- wave components [9],
Construction of complete systems based on the millimeter- wave array technology
requires not only source and detector arrays, but control components for such func-
tions as amplitude modulation, phase modulation, and beam steering. The first effort
y
Page 46 Third International Symposium on Space Terahertz Technology
at addressing this need by a semiconductor device array was an experimental demon-
stration of a phase shifter at 93 GHz [1]. In this work, phase control of a reflected
beam over a range of 70 degrees with 6.5 dB loss was achieved by a monolithic array
of Schottky diodes. The stacking of more than one array [1] should allow a phase
range of greater than 360 degrees to be achieved. Operated in a nonuniform phase
(bias) mode, such an array should be capable of (phased array) beam steering and
beam focusing.
In the current work, multi-function beam control arrays have been fabricated,
and have successfully demonstrated amplitude control of transmitted beams in the
W and D bands (75-170 GHz). While these arrays are designed to provide beam
control under DC bias operation, new designs for high-speed electronic and optical
control are under development. These arrays will fill a need for high-speed watt-level
beam switches in pulsed reflectometer systems under development for magnetic fusion
plasma diagnostics.
A second experimental accomplishment of the current work is the demonstration
in the 110-170 GHz (D band) frequency range of a new technique for the measurement
of the transmission phase as well as amplitude [11]. Transmission data can serve as a
means to extract ("de-embed") the grid parameters; phase information provides more
complete data to assist in this process.
Additional functions of the array beam controller yet to be tested include electron-
ically controlled steering and focusing of a reflected beam. These have application in
the areas of millimeter- wave electronic scanning radar and refiectometry, respectively.
II. Theory and Design
The beam control device consists of a monolithic two-dimensional array of varactor
diodes embedded in metal strips. Its quasi-optical behavior is represented by a shunt
impedance across a transmission-line representation of the beam. The impedance is
that of a series RLC circuit. The inductance, some fixed (undesired bias-independent)
capacitance, and a small resistance are due to the metalization grid structure; the
bulk of the capacitance is due to the diode; the bulk of the (undesired) resistance is
a parasitic effect of the diode and its Ohmic contact.
The initial objective of the current effort was the design of the proposed stacked
360 degree reflection phase shifter array [1]. In this approach, an effective load re-
actance of arbitrary value is obtained by the use of two diode grids, each of which
possesses a reactance range of at least -ZoaAs to Za a Aa, where Za a Aa is the plane wave
impedance of Gallium Arsenide (approximately 105 Cl), and the grids are separated
by an odd multiple of A/4.
The capability of the array to control beam transmittance is based on its ability
to be switched, under bias control, between a low and high impedance state. When
the array is biased at resonance, it appears to the beam as nearly a short circuit,
and reflects most of the beam back. When the array is biased far from resonance
(i.e. at high impedance), the beam is affected very little by it. In this case, the beam
transmittance is large.
Third International Symposium on Space Terahertz Technology Page 47
The dimensions and doping profile for the monolithic Gallium Arsenide Schottky
varactor diode were determined with the assistance of a finite difference solution
program for Poisson's equation in one dimension. This provided an estimate of the
diode C-V characteristics. Additional routines were employed to estimate tunneling,
avalanche, and thermionic emission currents. A hyperabrupt doping profile, similar to
that employed by W. Lam [1], was chosen to provide linearity of reflection phase versus
DC bias. Different doping levels were simulated to optimize the capacitance range
and breakdown voltage. A heterojunction barrier, pioneered at JPL for frequency
multiplier diodes, has been incorporated to suppress tunneling current under reverse
bias and thermionic emission under forward bias. The experimental results in this
paper were obtained with arrays fabricated from MBE wafers with the profile shown
in Fig. 1.
Design of the array requires the theoretical prediction of the electromagnetic (grid)
behavior as well as the C-V characteristics of the embedded varactor diode. For the
passive (electromagnetic) design, a simple method of moments analysis has been em-
ployed. This has provided a new model which includes the effect of the discontinuity
of the current at the site of the diode ("grid capacitance") [12] (Fig. 2). This model
somewhat underestimates the fixed (bias-independent) capacitance at the diode site,
since it idealizes the current discontinuity as that for a "gap" at the diode site when
the diode is (analytically) open circuited. Even so, the capacitive effect was found
to be considerable (5-10 fF for typical array geometries at 100 GHz), and sufficient
to reduce the simulated phase range of the stacked phase shifter from well above
360 degrees (with the "grid capacitance" not included) to a value considerably below
360 degrees (with it included). A rigorous simulation of the C-V characteristics of
the grid would require a three dimensional simulation which includes the effect of
the anode "finger" , absence of dielectric beyond the etched walls of the diode, etc. A
more precise determination of the grid inductance should be possible with the Hewlett
Packard High Frequency Structure Simulator electromagentic analysis program. Such
an analysis may provide a more precise estimate of the grid capacitance.
The grid impedance model and diode model programs were combined and incor-
porated into a quasi-optical (transmission line) circuit analysis program based on
the application of Kirchoff's laws at the dielectric (and grid) interfaces. Simulations
were performed to determine unit cell and strip dimensions which maximized the grid
impedance range. The grid capacitance effect was suppressed by use of a "rectangular
unit cell", in which diodes are effectively placed "in series" by spacing them closer
along the axis of current flow. Such a configuration has the additional benefits of
allowing use of a large (easier to reliably fabricate) diode and providing significantly
higher power-handling capability. The one drawback of this design is a somewhat
higher loss, since the effective diode resistance is a/b times the actual diode resis-
tance. The increased size of the diode will partially, but not completely, offset this
effect. A possible alternative for the reduction of grid capacitance is to employ a
narrower metal strip in the vicinity of the diode [13]. This was not feasible for the
current design, since the strip was too narrow to taper.
The simulations provided the optimized unit cell dimensions a=300 /xm, b=120
fxm, and w=7 fxm, with diode dimensions of 3 /xm x 13 fim. The diode area assumed
Page 48
Third International Symposium on Space Terahertz Technology
Al Ga As
O.S 0.5
(undoped)
Staircase approximation
to hyperabrupt profile
1S0A
5.0 E17 cm-3
100 A
4.0 El 7 cm-3
150 A
3.0 E17 cm-3
250A
2.0 E17 cm-3
350A
1.5 E17 cm-3
SOOA
1.0 El7cm-3
700A
7.5 E16 cm-3
1600A
5.0 E16 cm-3
3800A
total hyperabrupt
hyperabrupt n GaAs
2000A
2200A
6000A
Semi-insulating
GaAs substrate
16000A
Figure 1: MBE profile of the beam control array.
ei
r ^
-7
b *- diode
\
=T<4
Figure 2: Electromagnetic model of the diode array.
Third International Symposium on Space Terahertz Technology Page 49
in the simulations is 26 /im 2 , since the effective diode width after fabrication should
be about 1 /im less than the drawn dimension. Due to the small cell dimensions, the
device count per array is very high, about 12,000 for a full sized (2.5 cm x 1.8 cm)
array. To allow reasonable spacing between wirebonds, the array was layed out so
that each bias line connects to eight contiguous rows of 60 diodes each. Thus, each
bias line connects to 480 array diodes.
As previously stated, the array should show some capability for electronic beam
steering and focusing. The array model cannot predict the extent of these capabilities,
however, since it is applicable only to uniform array operation.
III. Fabrication and Testing
Array fabrication is based (with some modifications) on the self-aligned Aluminum
Schottky diode process developed by C. Zah for millimeter- wave imaging diodes [4].
Most of the processing was performed at the J PL Microdevices Laboratory, a facility
with state-of-the-art microfabrication capability.
Device isolation was performed, as in [4], by proton implantation, with an im-
plant mask of thick photoresist. A two-step implant of 4xl0 14 cm~ 2 at 200 keV and
4xl0 14 cm~ 2 at 100 keV provided good isolation. However, photoresist edge bead
resulted in the corners of the arrays being unisolated. Several of the arrays were
therefore re-implanted to isolate these areas. However, the capacitance range of the
diodes was greatly reduced after the second implant. It thus appears that some of
the implant penetrates the photoresist mask and into the active device. The arrays
which were not re-implanted had some unuseable areas due to the edge bead problem,
but have proved sufficient for the experimental proof-of-principle testing.
Due to the parallel connection of array diodes, short circuited devices must be dis-
connected from the array. An HP4145 Semiconductor Parameter Analyzer, HP9836
controller, and Electroglas 1034 wafer prober were combined into a system which
allowed automated testing and storage of I-V characteristics for every array device.
Device probing was facilitated by an extra metalization step which creates probe pads
connected to each device prior to grid (bias) metalization. With the short-circuited
devices identified, a microprobe was used to sever their connection to the array. Fol-
lowing final metalization, very few of the bias lines were short-circuited. Therefore,
further identification of short-circuited devices by "hotspot" detection [1] was not
neccessary. Photographs of a single array device and a small region of an array are
shown in Figs. 3 and 4, respectively. The final steps for completion of the array are
the attachment and wirebonding of the array to a printed circuit "bias" board, and
the attachment of bias wires to the board. Series resistors of 220 SI are added to the
bias wires to prevent damage if an array device becomes short-circuited.
TV. Results
Millimeter- wave transmission testing was performed with the system shown schemat-
ically in Fig. 5. A small array (1.8 cm x 1.0 cm, 4800 diodes) successfully demon-
strated transmitted amplitude control throughout the W and D frequency bands
Page 50
Third International Symposium on Space Terahertz Technology
Figure 3: Photograph of a single device in a beam control array.
Figure 4: Photograph of a small section of the beam control array. Unit
cell dimensions are 300 [im x 120 fim. The vertical strips serve as the
"antenna" elements, while the horizontal strips provide bias voltage
to the diodes. The marks on the large rectangular test pads are from
automated device probing.
Third International Symposium on Space Terahertz Technology p „
(75-170 GHz). Results at 99, 132, and 165 GHz are shown in Fig. 6. Substantial
amplitude control was obtained (except near resonance), despite leakage currents on a
number of the bias lines preventing most of the bias from showing up their associated
sections of the array. In addition to amplitude control under DC bias, the array suc-
cessfully demonstrated low frequency (200 kHz) modulation of a 165 GHz beam (Fig.
7). The modulation frequency was limited by the bandwidth of the detected beam
amplifier. Further testing will be performed to determine the maximum modulation
frequency of the array.
To allow further verification of the grid behavior, a technique has been developed
to obtain the phase, as well as amplitude, of the transmitted beam [11]. The method
involves tilting the incident beam, so that a portion of it is aligned orghogonal to the
operational ("active") axis of the grid. The orthogonal beam component is employed
as a reference, with the polarization of the transmitted beam providing the relative
phase of the beam in the active versus orthogonal ("passive") axis. Since the phase
of the passive axis is highly predictable, the absolute phase of the transmitted beam
in the operating axis can be determined by this method. The method was verified
by application to strip arrays, whose theoretical behavior can be well predicted and
compared to the experimental results. The method was then applied to obtain the
transmission coefficient of the beam control array as a function of frequency and bias.
This was done for an 8600 diode beam control array. Estimates of parameters for the
series RLC model of the grid were obtained by varying the parameters until a good
match between the theoretical and experimental transmission curves was obtained.
This prov ided a fairly precise estimate of grid resistance and resonant frequency f rea =
(2iry/LC)~ 1 . For the array tested, the grid resistance is approximately 40 ft over the
entire bias range, and the resonant frequency versus bias is shown in Fig. 8. The
individual value of L (or C) has some range in which the theoretical and experimental
curves agree. This range is centered at an inductance value of L= 160 pH, with the
capacitance ranging from 5.2 fF to 13.9 fF as a function of bias for this value of
L. The results indicate a cutoff frequency for the beam control array f c = (C~} n —
C~\ x )(2irR)~ l of about 400 GHz. It appears that both the grid inductance and
capacitance are considerably lower (about 35 %) than predicted. The deviation in
the inductance value is probably due to the idealization of the inductive effect as that
of a uniform (and narrower than actually fabricated) vertical strip. Simulation by the
High Frequency Structure Simulator will provide a more definitive verification of this.
The deviation in grid capacitance is with respect to the values expected based on 1
MHz C-V measurements of sample diodes from the same wafer. This discrepancy
requires further investigation.
Initial reflection tests have been performed for "calibration" devices (strip grids),
with use of a focusing lens [14] and the polarization technique [11]. The phase of the
reflection coefficient has been successfully obtained by this method. Since the grid
parameters of the currently fabricated arrays are largely now known, the reflection
phase shift for a stack of two of these grids can be well-predicted. Predicted results at
128.3 GHz (at which the thickness of the GaAs can be made the desired odd multiple
of A/4), along with the originally simulated behavior at the design frequency of 99 GHz
rescaled to 128.3 GHz, are shown in Fig. 9. The lower than desired C max /C m ,„ ratio
Page 52
Third International Symposium on Space Terahertz Technology
diode grid
aperture ^
diode
detector
>-w-
i
m
■ i<
• i.
■ i
1
frequency ^
meter absorber ^
plate
5£
Grid bias or
modulation
preamplifier ^^
Figure 5: Schematic diagram of the test system for transmitted beam
testing.
0.8 -
o
c
2 0.6
E
W
C
CO
0.4
0.2
i i • • • r
i i i i i i i i 1 1 i i i i I i i i i i i i i i
i ■ ■ ' ' i
165 GHz
99 GHz
■ 2.5 -2 -1.5 -1 -0.5 0.5 1
Array Bias Voltage (V)
Figure 6: Experimental beam transmittance at 99 GHz, 132 GHz, and
165 GHz as a function of DC bias applied to the array. The basic
form of the curves can be understood by the fact that the array is
capacitive at 99 GHz, resonant at 132 GHz, and inductive at 165
GHz.
Third International Symposium on Space Terahertz Technology
Page 53
AAW
Detected output
Bias modulation
j2us
Figure 7: Detected output versus array input voltage for a sinusoidal
modulation of the array at 200 kHz. The lower waveform is the array
bias modulation voltage, whose range was -3V to +1V. The upper
waveform is the detected output from the transmitted beam.
N
K
O
>>
o
c
3
cr
u
c
a
c
o
i — i — i — i — I — i — i — i — i — 1 — i — i — I — i — i — I — i — i — \ — l — i — I — i — r
-4
-2 -1
Bias (V)
Figure 8: Resonant frequency versus DC bias for the beam control array.
Page 54
Third International Symposium on Space Terahertz Technology
1 I — i — n — r— | — i — i — i — i — [— i — i — i — r
3
Q.
E
<
-i — n — i — I — I — i — i — i — 1 — i — i — i — r
Q I I I I i_l I I I I I I I I i_l i l_j i I i__i i i I ■ ■ ■ ■
-5 -4 -3 -2-10 1
Bias (V)
6.28
CO
c
3.925
CO
■a
CO
k.
1.57
CD
CO
CO
■0.785
•3.14
~ I I I I i i I r
I I I — I — I — I — I — I — I I I I | — I — I — I"
i ■ i i i i
-5 -4
3 -2 -1
Bias (V)
Figure 9: Predicted reflection coefficient for a two layer stacked phase
shifter with fused silica "window" at 128.3 GHz. Lines with no mark-
ers represent the predictions based on originally simulated grid pa-
rameters. Lines with markers represent the predictions based on es-
timates of the grid parameters based on transmission measurements.
Third International Symposium on Space Terahertz Technology Page 55
results in a phase range less than 360 degrees, contrary to the original simulation.
This, however, does not preclude the possibility of beam steering to small angles.
The higher than desired grid resistance results in a reflectance which is lower and is
much more variable over the bias (phase) range. A preliminary attempt at steering
of a transmitted beam with the current array was unsuccessful. This may be due
to an inability of the grid to produce the strong amplitude variation with position
associated with the desired phase distribution. Since steering to a fixed angle has
been successfully demonstrated by a (low-loss) passive grid [15], the beam control
array's ability to steer a reflected beam is likely to be governed largely by the grid
loss as well as phase shift range. To obtain higher performance for the beam control
functions, new arrays are being fabricated which should possess a larger C ma x/C m ,„
ratio and Ohmic contact resistance. In addition, we are considering the stacking
of a large number of arrays of the current design and operating them in the high
impedance region to accomplish transmitted beam steering.
V. Conclusions
A "second generation" millimeter-wave beam control array device has been con-
structed. This array has successfully demonstrated a new function by a millimeter-
wave quasi-optical array, that of beam transmittance control. Phase of the trans-
mitted beam has also been measured by a newly-developed technique. Reflection
measurements, which will test the arrays as phase shifters, beam steerers, and beam
focusers, will be performed soon. In addition, new arrays with a modified doping
profile for higher C max arid lower resistance are being fabricated. These "higher per-
formance" arrays should provide, for example, a greater "contrast ratio" (maximum
to minimum transmittance) when the array is operated as a beam modulator.
New array designs have been initiated for high-speed (under 200 psec) electronic
and optical beam control. For electronic control, the bias lines of the beam control
array are being designed to function as high-speed guided-wave paths. For opti-
cal control, monolithic arrays of photoconductive switch devices are being fabricated.
New concepts for barrier varactor photodiodes are under study for application toward
optically-controlled modulator and beam steering arrays. High-speed beam switch-
ing arrays have immediate application in plasma diagnostic reflectometry, and have
potential additional application for such functions as electronic input beam chopping
in high-speed imaging systems. Longer term possibilities include exciting possibilities
such as amplifying beam steeres with two-axis scan control.
Acknowledgements
Work supported by Northrop Corporation/ University of California MICRO pro-
gram, Department of Energy, and the Army Research Office.
The authors wish to thank the personnel of the JPL Microelectronic Devices
Laboratory for their assistance toward the fabrication of the beam control arrays. In
particular, we wish to thank R. Peter Smith, Suzanne Martin, Chuck Manning, Rich
Muller, Judy Podosek, and Doug Waltman.
Page 56 Third International Symposium on Space Terahertz Technology
Essential to the project has been the availability of high-quality MBE wafers gen-
erously provided by Prof. C. Jou (National Chiao Tung University, Taiwan), Prof.
M. Spencer (Howard University), John Liu (JPL), as well as Bellcore, Inc. Assis-
tance in this area has also been provided by Larry Kapitan (formerly with Northeast
Semiconductor, currently with QED).
The stepping wafer prober, along with technical assistance, was generously pro-
vided by Edith Baltram, Jack Hayden, Frank Freeman, Soo Kok Leng, and Gary
Castleman (Hewlett-Packard, Northwest Integrated Circuits Division).
We gratefully acknowledge the essential assistance in Ohmic contact alloying by
Dr. Marko Sokolich of Hughes.
Indispensible assistance with millimeter- wave measurement systems was provided
by Matt Espiau and Misti Christianson of the UCLA Millimeter- Wave Laboratory.
Additional individuals who provided essential assistance for this project include
Wayne Lam (TRW), Charles Meng, Prof. D.S. Pan, Prof. H.R. Fetterman (UCLA),
Mike DiLisio (Cal Tech), Prof. R.J. Hwu (University of Utah), Clarence Becwar
(Becwar Engineering), and Rene Bernescot (Rockwell).
References
(1) W.W. Lam, C.F. Jou, N.C. Luhmann,Jr., and D.B. Rutledge, "Millimeter-wave
diode-grid phase shifters," IEEE Trans. Microwave Theory Tech., 36, No. 5, p.
902, 1988.
(2) C.F. Jou, W.W. Lam, H.Z. Chen, K.S. Stolt, N.C. Luhmann,Jr., and D.B.
Rutledge, "Millimeter Wave Diode-grid Frequency Doubler," IEEE Trans, on
Microwave Theory and Techniques, 36, No. 11, 1988.
(3) H-X. King, X-H. Qin, W. Wu, L.B. Sjogren, E. Chung, N.C. Luhmann, Jr.,
W.A. Peebles, "Monolithic Millimeter- Wave Quasi-Optical Frequency Multi-
plier Arrays", presented at, 1991 International Semiconductor Device Research
Symposium, pp. 68-72, December, 1991.
(4) C. Zah, D.P. Kasilingam, J.S. Smith, D.B. Rutledge, T. Wang, and S.E. Schwartz,
"Millimeter- wave Monolithic Schottky Diode Imaging Arrays", Intl. J. of In-
frared and Millimeter Waves, 6, pp. 981-997, 1985.
(5) Z.B. Popovic, R.M. Weikle, M. Kim, K.A. Potter, and D.B. Rutledge, "Bar Grid
oscillators," IEEE Trans. Microwave Theory Tech. , 38, No. 3, p. 225-230,
March,1990.
(6) Z.B. Popovic, R.M. Weikle, M. Kim, and D.B. Rutledge, "A 100-MESFET
Planar Grid Oscillator," IEEE Trans. Microwave Theory Tech., 39, No. 2 pp.
193-200, February, 1991.
(7) J.B. Hacker, R.M. Weikle III, M. Kim, D.B. Rutledge, "A 100 Element Schottky
Diode Grid Mixer", 1991 IEEE AP-S symposium digest.
Third International Symposium on Space Terahertz Technology Page 57
(8) M. Kim, J.J. Rosenberg, R. P. Smith, R. M. Weikle III, J.B. Hacker, M.P. DeLi-
sio, D.B. Rutledge, "A Grid Amplifier", IEEE Microwave and Guided Wave
Letters, I, No. 11 pp. 322-324, November, 1991.
(9) R.J. Hwu, C.F. Jou, N.C. Luhmann,Jr., M. Kim, W.W. Lam, Z.B. Popovic, and
D.B. Rutledge, "Array concepts for solid state and vacuum microelectronics
millimeter wave generation," IEEE Trans. Elec. Dev., 36, No. 11, 1989.
(10) L.B. Sjogren, R.J. Hwu, H-X. King, W. Wu, X-H. Qin, N.C. Luhmann, Jr., M.
Kim, D.B. Rutledge, "Development of a 94 GHz Monolithic Quasi-Optical 360
Degree Phase Shifter," Proc. of the 15th Intl. Conf. on Infrared and Millimeter
Waves, pp. 696-698, 1990.
(11) L.B. Sjogren, et. al., "A Technique for the Measurement of Complex Transmis-
sion Coefficient of Millimeter- Wave Grid Arrays", to be submitted, Microwave
and Optical Technology Letters, 1992.
(12) L.B. Sjogren and N.C. Luhmann, Jr., "An Impedance Model for the Quasi-
Optical Diode Array", IEEE Microwave and Guided Wave Letters, 1, No. 10,
pp. 297-299, October, 1991.
(13) H-X. King, L.B. Sjogren, N.C. Luhmann, Jr., D.B. Rutledge, "New Concepts
for High Frequency and High Power Frequency Multipliers and Their Impact
on Quasi-Optical Monolithic Array Design", Int. J. of Infrared and Millimeter
Waves, Feb. 1992.
(14) David R. Gagnon," Highly Sensitive Measurements With a Lens-Focused Re-
flectometer" , IEEE Transactions on Microwave Theory and Techniques, 39, No.
12, pp. 2237-2240, December, 1991.
(15) Moonil Kim, Robert M. Weikle III, Jonathan B. Hacker, David B. Rutledge,
" Beam Diffraction by a Planar Grid Structure at 93 GHz", 1991 IEEE APS
Symposium.
Page 58 Third International Symposium on Space Terahertz Technology
5Z-33
N93~t7732
A Study of Subterahertz HEMT Monolithic
Oscillators *
Youngwoo Kwon and Dimitris Pavlidis
Center for Space Terahertz Technology
Solid State Electronics Laboratory
Department of Electrical Engineering and Computer Science
The University of Michigan, Ann Arbor, MI 48109-2122, USA
^_ " Abstract
A detailed study of monolithic InP-based HEMT oscillators for subterahertz
operation is presented. InAlAs/InGaAs HEMT's have been optimized for high
frequency operation and showed very high maximum oscillation frequencies (f maa . )
of 310 GHz using offset self-aligned T-gate technology. Power characteristics of
HEMT oscillators are reported. An oscillation power of more than 10 mW was
evaluated by large-signal analysis at 320 GHz using HEMT's with f max = 450
GHz, V&r = 10 V and a gate width (W 3 ) of 8 x 22.5 /zm. Oscillator topology
studies showed that complex feedback schemes such as dual and active feedback
enhance the negative resistance. Push- push oscillator designs based on harmonic
signal generation can finally be used to overcome the frequency barrier imposed by
f
1 Introduction
InAlAs/InGaAs HEMT's have shown excellent high frequency characteristics and
operation capability as discrete devices. A current gain cut-off frequency (fy) of 305
GHz [1] and a maximum oscillation frequency of 455 GHz have been reported using
heterostructures of this type [2]. These very encouraging discrete device results indicate
that InP-based HEMT's can be used to realize monolithic circuits with operation
frequency well into the millimeter-wave region. A number of such monolithic integrated
"Work supported by NASA under contract NAGW-1334
Third International Symposium on Space Terahertz Technology Page 59
circuits have recently been demonstrated by the authors. These include monolithic
HEMT mixers at 94 GHz showing conversion gain of 1 dB [3] and HEMT doublers
at 180 GHz with a conversion loss of 6 dB [4]. Monolithic HEMT oscillators also have
been realized by the authors up to W-band showing more than 1 mW power with devices
having 36 jj,m gate periphery [5].
Another possibility opened to HEMT technology is its use for space-based remote
sensing and radiometry, where fundamental sources are required to operate above 100
GHz. A first detailed study concerning the power characteristics and the upper frequency
limit of InAlAs/InGaAs HEMT's when used as oscillators, has recently been presented
by the authors [6]. This paper provides further details on related issues of HEMT use for
signal generation. It addresses first the ways of further optimization of InP-based HEMT
technology in view of obtaining enhanced f mai performance (Section 2). Power and
frequency characteristics of monolithic oscillators evaluated with the help of a large-signal
analysis are presented in Section 3. Finally, specific designs and topologies of 160 GHz
fundamental monolithic HEMT oscillators are discussed in Section 4.
2 Device Optimization for High f max
A very high i max of several hundred gigahertz is necessary to guarantee the device
operation as oscillator at millimeter-wave frequencies. Optimization for high i max can
be achieved by reducing the parasitic resistances and capacitances of the HEMT. The
parasitic source resistance (R s ) consists of two parts : one coming from the contact region
(R c ) and the other from the ungated region between the gate and source. In an attempt
to minimize the ungated region resistance, a self-aligned gate technology has been applied
to InAlAs/InGaAs HEMT's [7]. The ungated region between source and gate has been
reduced in this case to less than 0.2 pun and the source access resistance was minimized,
resulting in a very high extrinsic (j of 250 GHz. Although f max is directly proportional to
Page 60
Third International Symposium on Space Terahertz Technology
Lgs=0 .2\lxa
(constant)
Lgd (variable)
Figure 1: Schematic view of the self-aligned offset T gate HEMT
fr, it was limited in this case due to the high output conductance (Gds) and gate-to-drain
capacitance (C g d); this was caused by the proximity of the gate and drain making G,i s
and Cgd higher than in HEMT's fabricated by conventional technology.
A better insight to the problem can be obtained by examining the f max expression
which is given by [8]:
fn
G d s
R s + R„
4 Co
= {^(On/fc + , * "o ) + 77^(1 + 2-5^)(l + G m R s f)-h
Cgd
c e
It l G m v ' l/G m + R s 5 C gs vy gs
It is obvious from Eq. (1) that a high f mai :/fr ratio can be achieved by increasing both
C g s /Cgd and Gm/Gds- These two ratios are related to L 3 d/L 5i , where L gs is the distance
between gate and source, and L g d is the distance between gate and drain. L g d/L gs can be
increased by offsetting the gate instead of placing it at the center of the source-to-drain
region. The novel self-aligned offset T-gate developed by the authors [9] and employed
in the analysis presented here, allows one to satisfy these requirements. An additional
feature of this approach is that L g d and L gs can be controlled much more accurately in
this way than in processes where the gate has to be offset aligned between two ohmic
contacts. Various L g j values ranging from 0.2 f.im to 0.6 /zm were employed while L gs
Third International Symposium on Space Terahertz Technology
Page 61
Figure 2: SEM photograph of self-aligned offset T-gate (L g d = 0.6 /j.m)
was fixed at 0.2 fim as shown in Fig. 1.
The devices were fabricated following the self-aligned process described in [9]. The
SEM photograph of the completed gate after the ohmic metal deposition is shown in
Fig. 2. The highest f mar values were obtained with L g d = 0.4 /.im and the corresponding
microwave results are shown in Fig. 3. fr ' s in this case around 150 GHz and f m(ir is
greater than 300 GHz. By increasing L g d further, the value of f max /fr increases due to
the higher C gs /C g d and G m /Gd s ratios. However, the magnitude of fy becomes smaller
with L g d due to the increased gate length and source-to-drain spacing. The increase of
fmai/fr ratio with L g d is thus compensated by the decrease of fr and the maximum l max
occurs for L g d = 0.4 /j,m.
Further f moI optimization is expected by reducing the gate length of the devices which
had large offsets; due to the increased number of line scans for highly offset gales, the
gate length becomes larger than in the case of the symmetric/centered realizations. This
Page 62
Third International Symposium on Space Terahertz Technology
_ 350
X
S 300 h
o
C
cr
250 -
200 -
? 150 -
3
O
100
1
1 1 1
max >.
I
-
\. f *
-
1
I I I
0.1
0.2 0.3 0.4 0.5
Lgd (|Lim)
0.6 0.7
Figure 3: Microwave data of offset self-aligned InAlAs/InGaAs HEMT's (h g d — 0.4 fxm).
The results show an ij of 150 GHz and an f mar of 310 GHz, corresponding to a high
imaxlh ratio of 2.0
drawback can, however, be eliminated by a better optimization of the doses used for the
footprint and side lobes of the gate.
3 Evaluation of Oscillator Characteristics Using
Large- Signal Analysis
The design of high frequency oscillators is generally based on either small-signal S-
parameters or measured large-signal S-parameters. The small signal S-parameters predict
the initial conditions necessary for oscillation build up. However, the steady-state oscilla-
tion condition can not be accurately predicted from small-signal S-parameters. Designs
using measured large-signal S-parameters present also certain difficulties arising from
measurement accuracy and differences between measured and simulated conditions.
Third International Symposium on Space Terahertz Technology
Page 63
G(jw,A)
V
cir
Ideal
Coupler
V
ref
7777
Figure 4: The schematic circuit setup for the large-signal oscillator analysis
An all frequency-domain large signal oscillator analysis method has been developed
in view of evaluating the HEMT potential as oscillator. It employs small-signal
S-parameters and a harmonic balance routine with 2-D interpolation functions. The
method predicts the operation frequency, output power and optimum load termination
conditions.
A special circuit set-up is used to perform the large-signal oscillator analysis (Fig. 4).
It consists of an amplifying unit and frequency-selective feedback loop together with an
excitation signal. The HEMT is used in common source topology and is considered as an
amplifying unit with power-dependent gain saturation characteristics. An ideal coupler
is placed between the amplifying unit and feedback loop to initiate and monitor the
oscillation (V ctr ). The excitation signal (V re /) is increased from a small signal level until
the gain of the HEMT saturates and the steady-state oscillation condition is reached.
The large-signal oscillator analysis method has been applied to study the oscillation
power dependence on the termination impedance of common source InAlAs/InGaAs
HEMT oscillators. The simulation results are shown in Fig. 5 for a HEMT with 2
Page 64
Third International Symposium on Space Terahertz Technology
S
n
s
u
t
o
Oh
o
03
i— i
o
O
I I I 1 l" I I I |
104 GHz
i nrr"T"r'i i i
121GHz
199 GHz
74 GHz
-i i i i 1 1
1 10
Termination Impedance (Q)
100
Figure 5: The oscillation power dependence on the termination impedance at various
frequencies for a HEMT (L 3 = 0.1 fim) with f max = 200 GHz and gate periphery of 2 x
45 \im
x 45 fim gate periphery and i max of 200 GHz. The analysis shows that an optimum
output power level is obtained when the load impedance is of the order of 1/2 to 1/4
of the small signal negative resistance at frequencies which are sufficiently away from
fmox- At very high frequencies, the load impedance determined by the above criteria is
reduced to very small values (below 5 Q) which are difficult to implement in monolithic
form. This termination load requirement sets the practical limit of upper frequency at
which the oscillator circuit can be implemented. These effects were studied and design
criteria were established on the basis of practical realization constraints imposed by load
terminations which should exceed 5 Q.
The available power was evaluated at different frequencies using optimum termination
Third International Symposium on Space Terahertz Technology
Page 65
10"
I
%
£
a
o
i
o
CO
O
10'
10 u
10
-1
10
-2
G-
f =450GHz \
tfp8X22.5um ^
Vbr=10V
O-
f =450GHz
tfg=2X22.5um
Vbr=10V
<9«
\
a
f m =200GHz
tfp2X45um
Vbr=3V
f mM =450GHz
T*g^2X22.5um
Vbr=3V
■ ■ ■ i
50
500
100
R~equency(GHz)
Figure 6: Power delivered by InAlAs/InGaAs HEMT oscillators as a function of
frequency, gate periphery (W 5 ), maximum frequency of oscillation (f ma x) and breakdown
voltage (Vbr)
conditions. Three parameters are used for the simulation: gate periphery (W H ),
maximum frequency of oscillation (fmax) and breakdown voltage (V(, r ). The results are
shown in Fig. 6. The oscillation power decreases first slowly and shows a more dramatic
degradation at high frequencies close to f max . This corresponds to the degradation of
maximum available gain (G max ) and negative resistance (Rneg) at high frequencies. The
reduced R ne g imposes a smaller value of termination load with the result of less power
delivered to the load. The overall characteristics suggest that generation of reasonable
power levels is feasible up to a frequency of 2/3 of f max . From Fig. 6, it is obvious that
a higher f max ensures large oscillation power for a given frequency. More than 15 mW
of output power can be expected up to 300 GHz out of 0.1 /zm HEMT's with eight 22.5
Page 66 Third International Symposium on Space Terahertz Technology
fim gate fingers assuming an i max of 450 GHz and V(, r of 10 V. This prediction doesn't
include any parasitic effects coming from mismatches, losses of transmission lines and
source grounding.
Fig. 6 also shows the characteristics of devices with different gate widths in view of
studying the effect of gate periphery on the output power level. Larger devices provide
higher oscillation power, but they are harder to implement in oscillator circuits. This
is due to their lower induced negative resistance, which implies the need for very small
termination impedance. As a result, the oscillation power degrades fast at the high
frequency end of operation bandwidth. A compromise has consequently to be made
between the oscillation power and ease of realization when choosing the device periphery.
Another important parameter in the oscillator evaluation is the breakdown voltage.
A higher breakdown voltage allows one to bias the transistor at larger drain bias and
thus to apply higher DC power to the device. The RF power generated from the
device is proportional to the DC power and increases consequently with higher V(, r .
Under conditions of optimum biasing for power, the maximum voltage swing is ~
V& r /2. By increasing the breakdown voltage of the HEMT's from 3 V to 10 V, it was
found that the RF power increased by approximately 6 dB. Breakdown improvements
in InAlAs/InGaAs HEMT's have recently been reported by Matloubian et al [10] and
validate this assumption. Further work is, however, necessary to justify this possibility,
especially at millimeter-wave frequencies. Similar improvements can also be made by
increasing the current density of the device and may be even easier to achieve in
InAlAs/InGaAs HEMT's with the help, for example, of multi-heterojunction designs
[11].
It should finally be noted that the simulation results in Fig. 6 were obtained using
a simple series feedback topology and the evaluated oscillation power values do not
therefore necessarily reflect the maximum power capability of the devices.
Overall, the HEMT's can be optimized for generation of adequate power levels at
Third International Symposium on Space Terahertz Technology
Page 67
0.0
£ -50.0
JS
Q,
<u
o
c
(0
J_>
.52-100.0
w
OS
_>
(0
60
4M50.0
-200.0
Single Series
. Feedback ~^^
\ ./^Active ?
c\ Feedback
a
Dual
/ Feedback
i
i i i
154 156 158 160 162 164 166 168 170
frequency (GHz)
Figure 7: Comparison of negative resistance for three different feedback schemes: 1)
single series feedback, 2) dual feedback, 3) active feedback
high frequencies with the help of: i) high f mox , ii) high V 6r , iii) high current density, iv)
choice of appropriate gate periphery.
4 Circuit Topologies for Subterahertz Monolithic
HEMT Oscillators
As already discussed earlier on, the negative resistance available at subterahertz
frequencies is usually rather small. Furthermore, the induced negative resistance is
present over a narrow frequency range especially when the oscillation frequency is close
to the ( max of the device. The availability of small negative resistance values make the
design task very difficult. Therefore, appropriate topologies have to be selected such that
the negative resistance can be maximized over a wide frequency range.
Page 68
Third International Symposium on Space Terahertz Technology
WW
H
Impedance
Transformer
O
7777
Figure 8: Equivalent circuit schematic of active feedback oscillator for subterahertz
application
Dual feedback schemes can be used to improve Rn eg over a narrow frequency range. In
addition to the series feedback element from source to ground, a parallel feedback element
can be inserted between the gate and drain. The negative resistance of this topology is
compared with that of single series feedback topology in Fig. 7. As shown in this figure,
dual feedback circuits provide the possibility of obtaining larger negative resistance.
However, the negative resistance is present over a narrower range of frequencies than
in the case of single series feedback circuits.
The active feedback approach is another alternative for designing oscillators with
devices of small R„ ep . It uses a small FET as a phase shifting element between the
gate and drain (see Fig. 8). Since the feedback loop is provided by active rather than
passive elements, the feedback phase shift is fairly independent of frequency and therefore
oscillation is guaranteed over a wide range of frequencies. This can be verified from
Fig.7. Furthermore, the active feedback approach is less sensitive to the parasitica coming
from passive elements and interconnects because it does not strongly depend on passive
Third International Symposium on Space Terahertz Technology
Page 69
*
VgSQ-
\
Impedance
Transformer
RFOut
-o
6
Yds
Figure 9: Equivalent circuit schematic of push-push feedback oscillator for subterahertz
signal generation
circuitry for inducing the oscillation. It is therefore very suitable for high frequency
circuit applications, where accurate modeling of passive circuit elements is not really
available.
Another very interesting topology is the push-push configuration. The circuit
schematic is shown in Fig. 9. It consists of two subcircuits combined in push-push
arrangement. Each circuit oscillates at half the output frequency. The second harmonics
are here combined in phase at the output terminal, while the first harmonics cancel
each other. This configuration has the advantage of essentially doubling the operation
frequency of the discrete devices. Thus, the circuit may operate beyond the frequency
limit imposed by f max . This topology also, provides the possibility of lowering the
phase noise because all the odd harmonics and associated noise are canceled at the
output. Good balance between the two subcircuits has to be maintained for the successful
operation of this circuit, but mutual interaction between the two devices is expected to
make this requirement less stringent. Monolithic technology provides additional means
Page 70 Third International Symposium on Space Terahertz Technology
of achieving perfect balance. The monolithic push-push HEMT oscillator is therefore a
very promising candidate for satisfying the needs for subterahertz signal generation using
HEMT's.
5 Conclusions
The use of submicron InAlAs/InGaAs HEMT technology has been discussed in view
of the possibility of realizing subterahertz oscillators. InP-based HEMT's have been
optimized for this purpose and showed very high f max of 310 GHz using offset self-aligned
T-gate technology. A gate-to-drain separation of 0.4 y.m was used in these devices.
A large signal modeling method has been developed and applied to the evaluation of
power characteristics of HEMT oscillators. The optimum termination loads have been
found to be 1/4 - 1/2 of the small-signal negative resistance of the devices. Upper
frequency limit criteria have been established and indicated the feasibility of signal
generation up to ~ 2/3 of the f max of the device. The large signal analysis has also
been used to evaluate the oscillation power of HEMT oscillators in the subterahertz
region. The use of HEMT's with f max = 450 GHz, V 6r = 10 V and W p = 8 x 22.5 [im
should allow an oscillation power of 15 mW at 320 GHz.
The topology study of subterahertz oscillation has shown that enhanced negative
resistance can be obtained by using complex feedback schemes such as the dual feedback
and active feedback scheme. The frequency barrier imposed by f ma x can be overcome by
harmonic oscillation operation as for example in the case of push-push oscillators.
Acknowledgment
The help of T. Brock, G. Munns and G. I. Ng in technology and material growth are
greatly appreciated.
Third International Symposium on Space Terahertz Techywlogy Page 71
References
[1] L. D. Nguyen, A. S. Brown, M. A. Thompson, L. M. Jelloian, L. E. Larson,
and M. Matloubian, "650-A Self-Aligned-Gate Pseudomorphic Alo.4sIno.52As /
Gao.20Ino.80As High Electron Mobility Transistors," IEEE Electron Dev. Lett., vol.
13, no. 3, pp. 143-145, March 1992.
[2] P. Ho, M. Y. Kao, P. C. Chao, K. H. G. Duh, J. M. Ballingall, S. T. Allen.
A. J. Tessmer, and P. M. Smith, "Extremely High Gain 0.15 \im Gate-Length
InAlAs/InGaAs/InP HEMTs," IEE Electron. Lett, vol. 27, pp. 325-327, 1990.
[3] Y. Kwon, D. Pavlidis, P. Marsh, G. I. Ng, and T. Brock, "Experimental
Characteristics and Performance Analysis of Monolithic InP-Based HEMT Mixers
at W-Band," To appear in IEEE Trans. Microwave Theory Tech., 1992.
[4] Y. Kwon, D. Pavlidis, P. Marsh, M. Tutt, G. I. Ng, and T. Brock, "180GHz
InAlAs/InGaAs HEMT Monolithic Integrated Frequency Doubler," in Tech. Digest
of 1991 IEEE GaAs IC Symposium, pp. 165-168, October 1991.
[5] Y. Kwon and D. Pavlidis, "Large Signal Analysis and Experimental Characteristics
of Monolithic InP-Based W-Band HEMT Oscillators," in Proceedings of the 21st
European Microwave Conference, pp. 161-166, September 1991.
[6] Y. Kwon, D. Pavlidis, and M. N. Tutt, "An Evaluation of HEMT Potential
for Millimeter- Wave Signal Sources Using Interpolation and Harmonic Balance
Techniques," IEEE Microwave and Guided Wave . Letters, vol. 1, pp. 365-367,
December 1991.
Page 72 Third International Symposium on Space Terahertz Technology
[7] U. K. Mishra, A. S. Brown, L. M. Jelloian, M. Thompson, L. D. Nguyen, and S. E.
Rosenbaum, "Novel High Performance Self-Aligned 0.15 Micron Long T-Gate," in
Tech. Digest of 1989 International Electron Device Meeting, pp. 101-104, December
1989.
[8] M. B. Das, "A High Aspect Ratio Design Approach to Millimeter- Wave HEMT
Structures," IEEE Trans, on Electron Devices, vol. ED-32, no. 1, pp. 11-17, January
1985.
[9] Y. Kwon, T. Brock, G. I. Ng, D. Pavlidis, G. 0. Munns, M. E. Sherwin, and G. I.
Haddad, "F mai -Enhancement in CBE-Grown InAlAs/InGaAs HEMT's Using Novel
Self-Aligned Offset-Gate Technology," in ^th Conf. on InP and Rel. Materials, April
1992.
[10] M. Matloubian, L. D. Nguyen, A. S. Brown, L. E. Larson, M. A. Melendes, and M. A.
Thompson, "High Power and High Efficiency AlInAs/GalnAs on InP HEMTs," in
1991 IEEE Int. Microwave Symp. Dig., pp. 721-724, June 1991.
[11] G. I. Ng, D. Pavlidis, M. Tutt, J.-E. Oh, and P. K. Bhattacharya, "Improved Strained
HEMT Characteristics Using Doouble-Heterojunction Ino.65Gao.35 As / Ino.52Alo.4sAs
Design," IEEE Electron Dev. Lett., vol. 10, no. 3, pp. 114-116, March 19S9.
Third International Symposium on Space Terahertz Technology Page 73
VARACTOR DIODES FOR MILLIMETER AND SUBMILLIMETER WAVELENGTHS
Brian J. Rizzi, Jeffrey L. Hesler, Hasan Dossal and Thomas W. Crowe
Semiconductor Device Laboratory J> y-*2A
Department of Electrical Engineering N 9 3 * 2« «S 3
Thornton Hall /%06'Zr
University of Virginia
Charlottesville, VA 22903-2442 £ *D^>
ABSTRACT
Whisker-contacted GaAs Schottky barrier varactor diodes are the most common high-
frequency multiplier element in use today. They are inherently simple devices that have very
high frequency response and have been used to supply local oscillator power for Schottky
heterodyne receivers to frequencies approaching 700 GHz. This paper discusses the
development of improved varactor diode technology for space based applications at
millimeter and submillimeter wavelengths.
I. INTRODUCTION
Whisker contacted GaAs Schottky varactor diodes are presently in use to supply local
oscillator power at frequencies as high as 700 GHz for ground based and airborne
applications [1,2,3]. These diodes are also used in the Microwave Limb Sounder on NASA's
Upper Atmosphere Research Satellite which is now monitoring global ozone depletion [4].
Although these devices have proven to be quite useful, there is great interest in developing
technologies that are more mechanically robust, have higher operating frequency and have
the potential to generate greater amounts of power. This paper will review recent work at the
University of Virginia on multiplier elements. This includes both the development of planar
Schottky varactors and investigation of new devices that have the potential for improved
performance.
Page 74 Third International Symposium on Space Terahertz Technology
Section II will review our first attempt to fabricate a planar varactor diode for use at
millimeter wavelengths. The preliminary design is presented and the limitations of this
structure are considered. A next generation device is then proposed. It is hoped that this new
device will become a standard replacement for a very successful and commonly used
whisker-contacted varactor diode (U. Va.-6P4). Through development of this device we hope
to demonstrate the potential of planar varactor technology and investigate the factors that
will most seriously degrade planar varactor performance at high frequency.
We are developing varactor diodes for a multiplier chain to 1 THz. This system will
incorporate two doublers (80 to 160 GHz and 160 to 320 GHz) and a tripler (320-960 GHz).
The doublers will use multiple diodes integrated on a single chip to enhance power handling
ability. These chips are designed to be used in a balanced doubler developed by Erickson
[5]. The prototype doubler design and some preliminary results are presented in section m.
The tripler to 1 THz will be extremely challenging. Fortunately there is a great deal of
effort being expended world-wide on new varactor structures which may be useful for this
work. In section IV we will consider a variety of possible technologies, with special
emphasis on an integrated 5-doped varactor diode pair and consideration of the new
heterojunction barrier varactors. Section V is a brief summary of this work.
n. Development of a Planar Varactor technology
As a first step in the development of planar varactor diodes we will fabricate planar
devices to replace two commonly used whiskered diodes. These are the 6P4 diode, which is
commonly used for doubling in the millimeter wavelength range, and the 2T2, which is used
to double and triple at submillimeter wavelengths. The parameters of both of these devices
are listed in Table I. The primary electrical benefits of the whiskered diode technology is the
Third International Symposium on Space Terahertz Technology
Page 75
low shunt capacitance of the whisker and the ability of the diode user to tune the whisker
inductance to optimize performance.
A scanning electron micrograph of a prototype planar varactor is shown in Fig. la. The
surface channel fabrication procedure has been described previously [6,7]. The nominal
parameters for this diode, designated SC6T1, are also listed in Table I. This diode was
designed as a replacement for the 6P4 diode, however, there are two problems. First, series
resistance is substantially higher than the 6P4's and second, the planar diode has a parasitic
shunt capacitance of 12 fF which is unacceptably high. As might be expected, preliminary
RF measurements have been disappointing. The excess series resistance is due to the use of
Table I: Schottky Varactor Diodes
Batch Type
Epitaxial Epitaxial Zero-bias Minimum
Anode Layer Layer Series Junction Junction Breakdown
Diameter Thickness Doping Resistance Capacitance Capacitance Voltage
(urn) (urn) (cm" 3 ) (Q) (fF) (fF) (V)
6P4 Whiskered 6
2T2 Whiskered 2.5
SC6T1 Planar 6.2
1.0
3xl0 16
9.5
20
5.5
20
0.59
lxlO 17
12
5.5
1.5-2.0
11
1.3
2xl0 16
20
20
4
30
S.I. CaAs
60 urn
30
50-150 un
SX GaAs
Fig. 1. a) A prototype planar varactor diode. The surface channel technology is used to
achieve isolation between the contact pads [6,7]. b) A sketch of the second
generation device which has smaller contact pads and variable finger length.
Page 76 Third International Symposium on Space Terahertz Technology
an epitaxial layer that is too thick and too lightly doped. Although this epitaxial layer yields a
higher breakdown voltage, the penalty in R s outweighs this benefit. Since the fabrication of
the SC6T1, new material has been obtained and diodes with characteristics closer to those of
the 6P4 will be fabricated.
The increased shunt capacitance of the planar diode is a serious problem. This
capacitance is due primarily to the fringing field between the contact pads through the high
dielectric constant GaAs substrate. To reduce this capacitance there are three options:
1) Reduce the pad dimensions,
2) Increase the pad separation (and therefore the finger length), and/or
3) Use a substrate with a lower dielectric constant
The first two improvements will be implemented in our next generation device, as shown in
Fig. lb. The primary limitations on pad dimensions are the ohmic contact resistance and the
need to make a reliable solder contact. Although the proposed pad width of 30 |im is about
the minimum size that most users feel comfortable soldering to, it is clear that if smaller pads
will lead to better performance, users will develop more elaborate soldering techniques.
However, the minimum pad size is also limited by our ohmic contacts. We use SnNi/Ni/Au
plating for our standard ohmic contact and reliably obtain resistivities of 10 -5 Qcm 2 or
slightly less. Thus, a 30 nm x 30p.m pad should have roughly one ohm of contact resistance.
Smaller pads will require a significantly improved ohmic contact technology.
The new mask set will have several finger lengths, from 50 - 150 ^im. This will allow
evaluation of RF performance as a function of pad-to-pad capacitance and finger inductance.
It is expected that one specific finger length will give optimum performance in a given
Third International Symposium on Space Terahertz Technology Page 77
multiplier mount at a given frequency. Thus, we expect that detailed RF evaluation of these
devices will yield important guidelines for future chip designs.
The use of a quartz substrate for planar Schottky diodes has been demonstrated for
mixer applications [7]. This has led to significant reductions in shunt capacitance which may
be important for multiplier applications. However, the thermal properties of the GaAs
Schottky diode on quartz are not well understood, and we have noticed that mixer diodes on
quartz substrates are more likely to show signs of heating effects than equivalent diodes on
GaAs. Since the removal of heat from the varactor diode is particularly important, it is not
clear if quartz substrates will yield an overall performance benefit. Our next batches of
planar varactors will have GaAs substrates. However, we also hope to investigate quartz and
perhaps sapphire substrates in the near future.
III. INTEGRATED SCHOTTKY VARACTORS FOR BALANCED DOUBLING
The first step in the proposed multiplier chain to 1 THz is a doubler from 80 to 160
GHz. Since there are sources available that can deliver large amounts of power at 80 GHz,
our goal is to develop a doubler that is fairly efficient, but, more importantly, can handle
large input powers. With this goal in mind, a planar chip was designed based on the balanced
doubler configuration of Erickson, which has generated up to 25mW at 160 GHz using two
whisker contacted diodes [5].
One benefit of the planar diode technology is the ability to integrate several diodes on a
chip to increase power handling ability. For example, when two diodes are placed in series
their individual areas can be doubled in order to maintain the same total series resistance and
junction capacitance as a single device. However, the series pair will have twice as much
reverse breakdown voltage. The increased area and breakdown voltage will yield improved
Page 78
Third International Symposium on Space Terahertz Technology
power handling ability. Two scanning electron micrographs of our prototype are shown in
Fig. 2. The chip consists of four varactor diodes, two for each leg of the balanced doubler.
The design parameters and dc characteristics of the prototype balanced doubler chips
are shown in Table II. Our goal was to achieve a reverse breakdown voltage of 20V for each
anode. Also, the anode diameters of 10 and 12 nm were chosen to achieve zero-bias junction
capacitances of 40 and 60 fF per anode. As is seen in the table, the first batch had excessive
series resistance and extra breakdown voltage. This is due to the low doping density and
thickness of the epitaxial layer. For the second batch this problem was corrected at the cost
of reduced breakdown voltage. However, this trade-off is expected to yield significantly
improved RF performance.
The capacitance-voltage (C-V) curves for a single diode and a diode series pair are
shown in Fig. 3, indicating the increased breakdown voltage of the diode pair.
Preliminary RF tests for the first prototype balanced doubler were performed by Dr.
Erickson at the University of Massachusetts and the results are presented in Table II. These
initial results are quite encouraging, but not yet competitive with the whiskered-diode
Fig. 2. SEM photographs of the prototype balanced doubler to 160 GHz.
Third International Symposium on Space Terahertz Technology
Page 79
Table II:
Prototype Balanced Doubler Chips
DC Characteristics
Batch
#
Epitaxial
Layer
Thickness
(um)
Epitaxial
Layer
Doping
(cm" 3 )
Anode
Diameter
(urn)
Pair
Series
Resistance
(«)
Pair
Breakdown
Voltage
(V)
1
2
1.3
1.2
1.8xl0 16
2.5xl0 16
10
10
20
14
45
35
Preliminary RF Data 1,
Batch
#
Input
Freq.
(GHz)
Output
Freq.
(GHz)
Input
Power
(mW)
Output
Power
(mW)
Efficiency
(%)
1
1
82
82
164
164
55
100
3
6
6
6
f Preliminary RF data supplied by N. Erickson, University of Massachusetts.
Measurements have not been performed with batch #2.
40-
30-
Cd(fF)
20-
10-
f.
r.
~i r
10 20
Reverse Voltage
I
30
Fig. 3. C-V Characteristics of the prototype balanced doubler chip for a single varactor
diode (dotted), a diode series pair (solid) and for a single diode with anode-to-
pad connection to eliminate the pad-to-pad shunt capacitance (dashed).
page 80 Third International Symposium on Space Terahertz Technology
results. Two changes in the chip design are planned to improve performance. The first is the
increase in epitaxial layer doping to reduce series resistance, as was achieved with batch #2.
The second is the reduction of pad-to-pad capacitance. The importance of this is
demonstrated by the third curve (dashed) in Fig. 3. This curve was measured from the anode
to ohmic contact pad on a diode that had no ringer, and therefore does not include the pad-
to-pad capacitance. This curve has much greater modulation and much lower minimum
capacitance. This clearly demonstrates that the pad-to-pad capacitance is having a major
effect on performance. The shunt capacitance of future chips will be reduced through a
redesign of the contact pads and possibly through the use of quartz substrates.
Once the first stage multiplier has been optimized, the next step is to design a chip for
the doubler to 320 GHz. Since the second stage will not have to handle as much power as the
first, we will be able to trade-off some power handling ability in order to increase cut-off
frequency. It is expected that the optimum diodes for this stage will have smaller anodes and
higher epitaxial layer doping density.
There is much work to be done on the integrated balanced doublers. However, the
prototype devices have yielded encouraging results, and the improvements necessary to
increase performance are clearly defined. Thus, we expect to achieve significantly improved
output powers at 160 GHz in the near future. Also, the lessons learned on the first stage
doubler will be applied to the second stage, so that development of the higher frequency
chips should progress more rapidly.
Third International Symposium on Space Terahertz Technology Page 81
IV. POTENTIAL VAR ACTORS FOR TRIPLING TO 1 THZ
The development of a tripler to 1 THz is an extremely challenging task. Fortunately
there are several device technologies that may yield suitable performance. We have chosen
to investigate five of these, each of which is discussed in the following sections. The
whiskered Schottky and planar Schottky are considered briefly and the new two-
dimensional-electron-gas/Schottky (2-DEG/Schottky), which is considered in detail in a
separate paper, is also only briefly overviewed. The other two technologies, the integrated
5-doped varactor pair and the heterojunction barrier varactor, are considered in more detail.
A. Whiskered Schottky Diodes
The most likely candidate for the first successful tripler to 1 THz is simply a standard
whisker contacted Schottky varactor. The 2T2 diode has already been successfully used in
triplers to 500-700 GHz and can probably be extended to the THz range. However, the
efficiency will certainly be decreased and it is not clear how much output power will be
achieved. A more optimized diode can probably be developed, perhaps with slightly higher
doping density and smaller diameter. Although this technology appears to be reaching
fundamental limitations [8], it should continue to be pursued because the probability of some
level of success is high.
B. Planar Schottky Varactors
There are two advantages of using a planar Schottky device; the elimination of the
fragile whisker contact and the opportunity to use several integrated diodes to increase power
handling ability or achieve a more beneficial C-V characteristic. The drawback is the
increased shunt capacitance that is inherent in the planar diode. There are several areas that
must be researched. As discussed previously, these include the redesign of the contact pads
C-X
Page 82 Third International Symposium on Space Terahertz Technology
and anode finger, and the use of low dielectric constant substrates. Also, the potential use of
two Schottky varactors in an anti-series combination to achieve a symmetric C-V
characteristic may have substantial benefits for tripling applications. It is not yet clear if
planar Schottky technology will be useful at 1 THz, however we hope to answer many
important questions through our development of planar diodes for lower frequencies.
C. The 2-DEG/Schottky Diode
This device consists of a metal contact to the edge of a two-dimensional-electron-gas
(2-DEG) formed at a heterointerface. The capacitance is between the Schottky metal and the
undepleted portion of the 2-DEG. The voltage on the Schottky metal modulates the depletion
depth in the 2-DEG, thereby varying the capacitance. This device should benefit from
increased electron mobility and perhaps higher electron saturation velocities compared to
bulk devices. This may lead to significantly improved high frequency performance. Also,
this is an inherently planar device. Prototype diodes have demonstrated excellent capacitance
modulation and high reverse breakdown voltages. This new device is discussed in greater
detail in a separate paper [9].
D. An Integrated S-Doped Diode Pair
A design for a planar chip with two integrated 8-doped varactor diodes in a back-to-
back configuration is shown in Fig. 4. The symmetric C-V characteristic of such a diode pair
will yield significant benefits for tripler applications since an idler circuit at the second
harmonic is not needed. The 8-doped diodes have been shown to have a sharp C-V
characteristic [10,1 1]; which is a significant advantage at high frequencies since the available
input power is quite low.
Third International Symposium on Space Terahertz Technology
Page 83
The planar tripler has been designed to produce a capacitance ratio (C max /Cnun) of 2.5,
with an estimated cut-off frequency of 6 THz. The material structure is described in Table
HI. The mask set and epitaxial material for this device are now being purchased.
E. Evaluation of the Heterostructure Barrier Varactors
In 1990 Rydberg et al. demonstrated that a thin layer of high band-gap material
sandwiched between two thicker layers of low band-gap material could yield a symmetric
C-V characteristic that is ideal for tripler applications [12]. This Quantum (or
Heterostructure) Barrier Varactor (QBV or HBV) has promise for high frequency multiplier
applications, and is now being investigated by several groups. The goal of our investigation
60 un
50-150 un
S.I. GaAs
I
- 1
30
un
/
"^
P
J
S
^.
r
\
j
a
S.I. GaAs
k
Fig. 4. A sketch of the proposed integrated 6-doped varactor pair. The finger length will
be variable on the mask set and the anode spacing has not yet been determined
Table III: Epitaxial Material for the 8-doped Diode Pair
Layer
Doping Thickness
Type
Density (um)
NGaAs
<lxl0 15 cm -3 0.05
Si atomic layer
3.4xl0 12 cm -2
NGaAs
2xl0 17 cm" 3 0.13
N+GaAs
>3xl0 18 cm -3 3.0
Al x Gai_ x As etch stop
undoped(x>0.5) 2.0-2.5
GaAs substrate
S.I.
page 84 Third International Symposium on Space Terahertz Technology
is to determine if HBVs offer significant improvement over standard Schottky technologies,
and, if so, to demonstrate such improvement. To determine the potential of these devices we
will discuss the design of HBVs that have characteristics similar to the state-of-the-art
whiskered varactors whose characteristics were presented in Table I.
A schematic band diagram of a zero-biased single barrier GaAs/AlGaAs/GaAs HBV is
shown in Fig. 5a. When a voltage is applied to the device a depletion region is created on
one side of the barrier which increases in length as the voltage is increased. The capacitance
of this device is approximated as
C = * (1)
Xb/eb + Xo.total/eM
where A is the device area, e is the permittivity of the barrier (B) and the modulation region
(M) materials, Xb is the barrier layer thickness and Xo.totai is the total depletion layer
thickness on both sides of the barrier as a function of voltage. The maximum capacitance
can be as high as EbA/Xb if there is negligible depletion in the modulation layers at zero-
bias. The series resistance for a single barrier HBV, including spreading, epilayer and ohmic
contact resistance, is estimated as
j 2Xjvi Re
Rl.s = Rl.spr + Rl.epi + Rl.ec = ^T + ^a" + X (2)
where d is the anode diameter, o is the conductivity of the substrate (S) and epilayer (E)
materials, Xm is the length of the n-type modulation regions, and Re is the specific resistivity
of the ohmic contact. It is important to note that the device area affects not only the junction
capacitance, but also the resistance of the ohmic contact Therefore, although we can reduce
the junction capacitance by shrinking the device area, this is not beneficial unless the ohmic
contact resistivity, R<., is low enough so that the third term in (2) remains negligible. For this
Third International Symposium on Space Terahertz Technology
Page 85
study we will assume a specific contact resistivity of 10 Qcm , which is consistent with
the best contacts reported in the literature.
An important parameter for all varactor diodes is the voltage at which the conduction
current becomes significant. For a standard Schottky varactor, impact ionization in the
depletion region determines the reverse breakdown voltage and thermionic emission over the
Schottky barrier determines the forward conduction current For the HBV either thermionic
emission or avalanche breakdown can play the critical role, depending on the device
parameters. Figure 5b shows an HBV band diagram with voltage applied, with a depletion
region on one side of the barrier and an accumulation region on the other. As the HBV is
biased, the accumulation region grows and therefore the effective barrier height, given by
AEg-qVacc, decreases. Simultaneously, the electric field strength in the depletion region
grows. Whether avalanche breakdown or thermionic current occurs first depends primarily
on the conduction band discontinuity AE C and the band-gap in the modulation region.
We would like to have a method to compare HBV diodes to standard varactors. A
simple computer model was developed in order to simulate the operation of the HBV under
applied bias. The simulation assumes that little current flows through the device, and
1
-►;
A. - -
AE _'-qv
C ^ ace
1
T
A
qv bar
i
L
«-x B -+
<- X M~
AE c
.T._ „ ^
t
1
'
^ E c
/■
" X
1
/////
AIGaAs n-GaAs
/////
n ++ -GaAs
_ _.T..-.~.
n^-GaAs n-GaAs
1
T...
D ^-\^
Fig. 5. The band-diagram of a simple Heterostructure Barrier Varactor (HBV), a) zero-
bias and b) bias applied.
Page 86 Third International Symposium on Space Terahertz Technology
calculates the quasi-equilibrium band diagram for different bias levels. The approximations
developed by Delagebeaudeuf et al. [13] are a relatively standard method to analyze a 2-
DEG at a heterostructure interface. For our case, these approximations were extended to
include additional energy levels due to the high doping density in the 2-DEG region. The
most crucial parameter to estimate is the maximum voltage that can be applied before
conduction current begins to degrade the multiplier efficiency. For this discussion we will
assume that impact ionization becomes important at the voltage where the electric field
strength exceeds a critical value. Since there is no experimental data from which to estimate
the critical field of HBVs, we have assumed that the critical field will be similar to that of
GaAs pn junctions [14]. For devices dominated by thermionic emission, the maximum
voltage was assumed to be that voltage at which
AE C - qV^ = nkT, (3)
and we have assumed a value of n=5 for this study.
For GaAs/AlGaAs devices the value of AE C is rather small (AE C =0.35 [15]) and
thermionic emission becomes important before impact ionization. Simulations show that for
low doping levels (less than about 10 16 cm" 3 ) these devices can have a maximum voltage
comparable to a standard Schottky varactor, however, Repj will be extremely large.
Increasing the doping level decreases Repi, but also decreases V max , which indicates that
single barrier GaAs/AlGaAs HBVs will have less power handling ability than standard
Schottky varactors. There are several possible solutions to this problem, two of which will be
considered in this paper; epitaxial stacking of barriers and the use of different material
systems. Epitaxial stacking divides the applied voltage among several barriers, thus
increasing the maximum device voltage. For an HBV with N barrieTS, VN max = NV lmax .
Third International Symposium on Space Terahertz Technology Page 87
As N is increased, it is best to increase the device area in order to maintain reasonable values
of junction capacitance and modulation layer resistance. Assuming that the area is increased
proportionally to the number of barriers, the series resistance of an N barrier HBV can be
expressed as,
R N.s=^^ R l,spr + 2N Rl.epi + ^ R l,oc- (4)
This equation shows that the increase in area has the important effect of reducing the
spreading and ohmic contact resistances. In fact, without using multiple barriers it would be
impossible to fabricate an HBV with reasonable capacitance and series resistance unless the
ohmic contact resistivity is exceptionally low.
Other material systems can have significantly higher values of AE C . For example, the
InGaAs/InAlAs system can yield barriers of near 0.8 eV, while the GaAs/GaN system allows
0.9 eV barriers. The computer simulation indicates that the maximum voltage in both of
these material systems is limited by impact ionization, rather than thermionic emission.
In the following paragraphs, the simulation results for single and multiple barrier HBVs
are discussed for the previously mentioned material systems. In order to compare the HBVs
with the 2T2 and 6P4 varactors, barriers are added until V max is greater than that of the
standard varactor. The area is then chosen so that C^ of the HBV is the same as the
standard varactor. A common varactor figure-of-merit used in our comparisons and listed in
Table IV is the dynamic cut-off frequency, given by [16]
tco " 2tcR s ' (5)
where Cma* and Cm^ are the maximum and minimum device capacitance.
Page 88
Third International Symposium on Space Terahertz Technology
GaAs/AlGaAs: Figure 6 shows the simulation results for GaAs/AlGaAs HBV's. As the
modulation doping density is increased, more barriers are needed to achieve the desired
maximum voltage and the series resistance is reduced due to the increase in both o"e and
device area. Table IV gives examples of GaAs/AlGaAs HBV's with V max , 0^ and R s
similar to the 2T2 and 6P4. However, the HBVs will have the added benefit of a symmetric
C-V curve.
InGaAs/InAlAs: With the InGaAs matched to InP (i.e. 53% In), Ino.32Alo.68 As will give a
AE C of about 0.8 eV with a 1% lattice mismatch. However, InGaAs has a narrower band gap
than GaAs, and will thus have a smaller critical field for impact ionization. In these
simulations, we used the critical field data versus doping for a Ge abrupt p-n junction [14]
since Ge and Ino.53Gao.47 As have similar bandgaps. Because of the lower critical field,
single barrier InGaAs/InAlAs HBV's will not have sufficient V max , and thus multiple
Table IV: Heterostructure Barrier Varactors
Material System
N
X B
N mo d
Diam.
"max
R,
*-niin
r
^-max
tco
(urn)
(cm" 3 )
(um)
(V)
(")
(fF)
(fF)
(THz)
GaAs/AlGaAs
1
0.02
5xl0 16
2.3
4.0
37
1.5
22
2.6
1
0.02
lxlO 17
1.7
2.6
21
1.5
12
4.4
1
0.02
5xI0 17
1.0
1.2
19
1.5
4
3.4
(6P4-like)
5
0.02
5xl0 16
9.0
20
7
4.5
66
4.6
(2T2-like)
5
0.02
lxlO 17
3.8
13.2
11
1.5
12
8.6
InGaAs/InAlAs
1
0.02
5xl0 16
2.6
6.7
25
1.5
32.3
4.0
1
0.02
lxlO 17
1.9
4.3
16
1.5
17.2
6.0
1
0.02
5xl0 17
1.2
2.8
16
1.5
6.4
5.1
(6P4-like)
3
0.02
5xl0 16
8.0
20.0
5.5
4.5
97
6.0
(2T2-like)
3
0.02
lxlO 17
3.4
12.9
9.5
1.5
17.2
10.4
GaAs/GaN
1
0.008
5xl0 16
6.0
19.2
13
4.5
367
2.7
1
0.008
lxlO 17
2.5
11.4
20.5
1.5
63
5.1
(6P4-like)
2
0.008
lxlO 17
6.1
23
5
4.5
190
7
(2T2-like)
2
0.008
2.3xl0 17
2.8
18.2
9.5
1.5
38.6
10.7
Third International Symposium on Space Terahertz Technology Page 89
barriers must be used. Figure 7 and Table IV show that devices similar to the 2T2 and 6P4
varactors can be achieved with three barriers.
GaAs/GaN: The GaAs/GaN material system has a high AE C and the critical field of GaAs.
The major disadvantage is that it is a relatively new material system on which little
experimentation has been performed [17]. Our simulations showed that impact ionization
will be the limiting factor for these devices. Single barrier GaAs/GaN HBV's have sufficient
V max , but tend to have higher R s than comparable Schottky varactors due to modulation
region resistance. Characteristics of single barrier GaAs/GaN HBV's are given in Table IV
for several dopings. By using higher N mo< j and multiple barriers, HBV's with low R s and
very little conduction current should be possible. Figure 8 and Table IV show that only two
barriers are required to achieve device characteristics similar to the 2T2 and 6P4.
V. Summary
Whisker-contacted GaAs Schottky barrier varactor diodes are the best multiplier
elements available for millimeter and submillimeter wavelength applications. However, the
development of planar diode technology and new devices promise to improve both system
reliability and performance. Our prototype planar Schottky varactors are not yet competitive
at millimeter wavelengths, but several straight-forward improvements in the chip design
should alleviate the problems of high series resistance and shunt capacitance. The ability to
integrate several varactor diodes onto a chip is being exploited to increase power handling
ability, and an integrated balanced doubler for millimeter wavelengths has been described.
The prototype devices have shown promising performance at 160 GHz and the second
generation chips have greatly improved dc characteristics. Both the single-diode planar
Schottky varactor and the balanced doubler will benefit from improved contact-pad/finger
Page 90
Third International Symposium on Space Terahertz Technology
geometries and lower ohmic contact resistances. The use of low dielectric substrates is also
being investigated.
Several device technologies may be useful for a proposed tripler to 1 THz. While the
whisker-contacted Schottky diode is likely to be the first device to yield reasonable output
power at this frequency, planar diodes and other device structures promise improved
performance. The 2-DEG Schottky, 8-doped Schottky and the quantum (heterostructure)
barrier varactors (QBV or HBV) are being investigated at U.Va. The 2-DEG Schottky
research is described elsewhere [9]. An integrated 5-doped varactor pair with symmetric C-V
was described and will be fabricated in the near future. The HBV devices are particularly
promising. Our simple analysis has shown that HBVs with parameters similar to the state-
of-the-art Schottky varactors can be designed if multiple barriers are used and ohmic contact
resistances are in the 10 _7 £2cm 2 range. These devices will also have the benefit of a
symmetric C-V curve. The use of InGaAs/InAlAs or GaAs/GaN for the HBVs promises the
best performance if epitaxial layers of the required quality can be obtained.
5el6
le17
5el6
1 el 7
2.3el7 5el7
N mod( OT - 3 )
2.3e17 5e17
N modK4)
Fig. 6 The number of barriers necessary to achieve a GaAslAlGaAs HBV with the same
maximum voltage as the 2T2 (left) and 6P4 (right) varactors as a function ofN mod .
Also, shown is the series resistance when the area is chosen to yield the same
minimum device capacitance as the standard varactors. Additional parameters are
listed in Table TV.
Third International Symposium on Space Terahertz Technology
Page 91
Se16
Iel7
2.3el7
N mod (a" -3 )
5el7
2.3elS
5el6
Iel7 ■ 2.3e!7
N mod ( cm_3 )
Fig. 7. The same as Fig. 6, but for InGaAslInAlAs.
5e17 1e18
N mod ( cm_3 )
mod
(cm-3)
Fig. 8. The same as Fig. 6, but for GaAslGaN.
Page 92 Third International Symposium on Space Terahertz Technology
ACKNOWLEDGEMENT
The authors would like to thank Dr. Neal Erickson (U. Mass, Amherst) for supplying
the initial RF measurements on the prototype balanced doubler and Dr. Peter Siegel (Jet
Propulsion Laboratory) for initial evaluation of the SC6T1 planar varactor. This work has
been supported by the National Science Foundation under Grant ECS-8720850, NASA and
the Jet Propulsion Laboratory.
REFERENCES
[I] R. Zimmermann, R. Zimmermann, and P. Zimmermann, "All Solid-State Radiometers for Environmental
Studies to 700 GHz," Third Int'l Symp. Space THz Tech., Ann Arbor, MI, March 1992.
[2] H. Nett, S. Crewell, K. Kunzi, "A 625-650 GHz Heterodyne Receiver for Airborne Operation," 16th Int'l
Conf. IR and MM Waves, Lausanne, Switzerland, August 1991.
[3] S. Crewell and H. Nett, "Measurements of the Single Sideband Suppression for a 650 GHz Heterodyne
Receiver," Third Int'l Symp. Space THz Tech., Ann Arbor, MI, March 1992.
[4] A.L. Riley, UARS Microwave Limb Sounder Instrument Description, Jet Propulsion Laboratory
Document D-1050, 1984.
[5] Nil. Erickson, "High Efficiency Submillimeter Frequency Multipliers," 1990 IEEE MTT-S Int'l.
Microwave Symp., Dallas, TX, May 1990.
[6] Wl. Bishop, K. McKinney, RJ. Mattauch, T.W. Crowe, and G. Green, "A Novel Whiskerless Schottky
Diode for Millimeter and Submillimeter Wave Applications," Proc. 1987 IEEE MTT-S Int'l. Symp., Las
Vegas, NV, pp. 607-610, June 1987.
[7] W.L. Bishop, E.R. Meiberg, RJ. Mattauch and T.W. Crowe, "A Micron Thickness, Planar Schottky
Barrier Diode Chip for Terahertz Applications with Theoretical Minimum Parasitic Capacitance," 1990
IEEE MTT-S InL Microwave Symp., Dallas, TX, May 1990.
[8] T.W. Crowe, W.C.B. Peatman and E.M. Winkler, "GaAs Schottky Barrier Varactor Diodes for
Submillimeter Wavelength Power Generation," Microwave and Optical Technology Letters, Special Issue
on Space THz Technology, Vol. 3, No. 1, pp. 49-53, Jan. 1991.
[9] W.C.B. Peatman, T.W. Crowe, and M. Shur, "A 2-DEG Varactor Diode for Millimeter and Submillimeter
Wave Multiplier Applications," Third Int'l Symp. Space THz Tech., Ann Arbor, MI, March 1992.
[10] T.W. Crowe, W.C.B. Peatman and W.L. Bishop, "GaAs Schottky Barrier Diodes for Space Based
Applications at Submillimeter Wavelengths," The First International Symposium Space Terahertz
Technology Proceedings, pp. 256-272, Ann Arbor, Michigan, March 1990.
[II] B.J. Rizzi, T.W. Crowe, and W.C.B. Peatman, "A 5-Doped Varactor Diode for Submillimeter
Wavelengths," The Digest of the 15th International Conference on Infrared and Millimeter Waves, pp.
478-480, Orlando, Dec. 1990.
[12] A. Rydberg, H. Gronqvist and E. Kollberg, "Millimeter- and Submillimeter- Wave Multipliers Using
Quantum-Barrier- Varactor (QBV) Diodes," IEEE Electron Device Letters, Vol. 11, No. 9, pp. 373-375,
1990.
[13] D. Delagebeaudeuf and N.T. Linn, "Metal-(n)AlGaAs-GaAs Two-Dimensional Electron Gas FET," IEEE
Trans. Electron Devices, Vol. ED-29, No. 6, pp. 955-960, 1982.
[14] S.M. Sze, Physics of Semiconductor Devices, Wiley, New York, p. 103, 1981.
[15] J. Batey and S.L. Wright, "Energy band alignment in GaAs:(Al,Ga)As heterostructures: The dependence
on alloy composition," J. Appl Phys., Vol. 59, No. 1, pp. 200-209, 1986.
[16] P. Penfield and RJ? Rafuse, Varactor Applications, MTT Press, p.86, 1962.
[17] G. Martin, S. Strite, J. Thornton, and H. Morkoc, "Electrical properties of GaAs/GaN/GaAs
semiconductor-insulator-semiconductor structures," Appl. Phys. Lett., Vol. 58, pp. 2375-2377, 1991.
2-
Third International Symposium on Space Terahertz Technology Page 93
A Schottky/2-DEG Varactor Diode for N93*27734
Millimeter and Submillimeter Wave Multiplier Applications
W. C. B. Peatman, T. W. Crowe, M. Shur, and B. Gelmont S& _^ *Z
Semiconductor Device Laboratory /, _ ;
Department of Electrical Engineering / (?0 b £- ^—^
University of Virginia
Charlottesville, VA 22903
ABSTRACT
A il
A new Schottky diode is investigated for use as a multiplier element in the millimeter and
submillimeter wavelength regions. The new diode is based on the Schottky contact at the edge
of a 2-dimensional electron gas (2-DEG). As a negative voltage is applied to the Schottky
contact, the depletion layer between the Schottky contact and the 2-DEG expands and the
junction capacitance decreases, resulting in a non-linear capacitance-voltage characteristic. In
this paper, we outline the theory, design, fabrication and evaluation of the new device. Recent
results include devices having cutoff frequencies of 1 THz and above. Preliminary multiplier
results are also presented.
I. BACKGROUND
Schottky barrier varactor diodes are used as frequency multiplier elements for local
oscillator (LO) sources for the millimeter and submillimeter wavelength region. These sources
are used in heterodyne receivers for a variety of applications including radio astronomy,
atmospheric studies and plasma diagnostics. For space-based receiver systems, the LO source
must be compact, lightweight and reliable; and power and cooling requirements must be
minimized. While molecular gas lasers have been used as LO sources for airborne
radioastronomy measurements at frequencies as high as 2.5 THz [1,2], the stringent
requirements for space applications will require the use of a solid-state LO source. Although
standard varactor diodes have been used to generate 0.7 mW at 474 GHz [3] and 0.2 mW at
640 GHz [4], these devices will not provide usable amounts of LO power above about 1 THz
[5]. Schottky barrier varistor diodes and quantum well oscillators have been proposed as
sources of LO power but these technologies will not provide sufficient power to drive the GaAs
Schottky barrier mixer diodes used in these receivers [6,7,8]. We report here on a new planar
varactor diode in which the Schottky contact is formed at the edge of the 2-dimensional
electron gas (2-DEG). This new device, which is essentially the 2-d analog of the standard (3-
d) Schottky diode, has unique properties and is a promising candidate for use in millimeter and
submillimeter wave multiplier applications [9,10]. In addition, it may be possible to use this
device to investigate conduction in a 2-d electron gas at frequencies significantly above 100
GHz.
Page 94
Third International Symposium on Space Terahertz Technology
Schottky Contact
Pad
Fig.l.
Schematic of the planar Schottky 12 -DEG varactor diode.
In Section II, the theory of the device and the design for multiplier applications is
reviewed. In Section IE, the fabrication of the devices is briefly described. The low frequency
evaluation is presented in Sections IV and V for the interdigital-type and the "refined
prototype" devices, respectively. The preliminary multiplier performance of the "refined
prototype" devices are presented in Section VI. Finally, a summary of the work and outlook for
future research is presented in Section VII.
II. THE SCHOTTKY/2-DEG DIODE
A. Overview
A sketch of the Schottky/2-DEG diode is shown in Fig. 1. Also shown is an expanded
view of the Schottky contact region. The chip dimensions are typically 100 um by 200 \xm by
about 50 fJ.m thick. Two device configurations are discussed in this paper. These are the
interdigitated contact device (also described in [9,10]) and the "refined prototype" device
which is similar to that shown in Fig. 1. The interdigitated devices have been realized in both
single and dual anode configurations, the latter being intended for symmetric C(V)
Third International Symposium on Space Terahertz Technology
Page 95
V a =0
V =V
v a v r
Fig. 2. Conduction band diagram of the Schottky/2-DEG diode. The 2-DEG is bounded on
the left by a Schottky contact and on the right by an ohmic contact. The depletion
depths for two applied voltages are shown (left). A qualitative sketch of the potential
well in the undepleted channel is shown (right) with the Fermi and lowest sub-band
energies indicated.
applications. These devices had Schottky contact widths in the range 150-350 fim. The
"refined prototype" devices are single anode devices (whose cathode is an ohmic contact) with
widths of about 100 |im. The isolation between pads is achieved by etching through to the
semi-insulating GaAs substrate everywhere except in the channel region and beneath the pads.
Further details of the fabrication process are outlined in Section HI.
B. Physics and Equivalent Circuit
The conduction band diagram of the Schottky/2-DEG diode is shown in Fig. 2. The
theory of the metal/2-DEG junction was first considered in [1 1] and more recently extended in
[12]. In [12], the junction capacitance was derived using a conformal mapping technique and
by making suitable assumptions about the boundary conditions. The capacitance-voltage
characteristic was given by the expression
(R2 + d| ep ) 1/2 +R 1
Cj(V) = ^L ln[-
%
(1)
where W is the width of the contact (see Fig. 1), e is the permittivity of GaAs and R is the
Page 96
Third International Symposium on Space Terahertz Technology
O-
C.(V)
H \-^W-
■O
'sh
Fig. 3. A simple equivalent circuit model of the Schottky/2-DEG diode.
half-height of the Schottky metalization. The depletion depth ddep of the 2-DEG is given as
2e(V bi -V a )
ddep =
qn s
(2)
where V bi is the built-in voltage (0.7-1.0 V), V a is the applied voltage (which is negative for
reverse-bias), and n s is the 2-DEG sheet charge density. For a 2-DEG sheet charge density of
10 12 cm -2 (10 13 cm -2 ) and assuming e of GaAs, the 2-DEG depletion depth is 3 |im (0.3 Jim)
at 20V reverse bias. The total capacitance is equal to the junction capacitance in parallel with
the shunt capacitance associated with the pad-to-pad fields and anode-to-ohmic fields.
The equivalent circuit is shown in Fig. 3. For simplicity, the junction conductance is
neglected. [The skin effect is probably not important for this device since the length and depth
of the 2-DEG are small]. Also, the inductive effect of charge carrier inertia is neglected
although this effect may be important at cryogenic temperatures due to the long momentum
relaxation time of the 2-DEG. The junction capacitance Cj can be estimated using Eq. 1. The
series resistance, R s , is composed of the resistance of the undepleted channel and the ohmic
contact resistance. The former is given by
r _ (L-ddep)
s q^n s W
(3)
where L is the channel length and m is the electron mobility in the undepleted channel. The
ohmic contact resistance is simply Roc = r sc/W where r^ is the specific contact resistivity
(normally specified in units Qmm) of a HEMT-like ohmic contact, and W is the device width
(in mm). The shunt capacitance term C S h includes pad-to-pad capacitance and Schottky-to-
ohmic capacitance. For high frequency design, C S h and the Rs must be minimized.
The theory of the junction breakdown has not been fully developed. However, Vb r in long
channel devices is assumed to be caused either by impact ionization or by tunneling.
Experimental observations of very large breakdown voltages in prototype devices [9] lead to
the conclusion that for short channel length devices, the breakdown voltage was limited by
punch-through. In this paper, those devices with channel lengths of 1 |im are probably punch-
through limited whereas the 2-3 ^m length devices may be limited by impact ionization or
Third International Symposium on Space Terahertz Technology Page 97
tunneling, or both. The breakdown voltage may also be limited by the geometry of the anode
metalization (for example, the half-height R) or by processing and material defects. A more
detailed discussion of the breakdown in these devices will be presented in a later paper.
C. Frequency and Power Limitations
Several factors limit the frequency response and power performance of the multiplier. The
frequency response may be limited by the dynamic cutoff frequency which is usually defined
[13] as
_ S m ax~Smin ,,v
Vco 27tR s ()
where S max (Smin) is 1/Qnin (1/QnaxX and R s is the series resistance of the varactor diode. It is
desirable for the device to have a v co value much higher than the operating frequency to ensure
that the multiplier efficiency is not degraded. To achieve high v co , the series resistance should
be as small as possible, the minimum capacitance (near breakdown) should be small and the
capacitance modulation ratio Qnax/Qnin should be large.
Another important quantity which may limit both the frequency response and maximum
output power is the finite velocity of the electrons traversing the modulation region (the
epilayer in GaAs diodes or the 2-DEG channel in the Schottky/2-DEG diode). Recently,
Kollberg et al [14] showed how the finite electron velocity limited the current in the 6P4 diode
used by Erickson [3]. Kollberg argued that the ac displacement current could not exceed the
saturation current which in turn is limited by the electron drift velocity. Using Monte Carlo
analysis, the effective velocity and the saturation current in the 6P4 diode were estimated (in
[14]) to be 2.4 x 10 7 cm/s and 44 mA, respectively. At input powers beyond that which causes
the current to saturate, the diode's rf impedance increases (since the current cannot). Kollberg
used this analysis to simulate the roll-off in efficiency with input power, which was observed
by Erickson.
The velocity saturation current is written here for the 2-d case as
I vs = qn s v e ffW (5)
where n s is the 2-dimensional sheet charge density, W is the contact width and v e ff is the
effective velocity of the electrons in the channel. As will be shown, a Schottky/2-DEG diode
with W = 100 |im and n s = 1.85 x 10 12 cm -2 has roughly the same characteristics as the 6P4
diode. Assuming the same effective velocity as was used for the 6P4, namely v e ff = 2.4 x 10 7
cm/s, I vs is about 70 mA.
The finite electron velocity may also limit the frequency response if the transit-time for
electrons traversing the modulation region is comparable to the period of the LO frequency.
The transit-time corner frequency has been proposed [10] as a useful design parameter for the
diode's frequency response, and is defined as
where L is the channel length (or the epilayer thickness in the standard diode). Note that the
Page 98 Third International Symposium on Space Terahertz Technology
frequency given by Eq. 6 is a corner frequency since, as the frequency increases beyond this
value, the maximum if modulation length shortens, resulting in a smaller capacitance ratio and
thus to a roll-off in the multiplier efficiency. For example, assuming an effective electron
velocity of 2.4 x 10 7 cm/s (as was used in [14]), a varactor diode with input frequency of 80
GHz will have a maximum modulation length of about 0.48 urn. In comparison, the 6P4 diode,
which is often used at this frequency, has an epilayer thickness (and maximum dc modulation
length) of about 1.0 um. Thus, both standard and 2-DEG diodes should be designed to achieve
a large capacitance modulation ratio within the length given by Eq. 6.
Having outlined the equations for the capacitance, resistance, punch-through voltage (Eq.
2 with ddep = L), the dynamic cutoff and the transit-time frequencies and the saturation current,
the Schottky/2-DEG diode may be designed for particular applications. This procedure is
simplistic but is similar to the design of the state-of-the-art varactors currently in use. This
design procedure was used for the "refined prototype" devices whose results are given in
Section V. Before presenting the device results, the fabrication will be briefly reviewed.
HI Fabrication
The prototype Schottky/2-DEG devices discussed here were fabricated on a
pseudomorphic Al .25Gaj5 As/In. isGa 85 As/ GaAs structure shown in Fig. 4. This structure was
grown by MBE and analyzed using the Van der Pauw method to determine the mobility and
sheet charge density. The electron sheet charge density at both 77 K and 300 K was
1.85xl0 12 cm -2 and the electron mobilities were 31,400 cm 2 /V-s and 6640 cm 2 /V-s at 77 K
1
GaAs
5 x 10 18 cm" 3
40 A
2
Alo.25Gao.75As
5 x 10 17 cm" 3
300 X
3
Si Atomic Plane
5 x 10 12 cm" 2
~
4
Alo.25Gao.75As
~
50A
5
In 15 Ga 85 As
—
150 A
6
GaAs
—
5000 A
7
SI GaAs Substrate
Fig. 4. AlGaAslInGaAslGaAs heterostructure used for the devices discussed here.
Third International Symposium on Space Terahertz Technology
Page 99
' "" ' j y' , '^ : .^-" :i ^"*'-"l> '• --''* L^Sr*-" 4.
W'r^' ■ ■> " -^
•
8.58kx 18ku 123
Fig. 5. Scanning electron micrographs of Schottky/2-DEG devices. The interdigitated
device (top) has anode width of 250 \xm and channel length of 2 Jim. The "refined
prototype" device (bottom) shown here is similar to the devices discussed in Section
V. Here, the anode width is 100 |im and the channel length is 5 \\m.
Page 100 Third International Symposium on Space Terahertz Technology
and 300 K, respectively. The supply and cap layers are substantially depleted to the 2-DEG
and/or the surface, to minimize parallel conduction. The ohmic contact consists of an
electroplated SnNi/Ni/Au trilayer which is alloyed at about 380°C. To form the Schottky
contact, a trench is etched through the 2-DEG layer and a Pt/Au contact is electroplated into
the trench. Next, the contact pads are plated and finally a 2-3 micron deep NaOH:H202 etch to
the SI-GaAs substrate is performed to isolate the two pads. All lithography levels are
performed using a Karl Suss MJB-3 (405 nm). An SEM photo of the interdigitated device is
shown in Fig. 5 (top). A sketch of a "refined prototype" is also shown (bottom). The rough
surface of this device was due to the isolation etch, performed using chlorine reactive ion
etching. A subsequent wet chemical etch reduced the surface roughness considerably.
IV. LOW FREQUENCY EVALUATION OF INTERDIGITATED DEVICES
The dc evaluation of the Schottky/2-DEG diodes include I(V), C(V) and reverse
breakdown voltage measurements. First, the interdigitated device results are presented (these
results were also presented in [10]). The forward and reverse I(V) of a single Schottky/2-DEG
device is shown in Fig. 6 (top), measured at room temperature. The exponential diode
characteristic is seen as the linear portion of this semi-log I(V) plot, in the range 0.35 - 0.7 V.
The "knee" voltage (at 1 |J.A) was 0.512V. The AV values for the current intervals 0.1 - l.OjiA,
1.0 - IO.O^iA and 10.0 - 100(iA are 74mV, 74mV and 81mV, corresponding to inverse slope
parameters, V , of 32.1mV, 32.1mV and 35.2mV, respectively. This corresponds to a diode
ideality factor of 1.26. The series resistance of this device was determined to be 56Q. The
expected 2-DEG channel resistance at room temperature is 6 Q.. Allowing for a pessimistic
value of the ohmic contact resistivity, r^ of 2.5 Qmm, the total series resistance expected for
this device was about 16 Q. The remaining 40 Q series resistance is most likely due to
insufficient plating of the ohmic contacts, as was substantiated by inspection using scanning
electron microscopy. The dual anode devices have no ohmic contact resistance. In these
devices, the I(V) is dominated by the characteristic of the reverse-biased junction and a series
resistance measurement cannot be made. However, using Eq. 3, the L = 2|im, W = 250um
device resistance is about 4 Q at 295K and about 1.0 Q at 100 K due to the increase in mobility
uponcooling.
The C(V) curves of the single and dual anode interdigitated devices are shown in Fig. 6
(bottom). The channel length (gap between fingers) is 2 |im (3 p.m) for the dual (single)
Schottky device. The anode widths were 250 jxm for both devices. As expected, the dual
Schottky device has a nearly symmetric C(V) characteristic and it's zero-bias capacitance is
about half that of the single anode device. Subtracting the pad-to-pad capacitance which was
measured to be 4 fF, this symmetric C(V) device had a dynamic cutoff frequency of about 1
THz at 295K and about 4 THz at 100 K. The velocity saturation current (Eq. 5 using
v e ff = 2.4 x 10 7 cm/s) is 178 mA. Also, the transit-time corner frequency of the 2 |im channel
length device is about 19 GHz, calculated using Eq. 6. The capacitance levels of the dual
anode device is probably too high for most multiplier applications. Nevertheless, multiplier
testing of these devices will be performed in the near future.
Third International Symposium on Space Terahertz Technology
Page 101
3- ie-06
4-1
C
d)
M
P
O
' ' ' '
10 15 20 23 30~~
Revarsa Voltage (V)
■
1.4
Forward Voltage (V)
Voltage (V)
Fig. 6. Forward and reverse I(V) of single anode interdigitated device (top) with
W = 250 |im and L = 3.0 |im. C(V) characteristics (bottom) ofL = 3.0 (im single and
L = 2.0 |im dual anode interdigitated devices of width 250 p.m.
Page 102
Third International Symposium on Space Terahertz Technology
40
35
30
b4
4-1
^^
25
<U
O
C
(0
20
■P
-H
u
15
a
nJ
u
10
5
I I
Data -»
Fit -
4 6 8 10
Reverse Voltage (V)
12
14
Fig. 7. C(V) of refined prototype device A. Also shown is a fit (Eq. 1) using R = 0.75 fim,
C sh = 8.0 fF, V bi = 0.7 V andW = 90 ^im.
V. LOW FREQUENCY EVALUATION OF REFINED PROTOTYPE DEVICES
The refined prototype devices had two anode width/channel length combinations. The
"A" devices had anode widths of 90 |im (on average) and channel lengths of 2.5 |im while the
"B" devices had anode widths of 80 [im and channel lengths of 3.0 (im. The C(V) characteristic
of device A is shown in Fig. 7. The theoretical capacitance, shown fitted to the data, agrees
well with the data except near zero-bias where the fit is lower than the data. The fit assumed a
reasonable value of the anode metal half-height (R = 0.75 |im) and a shunt capacitance of 8.0
fF. This value of C S h is higher than expected since the pad-to-pad capacitance was measured to
be about 2 fF. The additional shunt capacitance may be due in part to fields between the anode
and ohmic metals. This contribution to the capacitance is not easily determined and is also not
substracted for the v c0 calculations. The difference between the theory and the data near zero-
bias is either due to inaccuracy of Eq. 1 for the geometry of this device or to effects related to
the leakage current at low bias. We are currently investigating a more general theory of the
junction capacitance for devices of various geometries.
The forward I(V) as a function of temperature of device A is shown in Fig. 8. As the
temperature decreased, several changes occured. First, the entire I(V) curve shifted to higher
voltages, as expected due to the temperature dependence of the saturation current (the theory of
the thermionic saturation current of the Schottky/2-DEG junction is being investigated [15]).
At lower currents, the "leakage" current which has been routinely observed at room
temperature is seen to decrease substantially so that, at 220 K, it is much less than one
nanoamp. Finally, the strong temperature dependence of the series resistance is evident at high
currents.
Third International Symposium on Space Terahertz Technology
Page 103
le-01
le-02 -
le-03
le-04
le-05
4->
c
u
le-06
u
D
o
le-07
le-08
le-09
le-10
CO
6
O
O
C
(0
4->
CO
•H
CO
0)
PS
0.2 0.4 0.6 0.8
Forward Voltage (V)
1 1 1 1 1 1 r— i i
1
— r-
-
syyyjfow
.'//.■VA ft-W
.'//.•//'•'c'ty
■ '/■//•■■ .''9
.' '' •'///•' 'tiff
-
/////Mm
////////''•/&
23
K
40
K
'■ /'/.•//::•• i /if
/////// f/'/ff
60
K
80
K
—
.'• ,-'// ''/•''/ '■'■'if
<■' '/■■''/.'''■'■'//
< ' / ; / /.'a.' /if
100
K
~
120
K
' / - f / ' iliiff
140
K
160
K
-
■-■-' - — / •' '' /■' ':'/?/
180
K
•-"'" /' ■' •'//■' ■'////
200
K
y .^-^^ „__- — ■"'* .*' / / i / ' ? iff
220
K
,■' .-•'' """ -■' //•'.'•' i itif
240
K
' / ,-'•"' ..-•'"" /////if
260
K
—
280
K
.
1 1 1 1 1 1 1 1 1
299
K
1.0
1.2
100 150 200
Temperature (K)
Fig. 8. Forward I(V) (top) and series resistance (bottom) versus temperature of device A.
Page 104
Third International Symposium on Space Terahertz Technology
-p
c
CD
u
u
o
<v
co
u
CD
>
OS
le-04
1 1 1 1 1 1 1 —
— 1 1 1 1 T 1 1 1
1
j&5-' ■■'■■■ y y
/
j&r.-- .--'.--" y^ s
j4P-- .■■.■■ /
/
O* '
■■' •■' / / ••'' /
£?/ *
: : / / ■■ y
le-05
// / / / X
-
It! 1 * '.
: ■ / /
fin*' :
: / / S*
flih : .
' 1 f / J"
-
le-06
Ill i /
/ / J^S 40
K "
ki! '■ '• : :
J^V •-•'' 60
K
Ki! ! ■ ij.
p!>-~- 1 -'..
*//.-■■' 80
K -
//-•'' 100
K
le-07
y .--pit >-'2>> /
■/■/ 120
'' 140
K — -
K — - "
168
K
/ .■■'' J*i'l><r
190
K -
\ S ,-\'"'W^0
200
K
le-08
1 / / /j0-
220
240
K
K - — "
260
K
.' t / .'IB?*'
295
K -
le-09
/ / §
1 1 1 1 1 1 1
-1 1 1 .1,. J 1 L 1
1
Fig. 9.
2 4 6 8 10 12
Reverse Voltage (V)
Reverse I(V) versus temperature of device A.
14
16
18
The measured series resistance of device A is plotted as a function of temperature in Fig.
8 (bottom). The resistance decreased nearly linearly from 19.9 Q at 300 K to 10.5 Q. at 130 K,
and the resistance at 77 K was 8.9 Q. The 2-DEG resistances calculated using Eq. 3 with the
low field mobility values at 300 K and 77 K are 14.1 Q. and 3.0 Q, respectively. Thus, the
ohmic contact resistivity (for the width 90 Jim) is estimated to be about 0.6 Qmm. This value
of Tsc is much lower than was achieved on the earliest Schottky/2-DEG devices. The
improvement is probably a consequence of the higher doping at the heterostructure surface and
an improved ohmic plating and alloying procedure. Further reduction of r sc (perhaps to 0. 1
Qmm) should be possible using evaporated Ni/Ge/Au ohmic contacts.
The reverse I(V) of device A was measured as a function of temperature and is shown in
Fig. 9. As in the forward I(V), the leakage current is seen to decrease upon cooling. At the
highest reverse current (100 (j.A),"the reverse voltages decreased with temperature from 17 V at
300 K to about 9 V at 40 K. This temperature dependence of the breakdown voltage is
qualitatively consistent with impact ionization theory. Since the mean free path for electron
phonon interactions increases with decreasing temperature, the electrons can achieve higher
kinetic energies before phonon scattering occurs. Consequendy, as the temperature decreases,
electrons reach the impact ionization energy at lower field strengths (lower reverse voltages)
and the breakdown voltage decreases.
Third International Symposium on Space Terahertz Technology
Page 105
le-02
le-03
le-04
le-05
c
u le-06
u
5 le-07
le-08
le-09 -
1 1 1 1 1 1 1 1 1 1 1
-i r
i
/ ■ ' / ' / / /
/ / / / / / if
/ / •' ' / / 1
■■ ■■ ■' 1 i i
■■' •' ■' •' / : if
-
/ / / / ' ' 1
/flllli
-
! ••" / / / ///
299
K
/////:.'/
260
K
220
K — -
••' / / / / : '•
180
K — .
///////
140
K —
100
K
/ / / / / / !
60
K
//////
24
K —
//////
i i i i i i i i i i i
< i
i.
le-04
le-05
~ le-06
■p
c
m le-07
o
le-08
le-09
0.2 0.4 0.6 0.8 1.0 1.2 1.4
Forward Voltage (V)
ii i i i ■ —
"22K"
"5 OK"
"100K"
"142K"
"200K"
"263K"
"299K"
10 15 20
Reverse Voltage (V)
25
30
Fig. 10. Forward I(V) (top) and reverse I(V) (bottom) vs. temperature of planar GaAs
Schottky diode SC6T1 .
Page 106 Third International Symposium on Space Terahertz Technology
The temperature dependence of the breakdown voltage of the Schottky/2-DEG diode
raised the question of whether the breakdown voltage in standard GaAs varactors had a similar
temperature dependence. To investigate further, the I(V) of several planar GaAs varactor
diodes (UVA-SC6T1) were measured as a function of temperature. The doping concentration
and anode diameter for this diode were 2 x 10 16 cm -3 and 6 urn, respectively, compared to the
respective values 3 x 10 16 cm -3 and 6 Jim for the whisker-contacted UVA-6P4 diode. The
forward and reverse I(V) of one device is shown in Fig. 10. The forward I(V) is typical for a
low doped diode. Here, the knee voltage increased from 0.678 V at 299K to 1.015 V at 24K.
The inverse slope parameter decreased from 28.7 mV at 299 K to 12.7 mV at 24 K. The series
resistance was constant and equal to 30 Q (± 1 Q) over the range 24-299 K. The reverse I(V)
was measured from 10 nA to 10 ^lA. The breakdown voltage characteristics are all very sharp.
Of the three SC6T1 diodes tested, all had breakdown voltages of 18-19 V at 24K, although the
room temperature breakdown voltages were 26 V, 29 V and about 30 V. Thus, the average
reduction in Vbr was about 35 percent upon cooling, compared to a reduction of about 50
percent for the Schottky/2-DEG diode. A decreasing breakdown voltage in these devices upon
cooling may impact the low temperature multiplier performance. On the other hand, the large
decrease in the series resistance of the 2-DEG upon cooling should result in higher multiplier
efficiency.
Finally, the saturation drift current of device A was calculated, using Eq. 5 with
v e ff = 2.4 x 10 7 cm/s, to be 64 mA. This may be a conservative estimate since higher velocities
may be possible in the 2-DEG than in bulk GaAs. In any case, the Schottky/2-DEG diode of
roughly equivalent properties as those of the 6P4 diode has a significantly higher saturation
current (at least 64 mA) than does the 6P4 diode (44 mA according to [14]).
VI. PRELIMINARY MULTIPLIER RESULTS
The first rf measurements of a Schottky/2-DEG diode were performed at the National
Radio Astronomy Observatory using a modified NRAO tripler which was designed [16] for
whisker contacted diodes (such as the UVA-6P4). The Schottky/2-DEG diode chip was
soldered across the output waveguide. A klystron was used as a source and two power meters
were used to measure the power at the input and output ports of the multiplier. Further details
of the measurement setup may be found in [17].
The preliminary measurements were of device A having chip width and length
dimensions of 100 (im and 180 |im. The chip thickness was about 60 fim. After soldering, the
clearance between the back of the chip and the waveguide wall was estimated to be only 2-3
(jm, due in part to the thick solder "bump". The I(V) of the device was checked after soldering
and found to be the same as before soldering. The room temperature R s and Wy^ of this device
were 20 Q and 12 V, respectively.
The first multiplier test was to determine the optimum frequency. Using 50 mW input
power, the multiplier tuners and dc bias were optimized for maximum output power at several
frequencies over the range 70-79 GHz. The result is shown in Fig. 1 1 where the input power
was 50 mW. The best performance was obtained at 75 GHz where P out was 160 p.W at 225
GHz. The input return loss was measured at 75 GHz to be about -20 dB so the input tuning was
relatively well optimized. Next, the output power at 225 GHz was measured as a function of
Third International Symposium on Space Terahertz Technology
Page 107
200
180
160
3
140
u
120
0)
15
o
100
4->
3
a
4->
3
o
80
60
40
20
550
500
450
s
3
400
350
M
300
o
250
-p
3
-p
200
3
O
150
100
50
72 74 76
Input Frequency (GHz)
80
40 60 80
Input Power (mW)
100
120
Fig. 11. Tripler output power versus input frequency (top) with Pin = 50 mW and output
power versus input power (bottom) for tripling to 225 GHz.
Page 108 Third International Symposium on Space Terahertz Technology
the input power, as shown in Fig. 11 (bottom). P^ was varied over the range 20 -100 mW. As
the input power increased, the output power increased, reaching a maximum value of 500 |iW
at 100 mW at the input. This corresponds to an efficiency of 0.5 percent. In comparison,
Bradley [17] used this multiplier with a planar GaAs varactor of doping 1.1 x 10 17 cm -3 and
diameter 7 |im and obtained 3.7 mW output power and 4.1 percent efficiency at 219 GHz. This
first multiplier measurement of the Schottky/2-DEG diode is encouraging but much higher
performance should be possible. Even the current devices should yield higher multiplier
efficiencies if they are thinned to 25 [im or less and if a thinner solder layer is used to reduce
the shunt capacitance between the chip and the waveguide wall.
vn. Summary and future research
In summary, we have reported on the recent progress in the research of a novel
Schottky/2-DEG varactor diode. Observations of reduced breakdown voltages upon cooling in
both the standard GaAs and the novel Schottky/2-DEG diodes were in agreement with the
theory of impact ionization. The problem of current saturation was discussed and the
Schottky/2-DEG diode of width 90 Jim was found to have a significantly higher saturation
current than the comparable GaAs 6P4 varactor. Recent improvements to the design and
fabrication procedures have resulted in devices having lower series resistance and lower
capacitance. Both single anode and dual anode (with symmetric C(V)) devices have been
investigated. The cutoff frequency of the dual anode device was estimated to be 1 THz (4 THz)
at 300 K (100 K), whereas the single anode device had cutoff frequency of about 0.6-1.0 THz,
depending on the temperature. Preliminary multiplier measurements of a single anode device
were encouraging, resulting in 500 fiW output power at 225 GHz with 0.5 percent efficiency.
Ongoing research will include more extensive multiplier testing of both the single anode and
the symmetric C(V) devices. Also, shorter channel length devices with Ni/Ge/Au ohmic
contacts will be fabricated to achieve much higher cutoff and transit-time frequencies. In
addition, AlInAs/InGaAs/InP heterostructures will be investigated. Finally, the theories
relating to the junction capacitance and breakdown are being developed and the current
transport in these devices will be investigated using Monte Carlo simulations.
ACKNOWLEDGEMENTS
This work has been supported by the National Science Foundation under contracts ECS-
8720850 and ECS-91 13123 (contract monitors Dr. T. Hsiang and Dr. B. Clifton) and by. the
Office of Naval Research under contract #N00014-90-J-4006 (contract monitor Dr. Y.S. Park).
The authors thank R. Bradley and N. Horner of the National Radio Astronomy Observatory for
assistance with the multiplier measurements, and T. Hierl of Quantum Epitaxial Designs, Inc.
for providing the MBE material and Van der Pauw data.
Third International Symposium on Space Terahertz Technology Pagel09
REFERENCES
[I] H. P. Roser, "Heterodyne Spectroscopy for Submillimeter and Far-Infrared Wavelengths From 100
urn to 500 urn," Infrared Physics, Vol. 32, pp. 385-407, 1991.
[2] J. Zmuidzinas, A.L. Betz, and R.T. Boreiko, "A Corner-Reflector Mixer Mount for Far Infrared
Wavelengths," Infrared Phys., Vol. 29, No. 1, 119-131, 1989.
[3] N. Erickson, "High Efficiency Submillimeter Frequency Multipliers," The 1990 IEEE MTT-S Int.
Microwave Symposium, Dallas, TX, May 1990.
[4] P. Zimmermann, private communication, March and May 1990.
[5] T. W. Crowe, W.C.B. Peatman, and E. Winkler, "GaAs Schottky Barrier Varactor Diodes for
Submillimeter Wavelength Power Generation," Microwave and Optical Tech. Lett., Special Issue:
Space Terahertz Tech., Vol. 3, No. 1, January 1991.
[6] K. Benson and M.A. Frerking, "Theoretical Efficiency for Triplers Using Nonideal Varistor Diodes
at Submillimeter Wavelengths," IEEE Trans. Microwave Theory Tech., Vol. MTT-33, No. 12,
1367-1374, Dec. 1985.
[7] C. Kidner, I. Medhi, J. East, G. Haddad, "Performance Limitations of Resonant Tunneling Diodes,"
The First Int'l. Symposium on Space Terahertz Technology, Ann Arbor, MI, March 1990.
[8] H. Rothermel, T.G. Phillips, J. Keene, "A Solid-State Frequency Source for Radio Astronomy in
the 100 to 1000 GHz Range," Int. J. Infrared and Millimeter Waves, Vol. 10, No. 1, 83-100, 1989.
[9] W. C. B. Peatman, T. W. Crowe and M. Shur, "Design and Fabrication of Heterostructure Varactor
Diodes for Millimeter and Submillimeter Wave Multiplier Applications," Proc. IEEE/Cornell
Conf. on Advanced Concepts in High Speed Semic. Dev. and Circuits, Ithaca, NY, 1991.
[10] W.C.B. Peatman, T.W. Crowe and M. Shur, "A Novel Schottky/2-DEG Diode for Millimeter and
Submillimeter Wave Multiplier Applications," IEEE Electron Device Lett., Vol. 13, No. 1, pp. 11-
13, January 1992.
[II] S.G. Petrosyan and A. Ya Shik, "Contact Phenomena in a two-dimensional electron gas," Soviet
Physics Semicond., 23 (6), pp. 696-697, June 1989.
[12] B. Gelmont, M. Shur and C. Moglestue, "Theory of Junction Between Two-Dimensional Electron
Gas and P-Type Semiconductor," to be published, IEEE Trans. Electron Devices, Vol. 39, No. 5,
May 1992.
[13] P. Penfield and R. Rafuse, Varactor Applications, MIT Press, 1962, p. 86.
[14] E. Kollberg, T. Tolmunen, M. Frerking and J. East, "Current Saturation in Submillimeter Wave
Varactors," Proc. Second Int'l. Symp. Space Terahertz Technology, Pasadena, CA, pp. 307-322,
1991.
[15] B. Gelmont, M. Shur (unpublished).
[16] J. Archer, "An Efficient 200-290 GHz Frequency Tripler Incorporating A Novel Stripline
Structure," IEEE Trans. Micrwave Theory and Tech., Vol. 32, No.' 4, pp. 416-420, 1984.
[17] R. Bradley, Ph.D. Thesis, University of Virginia, pp. 42-65 May, 1992.
Page UO Third International Symposium on Space Terahertz Technology
3?~33 N9 3 „ a? 7 35
I
THERMIONIC EMISSION CURRENT IN A SINGLE BARRIER V ARACTOR
Hans Hjelmgren 3 ), Jack East b >, and Erik Kollberg 2 )
a) Applied Electron Physics, Chalmers University of Technology, S-412 96 Goteborg,
Sweden
b) Solid-State Electronics Laboratory, University of Michigan, Ann Arbor, MI 48109-
2122
/ Abstract — From I-V measurements on Single Barrier Varactors (SBV) at different
, temperatures we concluded that thermionic emission across the barrier of the actual device
: is mainly due to transport through the X band. The same structure was also modelled with
a one-dimensional drift-diffusion model, including a "boundary condition" for thermionic
emission across the heteroj unction interface. By including thermionic field emission
through the top of the triangular barrier of a biased diode and the effect of a non-abrupt
interface at the heterojunction, we obtained good agreement between the modelled and
measured I-V characteristics.
1. Introduction
SBV diodes have been proposed as an alternative to Schottky barrier diodes for harmonic
multiplier applications [1]. However, the device discussed in [1] showed a higher than
expected current. We will present experimental data and the results of a numerical model to
explain the current vs. voltage characteristics of SBV diodes.
2. Experimental results
The device consists of a 200 A wide undoped Alo^GagjAs barrier, surrounded by 5300 A
wide GaAs layers, doped to lxlO 17 cm -3 . On both sides of the AlGaAs layer there is a 50
A wide undoped GaAs spacer layer. If the effect of field emission is neglected the current is
mainly limited by thermionic emission across the barrier and can be described by the
Richardson law,
J = A*T 2 e-W v )/ kT (1)
where A* is the modified Richardson constant and <{>b is the bias dependent barrier height
Both can be determined from the experimental data by plotting InCJ/T 2 ) against 1000/T for
different voltages [2]. The intersection with the y-axis gives us the Richardson constant,
while the slope is proportional to the barrier height for that specific bias voltage. Since our
barrier is comparatively thin, the influence of tunneling is observable even at quite high
temperatures, causing a deviation from a straight line in Fig. 1. This makes the
determination of the Richardson constant quite cumbersome. It must be chosen in such a
way that the barrier heights at low voltages are close to the expected barrier height at zero
bias voltage. A Richardson constant of 0.30 Acm _2 K -2 , much lower than the value
expected for thermionic emission in GaAs but in fairly good agreement with that measured
by Solomon et al. [2], corresponds to a barrier height at low voltages of about 0.29 eV.
This low Richardson constant together with the high current indicates that the electrons are
transferring to the X valley within the barrier. The assumed offset in T conduction band
corresponds to 59 % of the difference in bandgap [3].
Third International Symposium on Space Terahertz Technology Page 111
3. Numerical results
A drift-diffusion model for one-dimensional heterojunction structures was also used to
study the device characteristics. It accounts for thermionic emission across the barrier by
calculating the current at the heterojunction interfaces from,
J = qTl 2 n 2 v r2 - A^erMYJ/kT (2)
where the first term describes the current from the AlGaAs side [4] while the second term is
the current from the GaAs side emitted above the barrier. Without this "bottleneck" for the
current, the drift-diffusion model results in a much too high current. Since a theoretical
determination of the actual emission constant across the interface is quite complicated, we
have used the experimentally determined Richardson constant in Fig. 1. The probability of
emission above the barrier depends on several parameters, such as thickness of the barrier,
crystal orientation, and roughness of the interfaces. The thermally emitted electrons were
assumed to be transferred to the X-valley when they reached the AlGaAs barrier, causing
an effective barrier height at zero bias of 0.29 eV. During a simulation the barrier height is
determined self consistently from the modelled conduction band, Fig. 2. Even at 300 K the
effect of thermally assisted tunneling is considerable, and it has to be included in the model.
Since the top of the biased barrier is triangular, the probability of tunneling as a function of
electric field and carrier energy can be estimated from the WKB approximation [5],
where E m ax is the electric field in the barrier and Ae is the energy distance from the top of
the barrier. The effect is included in the model by reducing the barrier height in eqn. (2)
with an amount Ae corresponding to a tunnelling probability of e -1 . We assumed the
tunneling process to be indirect tunneling in the AlGaAs X band [6], and used a transverse
electron mass in the X band of 0.20mo. Tunneling through the T band is less probable due
to the much higher barrier, and tunneling from X valleys with the longitudinal mass
perpendicular to the interface is less probable due to the higher mass. Apart from band
bending due to accumulation of electrons at the interface, and thermionic field emission, the
fact that the transition between the two materials is not completely abrupt will also result in
a field dependent barrier height [7]. In Fig. 2 we roughly modelled this effect by using a
grid distance of 25 A at the interface. The obtained I-V characteristic in Fig. 3 is very
sensitive to the grid distance at the right interface of the barrier, while its value elsewhere is
of less importance.
The model has also been used to predict the capacitance vs. voltage. The capacitance is
determined from the change of charge concentration in the depletion layer for an
incremental change in voltage. As can been seen in Fig. 4, the agreement between modelled
and measured C-V characteristics is fairly good. In order to get reasonably good accuracy
the grid-distance in the depletion layer and the voltage step should not be too large.
4. Conclusions
The presented expression for the current between the two grid-points adjacent to the
heterojunction interface (eqn. 2) includes effects caused by the existence of two interfaces
and other effects, which for a theoretical determination require quantum mechanical
calculations [8]. It also gives us a possibility to model the effect of a non-abrupt transition
between the two materials. Similar expressions, have been used before in drift-diffusion
models, but they are restricted to a single interface [3,4]. The drawback with the model
used here is that it relies on measured results in order to find a value for the Richardson
Page 112 Third International Symposium on Space Terahertz Technology
emission constant. However, since the actual emission constant may be quite difficult to
calculate, we considered it as a device parameter specific for the actual device dimensions
and growing conditions, which could be measured instead of calculated. The good
agreement with measured results indicates that the current is mainly due to thermionic
emission and thermionic field emission across the X-valley of the barrier.
The inaccuracy in the experimental determination of the Richardson constant can be reduced
by using slightly thicker barriers and by performing measurements at elevated temperatures
and lower bias voltages.
References
1 . A. Rydberg, H. Gronqvist, and E. Kollberg, "Millimeter- and submillimeter-wave
multipliers using quantum-barrier-varactor (QBV) diodes," IEEE El. Device Lett., vol.
11, pp. 373-375, 1990.
2. P. Solomon, S. Wright, and C. Lanza, "Perpendicular transport across (Al,Ga)As and
the r to X transition" Superlattices and Microstructwes, vol. 2, pp. 521-525, 1986.
3 . G. B. Tait and C. R. Westgate, "Electron transport in rectifying semiconductor alloy
ramp heterostructures," IEEE Trans. Electron Devices, vol. 38, pp. 1262-1270, 1991.
4. K. Horio and H. Yanai, "Numerical modeling of heteroj unctions including the
thermionic emission mechanism at the heterojunction interface," IEEE Trans. Electron
Devices, vol. 37, pp. 1093-1098, 1990.
5. E. H. Rhoderick and R. H. Williams, "Metal-Semiconductor Contacts, 2nd edition,"
Oxford, England: Claredon, 1988.
6. E. E. Mendez, E. Calleja, and W. L. Wang, 'Tunneling through indirect-gap
semiconductor barriers," Physical Rev. B, vol. 34, pp. 6026-6029, 1986.
7. S. C. Lee and G. L. Pearson, "Rectification in A^Ga^As-GaAs N-n heterojunction
devices," Solid-State Electronics, vol. 24, pp. 561-568, 1981.
8. M. Rossmanith, J. Leo, and K. von Klitzing, "Model of T to X transition in thermally
activated runnel currents across Al x Gai_ x As single barriers," J. Appl. Phys., vol. 69,
pp. 3641-3645, 1991.
Third International Symposium on Space Terahertz Technology
Page 113
1 2 3 4 5 6 7 8 9 10 11 12 13
1000/T, K-l
Fig. 1. Plotting of experimental data in a InCJ/T 2 ) vs. 1000/T diagram. The applied
voltages are 0.1 V, 0.5 V, 1.0 V, 1.5 V, 2.0 V, and 2.5 V (filled squares).
— o
■- i
- 2
100 200 300 400 500 600 700 800 900
Distance, nm
Fig. 2. Simulated conduction bands for different applied voltages (0.0 V, 1.0 V, and
2.0 V).
Page 114
Third International Symposium on Space Terahertz Technology
1.0000
Modelled
O Measured
0.0001
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Voltage, V
Fig. 3. Comparison between modelled and measured I-V characteristics at 300 K. The
effects of thermally assisted tunneling and a nonabrupt heterojunction interface are
included in the simulation.
0.0022
0.00
0.0018
<n 0.0016
e
^ 0.0014
o 0.0012
£ 0.001
o
* 0.0008
<o
u 0.0006
0.0004f
0.0002
— .— Modelled
O Measured
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
Voltage, V
Fig. 4. Comparison between modelled and measured C-V characteristics at 300 K.
Third International Symposium on Space Terahertz Technology Pagell5
N93-27736
PROGRESS ON SINGLE BARRIER VARACTORS FOR
SUB MILLIMETER WAVE POWER GENERATION S/o "^33
Svein M. Nilsen, Hans Gronqvist, Hans Hjelmgren, Anders Rydberg, cf
and Erik L. Kollberg I - '
The Millimeter Wave Group
Department of Applied Electron Physics, Chalmers University of Technology,
S -4 12 96 Goteborg, Sweden
ABSTRACT
Theoretical work on Single Barrier Varactor (SBV) diodes, indicate that the
efficiency for a multiplier has a maximum for a considerably smaller capacitance
variation than previously thought. The theoretical calculations are performed, both with
a simple theoretical model and a complete computer simulation using the method of
harmonic balance. Modelling of the SBV is carried out in two steps. First, the
semiconductor transport equations are solved simultaneously using a finite difference
scheme in one dimension. Secondly, the calculated IV-, and CV characteristics are input
to a multiplier simulator which calculates the optimum impedances, and output powers at
the frequencies of interest. Multiple barrier varactors can also be modelled in this way.
Several examples on how to design the semiconductor layers to obtain certain
characteristics are given. The calculated conversion efficiencies of the modelled
structures, in a multiplier circuit, are also presented. Computer simulations for a case
study of a 750 GHz multiplier show that InAs diodes perform favourably compared to
GaAs diodes. InAs and InGaAs SBV diodes have been fabricated and their current vs.
voltage characteristics are presented. In the InAs diode, was the large bandgap
semiconductor AlSb used as barrier. The InGaAs diode was grown lattice matched to an
InP substrate with InAlAs as a barrier material. The current density is greatly reduced
for these two material combinations, compared to that of GaAs/AlGaAs SBV diodes.
GaAs based diodes can be biased to higher voltages than InAs diodes.
1. INTRODUCTION
A SBV device, in its simplest form, consists of a thin layer of a large bandgap
material which acts as a barrier, and a thicker layer of a small bandgap material on each
side of the barrier. A cross section of a typical SBV diode is shown in Fig. 1. The low
bandgap material at both ends of the device is normally highly doped in order to make
possible, the formation of low resistance contacts. Provided the layer thicknesses and
doping concentrations are symmetrical around the barrier, the current I vs. voltage V and
the capacitance C vs. voltf.ge V will be be symmetrical around zero volt. An applied rf-
voltage will then generate only odd harmonics. This makes it possible to design higher
order multipliers with fewer idler circuits and thus less losses. Also the design
procedure and mechanical construction of a higher order multiplier, making use of a
SBV device as the non linear element will become much easier compared to that of one
using a Schottky diode. It is the purpose of this paper to give an overview of our work
on SBV diodes [1][2][15].
Page 116
Third International Symposium on Space Terahertz Technology
Ohmic contact
Substrate
Ohmic contact
Fig. 1 . Schematic cross section of a SBV mesa diode. LI is the highly doped contact
layer, L2 n and L4 n are the depletion regions on either side of the barrier, index
n denotes sublayers, L3 n is the barrier and L5 is part of the buffer/substrate
layer. In the simulations LI and L5 are set equal. Mesa diameter = a.
2. DIODES
2.1 Quality factor of SB V's
The cutoff frequency for a varactor, as defined in Eq. 1, is often used as a quality
factor, suggesting that C ma x should be as large as possible for a fixed C m i n .
fr =
1
1
1
2 k R s i \ Cmin Q
max
(1)
where Rslo is the dc series resistance of the device, C max and C m in are the calculated
device capacitances for accumulated and depleted low doped epilayers, L2 and L4 in
Fig.l, respectively.
However, a simple analysis show that an optimum C max /C m in ratio exist. Choosing a
CV characteristic as shown in Fig. 2 and assuming that all harmonics, except the first
and third harmonics are shortcircuited over the variable capacitance, only these two
frequency components have to be considered. The bias voltage is of course always zero
volts. This makes the analysis quite simple. In Fig. 3 we have plotted the efficiency vs.
C m ax/Cmin for different values of R s , assuming l/a)C m i n = 180 Q. The important
conclusion from this graph is that there is an optimum value for C ma x/C m in , and that
too large C max values will deteriorate the multiplier's efficiency. Table 1 shows input
and output impedances. Although this model is much simplified the results show that the
input is highly reactive with a fairly low resistance while the output seems to be easier to
match. This result will be confirmed below in the harmonic balance simulations
employing a modified version of the computer program by P.H. Siegel et.al. [8].
Third International Symposium on Space Terahertz Technology
PageU7
-max
C = (C max + Cmin)^ + [(Cmax - Cmin)/2] cos(V)
C = Co = (C max + Cmin)/2
Fig. 2 Approximate CV characteristic of SB V diode used in the simplified analysis.
Vmax is the maximum voltage across the diode.
(C /C . )
v max mm 7
Fig. 3 Efficiency vs. the C m ax/Cmin ratio for different values of R$.
Table 1 Calculated efficiencies, optimum input and output imedances for a SBV diode
tripler with the approximate cosine shaped CV characteristic for different series
resistances.
SOURCE IMPEDANCE
LOAD IMPEDANCE
RsQ
Tlmax %
Rl n
XI fl
R3ii
X3Q
2.5
65.5
8.6
91.0
29.5
34.0
5
46.7
10.9
90.0
30.4
33.0
10
27.2
15.7
95.0
30.2
36.0
20
11.8
25.0
90.0
23.0
33.0
40
2.9
44.1
106.0
14.0
38.0
Page 118
Third International Symposium on Space Terahertz Technology
2.2
MEASUREMENTS
2.2.1 DC measurements
We have fabricated several diodes based on the materials: GaAs, InGaAs and InAs,
see Table 2 for details. Our efforts at modelling these devices are however, more recent.
For the GaAs based diode we use a barrier of Al x Gai- x As, in the InGaAs case we use
AUnAs as a barrier. The latter diode is grown lattice matched on an InP substrate. For
the InAs diodes we use an AlSb barrier which gives a very high barrier, 1.3 eV for
indirect transitions. It should therefore be very efficient in blocking the current through
the varactor. The InAs device is grown on a GaAs substrate with thick GaAs and InAs
buffer layers.
Table 2 Material from which diodes of various sizes have been
fabricated.
Wafer #ID&
Crystal
Grower
Material system
Layer thickness (top/down) and
doping concentration
= 400 nm, Nd = 3.4* 10 18 [cm' 3 ]
= 533 nm, Nd = 1.0*10 17 [cm* 3 ]
= 5.3 nm, undoped
: 21.3 nm, undoped
= L22, L4i = L2i
1998 n m, Nd = 3.4* 10 18 [cm' 3 ]
#614
IMEC
GaAs/Al x Gai. x As/GaAs
x = 0.70 in L3
Buffer n + GaAs
Substrate: n + GaAs
LI
L2i
L22
L3 =
L42
L5 =
#1566
Chalmers
InAs/AlSb/InAs
Buffer n + InAs, and n + GaAs
Substrate: n + GaAs
Barrier thickness = L3
LI = 200 nm, Nd = 5.0* 10 18 [cm" 3 ]
L2i = 400 nm, Nd = 6.0*10 16 [cm" 3 ]
L22 = 5 nm, undoped
L3 = 20 nm, undoped
L42 = L22, L4i = L2i
L5 = 2000 nm, Nd = 5.0*10 18 [cm' 3 ]
200 nm, Nd = 5.0* 10 18 [cm' 3 ]
= 320 nm, Nd = 1.0*10 17 [cm" 3 ]
= 5 nm, undoped
= 25 nm, undoped
= L22,L4i = L2i
2000 nm, Nd = 5.0* 10 18 [cm' 3 ]
#1567
Chalmers
InAs/AlSb/InAs
Buffer n + InAs, and n + GaAs
Substrate: n + GaAs
Barrier thickness = L3
LI i
L2i
L2 2
L3 ■■
L42
L5 =
400 nm, Nd = 4.0* 10 18 [cm' 3 ]
400 nm, Nd = 6.0* 10 16 [cm" 3 ]
25nm,Nd = 6.0*10 16 [cm- 3 ]
#ST1
Tampere
Lattice matched
InGaAs/AlInAs/InGaAs
Buffer. n + InGaAs
Substrate: n + InP
Barrier thickness = L3
LI
L2
L3
L4
L5
= L2
= 1000 nm, Nd = 4.0* 10 18 [cm' 3 l
" •" " iz^,
. 200 nm, Nd = 2.6* 10 18 [cm" 3 ]
> 150 nm, Nd = 1.0* 10 17 [cm" 3 ]
: 14 nm, undoped
= L2
■ 3000 nm, Nd = 2.6*10 18 [cm' 3 ]
#1047
IMEC
InAs/AlSb/InAs
Buffer n + InAs, and n + GaAs
Substrate: n + GaAs
Barrier thickness = L3
Ll =
L2 =
L3 =
LA-
L5>
Third International Symposium on Space Terahertz Technology
Page 119
The influence of the epilayer design on the IV-, and CV characteristics has been
investigated theoretically in some detail.
The different SBV diodes fabricated by us have been DC characterized with a DC
parameter analyzer. Most of the measurements have been done on large area diodes,
100- 2500 |im 2 . A limited number of measurements have been made on small area,
7 |im 2 , diodes of the #614 type.
Diodes characterized are one GaAs varactor with an AlGaAs barrier, one diode grown
on InP with InGaAs and InAlAs epilayers lattice matched to the substrate and two InAs
diodes with AlSb barriers grown on GaAs substrates. The material from which we have
fabricated devices are listed in Table 2. DC characterization of the different SBV diodes
show that the current density is decreased as the barrier height is increased as expected,
see Fig. 4. The InAs/AlSb material system has a 1.3 eV barrier for indirect transitions
and = 2 eV for direct. The lowest current density is obtained for InAs/AlSb devices.
They also show a large variation in their characteristics, a factor of two or more in
current density over a distance of a few millimeters on the wafer. Whether this variation
is caused by the epilayers being severely lattice mismatched to the substrate or by some
growth parameter, is difficult to say. In comparison, diodes of the lattice matched
InGaAs/InAlAs materials grown on InP substrate as well as those of GaAs/AlGaAs
show current density variations of only a few percent over the same distance. It can also
be noted that the breakdown voltage of the InAs and InGaAs diodes is low, about 2
Volt, in comparison with the GaAs diode. For a varactor to be used in practice the
breakdown voltage needs to be higher since the pump voltage otherwise could destroy
the diode. One method of increasing the total breakdown voltage is to fabricate several
varactor diodes in series. Each of n identical diodes would then only need to sustain the
total pump voltage divided by n.
0,2 0,4 0,6 0,8
Voltage (V)
1,2 1,4
Fig. 4 Measured current vs. voltage for SBV diodes of size 25*25 jim 2 in different
material systems.
Page 120 Third International Symposium on Space Terahertz Technology
2.2.2 RF measurements
Two methods have been developed for measuring the capacitance of these diodes,
the results will be published elsewhere [14]. One method is similar to the one by
Tolmunen et. al. [15], the other method uses a 60 GHz coplanar probe to contact a mesa
diode surrounded by a large ring which is contacted by the two ground contacts of the
probe.
2.3 MODELLING DEVICES AND MULTIPLIER PERFORMANCE
2.3.1 CV and IV modelling
For calculating the CV-, and IV characteristics, the semiconductor transport
equations including the energy balance equations for holes and electrons are solved
simultaneously using a finite-difference scheme in a detailed one-dimensional energy
transport model for GaAs/Al x Gai_ x As heterojunction structures with 0.0 < x < 1.0.
The computer program is a modified version of the program developed by H. Hjelmgren
[3] [4]. The program can readily be modified further for other heterostructures. A non-
uniform grid-mesh is used to obtain the necessary accuracy and acceptable calculation
times. Position dependent barrier parameters were implemented following the procedure
of M. S. Lundstrom and RJ, Schuelke [5]. The model used is based on the paper by R.
Stratton [6]. The five variables are the electric potential, the quasi-Fermi levels for holes
and electrons, and hole and electron temperatures. The equation system has a 19
diagonals band-matrix which is solved by LU decomposition. To improve the
convergence properties of the program the energy balance equation for holes is not used,
hence the hole temperature is set to the lattice temperature for all applied voltages. Hot
electrons and lattice heating are not simulated, hence the five equation model is reduced
to a three equation (Poisson's and current continuity equations) model. Recombination
is modelled with Shockley Read Hall (SRH) recombination. The relaxation time was set
to 3.0 ps, the electron life time to 1.0 ns, and the conduction band-offset to 65 % in
most simulations for which x < 0.45. The contact resistivity of the two ohmic contacts
can be set to a value of choice. They are modelled by modifying the boundary conditions
for the potential. The thermoionic emission current, which limits the current through the
device, is then calculated from the equation for a reverse biased Schottky diode, using a
voltage dependent barrier-height obtained from the shape of the conduction band.
Tunneling current is assumed to be sufficiently small and thus neglected altogether.
The diode capacitance, C= dQ/dV == AQ/AV is determined numerically from the change
AQ in the stored electric charge in the semiconductor for an incremental change AV in the
applied voltage. The integration of charges is carried out to the middle of the barrier for
single barrier devices. To obtain the depletion capacitance one must correct for the
influence of the series resistance on the voltage across the depletion layer.
2.3.2 CV characteristics and conversion efficiencies of modelled structures.
It is of importance to understand how the CV characteristic can be modified by
changes to the design of the epitaxial structure. It is obvious that increasing the width of
the barrier will decrease C max , and that a wider L2 region, see Fig. 1, will allow a larger
capacitance swing, i.e. make C m i n smaller. Below we will investigate how the doping
profile of the diode affects the efficiency of a 3*60 GHz tripler. The effect of the plasma
resonance on the series resistance was not included in the multiplier calculations on
structures #100-103B, #1 12 and #1 14 at 3*60 GHz.
Third International Symposium on Space Terahertz Technology Page 121
The following rules are basic to the understanding of the behaviour of the CV
characteristic for different choices of doping concentrations, viz.
i . Doping the barrier will increase C max without influencing the capacitance value at
large biases, i.e. C m in is not much affected. This phenomenon is studied in structures
#100-103B. However, too many dopants in the barrier and its blocking capability may
be destroyed.
ii. A similiar effect, an increase in Cmax, can be obtained by introducing thin regions
with high doping concentration adjacent to the barrier on each side. One such example is
#1 12. This scheme may be less damaging to the barrier's blocking capability, but it may
also affect material quality in a negative way or make symmetrical growth more difficult.
iii. A reduced doping concentration in the L2 layer where the depletion occurs, will
cause a more rapid change in the capacitance vs. bias voltage. It will also lead to a larger
series resistance. The method cannot therefore be recommended as a stand alone
measure.
i v . To avoid an increase in the series resistance according to iii., a tapered doping
concentration in the L2 region will improve the situation. A comparison between #112
and #1 14 illustrates this phenomenon.
Please note that the choice of doping profile will also influence the series resistance. An
advantageous CV characteristic may offer an undesireably large series resistance.
The effect of doping the barrier on the CV characteristic is illustrated in Fig. 5 which
show the results for structures #100-103B. The increase in C ma x is as large as 70 %
when the doping is increased from 1*10*5 cm~3 to 1*10*8 cm'3. To see any effect at
all, the doping in the barrier needs to exceed 1*10 17 cm"3. The series resistance remains
the same of course, since it is essentially determined by the doping in region L2. In Fig.
6 is shown the conversion efficiency vs. input power (all input power is assumed to be
absorbed). It is interesting to note that the efficiency actually decreases as C ma x
increases. This is in agreement with the results presented in paragraph 2.1.
The effect of a high doping concentration near the barrier interface is illustrated by the
CV characteristic of #1 12 in Fig. 7. Notice that the CV is almost identical to the one of
#103. Since the series resistances are also very similar, the efficiencies vs. pump power
are virtually identical as can be seen in Fig. 6 and Fig. 8.
Structure #114 has a linearly graded doping concentration in the L2 and L4 regions,
starting with 1*10*5 cm* 3 near the barrier and increasing to 2*10*7 cm" 3 at a distance
of 385 nm from the barrier. The ratio C ma x/C m in is increased considerably compared to
that of a homogenously doped L2 region, but the series resistance is larger than for
#100-103B. The CV resembles mostly that of #103B. The capacitance swings down
faster, but have the same C m i n and a small reduction in efficiency.
Page 122
Third International Symposium on Space Terahertz Technology
4.0 10" 14
^
3.0 10"
14 _..
0.0 10°
1111
\
"•. \
\
' ■ ■ ■ ■ I
#100, 10 \im2
#101, 10 \irr\2
#102, 10 \im2
#103, 10 u,m2
#103B, 10 u.m2
l i i i i
2 3
Voltage [V]
Fig. 5 Calculated CV characteristics for symmetric GaAs/Alo.44Gao.56As structures
#100-103, #103B with different barrier doping. L2 = L4 = 390 nm with
1.0* 10 17 cm- 3 , L3 = 20 nm, for #100: Nd = 1.0* 10 15 cm-3, for #101:
Nd = 1.0*10 16 cm" 3 , for #102: Nd = 1.0*10 17 cm' 3 , for#103B:
Nd = 5.0* 10 17 cm" 3 , for #103: Nd = 1.0*10! 8 cm' 3 .
For all structures LI = L5= 100 nm, Nd = 3.4*10 18 cm' 3
Third International Symposium on Space Terahertz Technology
Page 123
#
o
c
U
c
o
>
c
a
20 30 40 50 60
Absorbed Power [ mW ]
Fig. 6 Calculated conversion efficiency vs. pump power for structures #100- 103 and
#103B at 3*60 GHz. It is the devices with undoped or low doped barrier
which have the highest conversion efficiency. The series resistance Rslo used
was between 32.9 and 33. 1 ohm (calculated by the simulation program) for all
five structures.
Page 124
Third International Symposium on Space Terahertz Technology
uu
o
c
(X
U
4.0 10"
3.0 10 14 -\
2.0 10" 14 H
1.0 10" 14 -I
0.0 10
2 3
Voltage [ V ]
Fig. 7 Calculated CV characteristics for symmetric GaAs/Alo.44Gao.56As structures
for different doping profiles in the L2 region meant to enhance the Cmax/C m i n
ratio.
#112 has L2i = L4i = 386 nm with Nd going from 1.0*10 17 cm* 3 to
0.8*10*7 cm"3 at the barrier interface, i.e. a weak gradient, L22 = L42 =
4 nm with Nd =1.0*10 17 cm"^ L3 consists of three pans, L3i = L33 = 7 nm
Nd = 1.0*10 18 cm-3, L32 = 6 nm Nd = 2.0*10 17 cm' 3 .
#114 has L2i = IA\ = 385 nm with Nd going from 2.0* 10 17 cm" 3 to
1.0* 10*5 C m"3 at the barrier interface, i.e. a steep gradient, L22 = L42 =
5 nm with Nd =5.0*10^ cm" 3 , L3 consists of three parts, L3i = L33 = 6 nm
Nd = 1.0*10 18 cm" 3 , L32 = 8 nm Nd = 5.0*10 16 cm" 3 .
LI = L5 = 100 nm with Nd = 3.4*10 18 cm" 3 for both structures.
Third International Symposium on Space Terahertz Technology
Page 125
Fig. 8
*
u
c
.22
o
<o
c
o
'5
>
e
o
U
20
15 -
10 -
10 20 30 40 50 60 70
Absorbed Power [ mW ]
Calculated conversion efficiency vs. pump power for structures #112 and
#1 14 at 3*60 GHz. The value of R s \ was 33.8 ohm for #1 12 and 44.2 ohm
for #114.
3 . A 750 GHz MULTIPLIER DESIGN STUDY
A study on different 750 GHz multiplier configurations using GaAs and InAs
SBV-diodes was performed. The aim of the study was to evaluate theoretically the
optimum frequency multiplier configuration for generating > 50 |J.W of output power at
750 GHz, assumed to be enough for pumping a 750 GHz SIS mixer, using a Single
Barrier Varactor (SBV) diode as the nonlinear element.
Different important diode parameters were investigated, such as the cutoff frequency
which is a direct function of the doping level Nj of the low doped epilayers, L2 and L4
in Fig.l, in the device. The cutoff frequency was calculated for GaAs and InAs SBV-
diodes of three different sizes, see Fig. 9. In these calculations of Cmax and cutoff
frequencies, it was assumed that the effective barrier width was three times that of L3
because of the distance necessary for a lineup of the fermi levels at the interface between
the low doped layer L2 and the undoped barrier L3. It can be seen that the optimum
doping concentration is l*10^cm"^ and 2*10^cm"3 for GaAs and InAs devices
respectively, both having a diameter of 2 ^m.
Page 126
Third International Symposium on Space Terahertz Technology
X
S
o
c
<u
3
u
u
c
>»
Q
Doping concentration Nde, [ cm-3 ]
Fig 9 Calculated cutoff frequency vs. doping level Nde f° r GaAs and InAs SBV-
diodes of different diameters a. Nds = 3.4*l()18cm"3 in LI and L5, Ndb =
l*10 17 cm- 3 in L3, LI = 100 nm, L5 = 100 nm, L2 varies with Nde, the
barrier width L3 is 20 nm.
GaAs: (□): a = 2 (lm., (O): a = 4 |im., (A): a = 20 |im.
InAs: (■): a = 2 um., (•): a = 4 |im., (A): a = 20 |im.
Another important parameter to consider is the effect of the plasma resonance
frequency on the series resistance of the diodes [7]. This is of great importance to
devices intended for THz frequencies. In Fig. 10 is plotted the calculated series
resistance for InAs SBV-diodes when the plasmaresonance is taken into account . It is
found that by using InAs instead of GaAs in the low doped epilayers and in the
substrate, the plasma resonance frequencies are shifted to higher frequencies. This is
due to the higher electrical conductivity a of InAs compared to that of GaAs. It can be
seen in Fig. 10 that a doping Nd in L2 and L4 of about 1*10*7 cm" 3 gives sufficient
clearence up to a frequency of 2 THz
Third International Symposium on Space Terahertz Technology
Page 127
10'
II III I I I I I II I
10 (
Frequency (GHz)
Fig. 10 Calculated series resistance vs. pump frequency for an InAs SBV-diode for
different doping levels Nde in the low doped epilayers L2 and L4. Doping
concentration Nds is 3.4*10*8 cm" 3 in LI and L5, N D is 1*10*7 cm" 3 in the
barrier L3, LI =100 nm, L5 = 100 nm, L2 = L4 = 533 nm, L3 = 20 nm,
diameter a = 3.57 jim.
(•): Nde is 1*10 16 cm'3., (■): Nde is 1*10 17 cm' 3 .,
(A): Nd e isl*10 18 cm" 3 .
The avalanche breakdown voltage for the low doped epilayers limits the maximum
pump power that can be used for the device. Using the optimum doping concentration
the breakdown voltage was calculated to be 15.7 V for GaAs and 1.2 V for InAs, and
having depletion lengths of 480 nm and 100 nm respectively. Due to the very small
breakdown voltage for the low doped InAs epilayers, L2 and L4 in Fig. 1, the doping of
the epilayers was reduced slightly to a constant value of 1*10*7 cm -3 which gives a
breakdown voltage of 2 V and a length of 179 nm for the L2 and L4 epilayers.
The structure chosen for further study in the InAs/AlSb/InAs material system was the
following: LI = 100 nm with Nd = 3.4*10 18 cm" 3 , L2 = 150 nm with Nd = 1*10 17
cm-3, L3 = 14 nm with Nd = 5*10 15 cm-3, L4 = L2> L5 > LI.
Since the barrier length is choosen to be 14 nm thick, the tunneling current may for such
a thin barrier not be negligable in reality, especially not at room temperature, why a
thicker barrier may be more appropriate. The total current through the SBV-diode
consists of (i) the dc -current, mainly thermionic current, and (ii) the displacement
current due to the rf-voltage variation over the depletion region. Increasing the dc-
current compared to the displacement current makes the tripler work more in a varistor
mode with a reduced efficiency, see Fig. 1 1.
Page 128
Third International Symposium on Space Terahertz Technology
It is assumed in the simulations that the same IV and CV characteristic calculated for a
GaAs based SBV-diode can also be used for InAs based diodes, of a similar design.
This assumption is due to the fact that the difference in conduction current, see Fig.4,
between InAs and GaAs diodes at a fixed bias voltage, has been found to have a
negligible influence on the efficiency, see Fig. 11.
Input power (mW)
Fig. 1 1 Calculated conversion efficiency versus input power, when varying the
calculated conductance, dl/dV, for the device in steps of (dl/dV)* 10 v , where
y is 0, 1, 2 or 3. Nds = 3.4*10 18 cm"3, Nde =1*10 17 cm" 3 , Ndb = 5*10 15
cm-3, LI = 100 nm, L5 = 100 nm, L2 = L4 = 150 nm, L3 = 14 nm, diameter
a = 2^m. (■): y=0.0, (□): y=1.0, (O): y=2.0, (A): y=3.0
Three different multiplier configurations for generating the desired output power of
> 50 }iW at 750 GHz have been investigated, see Fig. 12. In the calculations was the
effect on the series resistance from the plasma resonance included.
Due to the different frequencies for pumping the 750 GHz SBV-diode multipliers, see
Fig. 12, the expected maximum available input power differs considerably, as shown in
the figure. It should be observed that 11 mW of output power from a 250 GHz
Schottky-varactor diode tripler, see Fig. 12, is achieved at operating conditions close to
the Schottky diode burnout [9]. Thus a more realistic value of a maximum of 6 mW of
output power was used for the 250 GHz tripler. In the calculations the maximum
allowed input power was set to such a value that the rf-voltage Vrf over the device,
minus the rf-voltage over the series resistance, is less than or equal to the breakdown
voltage Vbr of the low doped epilayer L2.
Third International Symposium on Space Terahertz Technology Pagel29
150 GHz
75 GHz
Iz 0(2) — l>(rf) > 75 ° GHz
<160mW ^— * <20mW ^^>0.05mW
Schottky SBV
^-^ 250 GHz ^^
83.3 GHz 0( X 3 j — fc(x3 1) 750 GHz
<60mW V-/ <HmW ^~^>0.05mW
Schottky SBV
107GHz 0(x7j > 750GHz
<36mW >0.05mW
SBV
Fig. 12 The different 750 GHz multiplier configurations investigated in this study.
The conversion efficiency for 750 GHz SBV-diode multipliers using SBV-diodes
based on InAs and GaAs is shown in Fig. 13. Simulations for two different contact
resistances Re = 8*10"^ ohmem^ and 1*10"^ ohmem^ were used for the GaAs SBV-
diodes, in order to evaluate what implications the choice of material has on the efficiency
of the device. The higher electron mobility in InAs reduces the series resistance of the
diode, thus giving InAs SBV-diodes an advantage in efficiency for low input powers
over GaAs SBV-diodes, having the same ohmic contact resistance. This advantage in
efficiency is even larger when a more realistic ohmic contact resistance, Re, of 8* 10" 7
ohmem^ is used for the GaAs SBV-diodes. The efficiency of the multipliers increases
with increasing multiplication rate, see Fig. 13, in principle in the same way for the
different SBV-diodes. This is due to the fact that at lower frequencies the impedance
(resistance R v and reactance jX v ) of the variable capacitance increases, i.e. becomes
larger as compared to R s . The efficiency is proportional to R v //(Rv + Rs). see ref.[10].
The cutoff frequency is defined in Eq. 1 .
Page 130
Third International Symposium on Space Terahertz Technology
InAs
Rc=Rco
10"
SlO 1
Il0°t
'o : GaAs
§ io^R^Rc
c
•g 10 " 2
>
o 10
U
10
GaAs
: Rc=80 Rco
I I I I I IS
■ • • ■ '
I I ll
I I I I 1 1
10
-2
10" 1 10° 10 1
Input power (mW)
10'
Fig. 1 3 Calculated conversion efficiency versus input power for various multiplier
configurations, see Fig. 12, using InAs and GaAs SBV-diodes as the nonlinear
element Two GaAs SBV-diodes having different contact resistances Re,
(=Rco in the figure), are shown in the figure where Rco = 1*10~8 ohmem^.
Nds = 3.4*10 18 cm' 3 , Nde =1*10 17 cm" 3 , N = 5*10 15 cm" 3 , Ll =
100 nm, L5 = 100 nm, L2 = L4 = 150 nm, L3 = 14 nm, diameter a = 2 |im.
(O): GaAs SBV-diode tripler, (A): GaAs SBV-diode quintuples
(□): GaAs SBV-diode heptupler, (•): InAs SBV-diode tripler,
(A):. InAs SBV-diode quintupler and (B):InAs SBV-diode heptupler.
Using a larger diameter SBV-diode means of course a lower impedance, although it
can handle a larger input power before Vrf exceeds the breakdown voltage Vbr- A larger
diameter device will also withstand larger thermal heating, caused by the absorbed pump
power, before efficiency degradation occurs due to the reduced mobility of the electrons.
The mobility ji of the electrons is proportional to the temperature T of the material,
where for GaAs: |i ~ T 1 / 2 and for InAs: \i ~ T 3 / 2 [1 1][12]. Thus InAs is more sensitive
to thermal heating. However, since the breakdown voltage for InAs material is small
compared to GaAs, the level of the absorbed power is restricted from that point of view.
The influence of the thermal characteristics of the SBV-diode on the efficiency is not
taken into account in this study.
The calculated output power, using the values for the efficiency shown in Fig. 13, is
shown in Fig. 14. The simulated output power for a state of the an 2.8 fF Schottky-
varactor diode described in [13], is also shown in Fig. 14 where it can be seen that the
maximum output power for the Schottky-varactor diode clearly exceeds the output
power for GaAs SBV-diodes having a contact resistance of 8*10"^ ohmem^ (80Rco),
for the same input power. However, it should be noted that output powers comparable
to the Schottky-varactor diode tripler are achieved from the InAs SBV-diode multipliers
at much lower input power, even though the Schottky-varactor diode is capable of
generating at least twice the output power compared to the InAs SBV-diode before the
diode voltage exceeds the breakdown voltage for the device.
Third International Symposium on Space Terahertz Technology
Page 131
It should be noted that the calculations for the Schottky-varactor tripler assume optimum
match at all frequencies for the tripler, i.e. including the idler. For the SBV-diode tripler
is optimum match only considered for the input and output frequencies. Thus the
matching conditions are much easier to obtain for the SBV-diode compared to the
Schottky-varactor diode, in reality giving the SBV-diode an advantage in efficiency
compared to the Schottky-varactor diode. Also the unavoidable resistive losses at the
idler frequency in the Schottky-varactor diode tripler gives the SBV-diode tripler an even
greater advantage.
The expected losses in the 750 GHz tripler mount is assumed to be less than 10 dB.
However it can be seen from Fig. 14 that the only realistic SBV-diodes capable of
generate ^ 500 |iW, necessary for > 50 jiW of output power at 750 GHz are InAs based
diodes, even though this is achieved close to or at the breakdown voltage V Dr limit for
the InAs devices. Although the breakdown voltage V Dr and thereby also the maximum
allowed absorbed input power for the InAs SBV-diodes are much smaller than for the
GaAs SBV-diodes, it can be compensated for by connecting several InAs/AlSb
junctions in series, creating a multiple barrier InAs SBV-diode.
10 x
10°
£ 10 1
s
^ io' 2
S. 10
a- io
5 io - 5
10
Fig. 14
-6
I I I I I I I | I I I I I I I I | I III TTTTT- I I I I I I I I
ordiod
Schottky-varactor diode
InAs
Rc=Rco
: GaAs
Rc=Rco
[GaAs
Rc=80 Ri
10
IO' 1 10 v
Input power (mW)
10'
Calculated output power versus input power for various multiplier config-
urations, using InAs and GaAs SBV-diodes and a Schottky-varactor diode as
the nonlinear element in the multiplier. Two GaAs SBV-diodes having differ-
ent contact resistances Re, (=Rco in the figure), are shown in the figure where
Rco = 1*10-8 ohmcm2. Nds = 3.4*10 18 cm"3, Nde =l*10l.7,c m •
Nb = 5*10 15 cm" 3 , LI = 100 nm, L5 = 100 nm, L2 = L4 = 150 nm,
L3 = 14 nm and diameter a = 2 (im.
(O): GaAs SBV-diode tripler, (A)
(□): GaAs SBV-diode heptupler, (•)
(A): InAs SBV-diode quintuples (■)
GaAs SBV-diode quintuples
InAs SBV-diode tripler,
InAs SBV-diode heptupler, and
(ffl): Schottky-varactor tripler with Co = 2.8 fF and R s = 20 ohm [13].
p a gel32 Third International Symposium on Space Terahertz Technology
It should also be noted that the GaAs SBV-diode, Re = 8*10 -7 ohmem 2 , could in
principle be pumped by a much larger input power than 6 mW, used in Fig. 14, before
the diode voltage exceeds the breakdown voltage Vbr of the device.
The dotted line in Fig. 14 marks the minimum 50 |i\V output power limit, which has to
be exceeded, assuming no losses in the mount.
As can also be seen in Fig. 14, higher order InAs SBV-diode multipliers can be used for
generating the necessary 500 jiW of output power. The penalty for using a higher order
multiplier is more resistive losses due to a larger number of idler circuits.
4. CONCLUSIONS
Theoretical work on multipliers show a maximum efficiency for a lower Cmax to
Cmin ratio than expected. This has been shown both in simplified calculations as well as
in full harmonic balance simulations. These findings have several implications. First,
there is no need for a thin barrier. Secondly, a wider barrier reduces the current. This
also make simulations easier. The C vs. V characteristics have been simulated from
epilayer parameters and the impact of different doping structures on both the CV and the
efficiency for the multipier has been presented. The examples presented can be used as
design rules.. In a case study for a 750GHz multiplier an InAs SBV diode outperforms
a Schottky diode.
5. ACKNOWLEDGEMENTS
Mikael Ekenstedt, Dept. of Physics, Chalmers University of Technology for growing
wafers #1566 and #1567. This work has been supported by ESA/ESTEC under contract
7898/88/NL/PB(SC) and The Swedish National Board for Industrial and Technical
Development (NUTEK).
6. REFERENCES
[1] Hans Gronqvist, Erik Kollberg, Anders Rydberg, "Quantum-well and quantum-
barrier diodes for generating submillimeter wave power", Microwave and Optical
Technology Letters, Vol. 4, No. 1, pp 33-38, 1991.
[2] E. Kollberg, T. Tolmunen, M. Frerking, J. East, "Current saturation in sub-
millimeter wave varactors", to be published in IEEE Transactions on.Microwave Theory
and Techn., May 1992.
[3] Hans Hjelmgren, "Numerical modelling of hot electrons in n-GaAs Schottky barrier
diodes", IEEE Trans, on Electron Devices, vol. ED-37, No. 5, pp 1228-1234, May
1990.
[4] Hans Hjelmgren, Erik Kollberg, and Lennart Lundgren, "Numerical simulations of
the capacitance of forward-biased Schottky diodes", Solid-State Electronics Vol. 34,
No. 6, pp. 587-590, 1991
[5] M. S. Lundstrom and R.J, Schuelke, "Numerical analysis of heterostructure
semiconductor devices," IEEE Trans. Electron Dev., vol. ED-30,p.l 151-1 159, 1983.
Third International Symposium on Space Terahertz Technology Page 133
[6] R. Stratton, "Diffusion of hot and cold electrons in semiconductor barriers", Phys.
Rev., vol. 126, pp 2002-2014, 1962.
[7] K.S. Champlin and G. Eisenstein, "Cutoff frequency of submillimeter schottky-
barrier diodes," EEEE Trans, on Microwave Theory and Tech., vol. MTT-26, pp. 31 -
34, 1978.
[8] P.H. Siegel, A.R.Kerr and W. Hwang, "Topics in the optimization of millimeter-
wave mixers," NASA Technical Paper 2287, 1984.
[9] N.R. Erickson, "Very high efficiency frequency tripler for 100-300 GHz," Proc.
of the 10th Int. Conf. on Infrared and Millimeter Waves, pp. 54-55, 1985.
[10] P. Penfield Jr. and R.P. Rafuse, " Varactor applications ", Massachusets Institute
of Technology, Cambridge, Massachusets, USA, The MIT Press, 1962.
[1 1] S.M. Sze, "Physics of Semiconductors," John Wiley & Sons, pp. 28-29, 1981.
[12] Landolt-Bornstein, "Numerical data and functional relationships in science and
technology, Group III: Crystals and solid state physics," Volume 17 Semiconductors,
Springer- Verlag Berlin, pp.577, 1982.'
[13] A. Rydberg, B. N. Lyons and U.S. Lidholm, "On the development of a high
efficiency 750 GHz frequency tripler for THz heterodyne systems," To be published in
IEEE Trans, on Microwave Theory and Techn., May 1992.
[14] H. Gronqvist, S. Nilsen, A. Rydberg and E. Kollberg, "Characterizing highly
efficient millimeter wave Single Barrier Varactor multiplier diodes", to be presented at
The European Microwave Conference, Helsinki, Finland, 1992
[15] T. Tolmunen, S. Nilsen, O. Boric, M. Frerking and E. Kollberg, "Accurate
characterization of varactors with fF capacitance", Conference Digest, 16 th International
Conference on Infrared and Millimeter Waves, Lausanne, Switzerland, August 26-30
1991, pp 214-215
Pagel34 Third International Symposium on Space Terahertz Technology
Effect of Cooling on the Efficiency
of Schottky Varactor Frequency
Multipliers at Millimeter Waves
Jyrki Louhi 1 , Antti Raisanen 1 , Neal Erickson 2
1 Helsinki University of Technology, Radio Laboratory, SF-02150 Espoo, Finland
2 Five College Radio Astronomy Observatory, University of Massachusetts,
619 Lederle Graduate Research Center, Amherst, MA 01003, USA
Abstract
The efficiency of the Schottky diode multiplier can be increased by cooling the
diode to 77 K. The main reason for better efficiency is the increased mobility of the
free carriers. Because of that the series resistance decreases and a few dB higher
efficiency can be expected at low input power levels. At high output frequencies
and at high power levels the current saturation decreases the efficiency of the
multiplication. When the diode is cooled the maximum current of the diode
increases and much more output power can be expected. There are also slight
changes in the I — V characteristic and in the diode junction capacitance, but
they have a negligible effect on the efficiency of the multiplier.
1 Introduction
It is well known, that cooling a Schottky diode mixer improves its sensitivity, i.e.
reduces the mixer noise temperature. This is mainly due to the sharper I — V
characteristic at cryogenic temperatures, only partly due to the smaller series
resistance and lower metal losses in the mixer mount. In satellite applications
the heterodyne receiver is readily cooled passively to temperatures of 110. ..150
K. Also, a space qualified 80 K cooler is available. This makes it very reasonable
to consider the effect of cooling on the frequency multiplier performance. This is
especially important at submillimeter waves, where not enough power is available
from ordinary all-solid-state frequency multipliers.
Third International Symposium on Space Terahertz Technology
Page 135
2 Diode model and effect of cooling
A simple equivalent circuit of the Schottky diode contains three components: non-
linear junction capacitance, nonlinear junction conductance and series impedance
[1].
WW
Ci (V)
V
Figure 1: Simple equivalent circuit of the Schottky diode.
Capacitance
The basic model for the junction capacitance of the Schottky diode is
Co
Cj(V) =
y/l-V/fa'
(1)
where <£« is the built-in potential (about 1 V) and Co is diode capacitance, when
the voltage over the junction is zero. For very small submillimeter wave diodes
the edge effect must be included in the diode model as [2]
C 0) . ^.(i+'-OT
w
(V)
2-r a
w(V) =
\q-N D
(2)
(3)
where A is the anode area, e, is the dielectric constant of the semiconductor,
w(V) is the length of the depletion region, r is the anode radius, q is the charge
of an electron, No is the doping density in the semiconductor, k is Boltzmann's
constant and T is the temperature. In these models the junction capacitance is
very high near the contact potential <£w Physically this is impossible, and a bet-
ter model for junction capacitance must be calculated by using the drift-diffusion
model [3]. In any of the models, the primary mechanism for the efficiency of the
multiplier, the degree of capacitance nonlinearity, is not temperature dependent.
Thus, cooling has no effect on the diode's inherent capability to generate har-
monics. In the two simple models the only temperature dependent factor is 0^.
Page 136
Third International Symposium on Space Terahertz Technology
When the diode is cooled from 300 K to 77 K, the contact potential fa increases
by about 0.1 V [4]. Because fa varies only slightly when the diode is cooled, the
same operation point can be reached if the bias potential Vfcuw is also increased
as much as fa. In all, the effect of cooling on the junction capacitance is so small
that it has an almost negligible effect on the multiplier efficiency.
U.
o
o
c
(0
*>
+-t
o
to
a
10
o
0.6 0.8 1
Voltage [V]
1.2
Figure 2: The junction capacitance at temperatures 300 K (solid line) and at 77
K (dashed line).
Series impedance
When the nonlinearity of the epitaxial layer above the plasma resonance is not in-
cluded, the series impedance of the submillimeter wave Schottky diode is modeled
as [5]
Z.(u) = Z*(v) + Z^{u) + Z. kin (u) + R e
1
Z^{u>) =
1 + ;u// u, » u d
+ ] —
4-r [l+ju//u/, w d
_ P**b
1
w.) - ^.^-.^y/yzg^,
(4)
(5)
(6)
(7)
where R* is the contact resistance (about 1 fi), p is resistivity, t e (e//) is t e — u>(Vw«),
t e is the thickness of the epitaxial layer, b is the radius of the chip and 8 t is the
skin depth in the substrate given by
(8)
Third International Symposium on Space Terahertz Technology Pagel37
where fi is permeability of GaAs. Scattering frequency u t and dielectric relax-
ation frequency uj are
"> = -r~, (9)
«- = ~V' ( 10 )
where m* is the effective carrier mass and \i, is the carrier mobility. The resistivity
is
P=— ~ , (11)
q-n-n,
where n is the concentration of the free electrons in the conduction band.
In a semiconductor, the concentration of the free carriers n and the mobility of the
carriers /z, are the most important temperature dependent factors in equations
given above. In GaAs the donor binding energy Ed is so small and the concen-
tration of donors Nd is usually so high that the concentration of the free carriers
n is equivalent to Nd at all temperatures, where the diode should be used. At
a very cold temperature, below 10 K, the concentration drops, because there is
not enough thermal energy to ionize electrons to the conduction band, and so the
resistivity of GaAs becomes high. At very high temperatures the concentration
of the intrinsic carriers is higher than Nd, and thus n is also higher than No-
In GaAs the mobility of the free carriers can be calculated from the mobilities of
the various scattering processes by using the Matthiessen rule
i = E^ (12)
/*. Hi
In GaAs the most important scattering processes are the ionized impurity scatter-
ing, acoustic-mode scattering and polar-optical scattering. At room temperature,
the polar-optical scattering dominates. When GaAs is cooled, the mobility in-
creases until the mobility of the polar-optical scattering and the mobility of the
impurity scattering are equal. At that temperature, mobility \i, has a maximum,
and when the diode is cooled more the mobility decreases. When Nd is rather
low (1 • 10 16 cm -3 ) the optimum temperature is low (~ 50 K) and the mobility
greatly increases [6]. At very high doping concentration (2-10 17 cm~ 3 ) the optimum
temperature is higher (~ 150 K) and the mobility increases only a little when the
diode is cooled to 77 K.
When considering the effect of cooling on the series impedance of the Schottky
diode, it is simplest to consider first its effect on the DC resistance and then the
effect on the series impedance at high frequencies. When the diode is cooled to
77 K, the mobility of electrons increases and thus the resistivity of the epitaxial
layer decreases, which also decreases the DC resistance of the diode. When the
doping concentration of the epitaxial layer is low, the DC resistance decreases
significantly. (For diode UVA 6P2 the measured decrease is about 4.5 fi, from a
Page 138 Third International Symposium on Space Terahertz Technology
value of 10.5 ft to 6 ft; the calculated values agree very well, see Figure 3) When
the doping rate is higher the decrease of the resistance is not as large. (For diode
UVA 2T2 the calculated decrease is about 3.5 ft, from a value of 12 ft to 8.5 ft)
When considering the effect of the decreased series resistance on the efficiency of
the multiplication, it must be noticed that the resistance of the epitaxial layer Z ep i
is a function of the thickness of the layer. In an efficient reactive multiplication,
the voltage over the depletion region spends a substantial part of the pump cycle
in the low voltage region, where the contribution of Z^ in Z, is large, but a small
part of the pump cycle in the high reverse voltage region, where the contribution
of Z^ in Z, is small. When the diode is now cooled, the decrease of the series
resistance is smaller than the decrease of the DC resistance, but still the decrease
of resistance has a very strong positive effect on the efficiency of multiplication.
At high frequencies the series impedance of the Schottky diode is no longer purely
resistive, because of the plasma resonance and the skin effect. When the diode is
cooled, the plasma resonance frequency
w p = ,/u t -u; d = W — (13)
does not change, because it is independent of the electron mobility fi t . Because u>,
and u>4 are temperature dependent, the Q-factor of the resonance is also temper-
ature dependent, and when the diode is cooled to 77 K the Q is increased (Figure
3). Because the mobility in the substrate changes only very little when the diode
is cooled, the impedance of the substrate Zmi, and the impedance of the skin effect
are not changed significantly.
Third International Symposium on Space Terahertz Technology
Page 139
10 a E
10 a
CD
CO
l i
T 1 — I I I I I I
I I I
J I 1 I I 1 1 1 1 I I I I 1 I I 1
0"
10 «
Frequency [Hz]
10
13
Figure 3: The series resistance of diode UVA 6P2 at temperatures 300 K (solid
line) and 77 K (dashed line).
&
i 1 — i — i i m i r— t — i — i i i t f
10 ia
Frequency [HzJ
10"
[h P
Figure 4: The series reactance of diode UVA 6P2 at temperatures 300 K (solid
line) and 77 K (dashed line).
Page 140
Third International Symposium on Space Terahertz Technology
I-V characteristic
For a Schottky diode the I — V characteristic is assumed to be [4]
B = flycoth^)
q± I N D
(14)
(15)
(16)
where R** is modified Richardson's constant, h is h/2ir and h is Planck's constant.
There are two important factors of the I — V characteristic for the efficiency of the
frequency multiplication: the turn-up point of the I — V curve, and the steepness
of the I — V curve beyond that. When the Schottky diode is cooled, the possible
voltage range where the multiplication is mainly reactive, increases, and thus the
maximum efficiency can also increase. For a cooled diode the shape of the I — V
characteristic is also sharper, and therefore the resistive multiplication is slightly
more effective.
XlQ- 6
C
<D
C
C
3
U
0.5
Voltage [V]
1.5
Figure 5: The current- voltage characteristic at temperatures of 300 K (solid line)
and 77 K (dashed line).
Third International Symposium on Space Terahertz Technology Page 141
Current saturation
At a low electric field the electron drift velocity vj is directly related to the electric
field £ as
v d = H. • S. (17)
When the electric field increases the drift velocity also increases until the velocity
reaches a maximum value v mox (= 2.2 -lO 5 m/s at about 3.2 kV/cm in an intrinsic
case). In that situation the electron conduction current
i e = A ■ n - q ■ fi t • £ (18)
must be replaced by the maximum current
imax = A-n-q-Vna*. (19)
This current saturation causes a very significant decrease in the efficiency of the
multiplier at high power levels and also when the output frequency is high, because
the junction capacitance cannot be pumped with optimum current. The current
saturation seems to be the most important factor for a submillimeter wave fre-
quency multiplier, when the efficiency of the multiplication is considered. When
the diode is cooled, the maximum drift velocity increases [6] and because of that
the maximum electron current also increases. Therefore, when the diode is cooled
the effect of the current saturation is less significant. This increases the efficien-
cy especially at high power levels^ at high frequencies, and in the case of a high
multiplication factor.
The current saturation may be modelled by strongly current dependent series
resistance R,(i) above the maximum current. Kollberg et al. have presented the
following model [7]:
R.{i) = R,(DC) • a • i 6 , (20)
where a is a parameter, depending on the maximum current of the diode i max . The
meaning of the R,(i) is to modify the current waveform approximately as required
by causing a very strong increase in the series resistance when the current of the
diode is higher than the maximum current i max . The parameter a has been fitted
empirically to the measured results only in one case and must be estimated for
other diodes and frequencies. There seems to be no physical background for this
model, but so far no better model has been proposed.
3 Analysis of multipliers
At millimeter waves Schottky varactors are often driven into conduction, which
is only nearly optimal. In this case, the usefulness of classical theories [8] is poor
and harmonic balance analysis [9] should be used. One form of the harmonic
balance analysis is the multiple reflections technique, where the multiplier circuit
is divided into linear and nonlinear subcircuits, which are then analyzed in the
frequency and time domain.
Page 142
Third International Symposium on Space Terahertz Technology
Doubler for 160 GHz
Let us first consider the effect of cooling on a two diode balanced doubler for 160
GHz, because we have also experimental results for it [10].
Table 1: Parameters used for UVA 6P2.
Co
A
t.
N D
M
^max
300 K
21
33
1.0
3.5-10 16
0.61
44
77 K
21
33
1.0
3.5-10 16
1.40
66
fF
fiva. 2
fiva.
cm" 3
m/s
mA
The two diode construction has been analyzed both at 300 K and at 77 K. First,
the efficiency has been calculated with optimum embedding impedances. These
results have been plotted in Figure 6 (solid lines). Here the efficiency has only a
very poor correlation with the measurements (0 and X) because of the VSWR,
which is mainly caused by the fact that the embedding impedances are optimized
for high input power. When the doubler is then analyzed by using the optimum
embedding impedances for high input power at all input power levels, the corre-
lation is much better, especially when 0.5 dB losses in the input and 0.8 dB losses
in the output have been taken into account.
Table 2: Experimental output power versus temperature and input power (two
diodes).
Input power
10
33
50
100
mW
Temperature
300 K
1.6
9.0
13.9
22.0
mW
223 K
1.9
10.4
16.3
26.7
mW
77 K
2.2
12.8
18.7
30.7
mW
In order to understand better the agreement between the theory and experiment, it
is worth separating the effects of the decreased series impedance and the increased
current handling capability due to the cooling. First, if the current saturation is
omitted in the theoretical analysis, the effect of cooling is as follows. At low
input power levels when the multiplication is purely reactive, the decreased series
impedance causes a clear increase in the efficiency due to smaller losses in the
series impedance. According to simulations, the increase of the efficiency in the
above case at low input power levels is about 1.5 dB. However, when the input
power per diode is large (i.e. > 10 mW), the multiplication efficiency tends to
decrease with the increased input power due to the resistive multiplication. This
Third International Symposium on Space Terahertz Technology
Page 143
u
c
tu
80
70
60
50
40
30
20
10
-1 —
: 1
1 —
_
_
- jT
-
1 / '
/ /
- / / /<
II'
J, '
o o
X
-
1
-
-
_1
-- J 1
»
10 20 30 40
Input power per diode [mW]
50
Figure 6: The efficiency of the 160 GHz doubler at 300 K and at 77 K (above),
when using optimum impedances (solid line) and impedances optimum for high
power (dashed line). Measurement results, when 0.5 dB input losses and 0.8 dB
output losses have been taken into account, at 300 K (o) and at 77 K (x) have
also been plotted.
is because the voltage swing reaches the conduction region during every cycle.
The smaller the series impedance, the lower the input power needed to reach this
conduction, and thus, resistive multiplication. Therefore, the gain due to the
smaller series impedance is smaller at high input power levels than at low power
levels. According to the simulations, the efficiency increase due to the smaller
series impedance in the multiplier described above is only 0.5 dB at 50 mW input
power per diode.
When the current saturation is taken into account, but not the series resistance,
the positive effect of cooling is seen only at high power levels. This is because
the junction capacitance can be pumped at 77 K more effectively than at 300 K.
At small power levels the saturation, of course, does not play an important role.
According to our simulations, the higher current handling capability of the cooled
diode 6P2 improves the efficiency by 1 dB at 50 mW input power per diode.
These two effects of cooling together, the decreased series impedance and the
increased current handling capability, explain the experimentally verified 1.5 dB
increase in the multiplication efficiency of all power levels and therefore give some
kind of a proof of the current saturation in the diode at high input power levels.
Due to the higher efficiency at high input power levels, the maximum output power
is also increased by the same amount, which helps in pumping the following stage
in the multiplier chain producing submillimeter wave frequencies.
Page 144 Third International Symposium on Space Terahertz Technology
Multipliers for 1 THz
When constructing multiplier chains for 1 THz, a reasonable choice is first to
double the output frequency of a powerful W-band Gunn oscillator and then to
follow by a tripler and a doubler or by a doubler and a tripler. The latter choice
does not only depend on the varactor diodes but also on the technology to build
fine mechanical multiplier mounts.
In order to get some understanding of how much power could be available at
1 THz, the choice of a tripler to 500 GHz and doubler to 1 THz has been made
because this allows comparison with experimental results up to 500 GHz [10]. The
tripler for 500 GHz and the doubler for 1 THz can be analyzed the same way as
the doubler for 160 GHz, but now current saturation plays a very important role.
Because the presented model for current saturation has only poor correlation to
the physics, the results for high frequency multipliers should be considered only
qualitatively.
Some general aspects can still be presented. First, when the diode is cooled,
the maximum drift velocity increases, which also increases the maximum current.
Second, when the output frequency is high, the changes during the voltage swing
are very fast. In that situation, the current needed for optimum multiplication
is very high, and then current saturation plays a very important role by greatly
decreasing the efficiency. By cooling, the maximum current should increase and
the efficiency of the multiplication may increase by a few dB. Third, when the first
or second stage multiplier is cooled, the maximum input power for the last stage
multiplier increases, and so also the maximum output power for 1 THz increases.
Our simulations have indicated an increase of about 7 dB from 100 /iW to 500
/xW in the optimum situation.
4 Conclusions
Cooling of a Schottky varactor multiplier increases its efficiency by as much as
a few dB. Because of the smaller series impedance the efficiency of frequency
multiplication increases by 1-2 dB at small input power levels. At large input
power levels the efficiency increases by 2-10 dB due to the higher current handling
capability of the diode. A cooled multiplier can be readily used in satellite ap-
plications, where the receiver is cooled to 50. . .150 K. The positive effect of the
cooling should be utilized especially in submillimeter wave multipliers when the
output power necessary cannot be reached in any other way.
Even though the model of a Schottky diode is already rather complex, it should
be studied more. The main reason for a poor model is that current saturation is
poorly handled. Much more work must be done to model the saturation exactly.
Also much more experimental work on cooling multipliers is needed, before all the
effects of the cooling can be understood.
Third International Symposium on Space Terahertz Technology
Page 145
References
[1] Raisanen A.V., Sironen M.: Capability of Schottky-diode multipliers as local
oscillators at 1 THz. Microwave and Optical Technology Letters, vol. 4, no.
1, 1991, p. 29-33.
[2] Copeland J.A.: Diode edge effect on doping-profile measurements. IEEE
Transactions on Electron Devices, vol. ED-17, no. 5, 1970, p. 401-407.
[3] Hjelmgren H., Kollberg E., Lundgren L.: Numerical simulations of the ca-
pacitance of forward-biased Schottky-diodes. Solid-State Electronics, vol. 34,
no. 6, 1991, p. 587-590.
[4] Kollberg E.L., Zirath H., Jelenski A., Temperature-variable characteristics
and noise in metal-semiconductor junctions. IEEE Transactions on Mic-
rowave Theory and Techiques, vol. MTT-34, no. 9, 1986, p. 913-922.
[5] Crowe T.W.: GaAs Schottky barrier mixer diodes for the frequency range
1-10 THz. International Journal of Infrared and Millimeter Waves, vol. 10,
no. 7, 1989, p. 765-777.
[6] Ruch J.G., Fawcett W.: Temperature dependence of the transport properties
of Gallium Arsenide determined by a Monte Carlo method. Journal of Applied
Physics, vol. 41, no. 9, 1970, p. 3843-3849.
[7] Kollberg E., Tolmunen T., Frerking M., East J.: Current saturation in sub-
millimeter wave var actors. Proceedings of the 2nd International Symposium
on Space Terahertz Technology, 1991, p. 306-322.
[8] Penfield P., Rafuse R.P.: Varactor Applications, Cambridge, Mass., The MIT
Press, 1962.
[9] Siegel P.H., Kenr A.R., Hwang W.: Topics in the Optimization of Millimeter
- Wave Mixers, NASA Technical Paper 2287, 1984.
[10] Erickson N.: High efficiency submillimeter frequency multipliers. IEEE MTT-
S International Microwave Symposium Digest, vol III, Dallas, 1990, p. 1301-
1304.
Page 146 Third International Symposium on Space Terahertz Technology
-5/3.-23
Superlattice Barrier Varactors*
C. Raman, J. P. Sun, W. L. Chen, G. Munns,
J. East and G. Haddad
Solid State Electronics Laboratory
University of Michigan, Ann Arbor, Michigan
Abstract
SBV (Single Barrier Varactor) diodes have been proposed as alternatives to Schottky
barrier diodes for harmonic multiplier applications. However these show a higher current
than expected. The excess current is due to X valley transport in the barrier. We will
present experimental results showing that the use of a superlattice barrier and doping
spikes in the GaAs depletion regions on either side of the barrier can reduce the excess
current and improve the control of the capacitance vs. voltage characteristic.
The experimental results consist of data taken from two types of device structures.
The first test structure was used to study the performance of AlAs/GaAs superlattice
barriers. The wafer was fabricated into 90 micron diameter mesa diodes and the resulting
current vs. voltage characteristics were measured. A 10 period superlattice structure
with a total thickness of approximately 400 A worked well as an electron barrier. The
structure had a current density of about one A/cm 2 at one volt at room temperature. The
capacitance variation of these structures was small because of the design of the GaAs
cladding layers. The second test structure was used to study cladding layer designs.
These wafers were InGaAs and InAlAs layers lattice matched to an InP substrate. The
layers have n + doping spikes near the barrier to increase the zero bias capacitance and
control the shape of the capacitance vs. voltage characteristic. These structures have a
capacitance ratio of 5:1 and an abrupt change from maximum to minimum capacitance.
The measurements were made at 80 K. Based on the information obtained from these two
structures, we have designed a structure that combines the low current density barrier
with the improved cladding layers. The capacitance and current-voltage characteristics
from this structure are presented.
- 'This work was supported by the Center for Space Terahertz Technology under NASA Contract No.
N AG W- 1334 and by the URI-ARO Program Contract No. DAAL03-87-K-0007.
Third International Symposium on Space Terahertz Technology Page 147
Introduction
Varactor diodes are an important component of harmonic multipliers operating above
100 GHz. These multipliers are the primary source of power in the submillimeter wave
frequency range, where the diode predominantly used is a Schottky barrier device. How-
ever, a multiplier based on Schottky diodes suffers certain disadvantages. The circuit is
complex, with higher order conversion requiring matching at all frequencies-the input,
the output, as well as idler frequencies. The varactor must also be biased to increase the
voltage swing and prevent current flow over the barrier in the forward direction. These
problems can be overcome by using heteroj unction based varactor structures. Implemen-
tation of such novel diode structures is motivated by the symmetry of their capacitance
vs. voltage characteristic which permits odd harmonic conversion without the added
complexity of even harmonic idlers and bias circuitry .
The Single Barrier Varactor
One possible structure is the single barrier varactor, shown in Figure la. The cor-
responding energy band diagrams under thermal equilibrium and bias are shown in lb
and lc. The analogue of the metal-semiconductor electron energy barrier found in the
Schottky varactor diode is the conduction band offset between the GaAs region and the
wider bandgap Al x Ga 1 _ r As layer. This energy barrier acts to inhibit electron transport
through the structure and ideally should be as large as possible for the device impedance
to be purely reactive. When one side is biased relative to the other, the GaAs region
on the anode side is depleted of electrons and becomes positively charged, similar to the
semiconductor region of a Schottky varactor under reverse bias. Electrons accumulate
on the cathode side of the structure, forming a charge separation across the Al^Ga^^As
region. The relationship between the stored charge and the applied voltage is non-linear,
resulting in a non-constant device capacitance. Moreover, since the structure is symmet-
ric, reversing the sign of the applied voltage merely interchanges the roles of depletion
and accumulation regions and does not affect the device capacitance. Therefore the
capacitance-voltage characteristic has even symmetry. Early attempts at implementing
such varactors were limited in efficiency by high leakage current densities 1 . Details of
the leakage current analysis are described elsewhere in these proceedings 2 .
By replacing the single heterostructure by a superlattice barrier, we propose to reduce
the carrier transport through the device by increasing the effective barrier height seen
Anders Rydberg, Hans Gronqvist, and Erik Kollberg, Millimeter and Sub- Millimeter Wave Multi-
pliers Using Quantum Barrier Varactor (QBV) Diodes, IEEE Electron Device Letters, Vol. 11, No. 9.
Sept. 1990, pp. 373-375.
2 H. Hjelmgren, J. East and E. Kollberg, "Thermionic Emission Current in a Single Barrier Varactor,",
these proceedings.
Page 148 Third International Symposium on Space Terahertz Technology
by an electron. A doping profile modification can improve the capacitance vs. voltage
characteristic; however, a trade-off is seen between maintaining a high barrier height and
a good C-V profile.
The Superlattice Barrier Varactor
Although the SBV presents a symmetric C-V characteristic, its efficiency is degraded
due to the high leakage current associated with the device because a purely reactive
multiplier is more efficient than a resistive one. The true thermionic emission energy
barrier seen by an electron in the GaAs regions is not the T to T energy level offset but
the considerably lower T to X energy level difference, as seen in Figure 2a. By means of
a scattering process an electron incident on the barrier can pass through the Al x Gai_ x As
region into the X valley of the barrier material. Thermionic emission over this X level
requires less kinetic energy than emission over the T level and consequently a smaller
bias voltage is needed to turn on the current.
The X valley transport can be suppressed by replacing the single heterojunction
by a series of barriers interspersed with quantum wells, i.e., a superlattice. Such a
superlattice appears in Figure 2b. The well regions are thin, resulting in quantum
mechanical confinement. The energy mini-bands are shifted upward considerably with
respect to the T point in bulk GaAs. A simple calculation indicates that the energy
increase can be on the order of an electron volt. Unlike the SBV, there is no longer a
continuous X valley current path since the well energies are higher than the X levels.
The superlattice structure achieves a larger effective barrier than that of the SBV and
suppresses the leakage current.
A thin, highly doped layer between the superlattice and each adjacent N~ region can
be used to modify the capacitance. The so-called 5-doped regions (see Figure 3) are thin
and contain a large amount of charge so that at zero bias they remain mostly undepleted.
Flat band effects are minimized and the zero bias capacitance is increased. Varying the
bias slightly causes almost no change in the depletion width and the capacitance remains
constant. However, beyond a threshold voltage, the entire 8 region becomes depleted
and the lightly doped N~ region begins to deplete rapidly, causing a sharp drop in the
capacitance from its zero bias value.
Advantages and Disadvantages
In designing superlattice varactors, two material systems were considered: GaAs/AlAs
and InGaAs/InAlAs lattice matched to InP. Structure I, seen in Figure 4, shows a 10 pe-
riod, 20 A / 20 A GaAs/ Al As superlattice barrier varactor chosen for examination of the
barrier properties. The MBE grown wafer was processed on the front side by photolithog-
raphy. Contacts were made to the front and back by evaporating a Ni/Ge/Au/Ti/Au
Third International Symposium on Space Terahertz Technology Page 149
sequence and annealing at 405 degrees Celsius. Mesa diameters of 5 to 90 //m were then
chemically etched. The measured room temperature I-V curve in Figure 5 demonstrates
the effectiveness of the superlattice in keeping the current to a minimum. At a bias of
1 volt the current density is about 1 A/cm 2 , compared with about 150 A/cm 2 for the
•single barrier varactor. However, the ratio of maximum to minimum capacitance is
insufficient for any significant harmonic conversion.
The second structure, a 40 A / 40 A InGaAs/InAlAs superlattice, was designed with
doping spikes added, as outlined in Figure 6a. Figure 6b shows the C-V data, taken at
SO K to reduce the parallel conduction current 3 . A close agreement is observed with the
capacitance characteristic predicted by a self-consistent quantum mechanical and Poisson
solution for the charge and potential distribution throughout the superlattice varactor.
When a voltage of approximately 0.2 volts is applied, the capacitance drops sharply to
about one-fifth of its zero bias value, corroborating the theoretical expectation. Thus
doping profile modifications allow good control of the capacitance-voltage characteristic.
The final structure utilized the low leakage properties of the GaAs/AlAs superlattice
and incorporated the doping modifications that had been tested in the InGaAs/InAlAs
system. In structure III the same superlattice as Structure I (with one well layer removed
from the end for symmetry) was grown on a new varactor wafer which incorporated a
doping spike of sheet density lxlO 12 cm -2 and a more lightly doped N" region than the
previous GaAs/AlAs wafer. Figures 7 and 8 present the room temperature I-V and C-V
data from 90 micron diameter mesas fabricated on this wafer. Compared with structure
I, the leakage current density has increased. This current degradation can be attributed
to the following mechanism: the effective electron thermionic emission barrier is reduced
by band bending at the barrier edge due to the high electron density in the doping spike.
The capacitance and parallel device conductance were simultaneously measured on a
HP4275A LCR meter, where the bias was varied until the conductance exceeded the
measurement capability of the machine. As seen in Figure 8, the C-V profile exhibits a
swing ratio of close to 4. However, the zero bias capacitance was larger than the value
predicted by the quantum mechanical analysis by about a factor of 2. The reason for
this is not clearly understood at the present time.
Despite the fact that the current levels in the final structure were higher than ex-
pected, the superlattice varactor represents improvement over the single barrier varactor.
Figure 9 compares the current densities for the two devices at low and high voltages and
temperatures. For operation at low temperature or for small voltages the superlattice
barrier fares better than the SBV.
3 J. P. Sun, W. L. Chen, J. East and G. Haddad, C-V Characteristics of Quantum Well Varactors,
Proceedings of the 1991 International Semiconductor Devices Research Symposium.
Page 150 Third International Symposium on Space Terahertz Technology
Conclusions
We have demonstrated a superlattice barrier varactor in the GaAs/AlAs system which
promises to be useful in odd harmonic generation due to its reduced leakage properties
when compared with the SBV. Moreover, modification of the capacitance by means of
an appropriate doping profile has been demonstrated in the InGaAs/InAlAs material
system. A compromise appears, however, between minimizing leakage current levels in
the device and achieving a large maximum to minimum capacitance swing ratio. We
hope to begin RF testing of the superlattice barrier varactors in the near future.
Third International Symposium on Space Terahertz Technology
Page 151
a)
N + GaAs
yV" GaAs
N~ GaAs
T
i Al x Gai_ T As
N+ GaAs
b)
c)
+ + +
+ + +
r
FIGURE (1): (a) Single Barrier Varactor , (b) band diagram under
equilibrium, (c) band diagram under bias.
Page 152
Third International Symposium on Space Terahertz Technology
a)
Gamma
Level
X
Level
Electron sees
lower barrier
Electron
Confined Gamma Energy Levels
In The Well Regions
b)
X
X
X
Electron
FIGURE (2): (a) X valley transport in Single Barrier Varactor,
(b) Suppression of X transport by superlattice.
Third International Symposium on Space Terahertz Technology
Page 153
SPACER
LAYER
SUPER-
LATTICE
DOPING
SPIKE
N+
FIGURE (3): Superlattice with doping spikes
3000 A
^ ► -^-
3000 A
3000 A 3000 A
-►-^-
^^^£
18
5X10
lb
2 X 10
Sw><>$£
16
2 X 10
5X10' 9
GaAs
GaAs
GaAs
GaAs
20 A/20 A
Ten Period
Superlattice
FIGURE (4): Structure I: GaAs/ Al As superlattice varactor
Page 154
Third International Symposium on Space Terahertz Technology
xxxxxx
IF
(uA)
GRAPHICS PLOT xxxxxx
SL I-V 90jJ DIAM.
•2.000
200.0
■ 1
i ! ■ '
< . ■ i
i
!
40.00
/div
200.0
i
• i
; i
i i
_. ;. j
* i
j !
i ;
■ !
!
i
i
•
i
t
j
i
i
i
i
i
i
i
/
/
/-
/
j
;
• /
,/
i /
"" 1 /i
/ !
_. / !
/i
/l
^ —
- ■
. . . .
....
VF
2.000
.4000/div ( V)
FIGURE (5): Structure I: I-V characteristic
Third International Symposium on Space Terahertz Technology
Page 155
4000 A
1000 A
-■• ^
1000 A
4000 A
<
VAA^J
n
5 X 10
InGaAs
16
1 X 10
InGaAs
Delta doping
InGaAs 5 E17
per cm" 3 80 A
thick
It
J6
1 X 10
InGaAs
17
5 X 10
InGaAs
Undoped 100 A
InGaAs Spacer
Layer
Three Barrier/Two Well
40 A InGaAs/ 40 A InAlAs
Superlattice
3.5
-0.5-0.4-0-3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5
Voltage (V)
FIGURE (6): (a) Structure II: InGaAs/InAlAs superlattice varactor with
doping spikes, (b) Structure II: G-V characteristic
Page 156
Third International Symposium on Space Terahertz Technology
SUFERUTTICE IV (DELTA-DOPED): 90 MICRONS
5* 10J3
! ! ! ! ! ! ■ 1
= ! = •:!=/
1 ! I t | i !/
i i i i i i /
ill!!!/!
1 ! i ! i !/ I
! 1 ! ! ! A I
i S i i !/! i
! ! ! i/f ! !
2
i ! (^ i 5 i
5
<K
;;;; = • ^-*— " T" — ;
U
i 1 = ! js^\ \ ■ =
! ! ! * \ i ! i !
i i ^ i f i i i !
i i/i . i ! i ! i !
«/! ! 1 1 ! I 1 J
/ i 1 ! I ! I i i
/! f s ! i 1 i i i
-4* -aj
•u -a.1 as u az aj i*
vetoes m
FIGURE (7): Structure III: I-V characteristic
GAAS/AUS DEOA-OOFED S/t C-V: SO MICRON MESA DIAMETER
== 300
i
• i !
i
i
I/l
1/
j
u
u
z
{
J
5
5
£
I
j
\ t
j a 1 {
j
i\ l I
| t :
|
: I x *
1 ! K 5
i l i
t : :
j
| | j \[
Sj.
|f si |l|
VOUKZM
u u
FIGURE (8): Structure III: C-V characteristic
Third International Symposium on Space Terahertz Technology
Page 157
J (Amps/sqcm)
At V = 0.1 volts
At V = 0.4 volts
SBV
1.1 @300K
0.2 @ 100 K
9.4 @ 300 K
1.8 @ 100 K
AlAs/
GaAs S/L
0.7 @ 294 K
negligible
@90K
16.5 @ 294 K
0.2 @ 90 K
FIGURE (9): Comparison of improvements
Page 158 Third International Symposium on Space Terahertz Technology
A NEW FABRICATION TECHNIQUE FOR BACK-TO-BACK
S/3-33 . VARAC^RDZODES N93-277 3 9
/ (& &£)*t / ^- P eter Smith, Debabani Choudhury, Suzanne Martin, and M.A. Frerking,
John K. Liu, and Frank A. Grunthaner
.V
Center for Space Microelectronic Technology
Jet Propulsion Laboratory/California Institute of Technology
4800 Oak Grove Drive
Pasadena, CA 91109
Abstract: A new varactor diode process has been developed in which much of the
processing is done from the back of an extremely thin semiconductor wafer laminated to a
low-dielectric substrate. Back-to-back BNN diodes were fabricated with this technique;
excellent DC and low-frequecy capacitance measurements were obtained. Advantages of
the new technique relative to other techniques include greatly reduced frontside wafer
damage from exposure to process chemicals, improved capability to integrate devices
(e.g., for antenna patterns, transmission lines, or wafer-scale grids), and higher line yield.
BNN diodes fabricated with this technique exhibit approximately the expected capacitance-
voltage characteristics while showing leakage currents under 10 mA at voltages three
times that needed to deplete the varactor. This leakage is many orders of magnitude better
than comparable Schottky diodes.
Introduction:
Planar varactor diodes * are being developed in place of whisker-contacted devices in order
to improve the performance and ruggedness of spacebome submillimeter-wave heterodyne
receivers; at the same time thin heterostructure layers are being used to improve diode
performance, making processing more demanding. Such devices could be more useful if
integrated into relatively large arrays that could potentially be used for communications
systems. It is expected that such back-to-back multiplier diodes with Schottky contacts and
heterostructure barriers can be made to operate reasonably efficiently at frequencies over
one terahertz2>3,4 j^ m i s paper we present results from devices in which the isolation
was performed from the back. This technique simplifies processing, greatly increasing
yield and providing a much lower dielectric constant, and thus low loss, environment for
antennas and other circuit patterns.
Conventional multiplier diode isolation techniques have a number of problem areas.
Isolation implants are commonly used, but the removal of masking materials from the
wafer often presents difficulties. An alternative technique for isolating the active devices is
to perform a mesa etch, but connection of the contacts to the top of the mesa is problematic.
One approach attempted here requires relatively difficult planarization processing that can
potentially damage the thin barrier layer. Metal step coverage may be also a problem with
this approach. Air bridging can also be used with mesa isolation, but this also exposes the
top of the semiconductor material to a larger number of process steps. Some inactive
semiconductor material, with its associated high dielectric constant, is generally left in place
with all of these approaches.
We have developed an alternative processing technique that promises to be simpler and
more robust. In this process, no inactive semiconductor material is left, and the front of the
wafer is exposed to the absolute minimum of processing possible for a front- side back-to-
Third International Symposium on Space Terahertz Technology Page 159
back diode process. In our work, the remaining process steps were completed with the
devices laminated to a 3 mil quartz substrate. This quartz is the same as is typically used
for the crossed-field waveguide multiplier filter structures for which the devices were
intended. Clearly, there is a wide lattitude in substrate material, which adds a great deal of
flexibility to the design of submillimeter-wave components. The total number of steps is
relatively low, improving yield. Since thinned 1 or lifted off 5 devices never need to be
handled off of the substrate material, relatively large scale integration can potentially be
achieved with this process.
Initial fabrication runs have been successful. Back-to-back BNN diodes have been
fabricated using the new technique and then measured for DC and low-frequency
capacitance characteristics. Eight micron mesas with 1.5 and 3.75 micron wide Schottky
metalization showed good C-V and outstanding I-V characteristics. While the C-V pulses
were approximately two to three volts wide (full-width half-maximum), leakage currents
were as low as 50 nA with 10 volts between pads. Mesas as small as 1 micron were
successfully patterned.
Device Fabrication:
A. Semiconductor layer structure
Details regarding appropriate BNN layer structures have been addressed before2>3 A The
general approach for a GaAs-based BNN diode, from the top surface down, is to include:
(1 -optional) a thin GaAs cap layer, (2) an AlGaAs layer that is sufficiently thick to preclude
tunneling but sufficiently thin to allow a large capacitance per unit area - 15 to 20 nm of
Alo.45GaQ.55As is typical, (3) near the AlGaAs, a highly doped region in order to ensure
that the high capacitance mentioned above is achieved at zero voltage, (4) a moderately .
doped GaAs dnft/varactor region in which all of the doping can be depleted with little
parasitic conduction to the metal pads, and (5) a highly doped region that provides a low-
resistance path between the two Schottky pads.
The structure used for these devices is as follows: (1) a 2 nm GaAs cap, (2) a 15 nm
AlQ.45Gao.55As barrier, (3) a 3 nm GaAs spacer followed by SxlO^/cm^ silicon planar
doping, (4) a lxl0 17 /cm 3 by 125 nm drift region, and (5) a 5xl0 18 /cm 3 by 900 nm
conducting base. A 600 nm undoped Alo.45Gao.55As layer for use as an etch stop layer
was located just below the active device layers, although much thinner layers have also
been successfully used. The layer structure is shown schematically in Figure 1.
Capacitance-voltage measurements between large capacitance pads on another area of the
same wafer gave the data shown in Figure 2.
The process is shown in Figure 3. We processed GaAs wafer pieces about 1.7 cm on a
side that were laminated onto quartz pieces 76 microns (3 mils) thick by 2.5 cm in
diameter. The GaAs wafers were initially 510 microns (20 mils) thick.
The Schottky contacts are defined by exposing AZ 5214 with a standard image reversal
technique and then evaporating and lifting off Ti/Pt/Au. While we did not do a surface
isolation, the devices should show less leakage current and particularly less parasitic
capacitance if etching (or implanting) were done through (or into) the delta-doped layer.
Etching by using the metal as a mask has been done successfully but not on the wafers
processed in this work.
Page 160
Third International Symposium on Space Terahertz Technology
Undoped A^ 4S Ga 0S As
1 50 Angstroms
ua
us cap
r ,«,12, 2
1x1Cr 8 /cm 3 x1250
"" Angstroms GaAs
5x1 /cm
<pP5 CURSOR <-9. 88O0V . 8. 7S|*00. >
E-00
3.000
/dlv
.0000
-a. o
-**
V-
»■
5x1 ,8 /cm 3 x 9000
Angstroms GaAs
/
\
/
\
-^
/
I
I
/
V
r
Undoped A^ 4S Ga 0S ,As
Etch Stop
S
■vj
—».
-^
oo ? __„_... . ■ =-
Substrate
ooo
Figure 1: Epitaxially grown GaAs/AlGaAs layer
for BNN diodes used in this study. Wafer layers were
grown with molecular beam epitaxy. Doping is silicon.
B. Device Processing
Figure 2: Capacitance-voltage
characteristics measured at 1 MHz
using large dots.
Saadng waftr
(b)
Figure 3: (1-a) The Schottky contacts are deposited, and a thin isolation through the delta-doped region is
etched optionally, (2-a) front-side passivation is deposited, (3-b) the wafer is mounted face-down on a quartz
wafer with wax for thinning (this wafer must be suitably thin to serve as the final substrate for the device
and may be mounted in turn on a sturdier substrate), (4-b) conventional lapping and then selective thinning
to an AlGaAs etch stop layer located just below the active layer, (5-c) lithography for mesa isolation is
performed from the back of the wafer (at this point the wafer is sufficiently thin that the front-side metal can
be viewed from the back through the substrate), and (6-c) the wafer is etched, (7-d) thin Si3N4 or Si02 is
deposited in order to passivate the back and, if necessary, opened for bonding pads, and (8-d) the thin quartz
wafer is diced.
Third International Symposium on Space Terahertz Technology Page 161
The GaAs wafers were glued face down to a 75 micron thick quartz wafer, 2.5 cm in
diameter, with UV-curing glue. The GaAs wafers were then lapped down to a 100 micron
thickness using 5 micron, 1 micron, and finally 0.3 micron grit. Polishing was done with
a silica colloidal suspension and pad. Selective thinning to the etch stop layer (5, above)
was done using a 95/5 solution of ammonia and peroxide *. The peroxide/ammonia etch
undercut the edges of the wafer two to three hundred microns, two to three times the
vertical etch distance. Little damage of scale larger than pinholes of a micron or two in
diameter could be seen over most of the wafers processed. A brief dip in HC1 and water
was used to remove the remaining oxide layer.
It should be noted that the wafers were remarkably rugged. The single cracked wafer
observed so far occured during the lapping operation, which in our lab is performed using
a relatively violent vibration table. The wafer that broke lost some small portions of the
edge, but the area on which the GaAs was epoxied proved to be extremely tough despite
cracks through the quartz. The wafer was processed to completion despite the cracks, with
no allowances made to ease the normal rigors of microelectronic processing (spinning,
contact lithography, etc.).
Mesas were defined by standard positive photolithography and dry etching. Wet etching
was unsuccessful using either photoresist masks (excessive undercutting for the smaller, 1
micron mesas) or nitride masks (adhesion problems). The mesas were aligned to the
frontside metal with IR backlighting, although the semiconductor is thin enough to permit
aligning with optical backlighting if available.
Additional process steps have been done, including backside passivation with more ECR
nitride and etching of contact holes with a CF4/02 plasma. Clearly, the ability to deposit
more metal in order to further reduce loss and to make MIM capacitors with the frontside
metal and backside nitride is very attractive. Also, we have tried thinning the glue with
acetone prior to spinning; this results in a much thinner layer (roughly 10 microns) between
the GaAs wafer and the quartz substrate.
Figure 4: Backlit photograph of central portion of device. The small dark rectangle in the center is the
approximately 4x16 micron micron mesa. The larger, dark portions with the thinner protrusions (2
micron fingers) are the Ti/Pt/Au pads. The remaining area is quartz.
Page 162
Third International Symposium on Space Terahertz Technology
The ECR nitride process is then repeated on the back, with windows opened with CF4/02
RIE to the frontside metal for contacting. Other MMIC-style processing, including plating,
MIM capacitor formation, etc. could be performed at this point.
Figure 4 shows a photograph of a completed device.
Electrical Measurements
Devices were checked for DC and capacitance characteristics after processing. Figure 5
shows the 1 MHz capacitance of a 3.75 x 8 micron device. While the peak capacitance is
lower than expected from the capacitance data shown in Figure 2, it is thought that the
difference is probably due to the fact that the processed wafer piece is from the edge of the
MBE wafer. Figure 6 shows the DC current leakage before backside passivation. The
measured current is many orders of magnitude lower than would be observed with
comparable Schottky diodes, leading us to believe that the BNN is potentially superior as a
multiplied power source.
60 t
^ 50
^ 40
u
I 30
o
§. 20
CO
° 10
-I H- 1 1
-3 -2
-10 12 3 4 5
Voltage (V)
Figure 5: Measured C- V characteristics of a device with back-to-back 3.75 x 8 micron BNN diode.
IF
.-
ID. OO
/dlv
O
/
/
SO. OO
-io. OO
2. QOD/dlv
< V>
IO. OO
Figure 6: Measured DC leakage current for back-to-back 3.75 x 8 micron diode. Traces for positive and
negative voltages were taken by sweeping away from zero volts. Higher leakage currents can be measured
instantaneously if the voltage is not slowly swept; similar behavior commonly observed in MESFETs
leads us to believe that the method used above is the most relevant.
Third International Symposium on Space Terahertz Technology Page 163
Conclusions
We have demonstrated a new process for back-to-back diode fabrication using a BNN
structure. The process should make integration of submillimeter-wave diodes much easier
while simultaneously reducing RF losses through the elimination of all non-essential high-
dielectric semiconductor. The BNN diodes constructed with this process show a strikingly
low leakage current.
Acknowledgements
The research described in this paper was performed at the Center for Space
Microelectronics Technology, Jet Propulsion Laboratory, California Institute of
Technology and was sponsored by the National Aeronautics and Space Administration,
Office of Aeronautics, Space, and Technology, and by the Army Research Office.
References
1 W.L. Bishop, E.R. Meiburg, R.J. Mattauch, and T.W. Crowe, "A Micron Thickness
Planar Schottky Diode Chip for Terahertz Applications with Theoretical Minimum
Capacitance," 1990 IEEE MTT-S International Microwave Symposium Digest, p. 1305.
2 U. Lieneweg and U. Maserjian, "Harmonic generation in the near-millimeter-wave
range by thin-MOS structures," presented at Sixth International Conference on Infrared and
Millimeter Waves, Miami, FL, 1981.
3 U. Lieneweg, T. Tomunen, M. Frerking, and J. Maserjian, "Design of Planar varactor
Frequency Multiplier Devices with Blocking Barrier," submitted to IEEE MTT Special
Issue on Terahertz Technology, 1991.
4 E. Kollberg, T. Tolmunen, M. Frerking, and J. East, "Current Saturation in
Submillimeter Wave Varactors," submitted to thte IEEE MTT Special Issue on Terahertz
Technology, 1991.
5 E. Yablonovich, T. Gmitter, J.P. Harbison, and R. Bhat, "Extreme Selectivity in the
Lift-Off of Epitaxial GaAs Films," Appl. Phys. Lett. vol. 51, p. 2222 (1987).
6 Norland Optical Adhesive 61, Norland Products Inc., P.O. Box 145, New Brunswick,
NJ 18902.
Page 164 Third International Symposium on Space Terahertz Technology
/&>&* N9 3 -27 7 40
v •■
A 200 GHz TRIPLER USING SINGLE BARRIER VARACTOR
Debabani Choudhury, Margaret A. Frerking
and Paul D. Batelaan
Jet Propulsion Laboratory
California Institute of Technology
Pasadena, California 91109, USA
ABSTRACT
/>' The GaAs Schottky varactor diode is the non-linear device most commonly
used for submillimeter wave harmonic generation. Output power adequate to serve as a
local oscillator source for SIS tunnel junctions has been demonstrated with whisker-
contacted GaAs Schottky varactor multipliers in waveguide mounts up to about 800 GHz.
In this paper, we present results for a tripler to 200 GHz using a new multiplier
device, the single barrier varactor (SBV). This new varactor has a potential advantages
such as stronger non-linearities or special symmetry, which make it attractive for
submillimeter wave frequency multiplication.
The performance of a tripler using a SBV over a output frequency range
from 186 to 207 GHz has been measured in a crossed waveguide mount. The theoretical
performance of the device has been calculated using large signal analysis. A comparison
Third International Symposium on Space Terahertz Technology Pagel65
of theoretical and measured results and a discussion of various losses in the mount and
the varactor have also been presented.
INTRODUCTION
Heterodyne receivers are used for high spectral resolution shorter-
millimeter and sub-millimeter wave astrophysics and earth remote sensing
observations. Local oscillator, mixer and antenna are the critical components in a
receiver. One approach to provide sub-millimeter power is to use the combination of a
high-power millimeter-wave source with a harmonic multiplier for higher frequency
generation. Frequency multipliers use a non-linear device to generate harmonics of the
input frequency from a fundamental oscillator. Although the Manley-Rowe relations
show that an ideal harmonic genarator with 100% efficiency is possible with a varactor,
real multiplier circuits are limited by loss in the device and circuit and by impedance
matching limitations at the input, output, idler and harmonic frequencies [1,2]. As the
circuits become smaller with increase in frequency, impedances and losses become more
difficult to control.
To achieve the full capability of the diode, appropriate embedding
impedances must be provided by the multiplier mount. The impedances at the input and
output frequencies must be set to maximize coupling power into or out of the device. In
higher order multipliers, current flow at the intermediate harmonics (i.e. the idler
frequencies) will enhance harmonic conversion. Therefore, the diode must be terminated
with a lossless reactance at these frequencies. The embedding impedances are provided by
the multiplier mount. Nonlinearities symmetric about zero bias will generate only odd
Page 166
Third International Symposium on Space Terahertz Technology
harmonics, greatly simplifying the multiplier mount design. For instance a tripler
mount for a symmetric device will be equivalent to a doubler mount for a device without
symmetry. Similarly, a quintupler mount will be equivalent to a tripler mount, both
requiring one idler.
This paper presents the theoretical and experimental results of a 200 GHz
tripler using a single barrier varactor (SBV) as the nonlinear device.
MULTIPLIER DEVICE
The single barrier varactor diode used as the multiplier device in our
experiment, was developed at the Chalmers University of Technology [3,4]. These
Chalmers devices were fabricated with the epitaxial GaAs/Alo.7Gao.3As/GaAs material
grown as indicated in Fig.1. It is inherently a symmetric device. The Alo.7Gao.3As barrier
19980 A
{
4000 A
.5330 A
530 A .
530 A
5330 A
Whisker
t*^«K««^««&
*mmmm%m*
jTOTJTwr
HMMlitMMHtUtMMiMU
*V'V
<i'
AuGe/Ni/Au
GaAs
As
GaAs
,GaAs Substrate
(n-doped)
u_ AuGe/Ni/Au
Fig.1 : Schematic of the Chalmers Device
Third International Symposium on Space Terahertz Technology p a ge 167
which blocks the current flow is in the center having a thickness of 213 A. On either side
of the barrier, there is an undoped GaAs spacer having a thickness of 53 A. GaAs depletion
region (n=1x10i? cm-3 ) on either side has a thickness of 5330 A. Top and bottom
a
contacts are formed on highly doped GaAs regions (n=3.4x10 18 cm-3 ) US j n g 1000 A
O
AuGe, 200 A Ni and 1600 A Au. The top contact is made with a whisker and the bottom is
a large area ohmic contact. The Alo.7Gao.3As barrier will to a large extent prevent
electrons from passing through the structure. Thus the conduction current through the
device is very small. For moderate voltages, the conduction current is essentially caused
by thermionic emission. The width of the depleted part of the moderately doped epitaxial
layer will vary with bias voltage, thus forming a voltage dependent capacitance C(V).
When the diode is biased in the forward direction, the depleted region will appear on one
side of the barrier, and the depletion capacitance of the device will decrease with
increasing voltage. Since the diode is symmetric, a reverse bias will in the same way
cause a decrease in the capacitance value of the device. Hence, the maximum capacitance
is obtained for zero voltage and is determined by the thickness of the Alo.7Gao.3As
barrier. The minimum capacitance which occurs for maximum bias voltage, is
determined by the doping concentration and the extension of the moderately doped drift
region. For an appropriately designed device, similar capacitance swing with voltage as
for the Schottky-varactor diode is expected [5].
The losses due to the series resistance may be larger in the SBV diode than
in Schottky varactors, since the maximum current i m ax= CdV/dt will occur for V(t) = 0,
i.e. when both dV/dt and C are maximum and n-doped drift regions on both sides of the
barrier are undepleted and contribute to the series resistance. In addition, for small area
device, the ohmic contacts exhibits higher resistance than Schottky contacts. However
for a Schottky varactor tripler, the idler current at the second harmonic will degrade
Page 168
Third International Symposium on Space Terahertz Technology
the tripler performance, since any finite reactance termination will cause power losses
in the series resistance. For the SBV tripler, this particular problem is virtually non-
existent [4].
The Chalmers devices tested here, have a mesa height of about 2.5 microns
and area of 5x5 micron 2 . In order to evaluate the dc characteristics, the device has been
mounted in a coaxial mount as shown in Fig.2. The S-parameters are measured using a
HP 851 OB Network Analyzer. The K-connector provides 50 ohms up to the whisker to
allow accurate de-embedding of the mount [6]. The equivalent circuit of the diode
mounted in a co-axial mount is shown in Fig.3. Chalmers device was measured to have a
dc series resistance of 7 ohms. The measured C-V and l-V characteristics for the 5x5
micron 2 Chalmers device are shown in Fig.4. The measured maximum capacitance is
Fig.2 : Schematic diagram of the device mounted in a coaxial mount
Device
C(V)
r s Lw K-Connector
-'NA/* — 9 — ono 1 I
#. ^«(V) J^Cj
Fig. 3 : Equivalent circuit of a device in a coaxial mount
Third International Symposium on Space Terahertz Technology
Page 169
•6
-4-2024
DC Bias Voltage (V)
-4-2 2 4
DC Bias Voltage (V)
Fig.4 : Measured C-V and l-V characteristics of Chalmers device
65.6 fF and minimum capacitance is 12.4 fF. The figure of merit of the diode, which is
its cut-off frequency, is given by,
f c; L_{_J ^}
27tK s v^min ^max
Chalmers device has a cut-off frequency of 1200 GHz. The diodes are found to be damaged
when the dc voltage exceeded about 6 V.
LARGE SIGNAL ANALYSIS
The critical step in the multiplier analysis is to solve the voltage and
current waveforms of the nonlinear device which is pumped and biased in an arbitrary
embedding network. A common solution of this nonlinear problem is to use a type of
harmonic balance technique. Time-domain current and voltage solution are sought which
Page 170
Third International Symposium on Space Terahertz Technology
satisfy the diode conditions and frequency-domain solutions are sought which satisfy the
external circuit equations. In this work, a modified nonlinear program based on Siegel,
Kerr and Hwang [5] has been used in order to calculate the tripling efficiency of the
Chalmers single barrier varactor diodes. Fig. 5 shows the equivalent circuit of a
v B ~±
c <%>.
9<»d>/ »d
1
2e(2fp)
2>(3lp)
Ze(n»p)
Fig.5 : Eduivalent circuit of a multiplier
multiplier. Harmonic triplers with 186 GHz,192 GHz, 200 GHz, 207 GHz output
frequencies are calculated. We have optimized the impedance at the third harmonic
frequency. The idler and the higher harmonics are set to open circuits. Impedance up to
12th harmonic are analyzed. In the analyses, the measured C-V and l-V characteristic
shown in Fig. 4 have been used. Since the series resistance is important in the device
performance, the calculations are carried out for a range of resistances, 5 ohm, 10 ohm,
15 ohm and 20 ohm.
Fig.6 presents efficiency versus the input power for a SBV tripler to 192
GHz taking series resistance of the device as a parameter. In order to quantify the effect
of current flow in the device on the multiplier performance, we have calculated the
performance for a device with 5 ohm series resistance with no current. The theoretical
efficiency is found to degrade from 45% to 20% when the measured current is included.
Higher device series resistance degrades the tripler performance significantly as the
Third International Symposium on Space Terahertz Technology
Page 171
so T
3" 45 ..
192 GHz Output
R»=5 Ohma, msaa CV only
Ra=5 Ohms, meat IV.CV
... , R»=10 Ohma,maaa IV.CV
Ha=15 Ohma, mmat IV.CV
7»— -- — — "*— " R»=20 Ohma, meaa IV.CV
1 1 1
10 20 30 40 50 60
Input Power at Diode (mW)
70
80
Fig.6 : Calculated tripling efficiency for Chalmers device
series resistance of 20 ohm results in about a factor of four worst performance than a
series resistance of 5 ohm.
200 GHz WAVEGUIDE MOUNT
To achieve optimum performance of the device, it must be provided with
the appropriate circuit embedding impedances. The impedances at the input and output
frequencies must be set to maximize coupling power into or out of the device. In higher
order multipliers, current flow at the intermediate harmonics i.e. the idler frequencies
will enhance the harmonic conversion. Therefore, the diode must be terminated with a
lossless reactance at these frequencies.
An output of the large signal analysis, used to optimize the device, is the
C-3
Page 172
Third International Symposium on Space Terahertz Technology
required embedding impedance. In fig. 7, the real part of the optimum impedance is
120,.
100
_ 80 .
sa
60 .
40 ■
20 .
64 GHZ
90 mW hput powar
10 20 30
XI Raal Unpadanca (Ohm*)
"20 Ohms
■ IS Ohma
-10 Ohma
-4 Ohms
5 -
SO
45 ■ ■
40
IS
30
2S
20
15
10
5
192 GHz
50 tnW Input
10
20
30
X3 Real Impadanca (Ohma)
Fig. 7 : Input and output circuit optimum embedding impedances calculated
by large signal analysis
shown on the horizontal axis and the imaginary part on the vertical axis, parameterized
by the input power for both the input frequency (Ri t Xi) and the output frequency (R3,
X3). Impedances are plotted for different series resistances of the device. Input power
increases from mW to 50 mW. At low input power the real part is same as the device
series resistance. The input imaginary impedance is the impedance corresponding to the
maximum capacitance at input frequency. As the input power increases, the device
capacitance decreases increasing the impedance. The real impedances needed are in the
range from 7-30 Ohms. The imaginary impedances range from 40-100 Ohms for the
input circuit and from 15-42 Ohms for the output circuit respectively.
The embedding impedances are provided to the single barrier varactor
(SBV) device by a crossed waveguide mount. In addition the mount distributes the power.
A schematic drawing of the crossed waveguide mount is shown in Fig.8. The single
barrier varactor device is mounted spanning the output waveguide. The output waveguide
is actually oriented perpendicular to the plane of the paper. Power at the input frequency
travels down the input waveguide. A low pass filter consisting of Au metallization on the
Third International Symposium on Space Terahertz Technology
Page 173
quartz substrate couples the input power from the waveguide to the whisker contacted
Fig.8 : Schematic diagram of the device in the mount
SBV device located at the output waveguide. An E-plane tuner and a backshort at the input
waveguide provide adjustments to optimize the embedding impedance at the input
frequency. The output waveguide is cutoff at the input frequency preventing propagation
down it, thereby confining the input power to the vicinity of the SBV device. The tripled
power is coupled out the output waveguide. The embedding impedance at the output
frequency is adjusted by varying the whisker length and by a movable backshort. The low
pass filter prevents the output frequency from traveling to the input waveguide. A
scanning electron micrograph of the device in the mount is shown in Fig. 9.
EXPERIMENTAL RESULTS
The set-up for 200 GHz tripler measurements is shown schematically in
Fig. 10. A 60-70 GHz klystron is used as the pump source. The input power is monitored
Page 174
Third International Symposium on Space Terahertz Technology
Fig. 9 : Scanning Electron Micrograph of the device in the mount
Out Fin
Power
Meter
In Fin
"^
Waveguide
Switch
Power
Meter
mm-Wave source
Klystron
60 - 70 GHz
100 mW
Fig. 10 : Test setup for the 200 GHz tripler measurement
by a Anritsu power meter, calibrated to give the power at the input flange. The reflected
power is measured using a directional coupler coupled to a second power meter. The third
harmonic output power is measured by a third powermeter. We determine the loss at the
third harmonic in the waveguide from the output flange to the powermeter by a
substitution technique. The observed loss in the WR4 output waveguide is 0.032 dB/A.,
Third International Symposium on Space Terahertz Technology
Page 175
consistent with the resistive losses corresponding to the metal conductivity of 2x1 7
mho/m. The flange-to-flange efficiency is defined as the ratio of the power at output
flange to the power available at the input flange. Using various whisker lengths, the
efficiency and output power were measured between 1 86-207 GHz. Measurements were
taken with three different whisker lengths, 6 mil, 8.4 mil and 11 mil. It was seen that,
the Chalmers 25 micron 2 device, contacted with a 8.4 mils long whisker gives best
tripler performance. The measured efficiency versus input power for 186 GHz, 192
GHz, 196.5 GHz and 201 GHz output frequencies are shown in Fig.11 for a 8.4 mil long
8.4 mil whisker
10 20 30
Input Power (mW)
186 GHz
-■ 192 GHz
- — 196.5 GHz
-•—201 GHz
Fig. 11 : Measured efficiency versus input power plot for the tripler
whisker contact. The best performance has been achieved at 192 GHz, giving an
efficiency of more than 2% at 40 mW input power. This is similar to the results
demonstrated by Rydberg et al. using the same device [3].
To compare the experimental results to the performance predicted by
large signal analysis, the loss in the multiplier mount is assesed. Loss arises from
several mechanisms. At the input frequency, the finite conductivity of the waveguide and
Page 176
Third International Symposium on Space Terahertz Technology
dielectric loss in the filter are very small. The primary loss mechanism is due to the
impedance mismatch. By measuring reflected power at the input, less than 0.2 dB loss
due to mismatch was observed, over the frequency range tested.
At the output frequency, the impact of finite conductivity is higher. In
addition, losses due to imperfections in the backshort are critical. Other mechanisms
include the impedance mismatch and higher harmonic generation.
Some of these loss mechanisms at the output frequency have been modeled
using Hewlett Packard's Microwave Design System (MDS) package. Fig. 12 shows the
no loss
'fl sigma=2e7
Bl!
50 100
Backshort Position (mils)
150
Fig. 12 : Calculated transmission from the diode to the output flange of
the tripler mount
calculated transmission from diode in the tripler mount to the output flange of the mount
as a function of backshort position. When there is no loss, the transmission is 100% at
resonant positions of backshort. If a finite conductivity of 2x1 7 mho/m, the measured
conductivity of the WR4 output waveguide is included, the peak transmission reduces to
about 90%. If in addition, the backshort has a 10% loss, the transmission reduces to
Third International Symposium on Space Terahertz Technology
Page 177
about 60%. For a backshort loss of 25%, the transmission goes down to 40%. The losses
reduce the height of the peaks and the sharpness of the resonances out. In addition, the
antiresonant backshort positions donot give zero percent transmission.
The measured results are plotted in Fig. 13, which shows the relative
Loss in Output Waveguide
1 t-
Intensity
o o o o o
A A
£ 0.4.
1 0-3.
"3 0.2 .
" 0.1-
A/\s
■
.
50 100 150
Backshort Position (mils)
Fig. 13 : Plot of measured output power as a fumction of output backshort position
output power as a function of output backshort position. This looks qualitatively similar
to the theoretical results. The measured first peak match the theoretical results with
10% backshort loss, while the valley and second peak are closer to the 25% backshort
loss.
Based on these observations, the multiplier mount loss budget is
presented in Table-I. In the output circuit, the loss due to the finite conductivity is
estimated as 1 dB. Loss due to backshort is 3 dB. Impedance mismatch loss and loss due to
higher harmonics are not known. Therefore, the loss in the output circuit is estimated to
be more than 4 dB. Input circuit loss is estimated to be less than 0.2 dB. Using these loss
values, 0.2 dB at input and 4 dB at output, the measured flange-to-flange efficiency is
Page 178
Third International Symposium on Space Terahertz Technology
Table
Loss
Output Circuit
Finite Conductivity
1 dB
Backshort Lose
3dB
Impedance mismatch
?
Hlaher harmonics
?
Total
> 4dB
Input Circuit
Impedance Mismatch
0.2 dB
(Reflected power)
corrected to determine the efficiency at diode, which is plotted in Fig. 14. Superimposed
of diode efficiencies are the theoretical efficiency calculated from the measured CV and IV
curves for series resistances 10 ohm, 15 ohm and 20 ohm. At low input power the
measured efficiency follows the theoretical efficiency for a series resistance of about 12
ohm. The measured dc series resistance is 7 ohm. At 192 GHz, the series resistance is
Rs=10 Ohms
. .»— --— — * ••
Rs=15 Ohms
Rs=20 Oh ma
Meas ® 192 GHz
10 20 30 40 50
Input Power at Diode (mW)
60
Fig. 14 : Tripling efficiency at diode versus input power plot
Third International Symposium on Space Terahertz Technology Page 179
expected to be somewhat higher due to skin effect. At higher power the efficiency falls
off. This fall off corresponds to the power at which the device starts drawing significant
current. This suggests that the impact of the current flow in the device on the multiplier
performance is not well understood.
DISCUSSION AMP. CONCLUSION
The single barrier varactor diode has been shown to be able to provide
more than 5% efficiency as a 200 GHz tripler. About 2% flange-to-flange tripling
efficiency is obtained using the crossed-waveguide tripler mount for symmetric devices.
The multiplier mount has a 0.2 dB input circuit loss due to the impedance mismatch,
introduced by the reflected power. A total loss of more than 4 dB is estimated at the
output circuit. This includes the 1 dB loss due to finite conductivity of the waveguide and
3 dB loss due to the backshort.
Development of devices with lower leakage current will significantly
improve the tripler performance. Results can be further improved by reducing the
output circuit 4 dB loss by improving the imperfect backshort.
ACKNOWLEDGEMENTS
The research described in this paper was performed by the Center for
Space Microelectronics, Jet Propulsion Laboratory, California Institute of Technology
and was sponsored by the National Aeronautics and Space Administration, Office of
Page 180 Third International Symposium on Space Terahertz Technology
Aeronautics, Space and Technology. Authors would like to thank Prof. E.Kollberg for
kindly supplying the SBV devices and Mark Natzic for whisker contacting the devices into
the multiplier mount. Authors also would wish to thank T.J.Tolmunen for various useful
discussions and Hans Grdnqvist for giving some valuable information about the device.
REFERENCES
[1] T.J.Tolmunen and M.A.Frerking.'Theoretical Efficiency of Multiplier Devices',
Second International Symposium on Space Terahertz Technology, 197- 211,
(1991).
[2] P.W.Penfield and R.P.Rafuse.'Varactor Applications', The MIT Press, Cambridge,
MA, (1962).
[3] A.Rydberg, H.Gronqvist and E.Kollberg, 'Millimeter- and submillimeter-wave
Multipliers using Quantum Barrier varactor (QBV) Diodes', IEEE Electron
Device Letters, 373-375, vol.11, No. 9, September, (1990).
[4] E.Kollberg and A.Rydberg.'Quantum-Barrier-varactor Diodes for High-
Efficiency Millimeter-wave Multipliers', Electronics Letters, 1696-1698,
Vol.25, No.25, (1989).
[5] O. Boric, T.J.Tolmunen, E.K.Kollberg and M.A.Frerking.'The Anamolous
Capacitance of Quantum Well Double-Barrier Diodes', submitted to International
Journal of Infrared and Millimeter Waves (1992).
[6] M.Frerking and J.East.'Novel Varactors', submitted to the Proceeding of the IEEE
(1991).
[7] P.H.Siegel, A.R.Kerr and W.Hwang, 'Topics in the Optimization of Millimeter-
wave Mixers', NASA Tech. Paper 2287, (1984).
Third International Symposium on Space Terahertz Technology Page 181
A Submillimeter Tripler Using a Quasi-Waveguide Structure
Neal R. Erickson and German Cortes— Medellin -</^=^3 R
Five College Radio Astronomy Observatory
Department of Physics and Astronomy /fro •^'Ky'
University of Massachusetts V Q O ^ & %t t**
Amherst. MA 01003 ** ^ *> • 7 4 I
Abstract
A new type of frequency multiplier structure is being developed which is suitable for
application at frequencies above 1 THz. This structure preserves some of the properties of
waveguide for mode control, yet is not truly single mode. The device resembles a sectoral
horn, with a varactor diode mounted near the throat. Input and output coupling are
through the same aperture, requiring a quasi-optical diplexer. Initial tests are directed at
building a tripler at 500 GHz, for comparison with waveguide structures. The diplexer is a
blazed diffraction grating with appropriate focusing optics. Model studies show that the
impedance match to a varactor should be good, and initial tests of the beam patterns of the
prototype indicate that optical coupling efficiency should be very high. The structure also
has the potential for use as a fundamental mixer, or as a third harmonic mixer.
Introduction
As the operating frequencies of receiver systems shift toward ever higher frequencies,
it has become apparent that there is no clear upper limit to the application of single mode
waveguide [1,2]. However, the machining and assembly problems increase rapidly above
~300 GHz, and it seems clear that by some frequency of ~lTHz, the cost of waveguide
components will be too high for use in most systems. This limitation is purely a practical
one; the loss of waveguide does not seem to be a serious limit since submillimeter parts use
very short waveguide runs. However, the loss is high enough to be of some concern. Cube
corner mixers have been used in the higher frequency range, but suffer from a low beam
efficiency and a rather high embedding impedance. This paper suggests a new type of
mounting structure to replace waveguide at these frequencies, which combines some of the
advantages of both waveguide and quasi-optical structures.
Page 182 Third International Symposium on Space Terahertz Technology
An ideal mounting structure for a >lTHz multiplier should have the following
characteristics:
1. Ease of fabrication, both in the machining and the assembly with the diode.
2. Good mode control at all frequencies involved.
3. Resistive losses should be low.
4. Impedance level should match varactors, which tend to have a low real part and
require series inductance for matching.
5. The structure should be readily suited for whisker contacted diodes. Shape or size of
the chip should not be critical, and whiskers should not be too short.
6. If the device is optically coupled, the ports should be linearly polarized.
7. Beam width should be reasonably narrow to ease the design of coupling optics.
The structure being studied appears capable of satisfying all of these points, although
tests are still in progress. The relative ease of fabrication of the 500 GHz prototype
indicates that a scaled device at well over 1 THz should be practical. Additionally the
needed coupling optics have been developed, which appear capable of separating the input
and output beams with low loss as well as coupling them to a source and load.
Quasi— waveguide Mount
The structure as shown in Fig. 1, is essentially a pair of parallel metal plates
separated by less than A/2 at the highest frequency of interest, with two intersecting
sidewalls to guide the beam. The sidewalls intersect at 90° in the prototype, although other
angles may be equally good or better. The varactor is mounted between the top and bottom
plates, spaced from the vertex by less then 0.71A at the highest frequency. With these
constraints, the structure can couple only to a mode with the electric field uniform and
perpendicular to the plates, and the mode pattern in the H plane has single maximum. To
produce a convenient output beamwidth, the top and bottom plates are made nonparallel so
that the separation becomes a few wavelengths at the opening. The best situation for ease of
design would be to make the planes parallel at the diode, and then change the angle to begin
the flare, as in a typical waveguide— horn interface. However, this makes the structure
impractical to build at >lTHz, so the approach taken was to maintain a continuous flare,
which is slow enough to not greatly perturb the embedding conditions at the diode. If this
flare becomes too fast, the embedding impedances may change and the evanescent higher
modes at the diode are not sufficiently cut off before the separation becomes great enough for
Third International Symposium on Space Terahertz Technology
Page 183
them to propagate. The relatively large spacing of the plates (compared to reduced height
waveguide) makes mounting a diode and whisker easier, and reduces the losses. The wide
side wall spacing minimizes their contribution to the loss except near the vertex.
A model study was done to determine the embedding impedances in this structure. A
coaxial probe was introduced up to the effective terminal where the diode would be
contacted, and a "contact whisker" of various lengths and shapes extended to the opposite
wall. Various vertex distances within the constraint of the maximum spacing were also
tested. The general result is that any antenna within this structure acts as an inefficient
radiator so that it maintains the character of a lossy transmission line, with the loss
increasing with frequency. Thus it shows resonant behavior, which is pronounced when the
plate spacing is as wide as used here. The diode terminal impedance circles the Smith chart,
at a high value of p, with approximately periodic behavior dependent on the physical length
of the whisker, rather than the plate spacing. The match initially appears quite poor at all
but the highest frequencies, but it is found that through the choice of whisker length and
shape, it is possible to achieve a considerable range of impedances at three harmonically
related frequencies. By adding a short circuited transmission line in series with the contact
whisker (with two adjustable parameters, length and impedance), it is possible to produce an
even wider range of values. The net result is that a reasonable match may be made to the
impedance of a varactor, over a bandwidth of 5%, at all three frequencies required for a
tripler, including a low resistance at the idler. The best configuration occurs with the
WALL
COAXIAL TUNER
COAXIAL
DC BIAS PORT
EXPANDED
SIDE VIEW
(E PLANE)
DIODE
.7X
MAX
DIODE AND
ANTENNA
DIRECTION
OF BEAM
TOP VIEW
(H PLANE)
Fig. l. Top view and side view cross section of the quasi— waveguide tripler.
Page 184 Third International Symposium on Space Terahertz Technology
maximum plate separation and vertex distance allowable, but there is still some freedom in
the choice of whisker parameters (length, diameter and shape). No other flare angles were
tested, so this remains as another possible adjustment. While the bandwidth of this
particular choice of geometry is limited, it seems good enough to evaluate the potential of
the device.
Based on this design, a prototype tripler has been fabricated for an output frequency
of 500 GHz. Initial tests seemed best at a frequency where comparison with the results for
waveguide mounted devices is possible, while this frequency is high enough to permit a
realistic assessment of the fabrication difficulties. The structure is built as a split block,
with the bottom plane and side walls in one part, and a flat plate forming the other half.
The diode in this device is biased through a coaxial filter designed to present a short circuit
at all three frequencies involved. As in typical mixers and multipliers, the diode chip is
mounted on the end of the coaxial filter, forming much of the final section. The diode
chosen is U.Va. type 2T2 with Cj(0) = 6fF, R s = 12ft, and V b = 11V. The flare angle of
the plates is 9° , over a total length of 2.3 cm, so that the opening aperture is 3.6 mm. The
whisker is mounted on the end of a short circuited coaxial section providing the needed
reactances at the input and output.
While the bottom half of the mount was made by electroforming over the corner of a
cube, it could also easily be machined, except for the vertex itself, where a small radius has
little effect. Probably the most difficult machining area is in the coaxial bias filter, which
would be impractical at substantially higher frequencies. An alternative is to use a very thin
capacitor for rf bypass between the diode and the bottom plane; this would also require a
thin diode. A bias wire can then connect to a feedthrough at the vertex.
Coupling Optics and Diplexer
A quasi— optical device is only of value if it can be coupled efficiently with optics. In
this case the input power is likely to be derived from a waveguide mounted varactor
multiplier with a feed horn on the output, which may be approximated by a Gaussian beam
waist. The output load will almost certainly be an antenna coupled mixer but the details of
the pattern of such a mixer are presently unknown. We can only assume that a Gaussian
beam is suitable.
Third International Symposium on Space Terahertz Technology
Page 185
Beam patterns for this device are expected to be those of a uniformly illuminated
aperture in the E plane and one sector of that due to a square array of four antennas in the
H plane. The E plane pattern has a moderate sidelobe level reducing its coupling efficiency
to about 85% to a Gaussian mode, while the H plane pattern is well tapered with no
sidelobes, and thus couples with high efficiency. The beam is unusual in that the phase
centers for the two planes are far apart. The H plane originates essentially at the vertex,
while the E plane center is at the physical aperture. Thus the optics must be very
astigmatic. In addition, the beams at the input and output are very different in size,
particularly in the E plane.
L
01
3
O
a.
<u
■'A
O
a.
ex.
rrrr
1"
//
-a
-4
-
//
\
-
-6
i
1/
\
\
\
-
-B
•
11
1
\
\
•
la
■
1
li
\
\
•
12
.
1
1
\
\
•
14
"*
It
II
\
V
—
It
18
•
\
'/
\
\
i
\
-
2B
UXI
llll
llllllllllllllllllllMIU4lllll.il
I1U.
uu
•••■•1,
11JJ111
-3a -2a
Azimuth CDegreesJ
Fig. 2. Cuts through the principle planes of the beams in the E plane
(azimuth) and the H plane (elevation). Solid line is at 164 GHz, dashed
line is at 492 GHz. Beams are for the tripler without additional optics.
Page 186
Third International Symposium on Space Terahertz Technology
Tests have been performed using the varactor diode as a video detector at both
frequencies, with sources at 164 GHz using a doubled Gunn oscillator, and at 492 GHz using
the same oscillator multiplied by six. E and H plane cuts are shown for the prototype in
Fig. 2, for both frequencies of interest. Full contour maps show no additional features out of
the principle planes. These patterns confirm the theoretical predictions, but the H plane at
the output frequency is off center by about 9° . This is due to a small asymmetry in the
whisker location or shape, but otherwise the beam shape is as expected.
Frequency separation in the submillimeter may be done in several ways, but is
particularly easy for a tripler because of the large frequency ratio. While a perforated plate
high pass filter would work well [3], an easier device to fabricate is a diffraction grating.
With the correct grating period, the input frequency can be below the onset of diffraction so
that the grating behaves as a simple mirror, while the output can be scattered in a very
different direction in the first order. A particular advantage of this mode of operation is
that the input signal is well isolated from the output, which makes measurement of the
output power easier since filters are not needed. The efficiency of this scattering can be
made very high through the correct choice of reflection geometry and the blaze angle of the
grooves. The electric field must be perpendicular to the ruling direction for high efficiency.
A convenient configuration is with the grating tilted by 45° relative to the two beams,
reflecting the input through 90° and the output by 45° . This requires a blaze angle to the
grooves of 22.5° and a period of 0.86 mm. These optics are shown in Fig. 3. There is no
» output\
\ BEAM N
TRIPLER
INPUT
BEAM
CYLINDRICAL
LENS
CYLINDRICAL
BLAZED GRATING
Fig. 3. Diplexer and optics to separate and focus the beams. The grating
is cylindrical in the plane out of the figure.
Third International Symposium on Space Terahertz Technology
Page 187
scattering loss at the input, while the theoretical scattering into the two possible unwanted
orders at the output totals about 5%. Gratings may be curved in one dimension without loss
of function, so a cylindrical grating with a radius of curvature of 5.3 cm is used to eliminate
the the very rapid divergence in the H plane. For a highly curved surface such as this one at
an off— axis angle, there is higher loss at the edges due to the projected tilt of the grooves
relative to the polarization vector.
These optics have been tested with the prototype and show essentially the intended
function. The grating efficiency is in fact very high and a scan through two orders shows
only 3% of the power in zeroth order relative to the desired first order. The spurious second
order is exactly backscattered and is unmeasurable, but is predicted to be the same as the
zeroth order. The focusing action is very good in the H plane, producing a beam at the
input requiring only one further cylindrical focusing mirror or lens (in the other plane). At
the output frequency the beam can be made fairly symmetric with just the one mirror. One
additional complication is due to the off axis cylinder; the focal length of such a mirror
8
6
4
r\
0>
2
Q
8
w
-2
N
-4
<E
-6
I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I ' I I I I I I
I I I I I I I I I I I I I I I I I I I I I ■ ■ I ■ I ■ I ■ ■ I ' ■ ■ ' I ■ ■ ■ ■ I
-25 -28 -15 -18 -5
18 15 28 25
El CDeg)
3.8 p-
-3.8
-8 -6 -4-2 8 2
El CDeg5
Fig. 4. Beam pattern measured at the line focus of the cylindrical grating
at (a) 164 GHz and (b) 492 GHz. Contour interval is 3 dB. The focal distance
is different for the two frequencies.
Page 188 Third International Symposium on Space Terahertz Technology
depends on the incidence angle (in the plane of Fig. 3). Because the beams are broad in this
dimension, this is quite noticeable in the beam patterns, particularly for the input since this
beam is the widest, and the most off— axis. This effect is apparent in the contour map of the
input beam, shown in Fig. 4a, as measured near the refocusing point about 50 cm away. The
problem is much less apparent in the output beam, shown in Fig. 4b. The easiest cure is to
make the grating a different shape, but a periodic ruling can only be made on a few surfaces.
It appears that a conical grating can satisfy the requirements, if it represents only a small
distortion of a cylinder, but this solution remains to be tested. Optics beyond this point are
still being designed, but appear to be straightforward mirrors or lenses.
Conclusions
A new type of mounting structure is proposed for use in a frequency tripler, which is
relatively easily fabricated for frequencies above 1 THz. Beam patterns are suitable for
efficient coupling to an input source, although the optimum optical system remains to be
designed. The present use of a blazed diffraction grating appears to be an excellent means to
separate frequencies, with high input-output isolation. The behavior of the device seems
less suitable for use as a doubler, although this has not been explored in detail.
A second use might be as a mixer mount for a higher performance submillimeter
mixer than cube corner mixers. The application would be much more narrow band, but
offers more optimal impedances and a greatly reduced side lobe level. It might also be
practical to use for a third harmonic mixer, extending the frequency range of mixers using
multiplied sources. A recent analysis of a third harmonic mixer for 1 THz indicates that a
properly designed mixer should be quite competitive [4], but this analysis has not been
extended to the impedance environment presented by this structure.
References
1. B.L.A. Rydberg, B.N. Lyons, and U.S. Lidholm, "Multipliers for THz heterodyne
systems," Proceedings of the Second Int'l. Conf. on Space THz Tech., pp. 212-217, 1991.
2. N.R. Erickson, "Low noise 500—700 GHz receivers using single— diode harmonic mixers,"
Proceedings of the First Int'l Symposium on Space THz Tech., pp. 339-408, 1990.
3. P.H. Siegel, R.J. Dengler, and J.C. Chen, "THz dichroic plates for use at high angles of
incidence," IEEE Microwave and Guided Wave Lett., Vol. 1, pp. 8-9, 1991.
4. N.R. Erickson, "Low noise submillimeter receivers using single— diode harmonic mixers,"
to be published in Proceedings of the IEEE, Nov. 1992.
Third International Symposium on Space Terahertz Technology p a q e 189
A 380 GHz SIS Receiver using N 9 3 ">•% 7 JZ,
Nb/A10 x /Nb Junctions for a Radio Astronomical /&~ &\
Balloon-borne Experiment : PRONAOS /£>0£3®
P. Febvre* + , P. Feautrier**, C.Robert* J.C. Pernot**, /° &'
A. Germont*,M. Hanus**,R. Maoli*,M. Gheudin*,G. Beaudin*, P. Encrenaz**
*Observatoire de Paris-Meudon, DEMIRM - URA 336
5, Place Jules Janssen 92195 Meudon - France
** Ecole Normale Superieure, Laboraxoire de Radioastronomie,
24 rue Lhomond 75005 Paris - France
+ Now at Jet Propulsion Laboratory
MS. 168-314
4800 Oak Grove Drive
Pasadena, California 91109 USA
ABSTRACT
y ^ The superheterodyne detection technique used for the spectrometer instrument of the
PRONAOS project will provide a very high spectral resolution (Av/v = 1(H). The most critical
components are those located at the front-end of the receiver : their contribution dominates the total
noise of the receiver. Therefore it is important to perform accurate studies for specific components,
such as mixers and multipliers working in the submillimeter wave range.
Difficulties in generating enough local oscillator (L.O.) power at high frequencies make SIS
mixers very desirable for operation above 300 GHz. The low L.O. power requirements and the low
noise temperature of these mixers are the primary reason for buiding an SIS receiver.
This paper will report the successful fabrication of small (< 1 pm 2 ) Nb/Al-Ox/Nb junctions
and arrays with excellent I-V characteristics and very good reliability, resulting in a low noise
receiver performance measured in the 368/380 GHz frequency range.
Page 190 Third International Symposium on Space Terahertz Technology
I - INTRODUCTION
Observations from a stratospheric balloon are unobstructed by the atmosphere which is
opaque at submillimeter and far-infrared wavelengths from the ground. For this reason, a
submillimeter balloon-bome observatory is being developed under the responsibility of the "Centre
National d'Etudes Spatiales" (CNES), the French Space Agency.
It consists of a stabilized gondola supporting a 2 meter diameter telescope, associated
alternately with an infrared multiband spectrometer or a submillimeter heterodyne spectrometer
(SMH). This last instrument will be used to simultaneously detect the 368 GHz O2 line and the
380 GHz H2O line in the interstellar medium. It is scheduled to fly in fall 1994 using a 1,000,000
m 3 balloon at an altitude of 37 km.
Receivers using SIS tunnel junctions have shown better sensitivities than Schottky diode
receivers operated at millimeter and submillimeter wavelengths. Theoretically, sensitivities
approaching the quantum limit can be achieved [1].
Up to about 300 GHz the most sensitive receivers use waveguides and superconducting RF
tuning circuits integrated with the SIS junctions [2,3,4,5,6,7,8]. Above this frequency, two options
appear to be available. The first possibility is to design a waveguide mixer (with full-height or
reduced-height waveguide) using two tuners (i.e. generally a backshort and an E-plane tuner)
[9,10]. A DSB receiver noise temperature of 150 K at 345 GHz has been reported with this design
[9]. Another possibility is to use a quasioptical SIS mixer, which is very promising above 500 GHz
where very small waveguides are very difficult to machine [11,12,13,14]. This design is
compatible with tuning elements.
Finally we have chosen for our first experiments a waveguide design because it is better
understood than open- structure mixers.
II - RECEIVER DESCRIPTION
A block diagram of our submillimeter wave heterodyne spectrometer is shown in figure 1.
Rotation of a flat mirror set allows the calibration of the receiver by commuting the incoming beam
from the telescope between a hot and a cold load. Due to the short wavelengths, a quasioptical free
space propagation is adopted [15]. A Mach-Zehnder type diplexer is used for the 374 GHz local
oscillator signal injection into the SIS mixer. The L.O. source consists of a phase-locked 93.5 GHz
Gunn diode oscillator combined with two varactor diode doublers connected in series. The
intermediate frequency (I.F.) is chosen at 5.85 GHz to allow the simultaneous detection of the O2
line in the lower band at 368 GHz and the H2O line in the upper band at 380 GHz. The I.F. output
feeds a specially designed cooled low-noise HEMT amplifier with a gain of 30 dB. A noise
temperature of 18 K has been achieved at 5.85 GHz over a 700 MHz bandwidth at a temperature of
Third International Symposium on Space Terahertz Technology
Page 191
27 K [16]. The signal is then amplified at room- temperature and coupled to the acousto optical
spectrometer (AOS) subsystem with a resolution of 800 kHz in a 800 MHz bandwidth.
telescope
368 -380 GHz
Calibration! r.f
device
Quasi-optical
diplexer
— T"
374 GHz
SIS Mixer
&
Integrated I.F.
matching circuit
I.F.
— ►
HEMT
amplifier
Liquid He cryostat
5.85
GHz
I „i _. _ _ _. _ i. _ ii
374
GHz
r
Output
Second
Doubler
^187
■ GHz"
First
Doubler
93.5
GHz
I
Gunn
oscillator
Phase-Lock
Loop
Figure 1: Block diagram of the receiver front-end
used for PRONAOS
III - 380 GHz SUB MILLIMETER RECEIVER FRONT-END
III-l - SIS junctions fabrication procedure
We report here the fabrication process of Nb/Al-AlOx/Nb junctions with very sharp I-V
curves and a gap voltage for one junction about 2.9 mV at 4.2 K. A high gap voltage is known to
be necessary for good results at high frequencies ( above 300 GHz ). It is the reason why NbN
junctions are promising for very high frequencies (above 500 GHz ). The smallest junction area
achievable with our technology without deterioration of the I-V curve is 0.9 |im2. Our process has
already been described in a previous paper [17]. Some parameters have changed since this article to
obtain the desired junction area for the 380 GHz mixer.
The fabrication process is described on figure 2 .The Nb/Al-AlOx/Nb trilayer is deposited
on the whole substrate without breaking the vacuum in order to have a good barrier interface (see
fig. 2-a). The diameter of this substrate is one inch, and the thickness is 95 ± 5 fim. It is made of
fused quartz and is polished on one side. During the deposition the substrate is attached to a copper
heat sink cooled by a closed water circuit at 20 °C. The vacuum is made by a cryopump with a
background pressure typically under 5. 10"6 Pa. The Nb and Al films are sputter deposited by a DC
magnetron at an argon pressure of 1.1 Pa. The Nb base electrode (170 nm thick) and
Pagel92 Third International Symposium on Space Terahertz Technology
counterelectrode (100 nm thick) are evaporated at a rate of 1.9 nm/sec. The Al film ( 10 nm ) is
deposited with an oscillating substrate table at a rate of 0.2 nm/sec and is oxidized by introducing Ar
+ 10% 02 into the chamber for 20 to 30 min at 60 to 1000 Pa.
A positive photoresist is deposited and patterned to define the RF filter with an etching
technique. Nb and Al films are etched by reactive ion etching in SF6- Nb is etched with a 10 seem
SF6 flow at 0.7 Pa using 60 W of power. The corresponding etching rate is 200 nm/min. Al is
etched at lower pressure and higher power with an etching rate of 10 nm/min; the ST 6 flow is
5 seem, the pressure is 0.3 Pa and the power is 80 W. Under these conditions, the etching is
dominated by a mechanical action rather than a chemical effect like in the plasma etching method.
We observed that a CF4 gas does not etch Al even at low pressure. RIE with Ar has not been
selected, because it produces too much damage on the resist (with Ar, it is only a mechanical
etching process).
After removing the remaining photoresist in acetone (see fig. 2-b), a new resist layer is
deposited to define the junction area (see fig.2-c). This is the critical point of the process which
limits the smallest area achievable by this technology. Our mask aligner uses a 400 nm UV source
and is limited to 0.8 jxm resolution. In practice, it is impossible to define a diameter smaller than 1
|j.m ( ie. an area smaller than 0.9 jirn^ ). This resist is used to protect the upper layer of Nb etched
by RIE under the following conditions : 20 seem of SF^, 6 seem of O2, a pressure of 0.7 Pa and a
power of 60 W. If the etching rate (100 nm/min) is lower than for the trilayer etching (see fig. 2-a),
these conditions provide sloped edges which are easier to insulate without microshorts in the next
step. The etch stops at the AI2O3/AI barrier, because the etching rate of Al is very low with SF6/O2.
We use laser end point detection to avoid overetching ( it is necessary to have a sufficient thickness
of resist for the SiO lift-off ).
Once the upper Nb etched then a 300 nm layer of SiO is evaporated to insulate the junction
perimeter (see fig.2-d). The excess SiO is removed in acetone (lift-off). Then, the junctions in
series are connected together by a 300 nm layer of Nb sputter deposited with a rate of 1.3 nm/s
through another resist stencil. The excess Nb is finally lifted-off in acetone. Different experimental
investigations have been made to optimise each parameter. For example, the stresses in Nb films
have been minimized by changing the Ar pressure during the sputtering step. The stresses are
evaluated by optical interferometry. The Nb edge is another parameter we have studied. We
succeeded in obtaining sloped edges with a reasonable selectivity by using a mixture of SF6 and O2
at low pressure for the RIE. Finally, anodisation spectroscopy was an useful method to investigate
the quality of the interfaces Nb/Al and to understand the diffusion problem of Al into Nb; such a
diffusion process gives poor quality junctions .
Third International Symposium on Space Terahertz Technology
Page 193
2-
Then, the individual junctions ( 400 junctions per substrate of 1 inch diameter) are cut with a
dicing saw and cooled in liquid helium at 4.2 K to test their I-V characteristics. It is possible to test
6 junctions in one run. The junctions are connected with spring contacts on gold pads evaporated at
the ends of the R.F. filter. With this technique, we can contact the 6 junctions very quickly without
problem of series resistance on Nb surface.
(a)
V//////////////////////M/7A
\i^^^^^^
(b)
Base electrode
V////////////7777?7t
xwwwvwwwwwv
(c)
(d)
Resist
72.
KWVW\V^VVVW\\S\M
SiO
X
*■'*■].■! .' ' .
m^K\^\\^\\\N\S\V^5^
Nb contacts
(e)
^^^^^^
Figure 2: Fabrication process of Nb/AI-AIOx/Nb junctions
(a) Nb/AI-AIOx/Nb deposition. Definition of the base electrode by photolithography, (b)
Etching of the trilayer. (c) Etching of the upper electrode, (d) Self-aligned deposition of a
SiO insulating layer, (e) Nb interconnection layer.
Figure 3 gives an example of a typical I-V curve of an array of 2 junctions in series. The
area of each junction is 0.9 nm^, so the effective area of the array is about 0.45 jjirA
Pagel94 Third International Symposium on Space Terahertz Technology
Figure 3
20 uA / div
i
2 mV / div
A eff = 0-47 p.m2
RN = 150 ft
R 300 K = 182 fl
III-2 - Mixer design
a) General features
The SIS mixer block is based on the Ellison design [9]. It includes an electroformed
integrated dual-mode Potter hom [18] transformed by a circular to rectangular transition into a third-
height reduced waveguide [19] to increase R.F. bandwidth and decrease the characteristic
impedance at 150 Q . Superconducting coils (to suppress the Josephson Current), an I.F. matching
circuit and junction DC bias are integrated in the mixer block in order to facilitate the installation of
the SIS mixer in the laboratory cryogenerator or in the flight cryostat . This also allows better
reproducibility of mixer performance due to the optimization of the mixer mount for the SIS
junctions. Dimensions of the waveguide are 700 (im x 120 u.m and two contacting tuners (i.e. a
backshort and an E-plane tuner placed at Xg/2 towards the feedhom in front of the junction) provide
a large range of embedding impedances to the SIS junctions (see figure 4).
b) Mixer configuration
A low-pass microstrip filter designed on Touchtone [20] is fabricated by photolithography
on a 0.1 mm thick fused quartz substrate; its rejection is about 20 dB at 374 GHz. The metallization
is made of Nb like the SIS junction and this 1.8 mm long 0.2 mm wide substrate is only put down
in the mixer block channel on a thin silicon grease film for a better thermal contact. Mechanical
support is provided by this silicon grease film when cooled at 4 K and by the 25 \xm gold wires
contacting the filter to ground and the LF. output. This assembly allows numerous tries of different
junctions without breaking substrates. The LF. output gold wire is fixed with silver glue on the
low-pass filter at one end and directly on the LF. matching circuit at the other end. This matching
circuit formed on Duroi'd (£r=10.2) supports the junction DC bias too. This avoids the sudden
impedance change of a SMA connector, increases the LF. bandwith and decreases the LF. losses.
The DC bias includes two 10 kQ chip resistors (to prevent junction from being destroyed by voltage
spikes) followed by an insulated wire soldered at X/4 of a X/2 stub (see figure 5) to provide
approximately an open circuit at the LF. frequency of 5.85 GHz on a 700 MHz bandwidth. The
Third International Symposium on Space Terahertz Technology Page 195
25 u.m gold wire is the first part of the I.F. matching circuit, then a length of a micTostrip line
provides a real impedance transformed into 50 Q by a X/4 line (figure 5).
Figure 4
Miniature connectors
(for bias)
I.F. output
-§fj e
Pocket for
superconducting
coil
s
figure 5
Page 196 Third International Symposium on Space Terahertz Technology
The 1.8 cm diameter superconducting coils have been designed to produce 310 Gauss with
a current of 1 A. Each one is made of about 1200 turns of Niobium-Titanium superconducting wire.
Indeed, for circular junctions of surface S, the magnetic field suppressing the Josephson current is
given by :
2,23 . IP' 11
B(Gauss) = =
d.VS (for one flux quantum )
with : d = 2Xl+w where:
Xl = London penetration depth of Nb (m)
w = width of insulator between the two superconductors (m)
The area of the smallest junctions fabricated in the laboratory is about 1 \xm 2 . So, with a
pessimistic value of the London penetration depth (400 A), B = 255 Gauss ; the real value should
be lower. The coils are small, because the flight cryostat was specified for a smaller Schottky
mixer. Moreover, some constraints about the optical axis were already fixed in the flight cryostat.
The mixer block is a Faraday cage for these coils against electromagnetic spikes even if any external
magnetic field can penetrate into it .
IV - LABORATORY MEASUREMENT BENCH
Results shown further have been obtained on a laboratory bench with a 4 K cryogenerator
including two closed circuits of helium. The first one is a classical CTI 1020 compressor including
two stages at 50 K and 12 K. The second one is a Joule-Thomson expansion pumping on the 12 K
stage to reach 3 to 4 K on the "4 K" stage. Temperature can be quickly changed and stabilized by
varying the return helium pressure of the 4 K helium circuit [21]. A teflon corrugated window is
used on the room temperature shield for the quasioptical RF input. The heat flux entering the
cryogenerator is then reduced with an IR filter. It's a 80 Jim thick (one wavelength at 374 GHz) 48
mm diameter crystalline quartz plate mounted on the 50 K stage shield. Then a 0.8 mm thick
fluorogold window, 38 mm diameter, filters the far IR 50 K blackbody radiations. The SIS mixer is
on the 4 K stage at the focus of a cold corrugated teflon lens cooled by the same stage.
Mechanical contacting tuners are operated by vacuum feedthroughs and are manually
movable with micrometer drives when measuring receiver performance. Each electrical wire, I.F.
cable or tuner drive is thermalized at 12 K and 50 K to exhaust heat flows. Some miniature
connectors are used for the DC bias. A four points measurement of the I-V curve releases us from
any series resistance.
The I.F. output of the SIS mixer is connected to a semi-rigid cable followed by a coupler, an
isolator and the HEMT amplifier. This low-noise amplifier is installed on the 12 K stage, its output
cable is thermalized at 50 K before going out of the cryogenerator (see figure 6). The coupler is
Third International Symposium on Space Terahertz Technology
Page 197
used to inject an additive noise at the I.F. mixer output to know its match relatively to 50 Q. A
preliminary calibration without mixer allows us to calculate approximatively the mixer temperature
Tm and its conversion losses LM-
The socket ot the superconductive coils is installed on the 12 K stage to have a better thermal
contact between superconducting and copper wires and to prevent a heating of the 4 K stage. The
L.O. and signal injections are achieved by a quasi-optical diplexer. The coupling ratio for the L.O.
is higher than 90 %.
Corrugated n
teflon window ^
For additive
noise injection
Towards
room-temperature
amplifiers
Copper wire A
. Towards
\ micrometer
drive
380 GHz SIS mixer
(with I.F. matching circuit)
Quartz window: * Fluorogold window:
thickness: one wavelength thickness: one wavelength
Figure 6
Page 198 Third International Symposium on Space Terahertz Technology
V - LOCAL OSCILLATOR VARACTOR DIODE DOUBLERS
The structure of each doubler has already been described in a previous paper [22]. The
maximum efficiency found for the first doubler was about 18 % for an incident power of 15 mW
with a 5P8 diode of the University of Virginia. The input frequency was 91.6 GHz and the output
power was higher than 6 mW with a 50 mW input power. These results haven't been found again
with the other doubler block at 93.5 GHz. They were due to a very good coupling between the
diode and the waveguide by the whisker. More commonly, we can reach 3 to 4.5 mW with a good
reproducibility and with an input power of 50 mW at 93.5 GHz. A typical curve of our last results
is shown on figure 7.
93,5-187 GHz Doubler
Output Power
(5P8 diode)
(mW)
Efficiency
<; .
_ Q<?n
4,5 -
■l*'
Efficiency "'■--. ry^ _
- 8%
4 -
■ 1
$
- *
■
*
£l * * -
- 7%
3,5 -
i
■
■ 6%
3 -
" i
■
*
PDutput Power
- 5%
2,5 -
- 9
2 -
1
■
1
■ >
■ 4%
1,5 -
u /
- 3%
1 -
i ^/
- 2%
0,5 •
■ /
Whisker length -
- 1%
■ J§r
289 microns
o ntffrT., .. • ■
iiit
_ r&z.
10 20
30 40 50 60 70
Input Power (mW)
Figure 7
Concerning the second doubler, its input power (approximatively the output power of the
first doubler) is low and consequently its efficiency is relatively low. Indeed, we can see on
Third International Symposium on Space Terahertz Technology
Page 199
figure 7 that the efficiency of the first doubler is lower than 4 % at 187 GHz for an input power of
about 5 mW.
Partly due to the much higher frequency, we can foresee that the second doubler will not
produce so much power. Such a local oscillator cannot be used for a Schottky mixer. Nevertheless,
some diodes whose the maximum efficiency is obtained for a 3-4 mW input power like bbBNN
diodes could provide sufficient power to pump a Schottky mixer [23]. The best output power
obtained at 374 GHz is approximatively 30 p.W with the bolometer hom put directly across from the
second doubler horn, i.e. an efficiency lower than 1%. Two types of diodes have been tested, 2T8
and 2T9, they come from the University of Virginia and we can see on the following figure 8 that
the 2T8 diode provides more power than the 2T9 diode. This is partly due to its smaller capacitance
(4 £F versus 8 fF for the 2T9 diode).
Output Power
(microwatts)
187-374 GHz Doubler
30 -i
25 -
20 -
15 -
2T8 diode /
whisker length: /
146 microns /
10 -
♦
5 -
2T9 diode
whisker length:
150 microns
f) t
1
2 3
Input Power (mW)
ii| i i i i 1
4 5
Figure 8
Page 200 Third International Symposium on Space Terahertz Technology
To prevent the second doubler diode from being destroyed by voltage spikes and due to the
low input power, it has been short circuited in direct current instead of being reverse biased for an
optimum efficiency, this diminishes the output power. Moreover, input and output backshorts of
each doubler have been fixed or soldered which still damages performance. So the output power at
374 GHz is about 13 (iW. Other measurements have been made with a quasi-optical bench
composed of two corrugated lenses which is approximately the bench used for the measurements
of the SIS mixer. The output power is then 10 jiW. This local oscilltor signal is powerful enough,
even to pump four SIS junctions in series.
VI - RESULTS
Different types of junctions have been tested with 2, 3 or 4 junctions in series coming from
the same wafer. The best results obtained with each substrate are summarized in table I. The L.O.
frequency is 374 GHz, the I.F. center frequency is 5.85 GHz. Measurements have been made with
a 285 MHz I.F. bandwidth filter, we used the Y-factor method with 2 loads at 77 K and 295 K.
Q We can firstly point out the good match between calculated and measured values of the
magnetic field suppressing the Josephson current Ij. The product B(Ij = 0) x D is reported on the
following table II (for one flux quantum), where B(Ij = 0) is the magnetic field suppressing the
Josephson current and D the diameter of one junction. This product should be constant for junctions
fabricated on the same wafer according to the previous formula of III-2-b: d is a parameter
depending only on the oxidation time of aluminium in AI2O3. We see that B(lj=0)x D is nearly
constant to within about 10 %, this comes from the uncertainty of the junction areas. We can also
deduce the London penetration depth of our niobium films which is about 600 A.
□ Nevertheless, the Josephson Current is not always completely suppressed with one flux
quantum, because the areas of the junctions in series are slighdy different. The Josephson current
for each of the couple of junctions in series of one substrate is reported on figure 9.
The relative difference of the magnetic field suppressing the Josephson Current of each
junction taken individually is about 5 to 10 %, that means a relative difference of area between the
two junctions of 10 to 20 %. Such a difference is in good agreement with the accuracy of
photolithography to define small junction areas. For this reason, the current densities and the coRnC
products are not exactly the same for the different junctions of the table I. This corresponds to the
uncertainty of the value of the junction area.
Q The measurements of the required L.O. power are deduced from a preliminary calibration
of the L.O. output power as a function of the first doubler self-biased voltage. The required power
depends on the square of the number of junctions in series; four junctions in series should require
Third International Symposium on Space Terahertz Technology
Page 201
about four times as much power as two junctions in series. We observed a 3.7 dB difference
between expected and measured values which corresponds mainly to the R.F. mismatch at the
374 GHz frequency since we measured the incoming L.O. power. And we can see that the
difference of the conversion losses for these junctions is 3 dB, this point confirms the first one.
Junction
E380-1-8-2
E380- 1-6-5
E380- 1-8-4
E380-1-4-1
E380-1-8-6
Diameter (jim)
1.1
1.5
1.1
1.9
1.1
Number of
junctions in
series
2
3
2
4
2
Effective
surface (|J.m 2 )
0.47
0.59
0.47
0.71
0.47
Rn(Q)
143
137
150
113
143
coRnC
at 374 GHz
9.5
11.4
10
11.3
9.5
j c (A/cm 2 )
4600
3600
4600
4200
4400
L.O. power
(HW)
?
?
7
7.5
0.8
Magnetic field
applied (Gauss)
175
255
(2 flux quanta)
175
192
(2 flux quanta)
185
DSB receiver
temperature (K)
1200
470
360
525
310
Mixer noise
temperatureTM
(K)
?
200
195
225
155
Conversion losses
(dB)
?
11
9,1
11,8
8,8
Transmitted I.F.
power
between 10
and 40 %
= 90%
= 90%
= 98%
= 97%
Contribution of
amplifier to noise
>70%
57%
46%
57%
50%
table I
Junction
E380-1-8-2
E380- 1-6-5
E380-1-8-4
E380-1-4-1
E380- 1-8-6
B(Ij = 0)xD
(Gauss x (im)
193
191
193
182
203
table II
Page 202
Third International Symposium on Space Terahertz Technology
□ Relatively high conversion losses result in a contribution of 50 % for the HEMT
amplifier in the receiver noise temperature. These conversion losses include intrinsic conversion
losses increased by RF quasioptical injection, RF and I.F. mismatches, RF filter and I.F. matching
circuit losses. Differences of receiver noise are mainly due to miscellaneous conversion losses.
Indeed, some different effective areas of junction have been tested and the couple of tuners don't
enable to completely tune out the junction capacitance because the coRnC product is high (>8). So
the excess of conversion losses corresponds to a higher RF mismatch.
Some typical curves of different measured junctions are shown on figure 10.
dc Josephson Current
for each junction
(microamperes)
15
Junction E380 -1-8-2
Seff = 0.47 microns A 2
Normal resistance: 143 ohms
Substrate with
2 junctions in series
10 to 20 Gauss difference
50
100 150 200 250 300
Magnetic field (Gauss)
figure 9
Remark : Exact values of magnetic fields haven't been measured but calculated with current flowing
through the coils. Error is around ±5 %.
Third International Symposium on Space Terahertz Technology Page 203
The three curves shown below are some experimental curves digitalized by our data
acquisition system of different arrays of SIS junctions in series fabricated on the same wafer with
nearly the same normal resistances (about 150 Q). Fot this reason, the current densities are of the
same order of magnitude for each array.
Static I— V curves tor 2.3 and * SIS junctions in series
V (mV)
figure 10
10 12 I* 16 18
We can see on figure 1 1 the dc characteristic of the junction E380-1-8-6 (a). Also shown on
this figure is the I-V curve of the same junction pumped with the 374 GHz L.O.. The width of the
photon assisted step is 2.h.VL.o7e where h is the Planck constant , e is the electron charge and vl.o
is the frequency of the local oscillator.
60 -
40
20
V (mV)
figure 11
10
Page 204 Third International Symposium on Space Terahertz Technology
The static impedance at the bias point (approximatively 4 mVrthe middle of the first photon
step) is about 500 Gl so the I.F. circuit has been designed to match this impedance, we assumed that
it is close to the impedance at 5.85 GHz (Ri.r). The normal resistance is 143 Q and the range of the
quotient Ri.f./Rn has been found to be contained between 3 and 4.5, the value of Ri.f. being
adjusted by varying the L.O. power. We can see in table I that the I.F. match for the last four
junctions is good which validates our assumption.
On the contrary the high receiver noise temperature measured for the first junction was due
to a poor I.F. match, the I.F. impedance being unknown at that time.
Some I-V curves with and without suppression of the Josephson current are plotted on the
following figure 12. We can observe 3 Shapiro steps due to the coupling of the L.O. power with
the Josephson current. when it is not suppressed. The width of these steps is exactly one half of the
quasiparticle step due to the L.O. power. These sharp steps partly explain the instabilities observed
when the Josephson current is not. completely suppressed.
JUNCT
ON E380-186. T=4.38 K
80 -
60 -
/
40 -
2.h.i//e *s
2.h.u/2.e
\ Y
20 -
; J
*~ V
VITHOUT
JOSEPHSON CURRENT
VITH
-
- ~
. i i i 1 i i i 1 i i ..... ,
s
V (mV)
10
12
figure 12
Some dependances of different parameters are shown on the following figures. We can see
on figure 13 the influence of the magnetic field to the noise receiver.
The noise temperature begins to increase for a magnetic field lower than 170 Gauss which
corresponds to a residual Josephson Current of about 1 (iA providing an additive Josephson noise
coming with instabilities of the I.F. output power.
Third International Symposium on Space Terahertz Technology
Page 205
DSB receiver
noise temperature
(K)
450 t
400 --
350 --
300 --
Junction E380 -1-8-6
Seff = 0.47 microns A 2
Normal resistance: 143 ohms
250 --
200
Bias current: 10 microamperes
Physical junction temperature: 4.41 K
I.F. Bandwidth: 285 MHz
— i 1 1 1 1 1 r-
120 140 160 180
Magnetic field
(Gauss)
200
220
240
figure 13
The noise temperature is plotted as a function of bias current on figure 14. The receiver
noise temperature remains lower than 330 K with a relative variation of bias current of 20 % which
is adequate for our balloon-borne experiment where there is no remote control of the DC bias; all the
other parameters remained unchanged.
At last, the L.O. frequency was varied from 345 to 385 GHz (see figure 15) the receiver
noise temperature is higher at lower and higher frequencies than 374 GHz. This is partly due to the
narrow RF bandwidth of the Potter hom. We can point out that the receiver temperature is below
380 K in the frequency range from 355 to 385 GHz.
Influence of temperature was only observed with the junction E380- 1-6-5. With other
junctions, the mixer noise temperature has not decreased by cooling more the junction; this is
certainly due to a poor thermal contact with the silicon grease film.
Page 206
Third International Symposium on Space Terahertz Technology
DSB receiver
noise temperature
(K)
460 -r
Junction E380 -1-8-6
Seff = 0.47 microns A 2
Normal resistance: 143 ohms
Physical junction temperature:4.32 K
I.F. Bandwidth: 285 MHz
Magnetic Field - 210 Gauss
9.5 10
Bias current
(microamperes)
10.5
11
11.5
figure 14
DSB receiver
noise temperature
(K)
410 1..
3901
370 +
350
330
Junction E380 -1-8-6
Seff = 0.47 microns A 2
Normal resistance: 143 ohms
Bias current = 9.1-10.6 microamperes*'
Physical junction temperature :4.3 K
I.F. Bandwidth: 285 MHz
310
— I —
350
— I —
355
-+-
+
+
■-«a«
345
360 365 370
L.O. Frequency (GHz)
375
380
385
figure 15
Third International Symposium on Space Terahertz Technology Page 207
V - CONCLUSION
Some Nb/Al-Al203/Al SIS junctions with small areas and sharp I-V curves have been
successfully fabricated, dc measured and integrated in the mixer. The smallest area achievable with
our process is about 0.9 (im^. Arrays of two junctions with this area have been made, the effective
area is then around 0.45 (jm^. They are very stable according to some repeated thermal cycles: more
than 15 cycles have been completed between room temperature and 4 K temperature and no change
has been detected. This reliability is essential for space applications.
The 380 GHz SIS mixer was designed with an integrated I.F. matching circuit and two
integrated superconducting coils; it has been tested over a 40 GHz L.O. bandwidth. The best
receiver noise temperature (310 K DSB) has been measured with an array of a couple of junctions
in series having an effective surface of 0.47 |im2 and a normal resistance of 143 Q.. The L.O.
frequency was 374 GHz. The relatively high conversion losses (8.8 dB) reveal a R.F. mismatch. It
could be decreased by using junctions with lower capacitances (i.e. areas) and lower normal
resistances. Then the fabrication of SIS junctions with higher current densities is planned. The
lowest mixer noise temperature is around 155 K and some new junctions with lower normal
resistances should also reduce it So we are optimistic for the following.
The 374 GHz L.O. source has been made with a fundamental InP Gunn Oscillator at
93.5 GHz followed by two GaAs varactor doublers in series. This subsystem provides enough
power to drive the SIS mixer even with 4 junctions in series but a more powerful first multiplier
will be necessary to produce more power at higher frequencies (above 500 GHz) for the future.
A 6 GHz low-noise H.E.M.T. amplifier has been specifically designed for cryogenic
applications, it meets fully the specifications and will be used in connexion with the SIS mixer. The
contribution of the amplifier to the system noise is about 50 % due to the high conversion losses.
We hope that the new junctions will decrease this contribution.
Acknowledgments : We would like to thank Gilles Ruffle for his valuable aid and support,
Andre Deschamps for the data acquisition system and Olivier Perrin for the design of the doubler
blocks. We are especially grateful to Serge Lebourg and Jean Morin for their help on the mechanical
realizations for the measurement bench. We also wish to thank Veronique Serpette (Observatoire de
Paris) for the numerous photolithographies of the I.F. matching circuits. In addition we thank Marc
David for his assistance and support on cryogeny. Thanks also to Albert Brel, Annick Gassais and
Francoise Gadea for their technical help. We would also like to thank Matthew Carter and Jacques
Blondel of IRAM (Institut de Radioastronomie Millimetrique) for useful discussions.
Finally we are greatly indebted to William R. McGrath for his careful reading and numerous
comments on this article.
This work is supported by the Centre National d'Etudes Spatiales (CNES) and the
C.N.R.S. (URA 336)
Page 208 Third International Symposium on Space Terahertz Technology
REFERENCES
[I] J J*. Tucker "Quantum limited detection in tunnel junction mixers", TF.P.F. J. Quantum Electron, vol QE 15,
pp 1234-1258.Nov.1979
[2] S.K. Pan, A.R. Kerr, MJ. Feldman, A.W. Kleinsasser, J.W. Stasiak, R.L. Sandstrom and W.J. Gallagher "A
85-116 GHz SIS receiver using inductively shunted edge junctions", IEEE Trans. MTT, Vol.37,N 3, March 1989
[3] A.R. Kerr, S.K. Pan: "Some recent developments in the design of SIS mixers" Jnt. J. of Infrared and Millimeter
Waves, Vol.1 1,N° 10,1990
[4] R. Blundell, M. Carter and K.H. Gundlach:"A low-noise SIS receiver covering the frequency range 215-250 GHz,
InU. of Infrared and Millimeter Waves, Vol 9,1^4,1988
[5] BiM. Ellison and R.E. Miller: "A low-noise 230 GHz SIS receiver", InU. of Infrared and Millimeter Waves,
Vol.8,pp 609-625June 1987
[6] H.H.S. Javadi, W.R. McGrath,S.R. Cypher, B. Bumble, B.D. Hunt and H.G. Leduc:"Performance of SIS mixers
at 205 GHz employing submicron Nb and NbN tunnel junctions",Digest of the 15th International Conference on
Infrared and Millimeter Waves,December 1990
[7] D. Winkler, W.G. Ugras, A.H. Worsham and D.E. Prober, N.R. Erickson and P.F. Goldsmith: "A full-band
waveguide SIS receiver with integrated tuning", IEEE Trans, on Magnetics, ^1.27^*2, March 1991
[8] J.W. Kooi, M. Chan, T.G. Phillips, B. Bumble and H. G. Leduc:"A low-noise 230 GHz heterodyne receiver
employing 0.25 p.m 2 Area Nb/AlOx/Nb tunnel junctions", 2"d International Symposium on Space Terahertz
Technology, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, Feb. 1991
[9] B.N£llison, P.L.Schaffer, W.Schaal, D.Vail, R.E.MiUer : "A 345 GHz SIS receiver for radio astronomy ", IntJ.
of Infrared and Millimeter Waves, Vol. 10, N° 8, 1989
[10] C.E. Honingh, M.M.T.M. Dierichs, H.H.A. Schaeffer, T.M. Klapwijk and Th. de Graauw:"A 345 GHz
waveguide mixer with two mechanical tuners using an array of four Nb-Al-AbC^-Nb SIS junctions", 2 n d
International Symposium on Space Terahertz Technology, Jet Propulsion Laboratory, California Institute of
Technology, Pasadena, Feb. 1991
[II] M. Wengler, D.P. Woody, R.E. Miller, T.G. Phillips:"A low noise receiver for millimeter and submillimeter
wavelengths" Jnt. J. of Infrared and Millimeter Waves, Vol. 6, pp 697-706, 1985
[12] T.H. Buttgenbach, R.E. Miller, M.G. Wengler, D.M. Watson, T.G. Phillips:"A broad-band low-noise SIS
receiver for submillimiter astronomy", IEEE Trans. MTT, Vol. MTT 36, pp 1720-1726, Dec. 1985
[13] X.Li, P.L. Richards, F.L. Lloyd, "SIS quasiparticle mixers with bow-tie antennas", Int. J. of Infrared and
Millimeter Waves, Vol. 9, pp 101-103, 1988
[14] J. Zmuidzinas, H.G. Leduc, "Quasi-optical slot antenna SIS mixers", to be published in IEEE Trans. MTT,
1992 and 2 n{ * International Symposium on Space Terahertz Technology, Jet Propulsion Laboratory, California
Institute of Technology, Pasadena, Feb. 1991
[15] PJ.Goldsmith, " Quasi-optical techniques at millimeter and submillimeter wavelengths", Infrared and Millimeter
Waves vol 6 :System and components, KJ.Button (editor), Academic Press, New York, p 277-343, 1982 .
Third International Symposium on Space Terahertz Technology Page 209
[16] CRobert, M.Gheudin: " A 6 GHz HEMT low-noise cooled amplifier for a radioastronomical submillimeter
heterodyne receiver ", 15th International Conference on Infrared and Millimeter Waves, Conference digest, pp. 127-
128, Orlando, dec. 1990.
[17] PJeautrier, J.Blondel, MHanus, J.Y.Chenu, P.Encrenaz, M.Carter :"Low noise 80-1 15 GHz quasiparticle mixer
with small Nb/Al-Oxyde/Nb tunnel junctions", Int. J. of Infrared and Millimeter Waves, Vol. 1 1, No. 2, 1990 .
[18] H.M. Pickett, J.C. Hardy and J. Farhoomand: "Characterization of a Dual-Mode Horn for Submillimeter
Wavelengths", IEEE Trans, on MTT, vol. MTT-32,N 9 8,August 1984
[19] Mixer block constructed by Radiometer Physics, Meckenheim, Germany
[20] Touchtone CAD Software, Eesof
[21] J.CMarechal, J.CPernot, PJ-Encrenaz: "A 2K closed cycle cryogenerator", Conf. URSI, Granada, Sept. 1984.
[22] CPerrin, CRobert, RFeautrier, PJebvre, G.Beaudin, P.Encrenaz, M.Gheudin, Jiacroix, G.Montignac : "380
GHz receiver front-end for the balloon-borne radioastronomical experiment PRONAOS", 2nd. International
Symposium on Space Terahertz Technology, Jet Propulsion Laboratory, California Institute of Technology,
Pasadena, feb. 1991 .
[23] TJ.Tolmunen, M.A. Frerking: "Theoretical performances of novel multipliers at millimeter and submillimeter
wavelengths", Int. J. of Infrared and Millimeter Waves, Vol. 12, No. 10, 1991.
p 210 Third ^ternational Symposium on Space Terahertz Technology
N93-2? 743
//~3-5 A low noise 410-495 heterodyne two tuner mixer,
/ GpOo^l using submicron Nb/Al 2 3 /Nb Tunneljunctions
y G. DE LANGE*, C.E. HONINGH', M.M.T.M. DIERICHS*, R.A.
PANHUYZEN*, H.H.A. SCHAEFFER", T.M. KLAPWUK # , H. VAN DE
STADT*. M.W.M DE GRAAUW'. # Universitv of Groningen. Ntjenborgh 4. 9747
AG Groningen; "Space Research Organisation of the Netherlands. Landleven 12.
9747 AD Groning en.
A 410-495 GHz Heterodyne receiver, with an array of two Nb/Al 2 0-/Nb
tunneljunctions as mixing element is described. The noise temperature
of this receiver is below 230 K (DSB) over the whole frequency range,
and has lowest values of 160 K in the 435-460 GHz range.The
I calculated DSB mixergain over the whole frequency range varies from
>^-11.9 ± 0.6 dB to -12.6 ± 0.6 dB and the mixer noise is 90 ± 30 K.
Introduction
SIS-mixers are currently being used as heterodyne receivers up to submillimeter
wavelengths [1][2]. Two different types of mixers are used to achieve low noise
receivers: quasi-optical mixers, with a fixed tuned broadband planar antenna, and
waveguide mixers, with one or two tuning elements.
In this paper a two tuner waveguide mixer designed for the 400-500 GHz range is
described. The used mixing elements are two different arrays of two Nb/Al 2 3 /Nb
tunneljunctions. This mixer is the second frequency step towards our goal to make
a THz receiver (350-490-750-1000 GHz). At these high frequencies the dimensions
of a waveguide structure become very small (100-400 jim) and it is unclear how far
losses due to surface irregularities deteriorate the mixer performance.
The laboratory tests of the mixer show that in the 400 to 500 GHz frequency range,
waveguide mixers can still achieve low noise temperatures (as low as 160 K DSB),
and the flat response over the band indicates that the current design can be scaled
up to higher frequencies.
Third International Symposium on Space Terahertz Technology Page 211
The behaviour of SIS-mixers is well described by theory [3][4]. An extensive
comparison between theory and measurements has been performed at 345 GHz [5]
and this comparison is started in the 410-490 GHz range. First results show a good
agreement between calculated and measured IV-curves, from which the
electromagnetic environment of the junction is deduced.
This paper describes the design of the mixer, the results from the two different
arrays of junctions and some preliminary analysis of the mixer performance.
Receiver design
The mixer block design for 400-500 GHz is a scaled version of the 345 GHz mixer
described by Honingh et al.[6].
The mixer system is placed inside an Infrared Laboratories HD 3 cryostat. The
signal and LO-power enter the cryostat via a 1 mm thick HDP window of 3 cm
diameter. On the 77 K radiation shield a 200 nm thick quartz plate serves as heat
filter.
A diagonal horn with an aperture size of 4;5 mm, a length of 12 mm and a flare
angle of 11 degrees is used. Laboratory tests of this horn showed a good gaussian
beam-coupling (side lobes <-15db), equal beamwidth in E-,D- and H- planes and
a low cross-polarisation (<-15 dB). In front of the horn a F=31.4 mm HDP lens is
used. The lens is mounted in a holder which can be directly mounted on the
mixerblock. The mixerblock is cut in OFHC. The full height waveguide with
dimensions 0.44*0.22 mm, has a cut off frequency of 340 GHz. The waveguide
system has two moving shorts as tuning elements, each with a quarter wave choke
section to improve the quality of the short. In order to suppress the Josephson
currents in the junction, a coil with 10.000 turns of 0.1 mm Nb wire is placed
around the horn, in front of the mixerblock. The IF-chain consists of a Radiall
R443533 T-bias, a Pamtech LTE 1290 isolator and a Berkshire Technologies L-1.4-
30H IF-amplifier.
Thomson care l natron
410-495 GHZ
Radloneter Physics
76-82 (jhz fiunn-t
douto I er *
tr I p I er
]:-CD"l
® Mixer
© Pan tech I so I a tor
® Berkshire technologies
Lll 1.4-30H,4 0dB Amp I I f I r . ( 1 -2 GHZ)
® Radial I R443533 T-blas
© I'-'Ti .4 HDP lens.
© c'OOvhi Ouartz Window. 77K.
© Inn H.n.P. window. 30 OK
© Mlteq AM 3A 3 0db anpllfir CI -2 GHZ)
© Ml teq AM 2A 2 0db anp I I fir < 1 -2 GHZ)
O Tuner dr I ves
O Narda Directional coupler.
QRadlall Directional coupler
O HP 43ba Power neter
HP Vectra
Stanford lock- in
o
K&.L F I Iter Q
1 .4 GHZ, 100MHZ
©
HP . Powersensor
8481a
8484a
Third International Symposium on Space Terahertz Technology
Page 213
SIS-junctions
The fabrication of the used junction arrays, is described elsewhere [7]. Up to now
two different types of arrays have been used. The arrays differ in junction area,
current density, RF-filterstructure and gapvoltage. An overview is given in table 1.
Junction array
Q34
Q35
A (1 junction) (/xm 2 )
0.8
2
I c (A/cm 2 )
7000
12.000
R n (n)
100
22
RF-filter
Chebychev
YaX
Gap-voltage (mV)
2.7
2.4
Table 1. Overview of the 2 different junction arrays.
The lower gapvoltage of junctions Q35 is caused by a higher oxygen background
pressure during sputter deposition of the trilayer.
Measurement set-up
A schematic diagram of the measurement set-up is given in Fig. 1. The use of two
coherent sources, in combination with a spectrum analyser allow to analyse the
mixer performance at the LO and the upper and lower sideband frequencies. In the
measurements the carcinotron acted as LO-source.
The noise temperature of the mixer was measured by using the well known "Y-
factor" method. The measurements were corrected for the loss of the beamsplitter
(15 Mm mylar, 95% transmission).
Results
The DSB noise temperature of the receiver is shown in Fig. 2 for two different
junction arrays. Array Q34 (Rn= 100 n, A =0.8 /im 2 , Chebychev RF-filter) has noise
temperatures below 230 K over the whole 410-495 GHz range. The lowest value
measured was 160 K at 445 GHz. Array Q35 (Rn=22 n, 2 Mm 2 ,1/4 k RF-filter) has
noise temperatures between 260 K to 220 K in the frequency range 445-495 GHz.
The noise temperature increases sharply below 445 GHz, due to leakage of the RF-
filter.
Page 214
Third International Symposium on Space Terahertz Technology
SRON 410-490 GHZ 2-tuner mixer
Q35 Rn-22 Ohms, 1/4 Xfilter, 2gm 2
Q34 Rn-100 Ohms, Chebychev filter .8 urn 2
2-junction array
800
DSB Noise Temperature
□
700
junction
■■■□■■ Q35
600
A Q34
500
400
300
200
□.
" A- A.. A ..A.
a -aaa-
D
.A-
•□■•□ ■•□■ ^ .A-A--&
A . A a A
AA"
100
n .
I I I I
i.
i i i i
400 410 420 430 440 450 460 470 480 490 500
Frequency (GHz)
State University Groningen
SRON Groningen, The Netherlands
Fig. 2
Third International Symposium on Space Terahertz Technology
Page 215
1200
SRON 490 GHZ 2-tuner mixer
Tuned and instantaneous bandwidth
Q34 Rn-100 Ohms ,Chebychev-filter
DSB Noise Temperature
400 410 420 430 440 450 460 470 480 490 500
Frequency (GHz)
State University Groningen
SRON Groningen, The Netherlands
Fig. 3
Page 216
Third International Symposium on Space Terahertz Technology
The instantaneous bandwidth of junction Q34 is shown in Fig. 3. Here the frequency
was changed without adjusting the tuners, only the pump power was adjusted for
optimum H/C response.
The unpumped and pumped IV-curves of the two arrays at two different frequencies
are shown in Fig. 4. Array Q34 has a low leakage current, a gap voltage of 17 mV
(1 junction) and a well defined photon step above the gap. The gap voltage of array
Q35 is 2.4 mV and it is clearly observed that if the LO-power is radiated on the
junction the gap voltage decreases due to heating. The shape of the photon step
above the gap also indicates heating effects occur in the junction.
Pumped IV-curvea 410-490 GHz
junction Q34
DC current (uA)
100
Pumped IV-curves 430-490 GHz
junction Q35
DC current (uA)
50-
4 • • 10 a
Bias voltage (mV)
a 4 •
Bias voltage (mV)
Fig. 4 Pumped and unpumped IV-curves of the two junction arrays
Fig. 5 shows the IF power output at two LO-frequencies, when a hot or cold load
are placed in front of the junction. The smooth curves indicate that the Josephson
current is suppressed sufficiently. The structure (at 3 mV) in the IF-output seen at
495 GHz is due to the fact that the second photonstep from the negative bias
voltage range "creeps" into the first photonstep at the positive voltage range.
Third International Symposium on Space Terahertz Technology Page 217
Hot/Cold response 410 GHz Hot/Cold response 495 GHz
IF-output (uW)
2 4 • t 10 12
Bias voltage (mV)
IF-output (uW)
Biaa voltage (mV)
Fig. 5 IF-output with hot and cold signal
Analysis
In the analysis of the noise temperatures, the receiver is divided into three
elements: the RF-input, the mixer and the IF-output. A schematic diagram of the
whole receiver is given in Fig. 6. Each of these elements contributes to the total
receiever noise and gain.
1
Trf,Grf
mix
,g
mix
RF-input
Mixer
T/f,Gjf
IF-output
Fig. 6 Schematic representation of the noise and gain
contributions in the receiver
For the analysis of the contribution of the IF chain (T-bias, isolator, IF-amplifiers),
the shotnoise of an unpumped junction is used. With the known shotnoise of an
unpumped junction a Y-factor measurement on the IF-chain is performed. The total
power at the end of the IF-chain is given by (1). Here r is the reflection coefficient
between the 50 fi line and the junction. P^cP^ and P IF are the noise powers
Page 218
Third International Symposium on Space Terahertz Technology
coming from the junction, the isolator load and the IF-amplifier. G^G^, and G T _
bias are the gains from the various components. The noisepower from an unpumped
array of two junctions is given by (2), where B is the bandwidth, R dyn is the dynamic
resistance of the array and V is the voltage over the entire array.
PlF
Pout — {Pjunc^ + -Pjso/(1 — T) +
*- r isol*- r T— bia:
■)G[FGi so iGT-bias (1)
Pjunc = £</?> Rdyn = l^eB coth( : ^ ; )I(V)R dyn
8
8
Ak B T
(2)
IF-noise of unpumped junction
IF-output (uW)
4 a 8 10 12
Bias voltage (mV)
T-Bias
Isolator
amplifier
Schematic Diagram of junction and IF-
chain
Fig. 7 Measured and calculated
noisepower output of an unpumped
array of two junctions
Fig. 7 shows the experimental and fitted curves. The values for the gain and noise
contributions of the IF-chain are: G, F =88.4 ±0.1 dB, Tj F =4.8 ± 0.2 K.
The gain and noise contributions of the RF input (beamsplitter, HDP-window,
quartz-filter, lens, horn, waveguide and tuners) are difficult to estimate. Several
elements were measured seperately, but reflections at the horn waveguide transition
and losses in the waveguide and thf two tuners are difficult to find. The total gain
and noise at the RF-input are: Grf=0.77 ±0.1 and ^=56 ± 20 K.
Third International Symposium on Space Terahertz Technology
Page 219
The mixer noise and gain are now calculated with (3) and (4), where S? ut is
measured power difference at the IF-frequency and 5P in is the difference in input
power from the hot and the cold load. T^ is the measured total receiver noise
temperature.
SPout
{Jmix —
GRFGrF^Pi
(3)
n
Tmix = T rec GRF — TrfGrf —
Tip
(4)
mix
The gain and noise of the mixer are, just as the receiver noise temperature, nearly
constant over the 400-500 GHz band. Typical values for the contributions in the
receiver DSB gain and noise are: G mix = -12.5 dB ± 0.6 dB and T mix = 90 ± 30 K.
For a complete calculation of the mixer performance it is necessary to know the
embedding admittances at the LO and the upper and lower sideband frequencies.
These admittances are found by fitting a calculated pumped IV-curve to a measured
pumped IV-curve. An example of the quality of this fit at two different frequencies
is shown in Fig. 8.
Pumped IV-curve 410 GHz
G-0.67 B-0.42 (norm, to 1/100 Mho)
DC current (uA)
Pumped IV-curve 490 GHz
00.60 B-0.31 (norm, to 1/100 Mho)
DC current (uA)
too -
so
oatautatad
■aaauraa'
-
100 ■
•0
aaaaured
-
IS
Bias voltage (mV) Bias voltage (mV)
Fig. 8 Measured and calculated IV-curves. The embedding parameters
are shown in the header
Page 220
Third International Symposium on Space Terahertz Technology
It is observed that the quality of the fit is good, except for the discrepancy near the
gap, which is due to heating and difficult to model. The derived embedding
admittances indicate that the tuning elements are able to compensate the junction
capacitance. Unfortunately the sideband admittances are not yet calculated and a
full analysis cannot be performed at the moment. The result of a calculation of the.
mixergain, under the assumption that the sideband admittances are equal to the
LO-admitance, is shown in Fig. 10.
Calculated Gain 410 GHz
G-0.67 B-0.42 (norm, to 1/100 Mho)
Gain
0.12
= /l
■■— urad
0.1
■ n\
0.0t
■ \ \
0,08
■ 1 \
0.04
■ I \
0.02
■ I \
n
. J. \l^c\- . .
0,6 1 1,8 2
Bias voltage (mV)
2.8
Calculated gain 490 GHz
00.60 B-0.31 (norm, to 1/100 Mho)
Gain
0.08
0.0*
0.04 -
0,02
0,8 1 1,8 2
Bias voltage (mV)
2.8
Fig. 10 Measured and calculated mixer gain
In this figure the calculated and measured gain are normalized to each other. One
observes a big discrepancy between the measured and calculated gain, which
indicates that the LO and sideband frequencies differ significantly.
In both the calculated and the measured gain, some fine structure on the first
photonstep region is observed, indicating that the calculation method is working
properly, but the input parameters are wrong.
Summary
Measurements were performed in the 400-500 GHz range with a two tuner
waveguide mixer. The measured (receiver) noise temperatures are amongst the
lowest values measured at these frequencies. The results show that an array of two
Third International Symposium on Space Terahertz Technology Page 221
junctions is suitable in achieving a low noise receiver in the 400-500 GHz range. It
is also found that a qualitatively "bad" junction with a low gap-voltage can still serve
as a low noise mixer element. The preliminary comparisons between theory and
measurement show a good agreement between calculated and measured pumped
IV-curves. Gain calculations indicate that the measured noise temperatures are not
fully DSB measurements. Further analysis is needed to determine the USB and LSB
gain and noise contributions.
Acknowledgements:
This work was supported by ESA under contract No. 7898/88/NL/PB(SC), the
Stichting Technische Wetenschappen and the Stichting voor Fundamenteel
Onderzoek der Materie.
References
1 J.Zmuidzinas, H.G. LeDuc, "Quasi-Optical Slot Antenna SIS Mixer", Proceedings
of the Second International Symposium on THz Technology.
2 C.K. Walker, M.Chen, P.L Shafer, H.G. LeDuc, J.E. Carlstrom, T.G. Carlstrom,
T.G. Phillips, "A 492 GHz SIS Waveguid Receiver for Submillimeter Astronomy",
Int J. of IR and Millimeter Waves 1992.
3 J.R. Tucker, M.J. Feldman, "Quantum Detection at Millimeter Wavelengths" Rev.
Mod. Phys 57, 1055 (1985)
4 C.A. Mears, Qing Hu, P.L. Richards, A.H. Worsham, D.E. Prober, A.V. Raisanen,
"Quantum Limited Quasiparticles Mixers at 100 GHz", IEEE Trans. Magrt, vol 27,
2, 1991
5 C.E. Honingh, G. de Lange, M.M.T.M Dierichs, H.H.A. Schaeffer, J. Wezelman,
J. v.d. Kuur, Th. de Graauw, T.M. Klapwijk, "Comparison of Measured and
Predicted Performance of a SIS Waveguide Mixer at 345 GHz",. these proceedings
6 C.E. Honingh, unpublished results
7 M.M.T.M. Dierichs, unpublished results
Page 222 Third International Symposium on Space Terahertz Technology
^3~£3 N9 3 -2? 744
/£0£)32^~ Double Dipole Antenna SIS Receivers at 100 and 400 GHz
\*
A. Skalare* »', H. van de Stadt", Th. de Graauw", R. A. Panhuyzen"**,
M. M. T. M. Dierichs'"
* Dept. of Applied Electron Physics, Rannvagen 6,
Chalmers University of Technology, G5teborg, Sweden.
** Space Research Organization of the Netherlands (SRON),
Landleven 12, 9747 AD Groningen, the Netherlands.
* " " Dept. of Applied Physics and Materials Science Center, University of Groningen,
Nijenborgh 4, 9747 AG Groningen, the Netherlands
Abstract
Antenna patterns were measured between 95 and 120 GHz for a double dipole antenna
/ ellipsoidal lens combination. The structure produces a non-astigmatic beam with low
side lobe levels over that whole band. A heterodyne SIS receiver based on this concept
gave a best noise temperature of 145K DSB at 98 GHz. Measurements were also made
with a 400 GHz heterodyne SIS receiver, using a double dipole antenna in conjunction
with a hyperhemispherical lens. The best noise temperature was 220 K DSB at 402 GHz.
On-chip stubs were used to tune out the SIS junction capacitance.
Introduction
We here describe two SIS heterodyne receivers, both using double dipole antennas
[1,2,3,4] placed on the back plane of a thick dielectric lens. In both cases, the antenna
consisted of two half wave dipoles, connected by a stripline with the SIS mixer at the mid
point, Fig.l . The quartz chip with the antenna was mounted on the back plane of a thick
quartz. lens, and was backed by a quarter wave thick quartz slab and a reflector, Fig.2 .
Some early low frequency scale model measurements of this structure can be found in
[4].
Both receivers were designed in similar but not identical ways, and will be described
separately.
100 GHz Receiver Design
The size of the dipole antenna was chosen to give a center frequency of 100 GHz. The
11mm diameter quartz lens was polished to an ellipsoidal shape, designed to produce a
Third International Symposium on Space Terahertz Technology Page 223
diffraction limited main lobe. The polishing tool was a brass rod with an ellipsoidal hole
at one end, machined with a numerically controlled milling machine.
As shown in Fig.3 the mixer fixture pinched the lens between a copper back plate and
two flanges, which fitted into two grooves in the quartz. With the help of a small amount
of vacuum grease, this provided excellent 4K cooling of the lens and the mixer chip.
The mixer itself was a series array of two Nb-Al/AlOx-Nb SIS junctions with a normal
state resistance of 34 ohms. The size of each junction was 2 square microns. No attempts
were made to tune out the parasitic capacitance of the junctions.
The mixer chip was contacted by a flexible strip transmission line, Fig.4 , which was
soldered to a 85 mil output co-ax. The strip was cut to high accuracy from a
Kapton/copper laminated sheet, to give a characteristic impedance close to 50 ohms. As
shown in Fig.5 , the strip line would work well even for intermediate frequencies of up
to 8 GHz, where resonances in the lens fixture begin to appear.
100 GHz Measurements
The antenna pattern of the dipole/lens fixture was measured at room temperature with
a bismuth bolometer in place of the SIS junctions, Fig.6 . Both the E- and H-plane beam
profiles were of high quality over the whole 95-120 GHz band, in good agreement with
earlier scale model measurements [4].
Y-factor measurements with the SIS mixer in a 4.2K helium cryostat gave a best receiver
noise temperature of 145K DSB at 98 GHz, Fig.7 . The intermediate frequency amplifier
was a 1.5 GHz cooled Berkshire HEMT, with a noise temperature of 4K (from the
manufacturers data sheets). We believe that the noise performance can be improved
considerably by the use of on-chip tuning structures, but have not yet implemented this
in the 100 GHz receiver.
400 GHz Receiver Design
The 400 GHz receiver differed from the 100 GHz one in a few ways, namely :
1. We chose to use a hyperhemisperical lens, mainly because it can be manufactured
to higher tolerances than an ellipsoidal one.
2. The antenna was a scaled down version of the 100 GHz structure. The center
frequency was chosen to 310 GHz, which puts the upper frequency limit of the antenna
itself somewhere around 430 GHz (one octave bandwidth).
3. The SIS parasitic capacitances were tuned by on-chip stubs, as can be seen in
Figs.8 and 9 . A series array of two junctions, each of 3 square microns, was used.
4. We used two different IF amplifier chains, one with a 1.5 GHz FET with
approximately UK noise temperature, the other with a 4 GHz Berkshire Technology
Page 224 Third International Symposium on Space Terahertz Technology
HEMT with 4K noise temperature.
5. The Josephson effect was suppressed by a superconducting coil.
400 GHz Measurements
The video response of the receiver over the range 100-600 GHz was studied with a
Fourier transform spectrometer, Fig. 10 . Two peaks are visible in the diagram, one just
above 200 GHz and one close to 400 GHz. The approximate positions of the peaks could
be predicted from a simple circuit model, where each junction was represented by a
capacitance of 165 fF in parallel with its normal state resistance.
Two initial Y-factor measurements with the same mixer chip are shown in Fig. 11 . The
best performance was at the lower end of the available local oscillator band, with a
lowest noise temperature of 220K DSB.
Summary
The dipole / ellipsoidal lens configuration was investigated in terms both of antenna
pattern and of matching to SIS junctions. The pattern measurements showed low side
lobe levels, and a non-astigmatic beam over the whole band 95-120 GHz. The best
receiver noise temperature was 145K DSB at 98 GHz, a value we believe will be
improved with the use of integrated tuning structures.
The initial measurements with the other receiver, in which a double dipole antenna is
combined with a hyperhemispherical lens, yielded a best noise temperature of 220K
DSB at 402 GHz.
The Kapton laminate strip should function well up to 8 GHz as an intermediate
frequency connection to the mixer chips.
Acknowledgements
The authors extend their gratitude to Prof Dr Ir T. M. Klapwijk for his advice and for
useful discussions, to Mr. H. Schaeffer for his technical support, and to Mr. G. de Lange
and Ms. C. E. Honingh for their assistance in the noise temperature measurements. The
work presented here was supported financially by the European Space Agency (ESA),
through contract 7898/88/NL/PB(SC).
References
[1] P. T. Parrish, T. C. L. G. Sollner, R. H. Mathews, H. R. Fetterman, C. D. Parker,
P. E. Tannenwald, A. G. Cardiasmenos, "Printed Dipole-Schottky Diode Millimeter
Wave Antenna Array", SPIE Millimeter Wave Technology, Vol. 337, 1982, pp.49-52
Third International Symposium on Space Terahertz Technology Page 225
[2] W. Chew, H. R. Fetterman, "Printed Circuit Antennas with Integrated FET
Detectors for mm- Wave Quasi-Optics", IEEE Trans. Microwave Theory Tech., Vol.
MTT-37, No. 3, 1989.
[3] J. A. Taylor, T. C. L. G. Sollner, D. D. Parker, J. A. Calviello, "Planar Dipole-fed
Mixer Arrays for Imaging at Millimeter and Sub-Millimeter Wavelenghts", Proc. of the
1985 Int. Conf. on IR and mm-Waves, 1985, pp.197-188.
[4] A. Skalare, Th. de Graauw, H. van de Stadt, "A Planar Dipole Array Antenna
with an Elliptical Lens", Microwave and Optical Tech. Letters, Vol.4, No.l, Jan. 1991.
Page 226
Third International Symposium on Space Terahertz Technology
.4
^
.25
Fig.l : The geometry of the 100 GHz double dipole antenna. The dimensions are in units
of wavelength at the design frequency.
Fig. 2 : The antenna chip is placed between a quartz lens and a quarter wavelength thick
quartz slab with a reflector.
Third International Symposium on Space Terahertz Technology
Page 227
Front Plate
(Copper)
Quartz
Lens
PTFE
Screw
Back Plate
(Copper)
Flanges
Cooling
Strap
Side View
Back View
Fig.3 : The fixture that holds the lens and the mixer chip. The diameter of the lens in
this figure is 11mm.
85 mil co-ax
(50 Q) Solder
Strip Line
Contacting Tabs
(Copper foil)
Mixer Chip
Fig.4 : The flexible Kapton laminate strip line used for the DC & IF connections. The
tabs at the end of the strip are glued to the contact pads on th< chip with silver paint.
I
00
CM 1 R u
1 U FS H 06.737 O
riAnCH 5 1992 SHORT.
-1,8004 O 20,997 pF
CHIP RES.
Cor
Del
Gnt
Hid
START .130 OOO 000 GHz STOP 20.000 000 000 GHz
Fig.5 : The room temperature Sll reflection on the IF line in Figs. 3 & 4 with the mixer
chip replaced with a chip resistor (51 ohms). A time gate was applied around the co-ax
to strip transition, the strip itself and the resistor. Marker 1 is at 4 GHz and Marker 2
is at 8 GHz.
a-
■v
3
a
to
3
o
</>
£"
3
o
a
CO
-a
&
r>
3
a-
a
5"
88
3-
-«°
a.
E
E
E
E
E
E
95 GHz
100 GHz
105 GHz
110 GHz
115 GHz
120 GHz
H
H
H
H
H
H
3
e
3
o
3
tj
&
r>
J?
&
3-
N
«?
3-
3
O
Fig.6 : Antenna patterns of the double dipole / ellipsoidal lens combination. The
patterns were measured with a Bismuth bolometer, and the radial scale is 5 dB per
division.
5
K>
to
400
300
I
o
Q)
(0
Z
m
CO
Q
200
100
-I — I — I — I — I — I — I — I — I I I I I I ■ I ■
J I l_
95
100
105
1 10
1 15
120
125
3
3
sa
S"
3
a
•—1
en
3
©
«■•**
C
3
o
3
C/>
a
ro
re
Frequency (GHz)
Fig.7 : The double sideband noise temperature of the 100 GHz receiver. The data was
corrected for the transmission of the LO injection beam splitter (95%) .
3
o
3
O
£
Third International Symposium on Space Terahertz Technology
Page 231
Fig.8 : The 400 GHz mixer chip.
Fig.9 : The 400 GHz antenna chip, detail.
CARC.1
CARC. 2
2 s
100
200
300
400
500
600 GHz
Fig. 10 : The Fourier transform spectrogram (arbitrary units on the vertical scale).
"CARC2" marks the frequency range where noise temperature measurements have been
made.
3
3
S'
in
3
o
3
.CO
5
r>
re
3
3
O
5*
J5
a*
1500
¥ 1000
Q)
CA
z
m
0)
Q 500
4.3 GHz IF, T IF =4K
re
a
&
S°
3
tn
Q
(n
5'
3
o
a
</>
&
Cl
re
H
3
3-
3
D
o L —
400
420
440
460
480
500
Frequency (GHz)
Fig.ll : DSB Noise measurement with the 400 GHz receiver. The data was corrected for
the transmission of the LO injection beam splitter (65%) .
S 3
00
re
/
Page 234 Third International Symposium on Space Terahertz Technology
S/6)—£ 2--" Slot Antenna SIS Mixers for Submillimeter Wavelengths
/(p 533 Jonas Zmuidzinas*, H. G. LeDuc**, and J. A. Stern^j Q 3 * ^ ^ ^
\ ^
s^' We are developing improved versions of a slot antenna SIS mixer which we have
previously described [1]. The initial work demonstrated a double sideband noise
temperature of 420 K for a 500 GHz quasi-optical SIS mixer employing a twin-slot antenna
on a quartz dielectric substrate. A quartz hyperhemispherical lens is used to focus the
incoming radiation onto the twin-slot antenna. The advantages of a twin-slot antenna
include a low impedance (35 CI) and a clean, symmetric beam pattern into the dielectric with
a 70% efficiency. In our original mixer, the radiation was coupled from the two slot
antennas onto superconducting microstrip lines which fed the SIS junction. By performing
an impedance transformation using tapered lines and by feeding the radiation from the two
slots to the junction in parallel, the effective (real) impedance seen by the junction was
reduced to just 4 Q. This very low impedance allowed a junction area of 2.3 (im 2 to be
used at 500 GHz, which was manufactured using optical lithography. However, no
attempt was made to tune out the junction capacitance. We estimate that this capacitance
reduces the impedance coupling efficiency to T|z = 0.23, for our junction with cdRn C = 5.3
at 500 GHz.
The recent development [2] of techniques using electron-beam lithography to
manufacture junctions with very small areas (= 0.1 pirn 2 ) now allows considerably more
flexibility in the design of SIS mixer circuits. We have redesigned the slot-antenna mixer
to take advantage of this possibility. In particular, we have included a novel circuit which
allows the junction capacitance to be tune out over a broad bandwidth. For instance,
mixers designed for 800 GHz using NbN/MgO/NbN junctions with realistic parameters
achieve a 3 dB impedance bandwidth of nearly 400 GHz. Furthermore, our circuit uses
only short lengths of microstrip and should therefore be less sensitive to RF losses than
other designs. The improved impedance match should give a large reduction in noise
temperature as compared to our previous mixer. The new devices are currently under
fabrication. Further details of the design and any available experimental results will be
discussed.
*Jonas Zmuidzinas is with the Downs Laboratory of Physics, California Institute of
Technology.
**H. G. LeDuc and J. A. Stern are with the Jet Propulsion Laboratory, California
Institute of Technology.
Wj. Zmuidzinas and H. G. LeDuc, "Quasi-optical Slot Antenna SIS Mixers," IEEE
Trans. Microwave Theory Tech., in press, 1992.
[2lH. G. Leduc, A. Judas, S. R. Cypher, B. Bumble, B. D. Hunt, and J. A. Stern,
"Submicron Area NbN/MgO/NbN Tunnel Junctions for SIS mixer Applications," IEEE
Trans. Magn., MAG-27, 3192, 1991.
Third International Symposium on Space Terahertz Technology Page 235
A PLANAR QUASI-OPTICAL SIS RECEIVER ~
p 7
FOR ARRAY APPLICATIONS ' ~
Philip A. Stimson, Robert J. Dengler, Peter H. Siegel and Henry G. LeDuc
Jet Propulsion Laboratory, Pasadena, CA, 91109
Abstract — A novel planar, quasi-optical SIS receiver operating at 230 GHz is
described. The receiver consists of a 2x5 array of half wave dipole antennas
with ten niobium-aluminum oxide— niobium SIS junctions on a quartz dielec-
tric-filled parabola. The 1.4 GHz intermediate frequency is coupled from the
mixer via coplanar strip transmission lines and 4:1 balun transformers. The
receiver is operated at 4.2 K in a liquid helium immersion cryostat. We report
here accurate measurements of the performance of single receiver elements.
A mixer noise temperature of 89 K DSB, receiver noise temperature of 156 K
DSB and conversion loss of 8 dB into a matched load have been obtained.
INTRODUCTION
The quasiparticle superconductor-insulator-superconductor (SIS) mixer is the most
sensitive detector in the millimeter-wave region and forms the basis of most high quality
receivers for millimeter- wave astronomy [l]. The quantum limit for noise temperature (in
a SSB mixer) has essentially been reached at 100 GHz [2] [6], but at higher frequencies the
available performance is poorer, with 10 times the quantum limit being a more realistic
goal. This figure has recently been reported from the best waveguide mixers around
200 GHz [3] [4] [5]. The major cause of the performance reduction at high frequency is
the SIS junction capacitance, which presents a smaller parallel reactance and shunts
the quasiparticle response. Tuning structures can, in principle, alleviate this limitation
but are not yet well understood at higher frequencies [8]. The approach most often
used, and that used here, is to fabricate high current density junctions with exceedingly
small areas (< 1 /im 2 ) to reduce the capacitance. Another serious problem is control of
Josephson currents in the junction. Noise temperatures obtained with broadband hot
and cold loads may be highly inaccurate in the presence of Josephson currents. These
effects become more important as the frequency and/or bandwidth is increased. Other
Page 236
Third International Symposium on Space Terahertz Technology
problems include losses in conductors and dielectrics, the fabrication difficulties of small
waveguide components and difficulties in obtaining convenient local oscillators.
Quasi-optical receivers with planar circuit mixers are an attractive approach for sys-
tems at frequencies in the neighborhood of 1 THz [7] [8] [9]. They suffer the disadvantage
of being fixed tuned but provide the advantage of convenient monolithic fabrication.
Planar configurations are also a desirable approach to realizing array receivers.
In this paper, we report accurate measurements on a quasi-optical array-type receiver
at 230 GHz. We have been able to suppress Josephson currents almost completely, and our
intermediate frequency versus bias voltage curve exhibits the smooth oscillatory behavior
of the best waveguide mixers [3]. Our configuration is designed to allow an array of mixers
to be measured during one cool down cycle. We report here the performance of a single
array element. We will report on complete array performance in a separate paper. The
SIS junctions used for these experiments were nominally identical to those used in recent
waveguide receivers [3] [4], with which our results may be compared.
SIS JUNCTION FABRICATION
The junction wafer used for this receiver carries a 2x5 array of resonant dipole an-
tennas with 0.4x0.4 //m niobium-aluminum oxide-niobium SIS junctions at the termi-
nals. The junctions were fabricated using a self aligned lift-off trilayer process. The
niobium-aluminum oxide-niobium trilayer was sputtered onto the 0.25 mm thick, 17 mm
diameter quartz substrate through a photoresist stencil. The trilayer remaining af-
Fig. 1. The mixer block with the upper half removed. The central dielectric-filled parabola (dark),
containing 10 antenna and mixer elements, is surrounded by 10 IF baluns (light) and SSMA connectors
at the edge of the block.
Third International Symposium on Space Terahertz Technology
Page 237
ter lift-off formed half of each antenna and the ten coplanar strip transmission lines
used for the. IF. The junction mesa was patterned using electron beam lithography on
1200 A thick PMMA over a 4000 A thick polyimide layer, followed by evaporation of
500 A of chromium metal and lift-off. The chromium stencil was transferred to the poly-
imide underlayer by reactive ion etching in an oxygen plasma. The contact regions of the
trilayer were then protected with a resist stencil and the chromium/polyimide mask was
used to etch the junction. Thermal SiO was deposited using the same stencil to provide
electrical isolation of the base electrode and to provide dielectric for two RF blocking
capacitors located one quarter and three quarter wavelengths away from the junction
down the coplanar strips. The polyimide was then removed with dichloromethane. The
second half of the antennas was made by deposition of niobium and reactive ion etching.
RECEIVER DESIGN
The mixer block, shown in Figure 1, consists of the junction/antenna wafer, a quartz
reflector, and IF baluns and connectors mounted in a brass housing. The wafer is held
on the flat face of a quartz parabolic lens, whose rear surface is metalized. Incoming
radiation is reflected by the metal surface and focussed onto the antenna elements at
the center of the wafer. The configuration, called a Dielectric-Filled Parabola (DFP), is
analogous to a conventional parabolic dish antenna. The IF signals are coupled from the
wafer via coplanar strip transmission lines. Monolithic IF baluns transform the 200 $7
characteristic impedance of the coplanar strips to that of 50 fl coaxial transmission line.
Details of this design, including extensive low frequency modeling, are described by Siegel
et al. [10] [16]. A superconducting magnetic field coil is mounted on the block to suppress
Josephson currents in the junctions.
chopper -i
mixer block
12 IF cobles
bias
to flange
20 dB coupler isolator
5Z
to flange
atten
IF amplifier
38dB
©
Voltage tuned
oscillator
variable temp
IF load
Fig. 2. Schematic diagram of the IF system of the array receiver. The entire system is immersed in
liquid helium except for the 77 K load which is bolted to the liquid nitrogen shield of the cryostat.
Page 238
Third International Symposium on Space Terahertz Technology
The IF system shown in Figure 2, consists of ten IF cables routed through two 6-
position coaxial switches and one 2-position switch to a single amplifier chain. The
remaining two switch positions are used to connect a short and a variable temperature
IF load to the amplifier input. The load consists of a resistor terminating a stainless
steel coax cable on a thermally isolated plate which contains a heater resistor and diode
thermometer. The structure is enclosed in an indium sealed can. This permits accurate
calibration of the IF system and very accurate mixer measurements [ll]. An isolator
is used to reduce the SWR at the amplifier input and a directional coupler with cooled
attenuators allows signals to be injected into the IF system to measure the mixer reflection
coefficient. After removal from the cryostat the IF signal is further amplified and passed
through a variable center frequency 50 MHz wide filter and fed to a power detector. The
IF system noise temperature is approximately 7K at 1.4 GHz.
The optical system consists of a chopper mounted directly in front of the mixer, and
the hot and cold loads. When the chopper blade is closed the input beam is directed
onto a 4 K (cold) load mounted on the receiver plate; when it is open the beam passes
through a quartz window to a 77 K (hot) load mounted on the liquid nitrogen shield of the
cryostat. The loads are pyramidal absorbers manufactured from Eccosorb CR-110, which
is known to provide high absorption and low reflection at this frequency. Reflection from
a flat plate of CR-110 has been measured at less than -10 dB in this frequency range [12].
The window is exactly five wavelengths thick and passes almost all the incident 230 GHz
radiation. The theoretical transmittance is 0.999; we measured a transmittance of over
0.95. Local oscillator radiation is produced by a Gunn diode and Schottky diode doubler
and is injected through the back of the mixer block. No diplexer is required.
The entire receiver is immersed in liquid helium which eliminates heat sinking prob-
lems. The dielectric constant of the helium is 1.048 [15]. The switches, thermometers,
a -o.oo
0)
3
U
' ' ' ' I I I— I I— I I — I — L-
Pumped
Unpumped
-2024
Voltage (mV)
Fig. 3. Pumped and unpumped IV curves for a typical Nb-A10 r -Nb SIS junction used in the planar
receiver.
Third International Symposium on Space Terahertz Technology
Page 239
liquid level meter.. .etc., and all data aquisition is controlled by a computer.
MEASUREMENT TECHNIQUE
We use a variation of the technique of McGrath et al. [ll], to obtain mixer gain and
noise temperature. First, the IF system is calibrated by plotting the temperature of
the IF load as a function of the IF output power. This measures the IF system noise
temperature Tip. The receiver noise temperature Tr is measured using the hot and cold
loads (Th and Tc), the ratio of the IF output powers Y = Pifh/Pifc and Equation 1.
Tr =
Th-YTc
Y-l
T T (Txp + Tsr 2 )
Lm V
(1-P)
T ifh -TifcA / l ~7 2
Th-Tc )\l-V\
(1)
(2)
(3)
Next, the temperatures of the IF load, Tifh and TjfC) which produce output powers
Pifh and Pifc are calculated from the calibration, and the effective bath temperature
Ts determined by measuring the power output from the IF system with a shorted input.
The IF reflection coefficient of the mixer T (and of the load 7) is measured by injecting
a signal from a voltage tuned oscillator through the coupler and recording the difference
in reflection between the mixer and the short. The loss into a matched load and noise
temperature are then calculated from Equations 2 and 3.
-2 2
Voltage (mV)
Fig. 4. IF output power as a function of bias voltage for hot and cold load inputs. The curve exhibits a
smooth oscillatory behavior similar to that expected from theory with no sharp spikes or discontinuities
indicating excellent control of Josephson currents.
Page 240
Third International Symposium on Space Terahertz Technology
RESULTS
Typical pumped and unpumped IV characteristics are shown in Figure 3, and IF
output power as a function of bias voltage for hot and cold load inputs is shown in
Figure 4. A superconducting magnet was used to suppress Josephson currents. The curve
exhibits a smooth oscillatory behavior similar to that expected from theory [13] [14] with
no sharp spikes or discontinuities. The IF output power is expected to decline towards
zero bias; the fact that there is some power output at zero bias indicates some remaining
Josephson currents which were not fully suppressed. These remain visible on the IF curve
even though the IV curve appears smooth. Nevertheless, we believe that this is the best
IF behaviour reported from a planar quasi-optical SIS receiver.
The most recent experiments performed with this receiver used junctions with an
area of 0.2 /an 2 . The normal state resistance was 56 fi, the critical current density was
15kAcm -2 and the ujRC product was approximately 1.3. The mixer and receiver noise
temperatures and mixer conversion loss are plotted as a function of IF frequency in Fig-
ure 5. The LO frequency was 230 GHz. The best results are obtained at 1.35 GHz where
a Tm of 89 K DSB, a Tr of 156 K DSB and conversion losses of 8dB (into a matched
load) were measured. The IF mismatch is approximately 1 dB across the IF band. Esti-
mated uncertainties in the noise temperatures are ±5K, and in the loss, ±0.5 dB. These
values neglect any uncertainty due to RF load reflections or beam spillover. The largest
Y-factor was obtained on the first quasiparticle step below the energy gap, at a bias
voltage of approximately 2.3 mV. An inferior Y-factor was noted on the second step. The
500
oj 400
u
cd
U 300
J 1 * 200
en
"o 100
1.1
1.2
1.3
1.4
1.5
Frequency (GHz)
1.6
1.7
Fig. 5. Mixer and receiver noise temperatures and mixer loss as a function of IF frequency. The best
results are obtained at 1.35 GHz where a Tm of 89 K DSB, a Tr of 156 K DSB and conversion losses of
8 dB were measured.
Third International Symposium on Space Terahertz Technology Page 241
mixer noise temperature and conversion loss are seen to be essentially constant across
the IF band. Mixer noise temperature is referred to the optically coupled loads at the
system input and includes the effects of all components through to the IF connectors at
the output of the balun transformers. The receiver noise temperature follows the noise
behavior of the IF amplifier.
At each data point on the curves, the change in IF reflection coefficient, and the
change in bias point, caused by switching between the hot and cold loads was measured.
This is necessary to ensure that the observed Y-factor is not produced by different LO
pumping conditions, or change in bias point when observing the hot and cold loads.
Different pumping would be expected to change the junction output impedance and the
IV curve shape. The reflection coefficient change was verified to be less than 1 %, and
the change in bias voltage less than 0.02 mV. This indicates that the observed Y-factor
has no appreciable component due to these factors.
Recent results from waveguide mixers at similar frequencies using junctions with
similar specifications from the same fabrication process [3] [4] give mixer temperatures of
48 K DSB and 60 K SSB and conversion losses of 2dB. Our noise temperature results,
although a factor of two higher, are consistent with these values given the lack of tuning
capability inherent in our planar circuit.
CONCLUSION
We have demonstrated a planar quasi-optical SIS mixer and low noise receiver which
is suitable for array applications. Best performance of an individual element at 230 GHz
was a mixer noise temperature of 89 K DSB, a receiver temperature of 156 K DSB and a
conversion loss of 8 dB. The IF output shows a smooth variation with bias, indicating good
control of Josephson currents. The noise results are consistent with recent measurements
using similar junctions in waveguide receivers, and are only a factor of two higher. The
conversion loss is rather large, but consistent with other planar mixer values. We will
report on array performance in a future publication.
ACKNOWLEDGEMENT
We are extremely grateful for the constant assistance and encouragement of Dr. W.R. McGrath,
without whom this work could not have been completed. We also thank Dr. H.H.S. Javadi and Dr.
M.A. Frerking of JPL, and Dr. A.R. Kerr and Dr. S.K. Pan of NRAO for useful advice and discussions.
We acknowledge the support of Mr. B. Bumble, Dr. J. Stern and Mr. S.R. Cypher on junction
fabrication, Mr. H. Moham for fabricating the array mount, and Mr. R. McMillan for fabricating the
quartz parabola. This work was carried out at the Jet Propulsion Laboratory, California Institute of
Technology under contract with the National Aeronautics and Space Administration.
Page 242 Third International Symposium on Space Terahertz Technology
REFERENCES
[I] P.L. Richards and Q. Hu, Proceedings of the IEEE, vol. 77, 8, pp. 1233-1245 (1989).
[2] C.A. Mears, Q. Hu, P.L Richards, A.H. Worsham, D.E. Prober and A.V Raisanen, IEEE Transactions
on Magnetics, vol. 27, 2, pp. 3363-3369 (1991).
[3] W.R. McGrath, H.H.S. Javadi, S.R. Cypher, B. Bumble, B.D. Hunt and H.G. LeDuc, Second Inter-
national Symposium on Space Terahertz Technology, Pasadena, CA, Feb. 26-28 (1991), pp. 423-428.
[4] J.W. Kooi, M. Chan, T.G. Phillips, B. Bumble and H.G. LeDuc, Second International Symposium
on Space Terahertz Technology, Pasadena, CA, Feb. 26-28 (1991), pp. 459-472.
[5] A.W. Lichtenberger, D.M. Lea, A.C. Hicks, J.D. Prince, R. Densing, D. Petersen and B.S. Deaver,
Second International Symposium on Space Terahertz Technology, Pasadena, CA, Feb. 26-28 (1991), pp.
439-458.
[6] S.K. Pan, A.R. Kerr, M.J; Feldman, A.W. Kleinsasser, J.W. Stasiak, R.L. Sandstrom and W.J.
Gallagher, IEEE Transactions on Microwave Theory and Techniques, vol. 37, 3, pp. 580-592, (1989).
[7] J. Zmuidzinas and H.G. LeDuc, Second International Symposium on Space Terahertz Technology,
Pasadena, CA, Feb. 26-28 (1991), pp. 481-490.
[8] Q. Hu, C.A. Mears, P.L. Richards and F.L. Lloyd, IEEE Transactions on Magnetics, vol. 25, 2, pp.
1380-1383, (1989).
[9] T.H. Biittgenbach, R.E. Miller, M.J. Wengler, D.M. Watson and T.G. Philips, IEEE Transactions on
Microwave Theory and Techniques, vol. 36, 12, pp. 1720-1725 (1988).
[10] P.H. Siegel and R.J. Dengler, IEEE Transactions on Antennas and Propagation, vol. 39, 1, pp.
40-47 (1991).
[II] W.R. McGrath, A.V. Raisanen and P.L. Richards, International Journal of Infrared and Millimeter
Waves, vol. 7, 4, pp. 543-553 (1986).
[12] J.B. Peterson and P.L. Richards, International Journal of Infrared and Millimeter Waves, vol. 5, p.
1507, (1984).
[13] J.R. Tucker and M.J. Feldman, Rev. Modern Physics, vol. 57, pp. 1055-1113 (1985).
[14] J.R. Tucker, IEEE Journal of Quantum Electronics, vol. 15, 1234-1258 (1979).
[15] "Handbook of Chemistry and Physics", CRC Press, 56'th ed. (1976), p. E-55.
[16] P.H. Siegel, First International Symposium on Space Terahertz Technology, Ann Arbor, MI, Mar.
5-6, (1990) pp. 218-227.
Third International Symposium on Space Terahertz Technology Page 243
N93-27747
A Fixed Tuned Broadband Matching Structure for ~>£^/'~3^b
Submillimeter SIS Receivers
*
Thomas H. BiittgenbadT, Henry G. LeDuc**. /fuS, ON J
Paul D. Maker**, and T. G. Phillips* / £ £3£T
A /
We have designed, fabricated and tested a quasi optical spiral antenna mixer with a
Nb/A10x/Nb tunnel junction. This design incorporates a hybrid antenna fed by a planar
logarithmic spiral antenna to couple to the radiation field, as previously done with Pb based
devices, as well as a newly designed matching circuit. This matching circuit is a relatively
complex structure requiring several layers of photolithographic processing on top of the
actual tunneling device. Computer modeling of the device predicted the measured
bandwidth to within 8%, making scale model measurements unnecessary. We have
obtained a good match from 210 GHz to 460 GHz between the antenna and a relatively
large area (1.25 by 1.25 Jim^) tunnel junction with co Rn C ~ 2 - 4.4. This compares to
simple inductive stubs that attain only a few percent of total bandwidth in the submillimeter
band or inductively tuned SIS arrays with an upper limit of operating frequencies well
below the submillimeter band. Noise temperatures were measured at 345 GHz, 426 GHz
and 492 GHz yielding double sideband noise temperatures at 200 K, 220 K and 500 K,
respectively.
Thomas H. Biittgenbach and T. G. Phillips are with the Division of Physics,
Mathematics and Astronomy, California Institute of Technology.
Henry G. LeDuc and Paul D. Maker are with the Jet Propulsion Laboratory,
California Institute of Technology.
Page 244 Third International Symposium on Space Terahertz Technology
N93 . 27748
/6o£3<o
[A
Modelling and Performance of Nb SIS Mixers
in the 1.3mm and 0.8mm Bands
Karpov, A., Carter, M., Lazareff, B., Billon-Pierron, D.,
Gundlach, K.H.
Institut de Radioastronomie Millimetrique (IRAM)
300, Rue de la Piscine
38406 ST MARTIN D'HERES Cedex (FRANCE)
Abstract
We describe the modelling and subsequent improvements of SIS waveguide
mixers for the 200-270 and 330-370 GHz bands (Blundell, Carter, and Gund-
lach 1988, Carter et ad 1991). These mixers are constructed for use in re-
ceivers on IRAM radiotelescopes on Pico Veleta (Spain, Sierra Nevada) and
Plateau de Bure (French Alps), and must meet specific requirements.
The standard reduced height waveguide structure with suspended stripline is
first analyzed and a model is validated through comparison with scale model
and working scale measurements. In the first step, the intrinsic limitations
of the standard mixer structure are identified, and the parameters are op-
timized bearing in mind the radioastronomical applications. In the second
step, inductive tuning of the junctions is introduced and optimized for min-
imum noise and maximum bandwidth. In the 1.3mm band, a DSB receiver
temperature of less than 110K (minimum 80K) is measured from 180 through
260 GHz. In the 0.8mm band, a DSB receiver temperature of less than 250K
(minimum 175K) is obtained between 325 and 355 GHz. All these results
are obtained with room-temperature optics and a 4 GHz IF chain having a
500 MHz bandwidth and a noise temperature of 14K.
Design goals
A receiver for radioastronomical use should meet specific design goals besides
low noise at a particular frequency such as : reliability, ease of tuning, wide
tuning range, capability for SSB tuning (increasingly important when the
Third International Symposium on Space Terahertz Technology p age 245
atmospheric radiation is a significant contribution to the system noise), good
coupling to the antenna, wide IF bandwidth (especially for extragalactic
work):
Junctions
The mixers described in the this report use Nb/A^C^/Nb junctions fabri-
cated in the IRAM facility (Lehnert et al 1991) Two junction series arrays
are fabricated with an integrated IF filter. Trilayers are deposited into resist
stencils, followed by lift-off. This technique is thought to reduce the mechan-
ical sress in the trilayer (Yuda, Kuroda, and Nakano 1987). The substrate
is fused quartz 100/im thick. The base electrode is 130nm, the counter elec-
trode 30nm, and the wiring layer 240nm thick. The junctions are isolated by
anodisation, up to 10V, and a sputtered Si02 layer 200nm thick.
Individual junction areas used in this work are about 2/im 2 , but progress in
photolithography should allow use of smaller areas. The results presented
here are obtained with 2-junction arrays having a total normal-state resis-
tance Rn near 50ft. Some 4-junction arrays were also fabricated.
The normal resistance of the junctions can be adjusted after fabrication by
controlled thermal annealing (Lehnert et al 1992).
Mixers currently in use on IRAM telescopes
A simple equivalent circuit was used for the suspended-stripline, reduced-
height waveguide mixer mount (Karpov et ad 1992). The values of the circuit
elements were derived from electromagnetic theory. They were then vali-
dated by measuring the embedding impedance of the pure SIS junction as
measured via a coaxial probe on a scale model, and comparing the results
with model predictions. Figure 1 shows the good agreement between mod-
elled and measured values for the return loss.
Figure 2 shows the measured DSB receiver noise. The degradation of re-
ceiver noise at the band edges and in the vicinity of 220 GHz was caused by
degradations of coupling efficiency at corresponding frequencies, which are
intrinsic to the basic mixer structure. Because 220 GHz is an astronomically
important frequency : 13 CO(2 — 1), it was decided to adjust one parameter of
the mixer — the length of the last section of the suspended stripline filter —
to shift this problem to a lower frequency; see fig. 3.
Good agreement is found between predicted and measured values for several
parameters : optimum backshort position (fig. 4), upper sideband rejection
Page 246
Third International Symposium on Space Terahertz Technology
DSSll!
i — r
1.7S
-3.000 -
-14.00.
0O.00
i\nHi
\T
100K0' F7EJ-GHZ 120.0
■ SU
a ooo
-e.000
mm. a mix/
XbcaiiS mm 2.0
-U00
wnir
BO. 00
100.0 FSEO-GnZ
120.0
Figure 1. Comparison of measured and computed values
for the return loss between the embedding circuit and a 50fi
junction. Frequencies have been scaled to the 3mm range.
200
150
100 -
50
IF=(3.7-4.3)GHz T ampl=14 K
200 220 240
LOCAL OSCILLATOR FREQUENCY (CHj)
260
o SO
220 230 2<0
LOCAL OSCIUAIOR mtOUENCY (GHi)
Figure 2. DSB receiver noise ver-
sus LO frequency for the original
1.3mm mixer.
Figure 3. Same as Fig. 2 for mod-
ified mixer.
Third International Symposium on Space Terahertz Technology
Page 247
versus backshort position (fig. 5), DSB receiver noise versus backshort posi-
tion for fixed LO frequency (fig. 6). Note that at 230GHz, an SSB receiver
temperature of 135K results when the mixer is tuned for 10 dB rejection of
the USB, versus 215K SSB when tuned DSB.
0.5 -
_^_
1 ' ' *^" "
MEASURED
""^sC^*,^
1 ' ' F—l 1 ' *"
-
CALCULATED
^-.^
-
--v
_
1 .
~~^..„^
210
220
230
FREQUENCY
240
(GHz )
250
Figure 4. Backshort positions for
optimum performance at each fre-
quency. Comparison of predicted
and measured values.
-20
F=231 GHz
0.J 0.4 0.5
BACKSHORT POSITION (MM)
Figure 5. USB rejection versus
backshort position. Comparison of
predicted and measured values.
o.s
o
J00
200 -
100
0.2
— . — ' — • — • — i — ' — • — • — > — r — ' — '" •
MEASURED
MODEL
i ' -
"•-'■-^J/
.
F=231 GHZ
0.3 0.4 0.5
8ACKSH0RT POSITION (MM)
0.6
Figure 6. DSB receiver tem-
perature versus backshort position.
Comparison of predicted and mea-
sured values.
Next generation receivers in the laboratory.
The large capacitance of SIS junctions causes a large mismatch, especially at
the higher frequencies. A reactive tuning structure with a high transforma-
tion ratio is needed. A backshort can in principle accomplish this, but only
Page 248
Third International Symposium on Space Terahertz Technology
over a limited frequency range (unless one is ready to accept the complication
of a two-backshort structure), and the performance is critically sensitive to
the backshort losses.
Local tuning with superconducting circuit elements overcomes these limita-
tions. With the junction capacitance tuned out at least approximately over
the frequency range, the demand on backshort reactive tuning is considerably
diminished, and better performance can be obtained. We have designed op-
timized tuning structures for the 1.3mm and 0.8mm bands. Figure 7 shows
the predicted mismatch loss and the measured laboratory performance for
the 1.3mm receiver. Figure 8 shows the same quantities for the 0.8mm re-
ceiver. Both receivers were measured with a Mach-Zender diplexer for LO
injection, a room- temperature lens producing a beam matched to the f/10
Nasmyth focus of the IRAM 30-M telescope, and a 4 GHz IF chain having a
14 K input noise temp-erature. We plan to improve these receivers by using
cold optics.
200
150
100
MEASURED
CALCULATED
-*^*w\---
//
/
lF=(3.7-4.3)GHz Tampl=14K
400
300
200
IF=(3.7-4.3)GHz, T ampl=14 K
ISO 200 220 2«0 260 280
LOCAL OSCILLATOR FRECUENCY (GHz)
320 330 .340 350 ' 360
LOCAL OSCILLATOR FREQUENCY (CHz)
Figure 7. DSB receiver noise ver-
sus LO frequency for 1.3mm mixer
with inductive compensation. Com-
parison of predicted and measured
values.
Figure 8. Same as Fig. 7, for ihe
0.8mm receiver.
Figure 9 summarizes the performance of receivers now operating on the
IRAM telescopes, and of improved receivers being developed in the labora-
tory. Figure 10 illustrates the discovery of Aluminum fluoride in the evolved
star IRC-f-10216, made at the IRAm 30-M telescope with one of the two
1.3mm SIS receivers.
Conclusion.
A relatively simple equivalent circuit can be used successfully to model and
Third International Symposium on Space Terahertz Technology
Page 249
400
§
Q.
3
UJ
O
m
Q
UJ
o
a:
300
200
£ 100 -
-" 1 r "
L-^-
100
1111
200 300
LOCAL OSCILLATOR FREQUENCY (GHz)
Figure 9. DSB noise versus frequency for IRAM SIS re-
ceivers. Dotted lines : receivers, now on the telescopes; con-
tinuous lines : receivers in the laboratory.
0.4
0.3
0.2
0.1
i — i r
t — i — i — r
t — i — i r
-0.1
II. I
in
CO (J-2-1)
A1F (J=7-8)
r\^^v^^^
>A
>V^M^W^J^
I ! I I
J L
I
I I I I I I I » I
I
230500
230600 230700 230800
Rest Frequency (MHz)
230900
Figure 10. Detection of aluminum fluoride with one of the
two 1.3mm SIS receivers at the IRAM Pico Veleta radiote-
lescope (Cernicharo and Guelin 1987).
Page 250 Third International Symposium on Space Terahertz Technology
improve the suspended stripline mixer mount. The performance of such
a mixer using an SIS junction can be significantly improved by employing
inductive tuning of the junction capacitance. It is also noteworthy that
focusing the design effort on mismatch losses, and leaving aside intrinsic
conversion loss, we can show good correlation between computed mismatch
losses and measured receiver temperature, and that such modelling can serve
as an effective guide to improving significantly the receiver performance.
References
Blundell, R. Carter, M., and Gundlach, K.H. 1988 Int. J. of Infrared and
Millimeter Waves, 8, 361
Carter, M.C., Navarro, S., Karpov, A., Billon-Pierron, D., Lehnert, T.,
Gundlach, K.H. 1991 Proceedings 16 th Int. Conf. Infrared and Millimeter
Waves.
Cernicharo, J., and Guelin, M., 1987 Astronomy and Astrophysics 183, L10.
Karpov, A., Blondel, J., Mattiocco, F., and Lazareff, B., 1992 Journees
d'Etudes Micro-Ondes et Espace, Toulouse CNES/CNET.
Lehnert, T., Grassl, C, Gundlach, K.H., and Blondel, J. 1991 Supercond.
Sci. Technol. 4, 419
Yuda, M., Kuroda, K., and Nakano, J. 1987 Japan. J. Appl. Phys. 26, 166
Lehnert, T., Billon-Pierron, D., Grassl, C, and Gundlach, K.H. 1992 Iram
Working Report 210
Third International Symposium on Space Terahertz Technology Page 251
Comparison of Measured and Predicted Performance of a SIS Waveguide Mixer at
345GHz fl9 3«§5L£^9
CE.Honingh*, G.de Lange', M.M.T.M.Dierichs*, H.H.A. Schaeffer*, J.Wezelman*,
J.v.d.Kuur', Th.de Graauw # , and T.M.Klapwijk* ^ t ^
# Space Research Organization of the Netherlands (S.R.O.N.), Landleven 12, 9747
AD Groningen, The Netherlands
* Dept. Applied Physics and Materials Science Centre, University of Groningen,
Nijenborg 4, 9747 AG Groningen,
The Netherlands
Abstract
The measured gain and noise of a SIS waveguide mixer at 345 GHz have been
compared with theoretical values, calculated from the quantum mixer theory using a
three port model. As mixing element we use a series array of two Nb-Al 2 3 -Nb SIS
junctions. The area of each junction is 0.8 \xvtf- and the normal state resistance is 52
n. The embedding impedance of the mixer has been determined from the pumped
DC-IV curves of the junction and is compared to results from scale model measure-
ments (105 x). Good agreement was obtained. The measured mixer gain however is a
factor of 0.45 ± 0.5 lower than the theoretical predicted gain. The measured mixer
noise temperature is a factor of 4 - 5 higher than the calculated one. These discrepan- _
Page 252 Third International Symposium on Space Terahertz Technology
ties are independent on pump power and are valid for a broad range of tuning
conditions. C <\ \
Introduction and measurement set up.
This study is done as part of an ESA research contract to investigate the feasibility of
SIS-mixers as space qualified THz-mixers. Predictions of the mixer performance are
mainly based on the quantum mixer theory, by Tucker, reviewed in l . At lower
frequencies the validity of the theory has been investigated thoroughly 2 , and quan-
tum limited noise behaviour has been measured in very few cases 3 .
Our main purpose for this study is to identify sources of noise in the receiver and to
asses the quality of the tuning of the mixer. Receiver noise temperatures measured
with the Y-factor method are shown in Fig.l. An overview of the route that we follow
to obtain all information using only noise measurements is outlined in Fig.2.
Measurements were done with two different mixerblocks. One mixer block (TT) a
backshort and an E-plane tuner 4 , and another similar mixerblock (ST), without the
E-plane tuner. We use non-contacting backshorts with two quarter lambda high/low
impedance sections covered with an insulating Si0 2 layer of 200 run.
As mixing element we use an array of two Nb-Al 2 3 -Nb junctions in series, each with
an area of .8 Mm 2 and a normal state resistance of 52 n. The toRC product of the
array is approximately 5 at 350 GHz. All measurements have been done with a
magnetic field of two fluxquanta in the junctions and over an IF bandwidth (B) of 80
Third International Symposium on Space Terahertz Technology Page 253
MHz around 1.4 GHz.
Measured mixer data
The mixer gain (GMM) is calculated from the subtraction of the IF-output power in
response to a 300 K and a 77 K input load. GMM={Pout(300)-Pout(77)}/{Gif.Gf.-
APin}, where Gif is the gain of the IF-chain, Gf is the gain of the IR-filter at 77 K,
and APin is the difference in input power between a 300K and a 77K load on the 77
K radiation screen in the dewar.
To achieve the highest accuracy Gif is determined in situ by using the unpumped
mixer junctions as a calibrated noise source as a function of bias voltage 5 . The total
IF output power as a function of bias voltage is given by
Pif^iV) = G„ [leBG, J de (lOcot:h(-gl) (*£& ( V) *G X ) ' 2 +kB ( T iaol \T ir ( V) \ 2 * T jr ) }
and is fitted to the measured power. V is the biasvoltage, and I DC (V) is the unpum-
ped IV-curve. e is the electron charge, k is Boltzmanns' constant and T is the physical
temperature of the junction, taken to be 4.5 K. G, is the input impedance of the IF-
chain.
Gif is obtained with an accuracy of 5% from the slope of measured IF-power as a
function of bias voltage above aprroximately two times the gap voltage. The noise
temperature of the IF-stage is T IF + | r(V) 1 2 T isol , where T IF is the noise temperature
of the HEMT-amplifier (Berkshire Technologies) and T^, assembles the noise contri-
Page 254 Third International Symposium on Space Terahertz Technology
butions from the bath temperature, and possible contributions of imperfect isolation
between amplifier and mixer. r(V) is the reflection due to the impedance mismatch
between the IF-chain and the junction. Since Tif= 3 ±0.5 K and Tisol=5.5±0.5 K are
obtained from the fitting, the second term, which is essentially depedend of the
dynamical conductance of the junction array, can have a significant contribution.
APin is calculated from Plancks' law. The gain of the dewar window (Gw), the
beamsplitter (Gbs) and the IR-filter (Gf) have been measured separately with a
Michelson interferometer. Gbs = 0.89 ±1%, Gw=95±2% and Gf=95±l% for the fre-
quency of interest. In the calculation of the input power on the mixer it is assumed
that the window is at 300 K.
GMM is given in Fig. 3 as a function of bias voltage for both mixers.
Determination of the embedding impedance
Knowledge of the embedding impedance is crucial to the theoretical calculation of
the mixer performance of an SIS junction. For design purposes we used a 105 x scale
model of the mixer mount. The impedance measured on the final structure as a
function of backshort position and at optimum E-plane tuner position, is given by the
larger circle in Fig. 4.. The estimated geometrical capacitance of the junction array (22
fF) has been added in parallel to the impedance measured in the scale model.
The embedding impedance in the real mixer has been determined from the pumped
IV-curves. We regard the series array of two junctions as one equivalent junction. The
measured and calculated pumped curves are compared using the voltage match
method 6 , where both the embedding impedance and the pump power are adapted to
Third International Symposium on Space Terahertz Technology Page 255
give a best fit. A typical example of a measured and a fitted curve is given in Fig. 5.
The correspondence between the two curves was always very good execpt for a small
region at the quasi particle step above the gap voltage.
The embedding impedance has been determined for various backshort positions at
one (optimum) E-plane tuner position. The expected circle in the Smith chart is fitted
through the points in Fig. 4. The given points are lying in a very small part of one
half lambda cycle of the backhort. The pattern is repeated for the next half lamba
cycle, without a measureable increase in loss. To make that more clear the data of
two cycles are given as a function of backshort position in Fig 4. The data as predic-
ted by the scale model and a direct measurement of the coupled power (the pump
step height) are also given as a function of backshort position. The DC-current at a
biaspoint on the quasiparticle step has been normalized to one.
Comparison between measured and calculated mixer performance
The embedding impedances determined from the pumped IV-curves (and checked by
the scale model measurement) have been used to calculate the gain and the noise
behaviour of the mixer. We used the three port model in the low IF approximation,
justified by the &>RC-product of the junctions and the IF frequency of 1.4 GHz. The
terminations on the LO-port and at both side band ports were each determined
separately. They differed considerably as can be seen in Fig.6, giving the pumped IV-
curves at a single tuner setting for three different frequencies.
Page 256 Third International Symposium on Space Terahertz Technology
1GAIN
The calculated mixer gain (GMC)
GMC(V) = 4G L (G ush |Z 01 (V)| 2 + G Js „|Z . 1 (V)| 2 )
is given as a function of bias voltage in Fig.3. Z 0l and Zq , are the relevant elements
of the 3x3 conversion matrix 1 and Gusb and Glsb are the real parts of the terminating
impedances at both side band frequencies, as determined from the pumped IV-curve.
This gain is directly compared to the gain (GMM) determined from the measure-
ments in Fig.3.
The discrepancy between GMM and GMC is independent of LO-power and also
within a 15% error independent of the tuning conditions. It must be noted this has
only been checked for the points given in the Smith Chart of Fig. 4. Around those
points the fitting of the embedding impedances is the most accurate. For the most
inductive tuning points the discrepancy in the gain is larger. At those points the bias
supply seems to skip over the regions with negative differential resistance, deteriating
the DC-curve and IF-output. For points more to the edge of the Smith Chart the
amount of pump power necessary to get a well developed pump step is larger and the
gap of the superconductor decreases, making accurate fits more elaborate.
2 NOISE
To obtain a measure for the noise contribution of the mixer we compared the
Third International Symposium on Space Terahertz Technology Page 257
measured and the calculated total noise output of the receiver in an IF-bandwidth of
80 MHz.
The noise contribution of the mixer is calculated from the DC-IV curve and the
embedding impedances using the current correlation matrix 1 . As in the unpumped
case the junctions array is regarded as one equivalent junction obtained by dividing
both the measured current as the measured voltage by the number of junctions. The
mixer gain used in the calculation is the gain determined from the measurements .
This means that we attribute the discrepancy between GMC and GMM fully to the
loss/coupling efficieny of the lens/horn/waveguide at 4K in front of the mixe. The
calculated and the measured IF-output power as a function of bias voltage are given
in Fig. 7a. However to get the correspondence at the first pump step as shown in Fig.
7a , an extra input noise power kBT ex , with T ex = 80 ± 20 K, had to be added in the
calculation at both side bands in addition to the shotnoise and temperature noise
contribution . This value for T ex is again independent of pump power and tuning
conditions within the same restrictions to the tuning range as mentioned in the
calculation of the gain. The calculated and measured noise contributions of the
various parts of the receiver are given in terms of noise temperature in Fig. 7b.
The results in Fig. 7 are for the ST-mixerblock but a similar performance is found in
the TT-block. Though still within the error margin the deviation in the gain has a
tendency to be less in the ST-mixer compared to the TT-mixer, probably as a result of
the improved fabrication and the use of an integrated horn.
We verified that the extra noise contribution was not a real extra input signal due to
LO-signal at the side band frequencies by filtering the LO with a high Q Fabry-Perot
filter.
Page 258 Third International Symposium on Space Terahertz Technology
Discussion and Conclusions
We compared the performance of two types of waveguide SIS-mixers with the three
port quantum mixer theory. We have obtained good agreement between the scale
model measurements and impedances determined from pumped IV-curves. The
quality of the fittings is very high in the sensitive tuning region of the mixer.
However we observed a reproducible difference between the measured and the
calculated gain of both mixers. The difference can be explained partly by losses in the
lens and horn.
The performance of the backshort seems to be quite lossless regarding the good
agreement between the scale model measurements and the impedances fitted to the
pumped IV-curves.
The noise values are more than a factor of four higher than expected from theory.
This seems to be a general feature of mixers using a series array of junctions. Up to
now we did not yet have single junction mixerrs available.
We acknowledge the financial support of the European Space Agency for this work
under contract 7898/88/NL/PB(SC) and Herman v.d. Stadt for careful reading of this
summary and Anders Skalare in general.
Third International Symposium on Space Terahertz Technology Page 259
References
1 J.R.Tucker, and M J.Feldman, Rev.Mod.Phys. £7, 1055 (1985)
2.W.R.McGrath, P.L.Richards, D.W.Face, D.E.Prober, and F.L.LIoyd, J.Appl.Phys. £3_,
2479 (1988)
3.C.A.Mears, Qing Hu, P.L.Richards, A.H.Worsham, D.E.Prober, and A.V.Raisanen,
IEEE Trans.Magn. MAG-27, 3363 (1991)
4. B.N.Ellison, P.L.Schaffer, W.Schaal, D.Vail, and R.E.Miller, IntJ. of IR and MM-
wavesiQ, 937 (1989)
5. J.R. Tucker , IEEE J. Quantum Electron. QE-15, 1243 (1979)
6. A.Skalare, Int. J. of IR and MM waves K), 1339 (1989)
G
./
Page 260 Third International Symposium on Space Terahertz Technology
Captions
Fig. 1 Receiver noise temperature tor the two types of waveguide mixers measured
with the Y-factor method, corrected for the beamsplitter loss.
Fig. 2 Overview of the different input an output parameters in the process of compa-
ring the measured and calculated performance of the mixers.
The measurements yield Tree as result of a Y-factor (H/C) measurement. The gain
and noise of the IF(Gif,Tif) and of the mixer(Gm,Tm) are obtained from the absolute
IF-output power at different input loads, knowing the loss and the physical temper-
ture of the input window(Gw.Tw).
The embedding impedance of the junctions is determined either with use of a scale
model or by fitting the pumped IV-curves to the theory. When a scale model is used
the geometrical (and parasitic) capacitance of the junction has to be estimated se-
parately.
When the embedding impedance is known, the mixer performance is calculated as a
function of bias supply at different LO-power levels.
Fig. 3 Measured ( + ) and calculated ( • ) coupled gain for both waveguide mixers. The
TT-mixer has a 500-50 n transformer at the IF-port to enhance the gain.
Fig. 4 Embedding impedance as a function of backshort position, as calculated from
the scale model (-+-) and as determined from the pumped IV-curves (M,A). As a
direct measure of the coupled power the pumped step height at the optimum
biaspoint (-o-) is also given as a function of backshort position
Third International Symposium on Space Terahertz Technology Page 261
Fig. 5 Measured (-■-) pumped IV-curve at 351 GHz and calculated(-) curve using the
given fitting parameters for the embedding circuit. The admittances are normalized
to the 104 n.
Fig. 6 Detailed view of DC-IV curve of the series array of junctions, pumped at three
different frequencies. The tuning conditions and the pump power are identical at all
frequencies.
Fig. 7A Total measured ( + ,o) and calculated (-) IF-output power in a bandwidth(B)
of 80 MHz at two different input signals, as a function of bias voltage. For the
calculated IF-power an extra noise power of 80kB has been added to the input of the
mixer. The contributions of the shot noise and temperature noise of the junctions
(dPjunctie) and of the IF-stage (dPif) are given also.
Fig. 7B The total measured ( + ) and calculated ( • ) reciever noise temperature as a
function of bias voltage. For reference the contribution to the calculated reciever
noise temperature of the IF-stage (dTif), the junctions (dTm) and of the input losses
(dTw) are also given.
Page 262
Third International Symposium on Space Terahertz Technology
300
ffl
09
Q 200
-+- 1 tuner -A_ 2 tuner
o
o
u
h
100
A— A y
\
+-+-
^<^~*
-K
./
/
340 348 356 364 372 380
frequency (GHz)
Fig.1
GUIDE LINE
H/C
Receiver
(dewar)
Phot
Pcold
Tree
fif(V)
LQpm.Mxer, jp — f Gif(V)
V
eg
Scale-
model
DC-
IV
Tm(V)
Qm(W
Analysis
Pumped
rV-curves
Ylo,
Yusb,Ylsb
Plo
Three-port
mixer model,
Tucker theory
-* Ti
m
Fig. 2
"Vto/V««0
H
■t
&
o
C/>
ST-mlxer
Mitx*!- ■■In
TT-mixor
Kflm*r aala
1 .2
CKlo.
so o
Meat.
I**
1 -O
o.«
o.e
O.-*
y\
0.2
a.o '
■^
Mi
A
i .a
> .o
o.a
©.«
<>.■♦
o.a
o.o
SOO O
IP
Q
t/i
£'
3
o
a
»
o
H
a-
55
3
n
a-
a
o
I
«n*
Fig. 3
Page 264
Third International Symposium on Space Terahertz Technology
ZO-110 Q
Coupled power as a function
of backshort position
I COLD uODl
■» fi {Yemb from fit
Wc
In
1500 Bacfakpos. (>im> 2300
Fig. 4
G.emb=1.60
2
B.emb=-0.33
■ meas.
— calc.
Io>=1.41
lb
Io
•
i
\
•
^■i — _
(
>
VtyVgap
2
Fig. 5
Third International Symposium on Space Terahertz Technology
Page 265
2.0 3.0 4.0
Vb (mV)
3.0 6.0
Fig. 6
ST-mixer
2 nx. trriy (Nb-SIS)
ST-mixer
NoiM contribution* (II)
o
o
u
H
300
250
200-
150
100
50
i i
i <
i i
i i
\ i
ti
u
1
Vb/Vjtp
B
<JTm
— — <JTif
— ♦— Ttot
* Tree
dTw
Fig 7.
Page 266 Third International Symposium on Space Terahertz Technology
jho&B N93-2? 750
A Low-Noise 492 GHz SIS Waveguide Receiver
C. K. Walker'.t, J. w. Kooi' . M. Chan', H.G. LeDuc* , P.L. Schaffer' , J.E.
Carlstrom' , and T.G. Phillips 1
1 California Institute of Technology, Pasadena CA
2 Jet Propulsion Laboratory, Pasadena C A
f Presently at the University of Arizona, Tucson AZ
Abstract
In this paper we discuss the design and performance of an SIS waveguide
receiver which provides low noise performance from 375 to 510 GHz. At its design
frequency of 492 GHz the receiver has a double sideband noise temperature of -172
K. By using embedded magnetic field concentrators, we are able to effectively
suppress Josephson pair tunneling. Techniques for improving receiver performance
are discussed.
Introduction
Over the last decade SIS receivers have been replacing Schottky diode based
systems at millimeter and submillimeter wavelengths. SIS junctions have a lower
shot noise and a more pronounced D.C. nonlinearity than Schottky diodes, with the
result being that mixers constructed with them are more sensitive and require less
local oscillator power than their Schottky diode counterparts.
SIS mixers can be constructed using waveguide or an open structure geometry.
To date, at all frequencies where they have been built, SIS waveguide mixers provide
superior performance. The main advantage waveguide mixers have over open
structure designs is that adjustable backshorts can be readily incorporated into the
mixer block. Backshorts are usually needed to match the complex impedance,
although recently structures employing lithographically produced matching networks
for waveguide mounted junctions has proved highly successful (Kerr et al. 1987).
Open structure mixers typically utilize a combination of lenses and planar antenna
structures to couple the incoming radiation to the junction. With this combination of
components it is not easy to incorporate an adjustable backshort. Fixed tuned reactive
stubs can be fabricated along with the SIS device to tune out the junction's
capacitance. A significant advantage of waveguide designs is that well characterized,
efficient feedhoms can be used to couple waveguide modes to free space modes.
Third International Symposium on Space Terahertz Technology
Page 267
Until recently SIS waveguide receivers have been constructed with center
frequencies only as high as 345 GHz. In this paper we discuss the construction and
performance of an SIS waveguide receiver with a center frequency of 492 GHz. It is
now permanently installed on the Caltech Submillimeter Observatory and has been
used for astronomical observations since September 1991.
Receiver Construction
Optics
A block diagram of the 500 GHz receiver is shown in Figure 1 . The receiver's
optics are designed to provide a -10 db taper on the edge of the telescope's secondary
mirror. The beam from the secondary is reflected from an offset parabola, a flat, and
a final offset parabola before entering the cryostat. At the entrance to the cryostat a
thin (- 0.2 mil) mylar beamsplitter is mounted at 45* to the signal and local oscillator
beams. With this thickness of mylar only 0.5% of the signal and local oscillator beam
is reflected. Therefore only a tiny fraction of the signal is lost, while almost all the
local oscillator power is terminated in an absorbing load.
, par coked
MKxn.D«moo«»
IF. LNA
m '.0- 20O.HI
PGOUTEUPAMR
HITS))
IGOft TEW* AMP. CSERTQO
10TALPCWER
OUTRjr
L
| Cl.'SCXi
I «.i.0 5aw 4
I I^.^Tt.1 R^T— I
SO LAWOETECTCn
4f.- i.o a*
TO OSSPECTROUEIEfl
Figure 1
The cryostat vacuum window is a 1.0 mil mylar sheet. The windows on the 12
K and 77K shields are made from disks of fluorogold laminated to a quarter
wavelength thickness of black polyethylene. The polyethylene reduces reflection
losses. The overall thickness of the fluorogold and black polyethylene disks is 17
mils. The disks serve as near infrared blocking filters for the system.
The last optical component before the mixer block is a small, low density
Page 268
Third International Symposium on Space Terahertz Technology
polyethylene lens mounted in front of the feedhom. The lens is designed using the
equations of Silver (1966) and is located so that it provides frequency independent
illumination of the secondary (Goldsmith 1982).
E PONE T & WAVEGUIDE
TRANSFORMER SECTION
IF OUTPUT
IF MATCHING CIRCUIT
CORRUGATED
FEEOHORN SECTION
..aACKSHOflr TUNER
Diagram of Mixer Block
Figure 2
Mixer Block
A schematic representation of the mixer block is shown in Figure 2. An
exploded view of the completed block is shown in Figure 3. The basic design of the
waveguide portions of the block follows that of Ellison et al. (1987). The block is
divided into five sections along the longitudinal axis of the block. The first section
consists of a corrugated feedhom (Thomas 1978) which terminates in circular
waveguide. The beamwidth of the horn, at its -10 db points is 13.4*. In the second
section, the signal passes through a three step, circular to full-height rectangular,
quarterwave transformer. This section also contains the waveguide for the E-plane
tuner. The center of the E-plane tuner is located - Xg/2 in front of the SIS junction.
The third section contains the SIS device, the IF impedance matching circuit,
magnetic field concentrators, and the waveguide for the backshort. A face-on view of
this section is shown in Figure 4. The location of the main circuit elements are
indicated. The SIS junction substrate resides in a long 4.5 by 4 mil channel milled
across the center of the waveguide. When the junction substrate is placed in the
channel, it is oriented so that the junction itself is in the center of the waveguide
facing the oncoming radiation. A RF choke is fabricated on the substrate at the same
time the junction is made. The choke, shown in Figure 5, is made from a series of
high and low impedance sections of microstrip line. Each section is a Ay 4 in length at
the RF center frequency. The ground side of the junction is held in place with silver
paint. The "hot" side of the junction is connected to the IF matching circuit via a
short 1 .0 mil gold wire. The wire is silver painted to the last section of the RF choke
located on the junction substrate. To keep the silver paint from inadvertently shorting
the hot side of the junction substrate to the block, the block is designed so that the last
8 mils of the junction substrate is suspended in free space.
Third International Symposium on Space Terahertz Technology
Page 269
Matching and Bias.
Circuit
Corrugated
Lens Holder
E-flane
Backshort
Drive
Waveguide Transformer
and E-Ptane Tuner
Exploded View of the 490 Bloc k
Figure 3
IF Matchi
Circuit
Signal
Waveguide
Field
Concentrator
F Connector
as Lines
lunction
Channel
Field
Concentrator
Junction Block
Figure 4
A detailed discussion of the integrated IF matching circuit has been given by
Kooi et al. (1992). It utilizes a 5-pole Chebyshev low pass filter and transformer to
match the IF impedance of the SIS junction to the input impedance of the IF amplifier
(- 50 £1) over a 1 to 2 GHz frequency range. The IF impedance of the junction is
roughly 2.5 times the junction's normal state resistance. The matching circuit is
designed for an SIS IF impedance of - 160 CI.
The magnetic field concentrators and the core of the external electromagnet to
which they are connected are made out of Cryoperm, a material which retains a high
magnetic permeability even at 4 K. To reduce heat loading on the 4 K cold plate, the
magnet's coil is made from niobium wire. The backshorts are non-contacting and
were designed using the techniques discussed by Brewer and Rasianen (1982). Non-
I 3
TO
O
T
18
2 \*.zJU*JU*JU:.
2.7., 2.7 „„ 2.7 „, 2.7
AiA
prWTVTYTV^VTVTYTVTV
I 0.2 J 0.2 02 f 02 0.2 0.2 J 02 J 0.2 J
21 2.1 2 1 21 2.1 21 21 2 1 2*1 Z.I
X- Socio n through Junction Subslrae
Mounted in Mixer Bbck
.0 (
2.0
I
!§
2.5 j
4.5
■ Waveguide
Metal Film
. Fused Quartz
Note All dimensions are in mils
492 GHz Mask Dimensions
Figure 5
40.0
f
1.8
5 3
or B
~ o
v? 5'
• OQ
cr
o
r>
5T
sr
o
3
a
a
o
<
3
OQ
cr
er
cr
o
3
00
cr
to
C
5-
n
3-
3
o'
s
CO
3
■a
o
w
C
3
o
3
CO
TJ
a
o
re
H
3-
5
3
n
3-
3
O
•—4
Third International Symposium on Space Terahertz Technology Page 271
Junction Fabrication
The Nb/ALO x /Nb tunnel junctions used in this receiver are planar, submicron
area devices defined by electron beam lithography. The junctions are fabricated using
a variant of the self-aligned-liftoff trilayer process (Shoji et al. 1983) with
modifications for electron beam lithographic patterning of junction area. The
Nb/ALO x /Nb trilayer is deposited in a high vacuum sputter deposition system (base
pressure 1.3 X 10 Torr). The Nb/Al layers are dc/rf -magnetron sputtered from 75
mm diameter targets. The A10 x tunnel barrier on the trilayer is formed by an in situ
oxidation in an oxygen/argon mixture. A gold layer (» 30 nm) is deposited on the
trilayer to act as a contact layer. The junctions are patterned by forming submicron
holes in PMMA by electron beam lithography (JEOL JBX5) followed by the
deposition and lift-off of chromium metal. The chromium pattern is transferred to an
underlying polyamide film using oxygen reactive ion etching (RIE). Junctions are
formed by RIE using a gas mixture containing CCI2F2 (chosen for its highly
anisotropic etch characteristics) and electrically isolated using thermally evaporated
silicon monoxide. Contact wiring is deposited and patterned using RIE to complete
the device. Tunnel junctions with areas of 0.25 um^ and 0. 13 urn 2 were fabricated on
on the same wafer
Receiver Performance
Figure 6a is a plot of the I-V (current versus voltage) curve of the SIS junction
used in the receiver. The solid line is the I-V curve with no local oscillator power
applied. The curve has a sharp knee at the voltage (- 2.78 mV) corresponding to the
band gap energy of the niobium junction. At the knee, the leakage current through the
junction is about 5 uA. The normal state resistance of the junction is about 90 ft.
The dotted line is the I-V curve with local oscillator applied. With local oscillator
power applied a single wide photon step is observed below the knee of the curve. A
single wide photon step is observed below the knee of the curve. The width of this
step corresponds to the voltage {(to/*) of a 492 GHz photon (- 2 mV). Figure 6b is a
plot of IF power versus SIS bias voltage. The top curve is the IF power obtained with
the receiver looking into a room temperature load (Tpj - 280 K). The lower curve is
the IF power obtained with the receiver looking into a cold load (Tq - 80 K). As
expected, the IF conversion peak occurs at a bias voltage corresponding to the middle
of the first photon step below the gap voltage. The ratio of the IF power obtained with
the receiver looking into the room temperature load to the IF power obtained with it
looking into the cold load is a measure of the receiver's sensitivity and is often
referred to as the Y-factor. The receiver noise temperature is derived from the Y-
factor using the following relation.
To = T " • rcy
R Y- 1
Page 272
Third International Symposium on Space Terahertz Technology
At 492 GHz the highest Y-factor obtained with this receiver was 1 .84, which
corresponds to a double sideband receiver noise temperature of - 172 K. This value is
a true receiver noise temperature, no corrections have been made for signal losses in
the beamsplitter or input optics, losses resulting from impedance mismatches, or from
IF amplifier noise. The bias voltage and current used during this measurement were
2.3 mV and 12 uA. At this bias voltage the receiver noise temperature increased
when the LO power was reduced tp a level where the junction current became less
than - 9 uA. Similarly, if the LO power was increased such that the junction current
reached a value greater than 16 urn, the receiver sensitivity decreased.
V(mV)
SIS RECEIVER I/V CURVE
FIGURE 6a
T T
V(mV) 10
SIS RECEIVER IF POWER vs. BIAS VOLTAGE
FIGURE 6b
Third International Symposium on Space Terahertz Technology
Page 273
During these measurements the magnetic field strength was adjusted so as to
minimize the manifestation on the I-V characteristics of the Josephson pair tunneling
current. As the magnetic field strength was increased, the Josephson super-current
went through several minima. A minimum in the super-current occurs when one or
more magnetic flux quanta are present across the junction. Since more than one
minimum was observed, we conclude the magnetic circuit used in this design is
capable of placing several flux quanta across the junction. Without the magnetic field
the smooth IF power curves of Figure 6b become jagged. In Figure 7 we present IF
power curves made at 420 GHz. Figure 7a shows the smooth IF power curves that
can be obtained with an optimized magnetic field. In Figure 7b the same IF power
curves are plotted, but with the applied magnetic field less than optimum. In 7a and
7b we also plot the corresponding junction I-V curve. The small dips in the IF power
curve of Figure 7a become more prominent in Figure 7b, and occur at voltages where
Josephson steps are seen in the pumped I-V curve. The association between the dips
and the Josephson steps indicates that these structures are the result of Josephson
effect tunneling. These results show that, even at high frequencies, a magnetic field
can be used to effectively suppress Josephson pair currents in small area junctions.
i — r
V(mV) 1.0
JUNCTION I/V & IF POWER CURVES WITH MAGNETIC FIELD
FIGURE 7a
Page 274
Third International Symposium on Space Terahertz Technology
V(mV) 1.0 2.0 3.0 4.0 5.0
JUNCTION I/V & IF POWER CURVES WITH REDUCED MAGNETIC FIELD
FIGURE 7b
In Figure 8 we show spectrum analyzer measurements of the receiver bandpass
from 0.9 GHz to 2.1 GHz. The lower curve in the figure is the bandpass with the
receiver looking into a room temperature (280 K) load. The bandpass is fairly flat,
with a total power variation of only - 3.5 db from 1 to 2 GHz. The upper curve is a
plot of the receiver's Y-factor across the IF band. This curve is essentially flat,
indicating the receiver's sensitivity is constant across the band.
In Figure 9 we present a double sideband, 500 MHz wide spectrum taken with
the receiver on the Caltech Submillimeter Observatory toward the young stellar
source Orion IRC2. The center frequency of the spectrum is 492.16 GHz. An
acousto-optical spectrometer was used to produce the spectrum. An atomic line and
several molecular lines were observed. They are identified in Figure 9. The ordinate
is in units of antenna temperature (K) and the abscissa is in units of frequency (GHz).
The total on source integration time was - 1.3 minutes. At this frequency, the system
noise is dominated by the atmosphere. During the time this spectrum was taken the
single sideband noise temperature on the sky, which includes the noise contributions
of the atmosphere, telescope, and receiver, was - 5000 K.
Third International Symposium on Space Terahertz Technology
Page 275
ATTEN 2MB
n. 4.0<fim
2dEV
MKP 2.1 <fim
1500GHz
-**■**
"\
/
/
^>
\
/
\
1
\
\
\
START 900MHz
RBW 1.0MHZ
VBW 300Hz
STOP 2.1GHZ
SWP 10sac
Receiver Bandpass Plot
Figure 8
♦ 94.7
494.8
Upper SidcSud Frequency (GHz)
494.9 495 495. 1
1 I »
495.2
- r — ■ i 1 I i r
Orion IRc2
CH3OH
_i 1 ■ i_
491.9
492
492.1 492.2 492.3
Lower SideBmd Frequency (Ghk)
Figure 9
492.4
Page 276 Third International Symposium on Space Terahertz Technology
The receiver noise temperature, *R, is determined by a number of factors.
These include the mixer noise temperature, ( *M ), the conversion loss in the mixer
(Closs ), the noise temperature of the first IF amplifier (TlF), and the coupling
efficiency between the IF port of the junction and the input port of the first IF
amplifier ("HlF).
Receiver Performance Comparison
Parameter
230Nb
345Pb
492Nb
Rn(fl)
82
54
85
TR(K)
48
159
176
Tmix (K)
34
129
123
C. Loss (db)
3.1
8.3
8.9
Tmix Cor
26
91
95
TIF(K)
7.0
4.2
6.8
IFConnib.
14
30
53
Table 1
In Table 1 we list the values of these parameters for the receiver described in
this paper and for the two other SIS waveguide receivers presently in use at the
Caltech Submillimeter Observatory. The values of 'M, TV, Closs, and "-IF were
calculated using the shot noise technique described by Tucker and Feldman (1985).
The other two receivers in the table were designed with center frequencies of 230 and
345 GHz. All three receivers use the same basic double stub tuner design. Like the
492 GHz receiver, the 230 GHz system employs a niobium junction and the IF
matching circuit described by Kooi el al. (1992). The 345 GHz receiver uses a lead
based SIS junction fabricated by Ron Miller at AT&T Bell Laboratories. This
receiver uses an older, less efficient IF matching circuit. In the table °N refers to the
-rcor
normal state resistance of the SIS junction used in each receiver. l M refers to the
mixer temperature corrected for the different beam splitter thicknesses used in each
receiver. The value of Tr for the 230 GHz system is a factor of 3 to 4 less than the
Tr achieved with the 492 GHz system. The decrease in system performance at 492
GHz is due to a factor of - 3 increase in Closs and Tm. The performance of the 345
and 492 GHz receivers are comparable. This similarity in performance is most likely
due to the niobium junction in the 492 GHz receiver having a better high frequency
response than the lead junction used in the 345 GHz receiver.
Third International Symposium on Space Terahertz Technology
Page 277
Broadband Heterodyne Recaverg
1000
a?
CO
in
>^
$
100
; a sis quasi optic planar and hybrid antennas
[ A SIS Waveguide
< Schottky Corner Cube
- + Schottky Waveguide * a
Z &
+ a
I +■
a°*°° ^**^~-
a
^^^^"^
: *-- —
10 hv/kg t
"i i i
1 1 1 1_ .1 1 1
100
1000
Frequency [GHz]
FIGURE 10
In Figure 10 we compare the performance of the 492 GHz receiver to other
receivers operating at millimeter and submillimeter wavelengths. At each frequency
where they have been built, SIS waveguide receivers have proven to be the most
sensitive. These receivers typically operate with noise temperatures 10 to 14 times the
quantum noise limit (fv/IQ. The receiver reported in this paper continues this trend.
We have also tested the 492 GHz receiver at 376 GHz and 420 GHz. We obtained Tr
values of 238 and 212 K respectively. (These measurements were taken at sea level
where the physical temperature of the mixer was - 0.62 K higher than at the altitude
where the 172 K, 492 GHz noise temperature measurements were made.) These noise
temperatures include all the losses in the system. If we simply compensate for the
difference in the thickness of beamsplitter used in the measurements, then we infer a
noise temperature of 200 K at 376 GHz and 168 K at 420 GHz. These results suggest
the receiver will have its optimum performance between 420 and 492 GHz.
Improving Receiver Performance
Lower values of Tr would be achieved if we could reduce 'IF, increase ^H\ r
decrease either C-Loss and/or Tm. In a properly designed amplifier, the value of iff
is a function of the quality of the HEMT devices used and the desired IF bandwidth.
In the receiver described here the value of ^IF is almost unity, so there is not much to
be gained in improving the design of the IF matching circuit. Significantly lower
Page 278 Third International Symposium on Space Terahertz Technology
noise temperatures could be achieved if the conversion loss of the mixer were
reduced. Lower conversion losses can be obtained by improving the impedance
match between the junction and waveguide embedding impedence. In the current
design this match is achieved by adjusting the positions of the two backshorts. The
ultimate quality of the impedance match depends on the performance of the
backshorts. Since this receiver will be tuned several times a day, non-contacting
backshorts were used to reduce wear. The backshorts are insulated from the block by
a thin (- 16 urn) layer of mylar tape. The performance of the backshorts could be
improved (perhaps at the sacrifice of durability) by using a thinner insulating material.
Ellison et al. (1991) have recently demonstrated that a 1 to 2 urn layer of silicon
dioxide evaporated onto metallic backshorts serves as an effective insulating layer and
improves backshort performance. An improvement in the rf match could also be
achieved by reducing the capacitance of the SIS junction. The most direct way of
doing this is by reducing the size of the junction. If the normal state resistance of the
junction is to remain the same, the thickness of the insulating layer must be decreased
simultaneously. These last two statements are equivalent to saying we need a high
current density junction with a low coRC. The limiting factor in obtaining small, high
current density junctions is the junction fabrication process itself. However, over the
past few years great strides have been made in improving processing techniques.
There is every indication that this trend will continue. The effective capacitance that
the backshorts need to tune out can also be reduced by fabricating broadband,
inductive stubs on the SIS junction itself. This technique has worked well at
frequencies below 300 GHz (Kerr et al. (1987)). However, as one goes up in
frequency, the dimensional tolerances on these matching circuits becomes more
severe, making fabrication difficult. Even so, recent work by Buttgenbach et al..
(1992), and Jacobs K. et al. (1992) suggests that if a broadband rf matching network
is included, this technique can be used effectively even at submillimeter wavelengths.
Indeed, we plan to test a mixer block like the one discussed here with a broadband,
two section, RF transformer matched SIS junction in the near future. Another way to
reduce the effective capacitance is to use a series array of junctions. Two difficulties
with this technique are that the array elements must be similar (a fabrication
challenge) and the required LO power increases as the square of the number of array
elements. At submillimeter wavelengths, where it can be difficult to generate LO
power, the extra power needed to drive a series array can become prohibitive.
If there were no Josephson pair tunneling, one might expect that an SIS bias voltage
corresponding to the center of the first photon step below the gap would provide the
best receiver noise temperatures. However, we find that even with an optimized
magnetic field, the best receiver noise temperatures are found at bias voltages close to
the gap voltage. To some degree this result is due to the matching requirements of the
waveguide and IF matching network. However, this dependence of the receiver noise
temperature on bias voltage suggests that, while significantly suppressed by the
magnetic field, Josephson pair tunneling does add noise to the system. To avoid this
problem, SIS junctions with larger bandgap energies are needed.
Third International Symposium on Space Terahertz Technology Page 279
Summary
We have constructed an SIS waveguide receiver which provides low noise
performance from 375 GHz to 510 GHz. The receiver is a facility instrument at the
Caltech Submillimeter Observatory, where it has been in use since September 1991.
The SIS junction used in the receiver is a 0.25 \urr niobium trilayer device with a
current density of - 10 A cm . At its design frequency of 492 GHz, the receiver has
a double sideband noise temperature of - 172 K. By embedding magnetic field
concentrators in the mixer block, we are able to put several quanta of magnetic flux
across the SIS junction. By adjusting the strength of the magnetic field we are able to
effectively suppress Josephson pair tunneling. The success of this receiver suggests
that SIS waveguide receivers can provide low noise performance at even shorter
wavelengths.
This work was supported by NSF contract AST 9015755 to the CSO and a gift from
AT&T to purchase the spectrum analyzer used to perform some of the measurements.
References
Brewer, M. K. and Raisanen, A. V. 1982, EEE. Trans. Microwave Theory and
Techniques, 30, 708.
Buttgenbach, T. H. 1992, private communication.
Ellison, B. N., Little, L. T., Mann, C. M., and Matheson, D. N. 1991, Electronics
Letters, 27, 139.
Ellison, B. N. and Miller, R. E. 1987, Int. J. Infrared and Millimeter Waves, 6. 697.
Goldsmith, P.F. 1982, in. Infrared and Millimeter Waves
Kerr, A. R., Pan, S. K., and Feldman M. J. 1987, Int. J. Infrared and Millimeter
Waves, 9, 203.
Kooi, J. W., Chan, M., Phillips, T. G., Bumble, B., and LeDuc, H. G. 1992, preprint.
Shoji, A., Kosaka, F., Shinoki, M., Aoyagi, M. and Hayakawa, H. 1983, IEEE Trans.
Magnetism, 19, 827.
Silver, S. 1966, 1EE Electromagnetic Waves Series, 19, 95.
Thomas, B. M. 1978, IEEE Trans. Antennas and Propagation, 26, 367.
Jacobs K., Kotthaus U„ andVowinkel B., International Journal of Infrared and
Millimeter Waves, Vol. 13, No. 1, 1992
Page 280 Third International Symposium on Space Terahertz Technology
Sur&l'' N93-27751
/bo^ Slot-line end-fire antennas
^\ for THz frequencies
by
H. EkstrSm, S. Gearhart*. P. R. Acharya, H. Dave**,
G. Rebeiz*. S. Jacobsson, E. Kollberg, G. Chin**
Department of Applied Electron Physics
Chalmers University of Technology
S-412 96 Gfiteborg. Sweden
•NASA/Center for Space Terahertz Technology
Electrical Engineering and Computer Science Department
University of Michigan, Ann Arbor. MI 48109-2122, USA
••Planetary System Branch
NASA/Goddard Space Flight Center
Greenbelt. MD 20771. USA
/
ABSTRACT
Tapered slot-line endfire antennas, of BLTSA type, have been fabricated on 1.7 \im
thin S102/S13N4 (£ r = 4.5) dielectric membranes. The antenna patterns, in the E-, H-,
D- and D-cross planes, were measured at 270. 348, 370 and 802 GHz using bismuth
micro bolometer detectors. The antennas have approximately 12 dB directivity, and
the -10 dB beam widths are 50° and 55° in the E- and H-planes at 348 GHz.
respectively. The measurements at millimeter/ submillimeter wavelengths compare
well with scale measurements at 45 GHz as well as with theoretical predictions. The
overall results are encouraging and show that slot-line antennas can be fabricated for
use at THz frequencies. Furthermore, it is shown that the very thin SIO2/SI3N4
membranes are strong enough to be used In practical applications.
Third International Symposium on Space Terahertz Technology
Page 281
INTRODUCTION
Tapered slot-line antennas are often considered for integration in planar millime-
ter/submillimeter wave circuits: e.g. In quasi optical mixers. These antennas can be
operated over a wide bandwidth and radiate wide or narrow beams. Various types of
endfire slot-line antennas can be found in the literature; e.g. the linearly tapered slot-
line antenna (LTSA) 11], the exponentially tapered slot-line antenna "Vivaldi" [2] and
the constant width slot-line antenna (CWSA) [3]. A review of antennas suitable for
integration in circuits for millimeter and terahertz frequencies has been written by G.
Rebeiz[4l.
In this work we have studied yet another member belonging to the family of endfire
slot line antennas; namely the BLTSA (Broken Linearly Tapered Slotline Antenna)
15], Fig. 1. The BLTSA has the advantage, among the slot-line antennas, to require the
small substrate area. In addition, the BLTSA has been extensively studied by the
Chalmers group.
Antennas of endfire slot-line type require a certain optimum substrate thickness
t = 0.03 X(^Tf -iy l [31 to avoid pattern degradation and power loss due to surface
modes. In the millimeter wave range the thickness should be only a couple of mi-
crometers, hence the antenna must be fabricated on a thin dielectric membrane, a fact
which introduces delicate manufacturing problems. A further complication is intro-
duced by the fact that the membranes must be left unsupported in the endfire direc-
tion. Fig. 2.
00
3.75
*±5_»
1.5
Fig. 1. Dimensions of the BLTSA. All dimensions are normalised to the
vacuum wavelength.
Page 282
Third International Symposium on Space Terahertz Technology
Fig. 2. The endfire slot line antenna deposited on BLTSA on 1.7 ^m thin
Si02/Sl3N 4 (£r = 4.5) membrane. The Si frame supporting the membrane
has a thickness of 385 pm. Note that the membrane is unsupported in the
end-fire direction.
FABRICATION
The 1.7 nm thin dielectric membrane supporting the antenna consists of three layers;
thermally grown Si02. LPCVD deposited Si3N5 and SiC>2. With compressive oxide
and tensile nitride, the relative thickness of the layers could be selected to form a
slightly tensile, and consequently flat and rigid membrane. The membrane layers
were deposited on both sides of 385 (im thick silicon wafers. To form the membrane
region for the antennas, the silicon was etched in EDP from the backside of the wafer,
with the backside nitride and oxide layers patterned with the membrane layout and
used as etch mask. The nitride and oxide layers on the front side acted as etch stops
for EDP. A considerably manufacturing problem is due to the fact that the antenna
requires that the membrane is not supported by silicon in the endfire direction. In
order to simplify the photo lithography were the antennas fabricated on fully sup-
ported membranes. After the antennas were fabricated was the silicon support in the
endfire direction removed. The fragility of the membrane limits the area to roughly
3 x9 mm 2 . Thus, the maximum available membrane size limits this particular an-
tenna design to frequencies above 300 GHz.
MEASUREMENTS
Several scaled versions of BLTSA's were fabricated on 25.4 |im thick Kapton foil and
measured at 45 GHz. The best antenna design was then scaled to 348 and 802 GHz
(dimensions normalized to vacuum wavelength are given in Fig. 1). These antennas
were fabricated on the 1.7 urn thin dielectric membranes. Fig. 2.
Third International Symposium on Space Terahertz Technology
Page 283
Bismuth micro bolometers were used as detectors. A Gunn oscillator with a
tripler/quadrupler was use as signal source at 270 and 348 GHz, whereas an optically
pumped far infrared laser was used to generate the 802 GHz signal. The dynamic range
in the antenna pattern measurements was approximately 20 dB.
The E-, H-, D- and Dx- planes of the 348 GHz antenna were measured at four frequen-
cies: 270, 348. 370 and 802 GHz, respectively, whereas the 802 GHz design was only
measured at 802 GHz. Measurements of the 348 GHz design at 270 and 802 GHz, Figs.
3a, b and c, show the wide bandwidth of this type of antenna. At the design frequency
the -10 dB beam width varies between 43 and 55° in the measured planes. The D-plane
crosspol level is as high as -6 dB, which is a typical value for endfire slot-line anten-
nas. The directivity was calculated to approximately 12 dB. In the calculations the
lobes outside the measured range and the back lobes were set to -14 dB and -20 dB re-
spectively.
-H
D Dx
-25.0
i i i i i — i— i — i— i — i i i i i i i i i i
-60 -40 -20 20 40 60
Angle [deg]
Fig. 3a. Antenna patterns for the BLTSA designed for 348 GHz but
measured at 270 GHz
Page 284
Third International Symposium on Space Terahertz Technology
25.0 I
-I I > — ' — u
-60 -40 -20 20 40 60
Angle (deg]
Fig. 3b. Antenna patterns for the BLTSA designed for the 348 GHz design
measured at 348 GHz
H
D Dx
O.Oi •■ ■ ■ ■
-20
Angle (deg]
Fig. 3c. Antenna'patterns for the BLTSA designed for the 348 GHz design
measured at 802 GHz
At 802 GHz the antenna beam is narrower and more symmetric (39 °- 43° at -10 dB
level and the directivity is approximately 13 dB) with slightly lower sidelobes than
the pattern at 348 GHz, also the D-plane cross-pol level is lower (-8 dB). Fig. 4a. b. The
improved pattern at 802 GHz was expected since the membrane is relatively thicker
Third International Symposium on Space Terahertz Technology
Page 285
(expressed in wavelengths) and closer to the optimum thickness. In fact, the optimum
frequency for these membranes Is approximately 4 THz Thus, at THz frequencies. It is
expected that the antenna pattern would be even more symmetric and the D-plane
crosspol level could be as low as -10 to -15 dB below the co-polarized level.
0.0
E, 348GHz
H, 348GHz
E, 802GHz
H. 802GHz
-25.0
-60 -40 -20 20 40 60
Angle (degl
Fig. 4a. Measured E- and H plane of the 802 GHz design at 802 GHz. and the
348 GHz design at 348 GHz
0.0
-25.0
D, 348GHz v
Dx, 348GHz
_i — , — , — i — u
D, 802GHz :
Dx, 802GHz
-60 -40 -20 20 40 60
Angle [degj
Fig. 4b. Measured D- and Dx plane of the 802 GHz design at 802 GHz, and
the 348 GHz design at 348 GHz
Page 286
Third International Symposium on Space Terahertz Technology
The performance of the BLTSA end-fire antennas on 1.7 |jm thin membranes is sum-
marized in the table below.
Frequency GHz
270
348
370
802
E- plane -10 dB beam width
side lobe level dB
59°
-12
50°
-19
49°
-19
43°
-20
H-plane -10 dB beam width
side lobe level dB
64°
-8
55°
-11
53
-11
44°
-11
D-plane -10 dB beam width
side lobe level dB
50°
-7
43°
-10
42°
-10
39°
-12
D-cross lobe level dB
-4
-6
-6.5
-8
Directivity dB
11
12
12
13
Table 1 . Compiled measured antenna data. All antennas were fabricated
on a 1.7 urn thick SiC«2/Si3N4 membrane. The thick frames around 348
and 802 GHz indicate the design frequencies. The 348 GHz antenna was
also measured at 270. 370 and 802 GHz. The 802 GHz antenna was only
measured at 802 GHz,
Scale measurements at 45 GHz show that the E and H-plane patterns agree reasonably
well with the patterns at the higher frequencies. Fig. 5. These antennas were made on
25.4 pi thick Kapton™ foil (£^3.5). Note that, this substrate thickness is compara-
tively thicker (measured in wavelengths) at the scale frequency than the thickness of
the dielectric in the 270 - 802 GHz measurements.
— H-plane
E-plane
CQ
u
>
OS
-10
-20 20
Angle [deg]
Fig. 5. E and H-plane patterns of a scale model of the BLTSA. The
measurements are performed at 45 GHz.
Third International Symposium on Space Terahertz Technology
Page 287
THEORY
The end-fire slot-line antennas have been analyzed by using method previously
described by e.g. Janasawamy [6]. In this method, the antenna tapering is
approximated by a "stalrcase'-function in a number of steps of different widths but
equal lengths, as shown in figure 6. Thus, the slot-line antenna can now be treated as
a linear array of apertures each fed with different phases and amplitudes. Fur-
thermore, power conversion is applied to relate the amplitudes and fields in
neighbouring apertures. The characteristic impedance and the wavelength of each
slot is calculated by using a spectral domain technique [7]. The far field pattern of the
antenna is calculated by applying an appropriate Green's function and adding the
field contributions from all apertures. In these calculations we ignore the reflections
from the discontinuities in the antenna taper (which has been confirmed in computer
simulations).
The theoretical patterns agree reasonably well with measured patterns. Fig. 7, 8, con-
sidering the approximations in the calculations. The calculated and measured beam
width correspond very well, and it is possible to predict level and position of the side-
lobes within 2 dB and 5°. The biggest difference between the calculated and measured
patterns is in the crosspolarized D-plane. where the theory predicts a level 2-3 dB
higher than the measured level.
Fig. 6. Approximation of the slot-line antenna by an array of slots.
Page 288
Third International Symposium on Space Terahertz Technology
0.0
-25.0
-60 \/-40 -20 20 401/ 60
Angle [deg]
O.Or
-5.0
CQ
? -10.0
1
§2 -15.0
J2
OS
-20.0
-25.0
-60 -40 -20 20 40 60
Angle [deg]
0.0
- - - D (exp)
V - -Dx (expj)
-25.0
-60 -40 -20 20 40 60
Angle [degl
Fig. 7. Calculated (th) and measured (exp) antenna patterns at 348 GHz
Third International Symposium on Space Terahertz Technology
Page 289
0.0
-60 -40 -20 20 40 60
Angle (deg]
0.0
£ -20.0
-25.0
-1 — i — | — i — I — r— r
1 — I — I — I — I — I — I — I—
D (exp)
i Dx (exp i
-60 -40 -20 20 40 60
Angle (deg]
Fig. 8. Calculated (th) and measured (exp) patterns at 802 GHz
Page 290 Third Interna lional Symposium on Space Terahertz Technology
REFERENCES
[1] S.N. Prasad, and S. Mahapatra. A novel MIC slot-line aerial Proc. 9th Eur.
Microwave Conf.. pp. 120-124. Brighton. UK, 1979.
[2] P.J. Gibson, The Vivaldi aerial 9th Eur. Microwave Conf.. pp. 101-105. Brighton.
UK, 1979.
(3) K.S. Yngvesson. D.H. Schaubert. T.L. Korzeniowski, E.L. Kollberg, T. Thungren.
and J. Johansson. Endjire tapered slot-line antennas on dielectric substrates,
IEEE Trans. Antennas. Propagat.. vol. AP-33. N.12. ppl392-1400. 1985.
[41 G. Rebeiz. Millimeter-wave and terahertz tntegrated-clrcuit antennas, preprint,
to appear in the October issue on Space terahertz Technology in the IEEE-
Proceedings.
[51 P.R. Acharya. J. Johansson, and E.L. Kollberg. Slotltne antennas for millimetre
and sub millimetre wavelengths. Proc. 20th Eur. Microwave Conf.. pp. 353-358.
Budapest. Hungary. Sept 1990.
[61 R. Janasawamy, D. Schaubert. Analysis of Tapered Slotline Antenna, IEEE
Trans. Antennas Propgat.. vol. AP-35. no.9. pp. 1058-1065. Sept. 1987.
[7] G. Johansson, P.R. Acharya. J. Johansson. Determination of Slot Line
Characteristics, Technical Report No. 97L. 1991. Chalmers University of
Technology.
f
Third International Symposium on Space Terahertz technology Page 291
93-27752
QUASI-OPTICAL ANTENNA-MIXER- ARRAY DESIGN
FOR TERAHERTZ FREQUENCIES
Yong Guo
Department of Electrical and Computer Engineering
College of Engineering, Clemson University
Clemson, SC 29634
Kent A. Potter, David B. Rutledge
Division of Engineering and Applied Science
California Institute of Technology-
Pasadena, CA 91125
Abstract
A new quasi-optical antenna-mixer-array design for terahertz frequencies is presented
in this paper. In the design, antenna and mixer are combined into an entity, based on
the technology in which millimeter-wave horn antenna arrays have been fabricated in
silicon wafers. It consists of a set of forward- and backward-looking horns made with
a set of silicon wafers. The front side is used to receive incoming signal, and the back
side is used to feed local oscillator signal. Intermediate frequency is led out from the
side of the array. Signal received by the horn array is picked up by antenna probes
suspended on thin silicon-oxynitride membranes inside the horns. Mixer diodes will be
located on the membranes inside the horns. Modeling of such an antenna-mixer-array
design is done on a scaled model at microwave frequencies. The impedance matching,
-RF and LO isolation, and patterns of the array have been tested and analyzed.
I. Introduction
In submillimeter-wave and terahertz frequency systems, because of their much shorter
wavelengths compared with microwave systems, waveguide circuits become much
smaller, which makes them very difficult and expensive to build. However, quasi-
optical components provide a solution to this problem. Quasi-optical antenna-mixer-
array combines antennas and mixer circuits into a single entity. The design is based
on an existing technology by which dipole excited integrated-circuit horn antennas are
made in silicon [1]. The horn antennas consist of probes suspended on a thin oxyni-
tride membrane inside pyramidal horns which are chemically etched in silicon. The
antennas are free of dielectric losses and have plenty of space for electronic intercon-
nections between the probes. The aperture efficiency of these etched horn antennas
has been improved to 72% [2]. Recent research shows that the experimental results
agree well with the theoretical analysis, including radiation patterns and resonant
dipole impedances [3]. Various antenna probes inside the pyramidal horns have also
been studied [4].
This integration of antenna and mixer eliminates the need for RF and LO circuit-
fashion connections. Such construction offers a device with potential of smaller size,
Page 292
Third International Symposium on Space Terahertz Technology
lighter weight, more ruggedness and less cost, as compared to conventional methods.
Moreover, this kind of design potentially can be mass produced by standard integrated-
circuit technology. The applications of the integrated-circuit antenna-mixer array
include imaging systems, radars, and satellite communications.
II. Horn Structure and Mixer Circuitry Design
In order to avoid the difficulty to supply an LO power at a frequency close to RF,
a subharmonically-pumped antenna-mixer array is designed which is pumped by an
LO at only half of the RF frequency. Since the RF and the LO frequencies differ
by approximately a factor of two, in principle, it is easier to realize the isolation
between the RF and the LO. Furthermore, spurious responses associated with the
odd harmonics of the LO can be rejected by using an antiparallel diode pair. The
subharmonically-pumped horn-antenna-mixer array is shown in Figures 1 and 2. It
consists of a set of forward- and backward-looking horns facing back to each other
made silicon wafers. Every four RF horns are provided with one LO horn which is a
rectangular-shaped horn. The spacing between RF horns is 1A. By using the sub-array
concept, every four RF horns can be taken as a sub-array. Four RF horns, together
Trough
Horn
Membrane
Monopole probe
Coplanar strips Loaded dipole probe
(a)
(b)
Figure 1 The horn structure of the subharmonically-pumped mixer design, one LO horn
corresponding to four RF horns; (a) LO horns, the trough made of two silicon wafers is put
on the top of the horns, (b) RF horns, monopoles are used for the RF reception.
Third International Symposium on Space Terahertz Technology
Page 293
Dipole probe — -
pr? — Silicon oxynitride
r membrane
P=1.0Xrf
Li=0.26\rf
L=0.40Xrf
D=0.11Xlo
d: varies
Membrane wafer
Monopole probe
Mixer diode
position
(a)
(b)
Figure 2 (a) Cross-sectional view of the antenna-mixer array, (b) The mixer circuit design
for a unit cell; four monopoles for the RF reception and one dipole for the LO reception;
both RF and LO are detected by the mixer diodes located in the center of the unit cell; IF
is led out from the ends of the LO horn through a coplanar-strip transmission line.
with one LO horn, form a unit cell. This design will keep the best symmetry, and the
beam patterns of the sub-array will be improved by a factor of 4 compared with that
of the single RF horn. Since the size of the LO horns should be twice that of the RF
horns, half of the area on the LO side in each unit cell would be left unused. This
would cause strong reflection from the flat surface, namely, a 3 dB reflection loss. In
order to eliminate this 3 dB reflection loss, a structure is designed to be placed on the
top of the LO horns. The structure has a long trough on each row of the LO horns
and will fill up the space between the LO horns to converge the incoming energy into
the LO horns. Figure 1 shows the antenna-mixer array with 2x2 LO horns looking
from the LO side and 4x4 RF horns looking from the RF side. The cross-sectional
view is shown in Figure 2(a). The mixer circuit design is shown in Figure 2(b). Every
monopole from each of the four RF horns will couple the RF signal down to the LO
horn through a coplanar-strip transmission line. A dipole probe is employed to receive
the LO. The mixer diodes are located in the center of the LO membrane, and the IF
is led out from the ends of the LO horn. The dipole probe is loaded on the ends near
the sidewalls with a short stub, which, as a result, could compensate for the capacitive
characteristic impedance of the short dipole probe.
Page 294 Third International Symposium on Space Terahertz Technology
III. Modeling
The mixer array design has been tested on a scaled model sub-array, which has four
RF horns and one LO horn. The gain of such a sub-array will be increased by a factor
of 4 compared with a single horn. Based on this scaled model sub-array, antenna
impedances and receiving patterns were measured. Antenna probes are required not
only to couple the free-space wave energy to the mixer circuit but also to provide a
suitable impedance, the embedding impedance, to the mixer diodes. This impedance
over a wide frequency range is also important for mixer performance because various
frequency components exist in the mixer circuits. In order to achieve good isolation
between the RF and the LO, as well as to match the impedances of the RF and the
LO to the diode impedance, various mixer circuits have been tested. Trade-off has
been made among the impedance match and the isolation between the RF and the
LO so as to minimize losses.
The actual model consists of two square RF horns and a half rectangular LO horn,
a half of the designed unit-cell, sitting on a big copper-clad circuit board which was
used as an image plane. The monopole built in each of the RF horns will couple
the incoming signals to the loaded dipole in the LO horn through the coplanar- strip
transmission line. A small channel in the middle of the horns will let the monopole
probes go through between the LO and RF horns. The design frequency for the RF
is 10 GHz and the LO is 5 GHz, corresponding to the wavelength of 3 cm and 6 cm,
respectively. The opening of RF horns is 1 Ar F square, while the height of the LO
horn is Alo/2, and the LO horn width is 1 Alo- Mixer diodes are to be placed in the
center of the dipole probe in the LO horn. An SMA bulkhead feed-through connector is
soldered from the back of the circuit board to the place where the diodes are supposed
to be. The inner conductor of the connector is soldered to the dipole probe, and the
outside conductor is soldered to the circuit board used as a ground plane.
Measurements were done on an HP 8510 Network Analyzer and data were collected by
a PC. Full two^-port calibration was made in order to measure not only the reflection
coefficients but also the receiving properties and the isolation between the RF port
and LO port by measuring the transmission coefficient Si2- For this purpose, a broad-
band horn antenna is used as a transmitting horn which has a working frequency
range from 2 GHz to 18 GHz. The measured impedances are marked on the Smith
chart in Figure 3(a). Although the impedances were measured on only a half unit-cell,
consisting of two RF horns and a half LO horn, the impedances in a full unit-cell
can be easily obtained by doubling those measured impedances in the half unit-cell.
The impedances in Figure 3(a) are plotted by using Puff, a software CAD program [5].
Both the RF and the LO impedance should be matched to 50 Q. because each beam-
lead diode in the antiparallel diode pair has a resistance of about 100 fi. In the
graph, when the loading stub on the dipole decreases in length, the LO impedance at
5 GHz changes from the inductive to the capacitive impedance, passing the resonant
resistance at about 50 ft, which is a very good matching impedance for the diode pair.
This LO impedance of the circuit can be regarded as the LO dipole-probe impedance
Third International Symposium on Space Terahertz Technology
Page 295
parallel with the impedance of the coplanar transmission line plus the RF probes.
At 5 GHz, the impedance of the coplanar transmission line plus the RF probes is
very high as is illustrated by the LO frequency mark "5" when the entire LO dipole
probe is taken away. Hence, the resonant LO impedance is mainly determined by
the loaded dipole probe and is relatively independent of rest of the circuits. On the
other hand, the RF impedances at 10 GHz are pretty high and independent of the
loading-stub length changes. The average value of those RF impedances at 10 GHz is
about 84 + ;82 ft.
The normal-incident power receptions by the RF and the LO horns were tested over
a wide frequency range, from 2.0 GHz to 12.0 GHz. This was done by putting a wide-
band transmitting horn in front side of the RF horns or the LO horn. Figure 3(b)
shows the measured power received by the RF horns when the transmitting horn
is on that side (solid line) and by the LO horn when the transmitting horn is that
side (dashed line). At the LO frequency of 5 GHz, the difference between the LO and
the RF power is defined as the LO-RF isolation. The higher isolation, the lower the
coupling loss, under the condition that other parameters stay the same. Similarly,
the RF-LO isolation is the power difference between RF and LO at the RF frequency
■ LO at 5 GHz
• RF at 10 GHz
1:d=0.12ALO
2:d=0.08XLO
3:d=0.06XLO
4:d=0.02\LO
5: d=0.00 Xlo
D=0.00 Xlo
-10
-0.2
-60
-70,
H I I | I I I | I I I | I I I | I I T"
LO-RF isolation 13dB
_i i i i i i_
i i i
j i_
RF side
LO side
i I i
6 8
Frequency, GHz
10
12
(b)
Figure 3 (a) The circuit impedances is indicated on the Smith chart with respect to the
different loading-stub length d\ measured at 10 GHz for the RF and 5 GHz for the LO.
(b) Measured normal-incident power, received by the RF horns (solid line) and by the LO
horn (dashed line) with equal distance.
Page 296
Third International Symposium on Space Terahertz Technology
Loss component
Simple-probe design
Split-LO-probe design
RF mismatch, dB
1.7
4.9
LO mismatch, dB
0.0
0.2
RF-LO coupling, dB
1.3
0.0
LO-RF coupling, dB
0.2
0.2
Figure 4 Comparison of the impedance- mismatch losses and the coupling losses between the
two different mixer circuit designs.
of 10 GHz. The measured RF-LO isolation is 6 dB or a 25 % loss. If the LO probe
is split into two, by tuning the spacing between two LO probes, minimum RF-LO
isolation of 20 dB was achieved. As a comparison, of the simple-probe design and
the split-LO-probe design, the impedance-mismatch losses and the coupling losses of
the two designs are listed in the table in Figure 4. For both designs, the biggest loss
comes from the RF-impedance mismatch, 1.7 dB for the simple probe and 4.9 dB for
the split-LO-probe.
IV. Conclusion
A new antenna-mixer array design has been presented, which potentially can be made
and used at submillimeter and terahertz frequencies. Modeling work shows that, for
this design, compromises have to be made between the RF-impedance mismatch, the
LO-impedance mismatch and the RF-LO coupling losses. In some applications, if cer-
tain LO loss is tolerable, then lower RF loss can be achieved (which means lower con-
version loss) by sacrificing some LO impedance-mismatch losses and LO-RF coupling
losses. The mixer elements could be either Schottky diodes [6] or superconducting SIS
mixers [7].
V. Acknowledgements
We appreciate the support of Aerojet Elect roSystems Co., Azusa, CA. and the Army
Research Office through the Jet Propulsion Laboratory.
VI. References
[1] G. M. Rebeiz, D. P. Kasilingam, Y. Guo, P. A. Stimson, D. B. Rutledge, "Mono-
lithic Millimeter- Wave Two-Dimensional Horn Imaging Arrays," IEEE Transactions
on Antennas and Propagation, September, 1990.
[2] Y. Guo, K. Lee, P. Stimson, K. Potter, and D. Rutledge, "Aperture Efficiency
of Integrated- Circuit Horn Antennas," Microwave and Optical Technology Letters,
January, 1991.
[3] G. V. Eleftheriades, W. Y. Ali-Ahmad, L. P. Katehi, G. M. Rebeiz, " Millimeter-
Wave Integrated-Horn Antennas: Part I-Theory, Part II-Experiment " IEEE Trans-
actions on Antennas and Propagation, November, 1991.
Third International Symposium on Space Terahertz Technology Page 297
[4] Y. Guo, J.C. Chiao, K.A. Potter, D.B. Rutledge, "Probe Modeling for Millimeter-
Wave Integrated- Circuit Horn Antennas," submitted to the IEEE AP-S International
Symposium, July 18-25, 1992, Chicago, Illinois.
[5] R.C. Compton, S.W. Wedge, D.B. Rutledge, "PuiE Computer Aided Design for
Microwave Integrated Circuits," Caltech Press, January, 1990.
[6] T.W. Crowe, W.C.B. Peatman, "GaAs Schottky Diodes for Mixing applications
Beyond ITHz," Second International Symposium on Space Terahertz Technology,
JPL, Pasadena, CA, February 26-28, 1991.
[7] M.J. Wengler, N. Dubash, G Pance, R.E. Miller, "High Gain and Noise in SIS
Mixers at the Submillimeter Wavelengths," Second International Symposium on Space
Terahertz Technology, JPL, Pasadena, CA, February 26-28, 1991.
Page 298 Third International Symposium on Space Terahertz Technology
5*7-32- N93-2? 753
ANALYSIS OF A NOVEL NON-CONTACTING
WAVEGUIDE BACKSHORT
T. M. Weller and L. P. B. Katehi,
University of Michigan NASA Center for Space Terahertz Technology
W. R. McGrath,
Jet Propulsion Laboratory Center for Space Microelectronics Technology
M^
ABSTRACT A new non-contacting waveguide backshort has been developed for mil-
limeter and submillimeter wave frequencies. The design consists of a metal bar with rect-
angular or circular holes cut into it, which is covered with a dielectric (mylar) layer to form
a snug fit with the walls of a waveguide. Hole geometries are adjusted to obtain a periodic
variation of the guide impedance on the correct length scale, in order to produce efficient
reflection of rf power. It is a mechanically rugged design which can be easily fabricated for
frequencies from 1 to 1000 GHz and is thus a sound alternative to the miniaturization of
conventional non-contacting shorts. To aid in high-frequency design, a rigorous full-wave
analysis has been completed which will allow variations of the size, number and spacing of
the holes to be easily analyzed. This paper will review the backshort design and the method
developed for theoretical characterization, followed by a comparison of the experimental and
numerical results. Low frequency models operating from 4-6 GHz are shown to demonstrate
return loss of > —0.2 dB over a 33% bandwidth. The theory is in good agreement with
measured data.
Third International Symposium on Space Terahertz Technology Page 299
INTRODUCTION
Waveguides are used in a wide variety of applications covering a frequency range from
1 GHz to over 600 GHz. These applications include radar, communications systems, mi-
crowave test equipment, and remote-sensing radiometers for atmospheric and astrophysical
studies. Components made from waveguides include transmission lines, directional couplers,
phase shifters, antennas, and heterodyne mixers, to name a few. In addition to the many
commercial applications of waveguides, NASA needs such components in radiometers oper-
ating up to 1200 GHz for future space missions, and the Department of Defense is interested
in submillimeter wave communications systems for frequencies near 1000 GHz.
One of the most frequent uses of waveguide is as a variable length transmission line.
These lines are used as tuning elements in more complex circuits. Such a line is formed by a
movable short circuit, or backshort, in the waveguide. A conventional approach is to use a
contacting backshort where a springy metallic material, such as beryllium copper, makes DC
contact with the broadwalls of the waveguide. The contacting area is critical, however, and
must be maximized to produce an acceptable short circuit. These backshorts are excellent in
that they provide a good short circuit over the entire waveguide band. The contacting areas
can degrade, however, due to wear from sliding friction. It is also extremely difficult to get
a uniform contact at frequencies above 300 GHz, where the waveguide dimensions become
0.5 mm x 0.25 mm for the 300-600 GHz band.
An alternative approach is the non-contacting backshort shown in Figure 1. A thin
dielectric layer (such as mylar) prevents contact and allows the backshort to slide smoothly.
In order to produce an rf short and, hence, a large reflection, this backshort has a series
of high- and low-impedance sections which are approximately -f in length, where \ g is the
guide wavelength. The rf impedance of this design is given approximately by [1]
Zrf = (|^-r Z low (1)
Page 300
Third International Symposium on Space Terahertz Technology
WAVEGUIDE
vN
\
MYLAR
INSULATOR
wmM%#M;M&WffimMWg%%%\
(FRONT)
v\
M?Mmm,*-m
mmmmmmmm\
\S HTfiH >J
LOW V fflGH ^ WAVEGUIDE
IMPEDANCE IMPEDANCE OPENING
SECTIONS SECTIONS
Figure 1: Cross sectional view of a conventional non-contacting backshort.
where Zi ow is the impedance of the thick (low-impedance) sections, Zhigh is the impedance
of the thin (high-impedance) sections, and n is the number of sections. Beginning near
100 GHz, the thin high- impedance sections become difficult to fabricate, and fabrication
may not even be feasible beyond 300 GHz. It would also be difficult to have the short slide
snugly inside the waveguide at these high frequencies, as the thin sections would be very
weak. To circumvent these problems, a novel non-contacting backshort design has recently
been developed [2, 3] which is suitable for millimeter and submillimeter wave operation.
It is a mechanically rugged design which can be easily fabricated for frequencies from 1
to 1000 GHz, and is thus a sound alternative to the miniaturization of conventional non-
contacting shorts. Previously, however, the new backshort was optimized empirically using
low-frequency models. This paper will discuss the new design and outline a new method
developed for theoretical characterization. The formulation is a rigorous full-wave analysis
which involves both mode-matching techniques and a coupled set of space domain integral
equations. A description of the experimental setup is included, followed by a comparison of
experimental and theoretical results. The new theoretical formulation fits these results well.
Third International Symposium on Space Terahertz Technology
Page 301
DIELECTRIC COVER
HOLES
(FRONT)
2«S
3^
DIELECTRIC COVER
METAL SHORT
(into waveguide)
Figure 2: The new non-contacting backshort design, shown with three rectangular holes. The size,
shape, and spacing of these holes are important in determining the rf properties of the short. S
is the spacing, L\ is the length, and Li is the width of each hole. The front of the backshort is
inserted into the waveguide opening.
NOVEL NON-CONTACTING BACKSHORT DESIGN
The novel non-contacting backshort has the merits of easy fabrication up to Thz frequen-
cies, flexibility of design, and very good performance over relatively broad bandwidths. The
important features are briefly reviewed here. In order to obtain a large reflection, a non-
contacting backshort must provide a periodic variation of guide impedance on the correct
length scale. This is accomplished in the new design by either rectangular or circular holes,
with the proper dimensions and spacing, cut into a metallic bar. A representative design is
shown in Figure 2. This bar is sized to fill the waveguide cross-section and slide smoothly
with a dielectric (mylar) insulator along the broadwalls. The holes replace the thin-metal,
high-impedance sections in the conventional design shown in Figure 1. Since the holes ex-
tend completely through the bar, this yields a higher impedance than the corresponding
sections in the conventional design. Thus, the high-to-low impedance ratio is larger in the
new design. In addition, the electromagnetic fields are concentrated near the central axis of
Page 302 Third International Symposium on Space Terahertz Technology
calculated return loss for a backshort with no holes, inserted about 4.5 inches into the end
of the waveguide, and covered with mylar (e r = 3.35). The waveguide opening is assumed to
present a Cl impedance (i.e., it is covered with a metallic plate). Although the gap height
is only 2.5% of the waveguide height, roughly 65% of the incident power is lost at resonance
due to finite conductor and dielectric loss. The utility of the holes, then, is to minimize or
eliminate these dropouts.
THEORETICAL CHARACTERIZATION
The theoretical characterization of the JPL backshort design is performed using a com-
bination of two well known full-wave analysis methods, namely mode-matching and the
application of equivalent magnetic currents in a space domain integral equation. In what
follows, the approach will be outlined and the major governing equations presented. It is
noted here that the symmetry of the backshort about the x-z plane (parallel to the plane of
the waveguide broadwalls) has been utilized to reduce the number of unknown parameters.
Furthermore, only rectangular holes (not round) have been considered in order to simplify
the analysis. Neither of these points, however, are necessary restrictions in the formulation.
A discussion of the analysis is aided by the schematic in Figure 4, which represents the
cross-sectional view of a backshort with two holes, inserted a distance d into the end of a
rectangular waveguide. The structure is symmetric about the x-z plane, with equal dielectric
regions (which are the dielectric covers shown in Figure 2) above and below the metal short.
The problem of interest is to determine the reflection coefficient for the dominant waveguide
mode, travelling in the +z direction, at the front of the backshort (z = 0).
The formulation is based on the decomposition of the problem into two primary compo-
nents. In the first part, we wish to compute the scattering matrix [S] at z = 0, as depicted in
Figure 4. As [S] represents simply the scattering at a waveguide discontinuity, the presence
of the holes may be neglected and thus becomes decoupled from the problem at hand. The
Third International Symposium on Space Terahertz Technology
Page 303
.3
6
3
o
-5.0
4.0 4.4 4.8 5.2 5.6 6.0
Frequency, GHz
Figure 3: Calculated return loss versus frequency for a backshort with no holes, where the gap
height is 2.5% of the total guide height.
the waveguide, such that the holes are effective in producing large correlated reflections, and
thus acting as an efficient rf short. The new design is also easy to fabricate and can be used
at any frequency between 1 GHz and 1000 GHz. For very high frequencies, above 300 GHz,
the metallic bar is a piece of shim stock polished to the correct thickness. The holes can
be formed by drilling, punching, or laser machining, or they can be etched using common
lithography techniques.
It is important to note that the holes are a critical factor in obtaining efficient reflection
from the non-contacting short. With the backshort inserted in the waveguide, a cavity forms
between the metal bar and the broadwall of the waveguide, in the region occupied by the
dielectric insulator. This cavity is terminated by the large discontinuities at the front of
the short and at the waveguide opening. (This is more clearly illustrated in Figure 1 for
the conventional design.) Deep dropouts in the return loss will occur at frequencies for
which this cavity resonates, even though the height of each gap may be only a small fraction
of the total waveguide height. The effect is well illustrated in Figure 3, which shows the
Page 304
Third International Symposium on Space Terahertz Technology
REGION
I
REGION (into waveguide)
-#-
WAVEGUIDE
FLANGE
[S]
M
upper
M
«pper
M
lower
HOLE1
M
lower
HOLE 2
(metal)
PLANE OF
(dielectric-filled gap)
Z=0
WAVEGUIDE
BROADWALL
7?
-ff-
SYMMETRY
k
Z=d
Figure 4: Cross- sectional schematic diagram (not to scale) of a two-hole non-contacting backshort,
inserted a distance d into the end of a waveguide. The waveguide broadwalls are on the top and
bottom in the figure.
well documented mode-matching method, which has been used to solve a variety of wave-
guide problems [4, 5, 6] is applied to determine [S], With this method, the fields at each
side of the reference plane (2 = 0) are expanded in infinite series of orthogonal mode pairs
(e.g. TE-to-z and TM-to-z), and continuity of the tangential electric and magnetic fields
is enforced to determine the scattered field amplitudes. This results in the following set of
generalized equations,
00 00 00 00
£(«.' + O^f '' + I>m + &)W = IK' + tf W + IK' + %)W (2)
n,m
n,m
E(«i - «)*? •' + IK - &)$«>' = £ -k" - 6J')*?." + E -(«" - tf)^" 0)
n,m
n,tn
n,m
where (2) satisfies continuity of tangential E and (3) satisfies continuity of tangential H. In
the above, a and b represent the coefficients for waves travelling toward and away from the
reference plane, respectively. The subscripts e and m are for TE-to-z and TM-to-z, while
Third International Symposium on Space Terahertz Technology Page 305
the superscripts denote the field type (electric or magnetic) as well as the region to which
they pertain (to the left or right of the reference plane). The vector functions $ contain
the appropriate constants and x- and y-dependencies for the transverse components of the
respective fields. At this point inner-products are formed using $f ,J and $^ ,J with (2),
and $^- /7 and $™' 7/ with (3). As these inner-products involve integration over the guide
cross-section, a system is linear equations results due to the orthogonality of the modal
components. This system is assembled into a matrix representation and, after inversion, the
solution is expressed as
{b} = {a} T [S] (4)
With [S] determined, the unknown scattered- field amplitudes {b} are found from (4) given
the known incident-field amplitudes, {a}. It is noted that the presence of a termination at
z = d (see Figure 4) is easily accounted for by assigning
b 1 = a I Sn+a I S 1 2(I-r L S 22 )- 1 T L S 21 , (5)
where I is the identity matrix and [Tl] is a matrix which accounts for the reflection at
z = d. As shown in the figure, we assume that the waveguide opening is terminated in a
complex load Zl for simplification. (This approximation is necessary because the conditions
outside the short are difficult to control experimentally and, likewise, difficult to accurately
characterize analytically. This will be discussed further in the section on results.) The matrix
[Ti] is thus a diagonal matrix of elements
< r ^ - ferf °- Hd < 6 >
In (6), Z x g and Y z are the guide impedance and propagation constant, respectively, for the
i th TE/TM mode. Conductor and dielectric loss may be included in the factor Y z -
The second principle step in the formulation is to apply the space domain integral equa-
tion technique to solve the boundary value problem at the aperture of each of the holes. The
Page 306 Third International Symposium on Space Terahertz Technology
introduction of the equivalent magnetic currents, M upper and M loweT (see Figure 4), allows
the hole openings to be closed by an imaginary metallic surface, provided that no natural
boundary conditions are violated. This is a crucial step in that it transforms the backshort
structure into a combination of a simple rectangular waveguide, which is the dielectric-filled
gap region, and a series of isolated metallic cavities, which are the holes. These unknown
magnetic currents radiate electromagnetic fields in the dielectric region, and a modified form
of (5) therefore results when treating a backshort with holes. The new expression is
b = a Sn + a S\2(I — TlS22)~ ^lS 2 \ +
{(F>T L + F<)S 22 (I - T L S 22 )- 1 T L + F>T L + F<}S 21 . (7)
Note that the only unknown variables in this equation are F K and F > , as the components
of the matrix [S] and [Tl] have previously been determined. These unknown components
are functions of the imposed equivalent magnetic currents.
The solution for the unknown surface currents is uniquely determined by enforcing con-
tinuity of the total tangential fields across the hole apertures. This insures that the natural
boundary conditions of the original problem are preserved. Continuity of the tangential
electric field is satisfied immediately by setting M upper = -M lower = M. Assuming a
backshort with N holes, continuity of the magnetic field at the k th hole leads to the following
space domain integral equation (SDIE) in the unknown M:
— n x ti = n x [Ji +
N I r r
d S '(^-(k 2 I+VV) ■G B )'Mn) +
n=l
5„ we//
f f <fa'(^L_(*'/ + w) • G c ) ■ M k ) (8)
J Js k uje ft
In the above, H ,nc represents the known incident magnetic field, which results from scattering
of the incoming wave at the waveguide step discontinuity (the reference plane). It is expressed
Third International Symposium on Space Terahertz Technology Page 307
in terms of TE and TM modes, the coefficients being given by
a 11 = a I S 12 (I-T L S 22 )- 1 T L
b n = a n S 22 (9)
for +z and —z travelling waves, respectively. Gb and Gc represent the dyadic Green's func-
tions for an infinite rectangular waveguide and a metallic cavity, respectively. Closed-form
expressions for these functions can be derived using well established boundary value for-
mulations [7]. The use of an infinite- waveguide potential in the dielectric-filled gap region,
which does not account for the actual finite length of uniform guide, is possible by consid-
ering the fields to be a superposition of primary and scattered components. The primary
components satisfy boundary conditions at the source, and radiate away from M in the
presence of matched conditions in either direction. These components are precisely those of
the second term on the right hand side of (8). The scattered components are required to
satisfy the boundary conditions away from the source, at the discontinuities at z = 0, d and
are also functions of M. Expressions for these fields, which are represented by H scat in (8),
are similar in form to the primary components but also include factors from the scattering
matrix [S] and the matrix [r^].
The final step in the formulation is to solve the coupled set of integral equations which
results from enforcing (8) over all N holes. This set may be reduced to a system of linear
equations by applying the method of moments (Galerkin's method) [8]. This approach has
been proven to yield excellent results and the convergence characteristics have been well
documented [9, 10, 11]. Using this procedure, the aperture of each hole is first divided into
discrete subsections using a rectangular grid. The unknown currents are then expanded in
terms of overlapping subsectional rooftop basis functions of the form,
M = Y.{^tM x ')<f>^') + ^t j <j> j {x')f i {z')) (10)
Page 308
Third International Symposium on Space Terahertz Technology
sin[k (u>'-u> n .i)) . r , <w ' <w
<t> n (w') = <
1 if u; n _i <w'< w n +i
else
where Mfj and M£ are constant coefficients, / n is the length of the n th subsection in the
^-direction, and k is the wave number in the medium. This expansion is inserted into the
integral equation, and inner-products are then formed using weighting functions which are
identical to the basis functions. The coupled equations are thereby reduced to the following
matrix form:
(Y xx ) (Y X2 )
\
t
{M x }
\
(11)
(Y«) (Y„) J \ {M*} )
where (Y^(C, £ — x, z)) represents blocks of an admittance matrix. The unknown current co-
efficient vectors {Mfj} and {M£} are then determined by solving (11). With M determined,
all elements of (7) may be computed and the solution is complete.
MEASUREMENT TECHNIQUES
The backshort design was optimized by testing the performance in WR-187 band wave-
guide (3.16 GHz - 6.32 GHz), for which the dimensions are 47.5 mm x 22.1 mm. The
dielectric layer around the metal short was formed by stacking sheets of mylar tape. The
magnitude and phase of the reflection coefficient were measured with an HP 8510B Vector
Network Analyzer. A commercially available coaxial-to-waveguide transition connected the
waveguide to the network analyzer. This measurement system was calibrated using two off-
set contacting shorts set at -^ and -g 4 , and a sliding waveguide load. Subsequent verification
using a contacting short indicated a measurement error of about ±0.2 dB in the magnitude
measurement.
Third International Symposium on Space Terahertz Technology Page 309
RESULTS AND DISCUSSION
This section presents examples of measured data and analytical calculations. It will also
address some conclusions drawn regarding the theoretical characterization and performance
of the new design. Regarding the numerical aspects, the code developed to calculate the
scattering matrix [S] at the waveguide discontinuity agreed very well with results found in
[12]. In particular, results were compared for the reflection coefficient from asymmetric (i.e.,
single-step) E-plane and H-plane waveguide junctions. The validation of the remainder of
the theoretical formulation and the associated software was completed by comparison with
measured data. Part of this validation included a study of convergence as a function of
the hole (aperture) mesh size and the number of modes used in the dyadic Green's function
expansions. It was found that using subsections which are approximately ^f on a side, where
\ 9 is the guide wavelength, yields a good compromise between accuracy and the requirements
on storage and computation time. The number of modes for the Green's functions is kept
> 600.
The measurements performed to investigate the new design involved many variations on
the size, shape, number, and spacing of the holes cut into the metal bar [2]. An additional
test variable was the number of stacked mylar sheets used to form the dielectric layer. In
many cases, the height of the backshort was such that a relatively large space was left between
either side of the metal bar and the waveguide broadwalls. This large gap, combined with
the variations in the mylar thickness, are used to. help understand the effect that typical
machining tolerances will have for operation at 200-300 GHz and above.
Results which are typical of the best performance to date are given in Figure 5b. This
data is for a backshort with three rectangular holes, each with dimensions L\ = 19.3 mm,
L? = 28.4 mm and spacing S = 8.7 mm. The width and height of the bar are 47.5 mm
and 19.7 mm, respectively, leaving a gap of 1.2 mm between the bar and the waveguide
broadwalls. The measured results in Figure 5b were obtained using a total mylar thickness
Page 310
Third International Symposium on Space Terahertz Technology
CO
O
O
Hi
_l
LL
UJ
GC
CD
;o
Z
O
O
UJ -1
U-
LU
OQ
z
o
O
UJ
_l
LL
HI
X
■1 -
-2
4
(c)
4.5
5.5
FREQUENCY [GHz]
Figure 5: a) Reflected power measured from a solid bar without holes. This does not make a
good backshort due to the several large dropouts across the frequency band, b) Reflected power
measured from a backshort with three rectangular holes. The mylar is 0.89 mm thick. Excellent
performance is obtained over a broad bandwidth, c) Same backshort as in (b), but mylar thickness
has been reduced to 0.64 mm.
Third International Symposium on Space Terahertz Technology Page 311
of 0.89 mm. The reflection coefficient in this case is greater than —0.2 dB (0.95 reflected
power) over a 33% bandwidth centered around 4.8 GHz. For comparison, the measured
results for the same backshort without holes are shown in Figure 5a. This data clearly
illustrates the improvement from the holes. The complex structure of this response, relative
to that shown in Figure 3, is caused by asymmetrical positioning of the bar inside the
waveguide. Other measurements made with the gap completely filled by dielectric, which
forced a near-symmetric positioning of the bar, agreed very well with our theory and were of
the form shown in Figure 3. The effect of reducing the mylar thickness is seen in Figure 5c,
which gives measured data using 0.64 mm of mylar. The large dropout near 5.8 GHz has
been shifted out of band, due to the decrease in the. effective dielectric constant. This
response is comparable to that obtained for the conventional type of backshort shown in
Figure 1. As expected, increasing the mylar thickness (and thus increasing the effective
dielectric constant) moved the large dropouts lower in frequency.
Performance similar to that with rectangular holes could be obtained using circular holes.
Results obtained with 3 circular holes and a mylar thickness of 0.89 mm demonstrated greater
than 95% reflected power over a 32% bandwidth centered around 4.75 GHz. This is encour-
aging since round holes are easier to fabricate than rectangular holes for high frequencies.
Many other variations of the backshort parameters were tested. Also, the small dips
around 4.5 GHz in Figure 5b, and those seen in Figure 5c, are currently being investigated.
As noted for the plain metal bar, these dips may result from asymmetrical positioning of the
backshort inside the waveguide [13]. A more extensive discussion of the systematic parameter
variations, measurement" observations, and comparisons with theory will be given at a later
date. Some millimeter wave tests have also been performed and are discussed elsewhere [2].
In order to theoretically model the backshort performance, appropriate values were re-
quired for an effective dielectric constant, e r , and the terminating load impedance for the
waveguide opening (Zl in Figure 4). The problem of the dielectric constant arises because
Page 312 Third International Symposium on Space Terahertz Technology
the gap above and below the metal bar is only partially, and non-uniformly, filled by the
mylar sheets. The transverse resonance technique [4] may be used to approximate e r by
solving the exact inhomogeneous problem, and then assuming the entire guide is filled with
some "average" material. (For the inhomogeneous case, a two-layer guide is assumed, where
one layer is air-filled and the other is mylar-filled and of a thickness equal to the total mylar
thickness.) A simpler approach, which yields higher values for the dielectric constant, is to
merely compute e r based on the percentage of mylar relative to the total gap height. By
numerical experimentation, it was found that the best approximation lies nearly midway
between the two values. It is noted that obtaining an exact solution for a layered waveguide
is not justified due to the unpredictable spacing of the various mylar sheets.
The other issue was determining the correct value for the load impedance, Zl, to use
in the calculations. Although an exact analysis is a formidable task, an approximate load
impedance can be obtained to adequately model the waveguide discontinuity. This is done
by first considering the exact impedance for a very thin aperture opening onto an infinite
ground plane. For typical gap dimensions used here, this is a large, capacitive value, with
real and imaginary parts which are both 3-5 times the dominant waveguide mode impedance.
The extension of the backshort beyond the waveguide opening, however, will provide a better
transition to free-space and effectively lower this impedance toward a better match. In many
test cases, use of a normalized load impedance of Zl « 2.5 ± .5 + j\ in equation (6)
yielded good agreement with the measured results. An important point, however, is that the
predicted performance is nearly independent of Zl in precisely the frequency bands where
the backshort works well. This is as expected, as most of the incident power is reflected and
never reaches the end of the guide. It follows that changes in Zl, do modify the dropout
regions in the return loss, as these dropouts result from out-of-band power leaking past the
backshort. (The performance of the bar without holes is likewise strongly dependent on Zl-
This fact can be used in determining an appropriate impedance for a given geometry, by
Third International Symposium on Space Terahertz Technology
Page 313
1.0 r
2
0.0
-1.0 -
1 ■■■ r
■ ■ i ■ •
■ 1 1 1 1 ■ ■
Measured
--•a.--. Calculated
O A * . . ..I. ........ I...... . . . I . . . . f . ...I i . .
4.0 4.4 4.8 5.2 5.6 6.0
Frequency, GHz
Figure 6: Measured data and calculated performance for the backshort with three rectangular holes.
The mylar thickness is 0.89 mm.
comparing measured and theoretical results for various load values.)
A comparison between measured data and calculated performance is given in Figure 6.
These results are for the backshort with three rectangular holes, using a mylar thickness of
0.89 mm. Very good agreement has been obtained. The broadening of the dropouts in the
measured data, relative to the calculated results, is believed to be due Zl and to loss in
the measurement system which is not accounted for by the theory. The bandwidth is very
accurately predicted, however, such that it should now be possible to design and analyze
these backshorts for specific applications.
CONCLUSIONS
In summary, we have developed a theoretical analysis to predict the rf performance of
a new non-contacting waveguide backshort. This backshort consists of a metallic bar with
rectangular or circular holes which enhance the reflections of rf power. The simplicity of this
design allows it to be easily scaled to millimeter wave and submillimeter wave frequencies.
The new theoretical development is a rigorous full-wave analysis which employs a coupled
set of space-domain integral equations and mode-matching techniques. Comparison between
Page 314 Third International Symposium on Space Terahertz Technology
theory and experiment on model backshorts optimized for best performance at 4-6 GHz show
very good agreement.
ACKNOWLED GEMENTS
This work was supported in part by the Jet Propulsion Laboratory, California Institute
of Technology, under contract with the National Aeronautics and Space Administration, and
the Innovative Science and Technology Office of the Strategic Defense Initiative Organization,
and by the University of Michigan NASA Center for Space Terahertz Technology.
References
[1] Collin, R. E. Foundations for Microwave Engineering, New York: McGraw-Hill, 1966,
pp. 259-262.
[2] McGrath, W. R. "A Novel Non-Contacting Waveguide Backshort for Millimeter and
Submillimeter Wave Frequencies," Conference Proceedings of the Second National Tech-
nology Transfer Conference, NASA Conference Publication 3136, Vol. 1, pp. 161-168,
December 1991.
[3] McGrath, W. R. "Non-contacting Waveguide Backshort," U.S. Patent pending.
[4] Itoh, Tatsuo (editor). Numerical Techniques for Microwave and Millimeter- Wave Pas-
sive Structures, John Wiley & Sons, 19S9.
[5] Eleftheriades, G. V., Ali-Ahmad, W. Y., Katehi, P. B., and Rebeiz, G. M. "Millimeter-
Wave Integrated- Horn Antennas: Part I - Theory", IEEE Trans. A. P. , vol. 39 , No.
11, November 1991, pp. 1575-1581.
[6] Masterman, P. H. and Clarricoats, P. J. B. "Computer field-matching solution of wave-
guide transverse discontinuities", Proc. IEE, vol. 118 , No. 1, January 1971, pp. 51-63.
Third International Symposium on Space Terahertz Technology Page 315
[7] Collin, R. E. Field Theory of Guided Waves, Piscataway, NJ: IEEE Press, 1991, pp.
78-86.
[8] Harrington, R. F. Field Computations by Moment Methods, New York: Macmillan,
1968.
[9] Dib, N. I., Katehi, P. B., Ponchak, G. E., and Simons, R. N. "Theoretical and Experi-
mental Characterization of Coplanar Waveguide Discontinuities for Filter Applications,"
IEEE Trans. MTT , vol. 39 , No. 5, May 1991, pp. 873-882.
[10] Dib, N. I., Katehi, P. B. "Modeling of Shielded CPW Discontinuities Using the Space
Domain Integral Equation (SDIE)," Journal of Electromagnetic Waves and Applica-
tions, vol. 5 , No. 4/5, 1991, pp. 503-523.
[11] Dunleavy, L. P. "Discontinuity characterization in shielded microstrip: A theoretical
and experimental study," Ph.D. Thesis, Radiation Laboratory, University of Michigan,
Ann Arbor, 1988.
[12] Marcuvitz, N. Waveguide Handbook, vol. 10 of MIT Rad. Lab. Series, New York:
McGraw-Hill, 1948.
[13] Kerr, A. R. "An Adjustable Short-Circuit for Millimeter Waveguides," Electronics Divi-
. sion Internal Report No. 280, National Radio Astronomy Observatory, Charlottesville,
Virginia, July, 198S.
Page 316 Third International Symposium on Space Terahertz Technology
9 3 -. 2? •? 5 4
SILICON MICROM ACHINED WAVEGUIDES FOR
MILLIMETER AND SUBMILLIMETER WAVELENGTHS
MarkusYap, 1 Yu-Chong Tai, 1 William R. McGrath, 2 Christopher Walker 3
1. Department of Electrical Engineering, California Institute of Technology,
Pasadena, CA91125
2. Center for Space Microelectronics Technology, Jet Propulsion Laboratory,
California Institute of Technology, Pasadena, CA 91109
3. Department of Astronomy, University of Arizona, Tucson, AZ 85726
Abstract ~ The majority of radio receivers, transmitters, and components operating
at millimeter and submillimeter wavelengths utilize rectangular waveguides in some
form. However, conventional machining techniques for waveguides operating above a
few hundred GHz are complicated and costly. This paper reports on the development
of silicon micromachining techniques to create silicon-based waveguide circuits
which can operate at millimeter and submillimeter wavelengths. As a first step,
rectangular WR-10 waveguide structures have been fabricated from (110) silicon
wafers using micromachining techniques. The waveguide is split along the broad wall.
Each half is formed by first etching a channel completely through a wafer. Potassium
hydroxide is used to etch smooth mirror-like vertical walls and LPCVD silicon nitride
is used as a masking layer. This wafer is then bonded to another flat wafer using a
polyimide bonding technique and diced into the U-shaped half waveguides. Finally a
gold layer is applied to the waveguide walls. Insertion loss measurements show
losses comparable to those of standard metal waveguides. It is suggested that
active devices and planar circuits can be integrated with the waveguides, solving the
traditional mounting problems. Potential applications in Terahertz instrumentation
technology are further discussed.
Third International Symposium on Space Terahertz Technology
Page 317
I. Introduction
Rectangular waveguide is a well characterized transmission medium which is
used in a variety of complex rf components and circuits. Many sophisticated
applications including radar, communications systems, test instruments, and
heterodyne radiometers use waveguide components up to millimeter wave
frequencies. The long history of development of waveguide components provides a
broad base of knowledge to synthesize and evaluate new designs for higher
frequencies.
Waveguide is typically fabricated from metals such as brass and copper using
conventional machining techniques. However, at frequencies above a few hundred
GHz, waveguide becomes so small (less than 0.3 mm x 0.1S mm for 500 GHz -
1000 GHz waveguide) that fabrication utilizing these conventional techniques is time
consuming, costly and difficult In addition, mounting active and passive devices such
as mixer diodes, filters and planar probes on these waveguides is difficult.
A substantial research effort in recent years has been devoted to fabricating
micromechanical structures in silicon using micromachining techniques. Moveable
structures such as slider, gears, and spiral springs in the dimensional scale of 50-200
|im have been fabricated [1, 2]. We have taken a new approach in developing and
adapting silicon micromachining techniques to create silicon-based waveguide
circuits which can operate up to millimeter and submillimeter wavelengths.
As a first step we have started fabricating rectangular waveguides for frequencies
between 100 GHz and 1000 GHz. Here we only emphasize WR-10 waveguides
(operating at 75 GHz - 115 GHz) because it is compatible with our existing
measurement equipment. Conventional WR-10 waveguide is a rectangular channel
with inner wall dimensions of 0.1 x 0.05 inches. Our waveguide, however, is made of
two half sections split along the broadwall as shown in Fig. 1. The reason for
splitting the waveguide is to simplify the fabrication process and to facilitate
integration of planar circuits and devices, which is further discussed in Section IV.
Waveguide Channel
Top Si Wafer
onding Layer
Bottom Si Wafer
Fig. 1. a) The waveguide is split into 2 half sections,
b) One half section of a waveguide.
Page 318 Third International Symposium on Space Terahertz Technology
n. Fabrication Process
The fabrication process for the half sections with emphasis on the cross
section is shown in Fig. 2. A thick (0.05 inches) double-side polished silicon wafer
with (110) surface orientation is used. The major flat has a normal in the [111]
direction within 0.4°. After a standard piranha-bath cleaning, a 1000 A Low
Pressure Chemical Vapor Deposited (LPCVD) silicon nitride layer is deposited on
both sides of the wafer. Photolithographic techniques are then utilized to pattern
the waveguide etching windows. Photoresist is used as a masking layer for etching
the silicon nitride windows with an SFg plasma. The silicon nitride is used as an
etching mask to define b, the waveguide height, shown in Fig. lb. After removal of
the photoresist using acetone solvent as shown in Fig. 2a, the wafer is put in a
reflux system and etched in a water based solution of 40 % KOH at 80 °C. Figure
2b shows the wafer after it has been etched completely through to form half of the
waveguide. The etching rate of (110) silicon in this KOH solution is 2 jim/min and
the etching ratio of (110):(111) planes is 170:1. At this rate 0.05 inches (1270 p.m)
of silicon is etched thru in -11 hours. Following removal of the nitride mask using
hot hydrophosporic acid at 150 °C, a polyimide bonding technique is used to glue
these etched grooves to a smooth silicon wafer with an identical thickness (0.05
inches) as shown in Fig. 2c. The wafer is then diced into pieces of half waveguides.
Such a half waveguide is shown in Fig. 2d. Metalization is done by first depositing a
thin (200 A) chrome layer followed by a thicker (5000 A) gold layer on the waveguide
walls using vacuum evaporation. Further metalization is done by electroplating gold
to a thickness of ~3 \im to reduce rf conduction losses.
HI. Experimental Results
In order to perform insertion loss measurement, we designed a pair of brass
mounting blocks. The two waveguide . half-sections are put on the brass mounting
blocks and mated together. This allows the silicon waveguide to be connected to
microwave test equipment using conventional waveguide flanges. The silicon
waveguides are rugged and can be firmly clamped to metallic flanges. The insertion
loss of the WR-10 waveguide is measured over a frequency range of 75 GHz to 110
GHz. The measurement system is shown in Fig. 3. The source is a BWO which
produces several milliwatts. A reference sweep is first taken without the waveguide.
This is compared to a sweep with the waveguide inserted between the source and
detector. The insertion loss for a 2.5 cm long section of waveguide is shown in Fig. 4
(the small wiggles in these curves are noise and do not reflect any resonances in the
waveguide components). The measured loss is about 0.05 dB per wavelehgth (at 100
GHz) across most of the band. This is very good performance and is comparable to
the result for commercially available waveguide which shows a loss of about 0.024
dB per wavelength. The small difference of 0.026 dB per wavelength is most probably
due to differences in the quality of the gold plated surfaces. Our evaporated gold
showed small pits which were still present in the plated layer. Also there was no
gold on the ends of the silicon waveguide where contact was made to the metallic
flanges of the test equipment. We expect improvements in the gold surface to be
directly reflected in improvements in the rf losses.
Third International Symposium on Space Terahertz Technology
Page 329
(110) silicon (0.05 inches)
polyimide bonding layer
KOH etch
(a)
(b)
(c)
silicon nitride (1000 A)
(100) silicon (0.05 inches)
^-chrome (250 A)
pr gold (5000 A)
(d)
Fig. 2. A cross section view of the fabrication process.
BWO
75 GHz to 110 GHz
\
WAVEGUIDE UNDER
TEST
\
DETECTOR
2
r~>
10 dB
10 dB
SWEEPER
SCALAR
NETWORK
ANALYZER
Fig. 3. Block diagram of millimeter wave insertion loss test system.
Page 320
Third International Symposium on Space Terahertz Technology
90 95
FREQUENCY [GHz J
110
90 95
FREQUENCY [GHz]
110
Fig. 4. (a) Measured loss of a 2.5 cm long section of Si-based WR-10 waveguide.
The surface of the silicon was metallized with approximately 3 ujn of gold to
reduce rf losses, (b) Measured loss of a 2.5 cm long section of conventional
metallic waveguide.
IV. Discussion
Waveguide circuits are preferable at frequencies above 100 GHz since
waveguide has the advantage of adjustable rf tuning. This solves the difficulties of
accurately designing fixed-tuned planar microwave integrated circuits.
Unfortunately, millimeter and submillimeter waveguide components are hard to
manufacture by conventional machining techniques. We have shown here the
feasibility of making silicon waveguides. It is also possible to use silicon
micromachining techniques to fabricate other components such as: directional
couplers, waveguide transformers, waveguide-to-planar circuit transitions,
low-loss filters, rectangular and conical feedhorns, and dichroic plates. This wide
variety of waveguide components will become the building blocks for complicated
circuits. For example, complex mixer and frequency multiplier embedding circuits
can be built. These are important for ground-based and space-based radar,
communications, and remote-sensing applications.
Third International Symposium on Space Terahertz Technology Page 321
Silicon micromachined waveguide components have several important
advantages: 1) These structures are produced by projecting the desired
pattern onto silicon with photolithographic techniques. Therefore waveguides with
dimensions suitable for use above 100 GHz can be easily fabricated. 2)
Dimensional accuracy is in the order of a few microns, which is essential for the
fabrication of high-Q components. 3) The waveguide walls would be
atomically smooth, thereby minimizing rf losses [3]. 4) Several versions of a
single component (with variations of a critical parameter) can be produced at the
same time on a single wafer. This would allow for rapid optimization and reduced
cost compared to conventional machining techniques where only one variation at a
time is produced. 5) Most importantly, active and passive devices can be
integrated with the waveguide. For example, a thin (~ l^im) rf transparent silicon
nitride membrane can be fabricated across the end of the waveguide or parallel to
its length in the E-field direction. Active devices such as Schottky diodes and SIS
tunnel junctions as well as micromechanical rf tuning elements[l, 4] can then be
fabricated directly on the membrane as shown schematically in Fig. 5. This would
eliminate the long-standing problem of mounting the devices and would represent a
significant advance for waveguide technology.
Active Device
1(i SiN Membrane
, Etched
Waveguide
Silicon wafer with several \ ^x^LlsXVs. ^ RF Filter
etched waveguides and
devices
RFin
Fig.5 A schematic view of an integrated waveguide circuit. Several waveguide
components can be produced on a single wafer. Active devices and planar
circuits can be integrated directly on thin membranes spanning the
waveguide. Micromechanical rf tuning elements can also be included in the
waveguide.
Page 322 Third International Symposium on Space Terahertz Technology
Currently, we are fabricating WR-10 waveguides with SiN membranes in
between the two half pieces of the waveguide. Metalization of the half sections of
these waveguides will require selective plating of the silicon walls without plating
the silicon nitride membranes. Tungsten substitution of silicon in an LPCVD
environment [5] is proposed to meet this need. In this process, tungsten hexafluoride
(WFg) gas attacks silicon surface and a thin layer of tungsten atoms substitute for
silicon atoms on the surface. Further metalization can be done by electroplating the
tungsten surface with gold.
V . Summary
We have demonstrated a new approach in fabricating waveguide' circuits using
silicon micromachining technology. In particular, we have fabricated a 100 GHz silicon
rectangular waveguide. The insertion loss of 0.05 dB/X is comparable to a
commercially available metal waveguide. As we improve our plated gold quality, we
expect to improve the insertion loss. We have also proposed a new approach of
integrating active/passive devices and micromechanical rf tuning elements with
waveguide.
This work is supported in part by California Institute of Technology President's Fund
under grant PF-347 and the Jet Propulsion Laboratory, California Institute of
Technology, under contract with the National Aeronautics and Space Administration
and the Innovative Science and Technology Office of the Strategic Defense Initiative
Organization.
Third International Symposium on Space Terahertz Technology Page 323
References
[1]. L.-S. Fan, Y.-C. Tai, and R. S. Muller, "Integrated Movable MicromechanicaJ
Structures for Sensors and Actuators," IEEE Trans, on Electron Devices, vol.
35, pp. 724-730, June 1988.
[2]. M. Mehregany, K. J. Gabriel, and W. S. N. Trimmer, "Integrated Fabrication of
Polysilicon Mechanisms," IEEE Trans, on Electron Devices, vol. 35, pp.
719-730, June 1988.
[3]. F. J. Tischer, "Experimental Attenuation of Rectangular Waveguides at
Millimeter Wavelengths," IEEE Trans. Microwave Theory and Tech., vol.
MTT-27, pp. 31-37, January 1979.
[4]. V. M. Lubecke, W. R. McGrath, and D. B. Rutledge, " Sliding Backshorts for
Planar Circuits," International Journal of Infrared and Millimeter Waves, vol.
12, pp. 1387-1397, December 1991.
[5]. N. Kobayashi, M. Suzuki, and M. Saitou, "Tungsten Plug Technology Using
Substitution of W for Si," IEEE Trans, on Electron Devices, vol. 37, pp.
577-582, March 1990.
Page 324 Third International Symposium on Space Terahertz Technology
This is intended as a review paper only and it summarizes work which has been submitted for
publication in the IEEE Transactions on MTT .
-Z&^Z^ N93-27755
\\ PROGRESS IN INTEGRATED- CIRCUIT HORN
ANTENNAS FOR RECEIVER APPLICATIONS
PART 1: Antenna Design
George V. Eleftheriades, Walid Y. Ali- Ahmad, and Gabriel M. Rebeiz
NASA/Center for Space Terahertz Technology
Electrical Engineering and Computer Science Department
University of Michigan
Ann Arbor, MI 48109-2122
ABSTRACT
The purpose of this work is to present a systematic method for the design of multimode quasi-
integrated horn antennas. The design methodology is based on the Gaussian beam approach and
the structures are optimized for achieving maximum fundamental Gaussian coupling efficiency.
For this purpose, a hybrid technique is employed in which the integrated part of the antennas
is treated using full-wave analysis, whereas the machined part is treated using an approximate
method. This results in a simple and efficient design process. The developed design procedure
has been applied for the design of a 20dB, a 23dB and a 25dB quasi-integrated horn antennas,
all with a Gaussian coupling efficieny exceeding 97%. The designed antennas have been tested
and characterized using both full-wave analysis and 90GHz/370GHz measurements.
Third International Symposium on Space Terahertz Technology Page 325
I. QUASI-INTEGRATED HORN ANTENNA DESIGN : INTRODUCTION
The integrated-circuit horn antenna was introduced in [1] and analyzed using a full-wave analysis
technique in [2]. It consists of a dipole (or monopole) feed evaporated on a thin dielectric
membrane which is suspended in a pyramidal cavity etched in silicon or GaAs. Recently, this
antenna has been used in several millimeter and submillimeter-wave applications including a
double-polarized antenna design at 93GHz [4], a 256 element imaging array at 802GHz [5], and
a monopulse tracking system at 94GHz [6]. However, the wide flare-angle of the integrated-
circuit horn antenna, which is dictated by the anisotropic etching involved in its fabrication
(70° in silicon), limits its useful aperture size to 1.6A and its gain to 13dB. To this end the
quasi-integrated horn antenna was introduced [3], which consists of a machined small flare-
angle pyramidal section attached to the integrated portion (fig.l). The resulting structure is a
simple multimode pyramidal horn with circularly symmetric patterns, high gain, and low cross-
polarization, which is particularly attractive for submillimeter quasi-optical receiver applications.
The minimum machined dimension involved in its fabrication is around 1.5 A which enables its
fabrication to frequencies up to 2THz. The purpose of this paper is to describe a systematic
approach towards the design of these horn antennas, and to provide a full range of practical
quasi-integrated horn antenna designs along with their detailed radiation characteristics. Since a
very desirable property of antennas intended for use in quasi-optical systems is the high Gaussian
content of their radiated fields [7], the developed design methodology is based on the optimization
of the quasi-integrated horns for achieving maximum fundamental Gaussian coupling efficiency.
The Gaussian coupling efficiency is particularly important in quasi-optical receiver applications
because it directly influences the total system performance with a pronounced effect on the
receiver noise temperature [8].
II. MULTIMODE APERTURE ANALYSIS FOR MAXIMUM FUNDAMENTAL
COUPLING EFFICIENCY
Page 326
Third International Symposium on Space Terahertz Technology
Integrated on Si section
Machined gain and phasing section
Fig.l The general configuration of the quasi-integrated multimode horn antenna.
1.1
1.0
0.9
{=■
0.8
o
0.7
c
.2-
0.6
t£
4)
0.3
00
c
0.4
3
0.3
O
U
0.2
0.1
0.0
iitijiiffftrtiittiiiiifft r't i 'i i i t t r i i i t t I i
TTTTTTT
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
w^a
Fig. 2 The maximum Gaussian coupling efficiency as a function of the
w / a ratio for various aperture modes available for beamshaping (up
to TE m ,n/TM m ,n , m = 1, 3 . . . M, n = 0, 2 . . . N) .
Third International Symposium on Space Terahertz Technology Page 327
Consider a square aperture of side a in a ground-plane which is radiating in the half-space
z > 0. The transverse electric field of the aperture at z = can be expanded in terms of the
eigenfunctions of a square waveguide of the same side a :
M,N
Et„(x,y)= £{A mn e™(x,y) + C mn e™(x,y)} . (1)
m,n
In (1) it is assumed that only modes with indices (m = 1, 3, 5 ... A/ and n = 0, 2, 4, 6 ... N) are
present as is the case of a pyramidal horn which is either fed by a centered Hertzian dipole or by
a waveguide which supports only the dominant TEio mode [2]. We now proceed to determine the
modal coefficients A mn ,B mn so that the coupling between the aperture field and a fundamental
beam is maximized. If the copolarized and cross-polarized components of the aperture field are
defined to be the E x , ap and the E yap components respectively, then the transverse electric field
can be rewritten in the form :
E y ,a V {x, y) = £ <£„¥£„(*, y) , £ r , op (*, y) = £ «#„*;*(*, y) (2)
171,71 171,71
where the orthonormalized copolarized and cross-polarized hybrid modes ^f^m ^mn are:
▼ /.« / v V^^n , v m + n — 1 TY17T X > , TT7T1J . ,, *— i i .. *«. ,«.*
*mn(^y) = ^—^(-l ^- cos( )cos(— £ , \x\ < a/2, \y\ < a/2 (3)
a a a
**" n (x, y) = 2L_i _i)-V^ S in ) sin(— y -) , |x| < a/2, |y| < a/2 (4)
a a a
In (3) and (4) the origin of the Cartesian coordinates is located at the geometrical center of
the aperture and e n = 2 — S n0 is the Neumann number. The corresponding copolarized and
cross-polarized modal coefficients of (3-4) are related to the modal coefficients of (1) through:
,co _ n ^mn ~ rnAmn p _ TlA mn -\- TnL> mn
Vm 2 + n 2 V"i + n l
Now the coupling efficiency r](w ) of the aperture to a fundamental Gaussian beam of waist
radius w , which has its waist on the aperture is given by [11] :
r)(w ) =
M,N
£ d- n I mn (w )
m,n
2 2 M,N
/(^E(fcr+K p ni 2 )) (6)
Page 328
Third International Symposium on Space Terahertz Technology
here, I mn (w ) = II ^ n (x,y)exp(-{x 2 + y 2 )/w 2 ) dx dy.
J Japert.
(7)
We wish at this point to determine the modal coefficients <f" n and d% n so that the coupling
efficiency r](w ) is maximized. For this purpose, the application of Schwarz's inequality to (6)
immediately implies that the maximum coupling efficiency r) max (w ) occurs in the absence of
cross-polarization and is obtained from :
M,N
with the corresponding co-polarization modal coefficients determined by
d mn I Imn(Wo) = Constant.
(8)
(9)
The condition for vanishing cross-polarization is (see 5) : nA mn = —rnC m n ■
(10)
Therefore, for maximum fundamental Gaussian coupling efficiency the aperture modes should
add in phase and their relative magnitudes should satisfy conditions (9) and (10). The maximum
coupling efficiency r) max (w ) of equation (9) still depends on the waist radius w and it is shown in
figure 2 as a function of the ratio w /a for various indices (M,N). In table 1 we show the relative
magnitudes between the modes at the optimum w , opt /a ratio, for some practically encountered
aperture sets of modes.
Available modes (M,N)
(1,0)
(1,2)
(1,2)+TE 30
(3,2)
w ,o P t/a
0.43
0.34
0.32
0.29
cpl. efficiency : r) max
84%
98.5%
99.2%
99.7%
^12/^10
-
0.51
0.56
0.64
dio/diQ
-
-
0.11
0.17
df 2 /d\Q
-
-
-
-0.11
Table 1: Optimum parameters for maximum fundamental Gaussian coupling efficiency for certain
practically encountered aperture modes available for beamshaping.
Third International Symposium on Space Terahertz Technology Page 329
III. APPROXIMATE ANALYSIS OF THE MACHINED SECTION AND
DESCRIPTION OF THE DESIGN PROCESS
Consider the gradually-flared pyramidal machined section of axial length Lm and of half flare-
angle O (see fig. 1) which is assumed excited at its throat by M x N locally propagating modes.
Since the machined section is gradually flared and the incident modes propagating, reflections at
the throat are considered negligible and the corresponding transverse electric field is given by :
E t M*,v) = £ Ml^ £ (*, v) + Ce M (x,y)} • (ID
To a first order approximation we can assume that each mode preserves its carried power upon
propagating from the throat to the aperture. Also, each mode acquires a phase shift computed
by:
*m» = I " mn {z)dz (12)
JO
where (3 mn (z) is the local propagation constant of the mn'^-mode. The above phase shift has
been used extensively for the design of multimode horns [9-10] and it can be rigorously justified
through a coupled-mode analysis of gradually flared tapers [13]. The aperture field is assumed to
be modulated by a quadratic phase factor Ql t (x, y) of curvature Lt = a/(2 tan O ) with Lt being
the total virtual length of the taper. Under the above assumptions and neglecting reflections ,
the aperture field is simply given by :
M,N
E t , ap (x,y) = Q LT (x,y)E{K P n ^ TE (^y) + C a m P n e a rn P n TM (^y)} (13)
with the quadratically modulated aperture modal coefficients related to the throat modal coef-
ficients through :
4?n = A%jY£ TE /Y eM-J*mn) , <fr B = CJ* Jy£™ /Y exp(-;$ mn ) (14)
where Y^ n is the throat admittance for the mn th mode and Y is the free-space intrinsic admit-
tance which has been assigned to the aperture modes. Based on the above simplified analysis for
Page 330 Third International Symposium on Space Terahertz Technology
the machined section and on a full-wave analysis of the integrated portion a three-stage design
process has been established and is summarized below :
1. The integrated 70° flare-angle section of the antenna structure of figure 1 (including the
step discontinuity) is selected and analyzed independently of the machined section. For
this purpose, the dipole-fed integrated portion is assumed to be terminated by an infinite
square waveguide of side (a 3 + 2s) and is analyzed using the full-wave analysis technique
of [2] to obtain the throat modal coefficients A^ n , Cj£ n . The junction cross-section a s and
the step size s (see fig. 1) are selected so that the magnitudes of the radiating aperture
modal coefficients, as predicted by equations 5 and 14, satisfy the optimal conditions (9)
and (10) as closely as possible.
2. The infinite waveguide is now replaced by the gradually flared machined section and the
assumption is made that the modal coefficients at the throat of the machined section retain
their computed values of stage 1. This is a good approximation since the actual excited
modal coefficients are determined by the difference between the integrated portion flare-
angle and the machined section flare-angle and this difference is always dominated by the
large 70° flare-angle of the integrated portion [10]. The length Lm and the flare- angle 8 of
the machined section are then selected iteratively (using 12) so that the modal coefficients
dmn appear in phase on the radiating aperture. The shortest possible length is chosen in
order to achieve the maximum bandwidth.
3. Finally, the length and the flare-angle of the machined section are "fine-tuned" using the
full-wave analysis of [2] for the entire quasi-integrated horn antenna and again for achieving
maximum Gaussian coupling efficiency.
In table 2 we quantify several practical geometries of integrated portions which have resulted
from the first stage of the design process.
Third International Symposium on Space Terahertz Technology Page 331
IV. NUMERICAL AND EXPERIMENTAL RESULTS FOR SPECIFIC
QUASI-INTEGRATED HORN ANTENNA DESIGNS.
The algorithm of section III has been employed for the design of a 20dB, a 23dB and a 25dB
quasi-integrated horn antenna, all with a fundamental Gaussian coupling efficiency exceeding
97% and with a full-null beam efficiency around 99%. Although, in the design process the
analysis of the machined section is performed using the approximate method of section III, the
computation of the radiation characteristics of the finally designed horns is carried out using the
full-wave analysis technique of [2]. Furthermore, using this full-wave analysis along with 6GHz
scale-model measurements it was verified that the input impedance of the feeding strip-dipole in
the integrated portion of the horn is not affected by the attachment of the machined section [3].
This is due to the fact that the input impedance of the feeding strip is mainly determined by
its local geometrical environment which remains unaffected by the attachment of the machined
section. The input impedance for the integrated-circuit horn antennas has already been analyzed
theoretically and characterized experimentally in [2] where it was shown that by adjusting the
dipole position inside the horn, the input impedance can be matched to either Schottky or SIS
diodes. Therefore, the results of [2] are directly applicable to the case of the quasi-integrated
horn antennas as well.
A. 20dB quasi-integrated horn antenna.
The geometrical parameters for the 20dB realization are calculated to be (a, = 1.35A,s =
0.0, Lm = 7A, 9 = 9°, dp = 0.39A) and the numerically computed patterns from the third stage
of the design process along with the corresponding 90GHz measurements have been reported
in [3]. In fig. 3 the principal patterns are compared to the patterns obtained by analyzing
the machined section using the approximate method of section III. As shown, the approximate
model agrees well with both the full-wave analysis and the measurements thus verifying the
approximations used in the design process. The main radiation characteristics of this horn at
Page 332 Third International Symposium on Space Terahertz Technology
the design frequency and at the edges of the ±5% bandwidth are summarized in table 3. The
indicated 10-dB beamwidth fluctuation corresponds to the variation of the beamwidth in an
azimuthal far-field cut. The Gaussian-beam rolloff was calculated at the edges of the ±5%
bandwidth using the Gaussian-beam parameters which were calculated at the design frequency
f . The calculated phase center was found to be located at a distance of 1.5A from the horn
aperture for the E-plane and at 1.4A for the H-plane.
B. 23dB quasi-integrated horn antenna.
The optimized design parameters for a 23dB quasi-integrated horn are found to be (a, =
1.52A.5 = 0.17A, Lm = 13A,0 O = 8.5°, dp = 0.39A) and the computed principal patterns from
both the full- wave analysis of the entire antenna and from the approximate model of section III
are compared in figure 4 to corresponding 370GHz measurements. In figure 5 we include also
the computed from the full-wave analysis and the measured patterns for the 45°-plane. The
radiation characteristics of this horn are being summarized in table 4. For the 23dB horn the
phase center was calculated to be at 3.7A inside the horn for the E-plane and at 3.5A for the
H-plane.
C. 25dB Quasi-integrated horn antenna.
In order to evaluate the efficiency of the design process and to provide a full range of practical
designs, a 25dB quasi-integrated horn has also been designed and the computed geometrical
parameters are found to be: (a, = 1.52A,3 = 0.0A, Lm = 19.5A,0 O = 10°, dp = 0.39A). The
radiation patterns, as calculated from the full-wave analysis and shown in figure 6 still exhibit
excellent circular symmetry, low cross-polarization and suppressed sidelobes. The location of the '
phase center for this horn was computed to be at a distance of 13A from the aperture for the
E-plane and at 11 A for the H-plane. The rest of the main radiation characteristics of this horn
antenna are being tabulated in table 5.
Third International Symposium on Space Terahertz Technology
Page 333
Optimum
a s = 1.35A
5 = 0.0
a s = 1.52A
5 = 0.0
a s = 1.35A
s = 0.17A
a s = 1.57A
s = 0.0
Mi°2 ap l/Ko p l
0.56*
0.52
0.50
0.55
0.51
i«i/ra
0.114
-
0.11
0.117
0.146
arg^/A'S)
180°
200°
183°
182°
179°
\C%\/\A a &
2
4.5
4.4
5.1
4.3
Table 2: Comparison between the optimum aperture modal coefficients and the modal coefficients
launched at the aperture by four practical integrated portion sections. The exciting dipole is
positioned at a distance of 0.39A from the apex of the horn. * The optimum ratio |^n' ap |/l^iol
is 0.51 for the a„ = 1.35A geometry which only triggers the TEi ,TEi 2 /TMi 2 modes.
0.95/ o
/•
1.05/o
Gain
19.4dB
20dB
20.6dB
Aperture efficiency
60.6%
62.8%
65.4%
lOdB Beamwidth
37° ± 1°
34° ±1.2°
32° ±1.8°
Sidelobe-level (E-plane)
-23dB
-27dB
-26.3dB
Cross-pol.(45°)
-22.5dB
-22.7dB
-23dB
Beam-efficiency (to -lOdB)
85%
86%
86.5%
Gaussian Coupling
96.4%
97.3%
96.9%
Gaussian Coupling rolloff
95.5%
97.3%
96.5%
Table 3: The main radiation characteristics of the 20dB quasi-integrated horn antenna (see text).
0.965/ o
/•
1.035/„
Gain
22.2dB
22.8dB
23.6dB
Aperture efficiency
48.5%
52%
58.4%
lOdB Beamwidth
27.6 ±0.2°
25° ±1.1°
22.5° ±1.3°
Sidelobe-level (E-plane)
-28dB
-33dB
-29.8dB
Cross-pol.(45°)
-20.5dB
-21dB
-22dB
Beam-efficiency (to -lOdB)
86.6%
86%
86.6%
Gaussian Coupling
97.2%
97.3%
96.8%
Gaussian Coupling rolloff
96.3%
97.3%
96.0%
Table 4: The main radiation characteristics of the 23dB quasi-integrated horn antenna (see text).
Page 334
Third International Symposium on Space Terahertz Technology
-5
-10
ffl
T3
-15
.s
cd
04)
-20
<U
>
•a
«j
-?,5
<D
od
-30
-35
-40
vnm i ii I iiht i ii i m i n i ii
l
FULL-WAVE
90GHz-MEAS.
APPR. MODL.
'■■'■'
-90
-60
-30
T-rT-m-r-
urii r-r r r
30
60
Elavation angle (deg)
90
Fig.3 The E (right) and H-plane (left) patterns of the 20-dB quasi-
integrated horn. The 90GHz measured patterns are compared to the
full-wave analysis and the approximate analysis patterns. Detailed pat-
terns including cross-polarization are shown in [3].
o
u
«
a
o
S
3
in
o
a.
5
•2
T3
00
V
>
■a
OS
-5 -
■10
•15
-20 -
-25
-30
-35
'"• ' 7
^ ' '
.
\
'
H-PLANE f
\ E-PLANE
.
I
T- II « - 1 /
■
a 370GHzmeasur. P
l
Approx. model /
|
-
\ -
-
P
1 -
.
I
*~
j
-
I
?
1
1
1 n'll
;A,.fl.q.^i
-90
-60
-30
30
60
Elavation angle (deg)
90
Fig.4 The E (right) and H-plane (left) patterns of the
23-dB quasi-integrated horn. The 370GHz measured pat-
terns are compared to the full-wave analysis and the ap-
proximate analysis patterns.
PQ
CO
fcfl
cu
>
cu
-5
10
■15
E-plane(exp.)
- H-j>lane(exp.)
1_ '_*_«_*_" 45 -plancfexp. ; i
2 -20
-25
-30
-35
itfi i i i 1 1 1 m i rr 1 1 1 1 1 1 1 1 ii 1 1 i /m 1 1 ii 1 1 1 1 1 1 1 i 1 1 1 1 1 1 1 1 1 1 1 1 m
1 1 1 1 1 1 i 1 1 1 Ai i i/i i
I
_ 45°-plane(exp.).
_ 45°-plane(th.)
j 1 1 1 i/i 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ii i 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
-90 -60 -30 30 60 90
Elevation angle (degrees)
Fig.5 The measured at 370GHz E/H and 45°-plane pat-
terns vs. the full- wave patterns of the 23-dB quasi-
integrated horn.
Page 336
Third International Symposium on Space Terahertz Technology
0.965/ o
fo
1.035/ o
Gain
24.7dB
25.5dB
26.2dB
Aperture efficiency
36%
40%
44%
lOdB Beamwidth
21.6 ±0.8°
19.2° ±0.7°
17.5° ±0.5°
Sidelobe-level (E-plane)
-28.7dB
-30.8dB
-30.SdB
Cross-pol.(45°)
-22.6dB
-24dB
-24.7dB
Beam-efficiency (to -lOdB)
84.5%
85%
85%
Gaussian Coupling
97.1%
97.5%
97.4%
Gaussian Coupling rolloff
96.5%
97.5%
97.1%
Table 5: The main radiation characteristics of the 25dB quasi-integrated horn antenna (see text).
CQ
c
'«
4)
>
I T T T' f tllfffrT 1 ^! lltfTfllllll II I JW I
-10
-15
^ -20
-25 -
-30
-35
«*«**» ■ ■ * I ■ ■ 'i 'i ■ '■ ' • ' ■ ' ' ■ '* • • • ' * ■
-90
-60
-30
30
60
90
Elevation angle (degrees)
Figure 6: The calculated from the full- wave analysis patterns of the 25-dB quasi-integrated horn.
Third International Symposium on Space Terahertz Technology
Page 337
References
[1] G.M Rebeiz, D.P. Kasilingam, P.A. Stimson, Y. Guo and D.B. Rutledge, "Monolithic
millimeter- wave two-dimensional horn imaging arrays," IEEE Trans. Antennas Propagat..
vol. AP-28, pp. 1473-1482, Sept 1990.
[2] G.V. Eleftheriades, W.Y. Ali-Ahmad, L.P.B. Katehi, and G.M. Rebeiz, "Millimeter-wave
integrated-horn antennas Part I-Theory, and Part II- Experiment," IEEE Trans. Antennas
Propagat, vol. AP-39, pp. 1575-1586, Nov. 1991.
[3] G.V. Eleftheriades, W.Y. Ali-Ahmad, and G.M. Rebeiz, "A 20-dB Quasi-integrated horn
antenna," IEEE Microwave and Guided Wave Letters, vol. 2, pp. 73-75, Feb. 1992.
[4] W.Y. Ali-Ahmad and G.M. Rebeiz, "92GHz dual-polarized integrated horn antennas," IEEE
Trans. Antennas Propagat., vol. AP-39, pp. 820-825, June 1991.
[5] W.Y. Ali-Ahmad, Gabriel M. Rebeiz, Hermant Dave, and Gordon Chin "802 GHz Integrated
horn antennas imaging array," International Journal of Infrared and Millimeter Waves, vol.
12, No. 5.1991.
[6] C.C Ling and G.M Rebeiz, "94GHz Integrated monopulse antenna," IEEE AP-S Interna-
tional Symposium , Ontario, Canada, June 1991.
[7] P.F. Goldsmith, "Quasi-optical techniques at millimeter and submillimeter wavelengths," in
Infrared and Millimeter Waves, vol. 6, New York : Academic, 1982, pp. 277-243.
[8] E. N. Grossman, "The coupling of submillimeter corner-cube antennas to Gaussian beams,"
Infrared Phys., vol. 29, pp. 875-885, 1989.
[9] P.D. Potter, "A new horn antenna with suppressed sidelobes and equal beamwidths," Mi-
crowave J., vol. VI, pp. 71-78, June 1963.
[10] S.B. Cohn, "Flare-angle changes in a horn as a means of pattern control," Microwave Jour-
nal., vol. 13, pp. 41-46, Oct. 1970.
[11] G.V. Eleftheriades, and G.M. Rebeiz, "High-gain step-profiled integrated diagonal horn-
antennas," To appear in IEEE Trans. Microwave Theory Tech., mini special issue on Space
Terahertz Technology, May 1992.
[12] C. E. Profera, "Complex radiation patterns of dual mode pyramidal horns," IEEE Trans.
Antennas Propagat, vol. AP-25, pp. 436-438, May 1977.
[13] L. Solymar, "Spurious mode generation in nonuniform waveguide," IRE Trans. Microwave
Theory Tech., vol. MTT-7, pp. 379-383, 1959.
Page 338 Third International Symposium on Space Terahertz Technology
N93-2775 6
/
PROGRESS IN INTEGRATED- CIRCUIT HORN
ANTENNAS FOR RECEIVER APPLICATIONS
Part II: A 90 GHz Quasi-Integrated Horn Antenna Receiver
Walid Y. Ali- Ahmad, George V. Eleftheriades and Gabriel M. Rebeiz
NASA/Center for Space Terahertz Technology
Electrical Engineering and Computer Science Department
University of Michigan
Ann Arbor, MI 48109-2122
ABSTRACT
A receiver belonging to the family of integrated planar receivers has been developed at
90 GHz. It consists of a planar Schottky-diode placed at the feed of a dipole-probe suspended
inside an integrated horn antenna. The measured planar mixer single-sideband conversion
loss at 91.2 GHz (LO) with a 200 MHz IF frequency is S.3dB±0.3dB. The low cost of
fabrication and simplicity of this design makes it ideal for millimeter and submillimeter-
wave receivers.
\
Third International Symposium on Space Terahertz Technology Page 339
INTRODUCTION
Fundamental mixers are currently the front-ends components for all millimeter- wave receivers
above 100 GHz. The mixers use a Schottky-diode suspended in a machined waveguide with
an appropriate RF matching network. These components are expensive to manufacture
especially above 200 GHz where waveguide tolerances become severe. A low noise planar
receiver consisting of a planar Schottky diode integrated with an efficient planar antenna
is a needed alternative at millimeter- wave frequencies. Recent advances in planar Schottky
diodes resulted in excellence performance at 94 GHz with measured diode temperatures
competitive with whisker-contacted diodes [1]. In this work, a planar diode is combined
with an integrated horn antenna [2,3] to yield a 90 GHz receiver. The antenna feed-dipole
impedance can be designed to conjugate match the RF diode impedance [4]. This eliminates
the need for an RF matching network and thereby simplifies the mixer design. A machined
section is attached to the front of the integrated horn antenna to yield a multi-mode horn [5].
The planar configuration results in an inexpensive quasi-monolithic receiver with an expected
performance as good as the best waveguide receiver at 100 GHz.
MIXER DESIGN AND THEORETICAL PERFORMANCE
The length of the feed-dipole and its position inside the integrated horn antenna are designed
so that its impedance conjugate matches the RF diode impedance [4]. As a result, the planar
diode is epoxied right at the dipole apex. An RF choke is obtained by using two integrated
lumped capacitors on a coplanar stripline. The first capacitor is A /4 away from the dipole
feed and the second capacitor is A^/2 away from the first one. These capacitors introduce an
RF open circuit at the dipole feed and let the IF signal pass through the coplanar stripline
(Fig.l). The circuit is integrated on highly resistive Silicon in order to minimize any losses of
the IF signal on the surrounding dielectric substrate. A microstrip quarter-wave transformer
over a Duroid 5870 substrate [7] is used to match the 1.4 GHz IF diode output impedance
to 50J1. Fig. 2 shows the structure of the integrated horn antenna receiver. The machined
section, not shown in this figure, is attached to the front aperture of the horn antenna. Gold
Page 340
Third International Symposium on Space Terahertz Technology
is evaporated on all the horn walls except on the membrane wafer walls, in order not to
short-out the feed lines. The diode of choice to be used in this design is the UVa SC2R4
planar Schottky diode with 2.5/zm anode diameter, a 5-6fF zero- bias junction capacitance, a
12-13fF parasitic capacitance and a 5-6fi series resistance. A microwave model of the horn
receiver structure shown in Fig. 2 was built at 2.55 GHz in order to find the right feed-dipole
impedance to conjugate match the UVa diode RF impedance. A feed-dipole, which is 0.392A
long and positioned 0.38A from the apex of the horn, has an input impedance of 75+J55
fi with the membrane walls uncoated and with no diode chip modeled at the dipole feeds.
The input impedance dropped to 70+jlO fi due to the capacitive effect of the diode block
when it was modeled. A LldB-1.3dB power loss was found in the microwave model by
measuring the difference in powers detected by the feed- dipole for the case of coated and
uncoated membrane walls respectively. Table I shows the mixer theoretical performance for
the UVa diode at 91.2GHz(LO) and 91.4GHz(RF) for a bias of 0.65V and an available LO
power of 2dBm. The analysis was done using the reflection algorithm [6]. the variation in
conversion loss over 10% bandwidth is due to the variation in the feed dipole impedance.
Table I
f /F (GHz)
0.2
fftF(GHz)
91.4
Zd,p /e,fiF(^)
70+jlO
Zd,po/e,2HF(^)
14+jlO
^diode.RFK^)
62-J19
^diodcLoify
55-J49
% diode, IF (fy
86
Diode SC2R4 SSB Conversion loss(dB)
5.7
Diode SSB Conversion loss(dB) over 10% BW
5.7-6.2
Third International Symposium on Space Terahertz Technology
Page 341
to IF
Amplifier
IF Matching network
Figure 1: The mixer design consisting of the diode epoxied at the dipole feeds, the two
lumped capacitors forming the RF choke, and the microstrip line IF matching network.
35X
Front
wafers
IF Matching Network
(microstrip)
Back
wafers
&$%&>*'&.
MembraiiL i ' v '^* , *v ! ',& r v
11 u ' yy.wgw ' «
770nm
\
Figure 2: The integrated horn antenna receiver structure. The horn walls of the membrane
wafer are not coated with gold.
Page 342
Third International Symposium on Space Terahertz Technology
RECEIVER MEASUREMENTS
A quasi-integrated horn antenna receiver was built at 91.4 GHz with a UVa SC2R4 diode
epoxied at the dipole feeds. Video detection measurements were done at 91.4 GHz by shining
a known plane wave power density onto the multi-mode antenna and measuring the output
detected diode voltage using a lock-in amplifier. The diode theoretical video responsivity
vs. bias current is fitted to the measured data by using the parameters shown in table II to
model the receiver.
Table II
^aperture
e lossinwalls
^dipole
R*
C J
<^bi
c P
7?
-2.0dB
-1.2dB
70+jlO ft
6ft
5.5 fF
0.88 V
12.5 fF
1.14
In fig. 3, the measured video responsivity is equal to the ratio of the detected voltage across
the diode over the plane wave power incident on the aperture of the quasi-integrated horn
antenna. The diode parameters used in the model are those provided by University of
Virginia. Although the receiver was designed for a 1.4 GHz IF frequency, we found that
epoxy and solder at the junction between the duroid and the silicon substrate have added
a parasitic IF capacitance. The measurements were therefore done at 200 MHz where this
capacitance has negligeable effect. For the SSB conversion loss measurement, a calibrated
91.4 GHz RF plane wave and a 91.2 GHz LO were combined using a thin Mylar sheet and
shined on the receiver. Figure 4 shows the measured planar mixer SSB conversion loss,
defined as the received IF power divided by the RF power absorbed by the horn aperture
(plane wave power density x horn area x horn aperture efficiency). The SSB conversion
loss includes the 1.2dB loss in the uncoated membrane walls. An S.3dB SSB conversion
loss is measured at 91.4 GHz with 3.5dBm estimated LO power available at the feed-dipole
terminals. The coupling efficiency of the horn aperture to a plane wave is normalized out of
the measurement because in a receiver system the horn has a gaussian coupling efficiency of
97%. Also, the measured result can be directly compared to waveguide mixers performance
which have no antennas attached. The 8.3 dB SSB conversion loss compares favorably with
the best waveguide mixers performance (5.3±0.5dB) using the same diode [1].
Third International Symposium on Space Terahertz Technology
Page 343
10" 10"° 10" 10 -* 10
Bias Current, Amps
Figure 3: Measured and theoretical video responsivity at 91.4GHz.
25
*& 20
(n
O
15
o
£ io
c
o
PQ 5
in
CO
i i i i i i i i i i i i i i i i i i i i i i i i i i i i | i i i i
-^ SSB (Exp.)
Q I I I I l I l l l I I I I I I I I l l l I I I I I I l i i i I i i i l
-25 -20 -15 -10 -5 5 10
Estimated LO Power Available
at Feed-Dipole Terminals (dBm)
Figure 4: Measured planar mixer SSB conversion loss for the SC2R4 diode at 91.2 GHz
(LO). The measured values include a 1.2dB loss attributed to power loss in the horn walls.
Page 344 • Third, International Symposium on Space Terahertz Technology
CONCLUSION
A 90GHz quasi-integrated horn antenna receiver has been designed and tested. The measure-
ments show that this new receiver is a very good candidate for millimeter-wave applications.
DSB measurements are being done on a new improved receiver design and using the UVa
SC2T3 diode which has lower parasitic capacitance and series resistance than the UVa SC2R4
diode.
ACKNOWLEDGMENTS
This work was supported by the NASA/Center for Space Terahertz Technology at the Univer-
sity of Michigan, Ann Arbor. We thank G.V. Eleftheriades for the multi-mode horn design.
We thank Dr. Thomas W. Crowe and William L. Bishop at the University of Virginia, for
providing us with the diodes.
REFERENCES
[1] D.G. Garfield, R.J. Mattauch, and S. Weinreb,"RF Performance of a Novel Planar
Millimeter- Wave Diode Incorporating an Etched Surface Channel," Trans Microwave
Theory Tech., vol MTT-39, pp. 1-5, Jan 1991.
[2] G.M. Rebeiz, D.P. Kasilingan, P.A. Stimson, Y. Guo, and D.B. Rutledge, "Monolithic
millimeter-wave two-dimensional horn imaging arrays," IEEE Trans. Antennas Propag.,
vol. AP-28, Sept. 1990.
[3] W.Y. Ali-Ahmad, and G.M. Rebeiz, "92 GHz dual-polarized integrated horn antennas,"
IEEE Trans. Antennas Propag., vol. AP-39, June 1991.
[4] W.Y. Ali-Ahmad, G.V. Eleftheriades, L.P. Katehi, and G.M. Rebeiz,"Millimeter-Wave
Integrated Horn Antennas, Part II: Experiment," IEEE- Trans. Antennas Propagation,
vol. AP-39, pp. 1582-1587, Nov. 1991.
[5] G.V. Eleftheriades, W.Y. Ali-Ahmad, and G.M. Rebeiz, "A 20dB Quasi- Integrated Horn
Antenna," IEEE- Microwave Guided-Wave Lett., vol. 2, pp. 72-75, Feb. 1992.
[6] D.N. Held and A.R. Kerr, "Conversion loss and noise of microwave and millimeter-wave
receivers: Part I-Theory; Part II- Experiment," IEEE Trans. Microwave Theory Tech.,
vol. MTT-26, p.49-61, 197S.
[7] Duroid is a trademark of Rogers Corporation. We thank Rogers Co. for the donation of
the substrate.
Third International Symposium on Space Terahertz Technology Page 345
ZONE PLATE LENS ANTENNAS FOR /4> oSy£~
P /?
MILLIMETER AND SUBMILLIMETER WAVELENGTHS / '
Paul F. Goldsmith
Five College Radio Astronomy Observatory
Department of Physics and Astronomy
University of Massachusetts, Amherst MA 01003
and
Millitech Corporation
P.O. Box 109
South Deerfield, Massachusetts 01373
Abstract
Zone plate lenses are a type of focusing element which function essentially as
differential phase shifters, having a relatively few, coarsely quantized phase delays across
the incident beam of radiation. The major advantages are ease of fabrication and much
reduced thickness, compared to conventional refractive focusing elements. These
considerations are both of particular importance for the submillimeter range, in which
manufacturing tolerances for curved optical elements can pose significant problems, and
where the absorption of readily available dielectric materials is quite large. In this
presentation we briefly review the theory of zone plate lens operation, present a relatively
simple method for calculating the aperture efficiency of zone plate lenses used as antennas,
and show some theoretical and measured results in the 100 GHz range.
Page 346 Third International Symposium on Space Terahertz Technology
I. Lens Operation as a Phase Transformer
An ideal lens changes the radius of curvature of an incident beam of radiation
without affecting its amplitude distribution. This situation is illustrated in Figure 1, which
shows a phase transformer converting a diverging spherical wave (as might be produced by
a feed horn) into a plane wave. In the paraxial limit, the phase variation of a spherical
wave perpendicular to an axis from its focus is a quadratic function of the radius r from the
axis:
A<p(i) = tttVAR , (1)
where A is the wavelength and R is the radius of curvature. The phase variation is defined
in the sense that the phase delay increases with increasing distance from the axis of
propagation.
A lens modifies the radius of curvature of the beam by making use of different
propagation speeds in different media; these can be dielectrics for which the speed is less
than that in free space by a factor n, and arrays of waveguides or metal plates for which
the propagation speed is greater than c according to the general relationship
v = c/[l-(A/Aco)2]0' 5 • (2)
Reduced to simplest terms, a dielectric lens modifies the phase distribution by providing a
phase delay which decreases away from its axis according to
A^i(r) = -7rr2/AF , (3)
where F is the focal length of the lens. The output phase variation is just the sum of that
of the input beam given by (1) together with that of the lens given by (3), so that we
obtain
TnrVARout = 7rr2/AR in - ttt2/AF , (4a)
which leads immediately to the relationship
1/Rout = 1/Rin - 1/F . (4b)
Third International Symposium on Space Terahertz Technology Page 347
With the convention that Ri n = Di n , the distance of the focal point of the input beam from
the lens, and R ou t = — Dout (the distance to the focal point of the output beam), we
recover the familiar expression
1/Din + 1/Dout = 1/F • (4c)
Equation (4b) also applies to quasioptical Gaussian beams [1], while (4c), being dependent
on the assumption that the radius of curvature is equal to the distance from the focal point,
applies only to geometrical optics beams.
A converging dielectric lens of index of refraction n has a central thickness t c
determined by the maximum phase delay that is required:
Ap,nax= (27r/A)-t c -(n-l) . (5)
In the simplest approximation, we find that the central thickness of a lens of diameter D
and focal length F is
t c = D2/8(n - 1)F . (6)
II. LENS LOSS
A lens which operates perfectly as a phase transformer may still suffer loss as a
result of reflections at free space — dielectric interfaces, and absorption within the lens
itself. Reflection losses are typically a few percent per surface if no anti— reflection
treatment is employed. The basic technique to reduce reflections is to include a matching
layer, which may be a natural dielectric of the required index of refraction, or an artificial
dielectric as formed by cutting grooves. Both approaches have limitations due to
variations in incidence angle and polarization effects, but can reduce reflections
significantly.
Absorption cannot be eliminated, and depends on the lens thickness, together with
the material properties. We use the definition of the fractional power loss per unit distance
to be a, so that the input and output power (or power density) after traversing a path
through the dielectric of length L are related by
Page 348 Third International Symposium on Space Terahertz Technology
Pout = Pin-exp(-aL) , (7a)
where
a = 2* n tan 6/ A . (7b)
In the preceding equations, we have employed the usual definitions of the complex
dielectric constant c = c' + ie", the index of refraction n = ft', and the loss tangent,
tan 5= e"/e'.
At submillimeter wavelengths, information on dielectric properties is quite scarce
and often not entirely consistent. Some of this may be a result of variations in sample
properties, while measurement techniques and errors may also be playing a role. Teflon is
a low— loss dielectric widely used for millimeter and submillimeter wavelength lenses. This
material has an absorption coefficient which rises almost linearly with frequency, and is
approximately 0.042 cm"i at 300 GHz and 0.09 cm-' at 600 GHz [2]. Other measurements
give higher absorptions of 0.2 - 0.5 cnr* at a 900 GHz [3]. We adopt an absorption
coefficient of 0.1 cm* 1 at 600 GHz and a real part of the dielectric constant of 2.0.
Rexolite tm is often used at millimeter wavelengths due in part to its good mechanical
properties. It is relatively lossy in the submillimeter range; different measurements give a
= 0.70 - 1.0 cm-i at 600 GHz [4].
Taking (6) as giving a representative lens thickness, we find t c = 0.31 D 2 /F. For a
F = D = 5 cm lens at 600 GHz, we find t c = 1.55 cm; a perfectly phase correcting
plano-convex lens of the same focal length and diameter has t c = 1.475 cm. The
absorptive loss at the center of a F = D = 5 cm teflon lens (where most of the power is
concentrated) will thus be about 15 %. A comparable rexolite tm lens will have an
absorption at its center of approximately 60 %! Clearly, these numbers are large enough to
suggest the use of refractive optics. An alternative that merits serious consideration is the
zone plate lens, which can be made far thinner and thus have negligible absorption loss.
III. Zone Plate Lens Operation
The monotonically decreasing phase delay as a function of distance from the axis of
c-s-
Third International Symposium on Space Terahertz Technology Page 349
symmetry characteristic of typical dielectric lenses (Figure 2a) can be interrupted by a step
change in thickness. If at a particular frequency this produces an increase in the phase
delay equal to 2tt radians, it will nominally not have any effect on lens performance, except
for possible shadowing by step boundaries. Such devices are generally called zoned lenses,
or Fresnel lenses and are widely used to reduce the thickness of relatively large and thick
lenses employed at microwave frequencies. As shown in Figure 2b, they still have at least
one surface which has curved sections, so that manufacturing is, in fact, more difficult than
conventional lenses since the steps are an added complication.
Zone plate lenses represent a more radical approach, in that, as illustrated in Figure
2c, they are designed using only surfaces perpendicular to the axis of propagation. This is
not at all a new concept, deriving quite directly from concept of Fresnel zones in diffraction
theory. A number of references which discuss theoretical and experimental aspects of zone
plate lenses are given in [5]. The closely related zone plate reflector antenna is discussed in
references [6].
Rather than attempting to achieve the desired phase error function (3), the zone
plate lens allows the phase error to increase as a quadratic function of distance from the
axis. When it has reached a certain point, the lens thickness is reduced to bring the phase
error to zero. If we define the maximum allowed phase error to be 27r/p, the axial size of
the step is given by
t!=A/(n-l)p . (8)
If this procedure were continued indefinitely, we would merely have a stepped
approximation to a conventional lens, which would not be particularly thin. The zone
plate lens is distinguished by the technique of increasing the phase delay by 2ir radians at
the design frequency, at points where the previously described procedure of reducing the
phase delay would lead to an accumulated phase error of 2t radians. The maximum
change in thickness of the zone plate lens is
At.ax = [(p-l)/p]-(A/(n-l) , (9)
since at the next step the thickness returns to its original value rather than to
p-ti = A/(n — 1). The total zone plate lens thickness is given by At max + t m i n , where the
latter is the minimum thickness required for mechanical integrity. The total thickness is
Page 350 Third International Symposium on Space Terahertz Technology
thus on the order of a wavelength, far less than required for unzoned lenses. The
absorption loss for a material with a proportional to frequency will thus be a constant,
providing one of the most important advantages of zone plate lenses at submillimeter
wavelengths.
The radii at which the zones occur are obtained by requiring that the total phase for
rays representing a plane wave converged to focal point be constant for all initial radii from
the axis of symmetry. We ignore any phase shift of constant thickness component of lens,
and take the phase shift of the zone plate lens to be
A^ p i = -27Tk/p , (10)
where k is an index which increases by unity at each zone boundary. As illustrated in
Figure 3, the radius of zone k is denoted rk and the distance from the lens at this radius to
the focal point Rk, so that the path phase difference between axial ray and an arbitrary ray
is
Aip = (27r/A)[R k - F] . (11)
The two previous equations can be combined with the constant total phase condition to
determine Rk. With the additional approximation of ignoring effect of changes in the lens
thickness on Rk, we obtain Rk = [rk 2 + F 2 ]°« 5 , which gives the relation
r k =[2kFA/p + (kA/p)2]o.5 . (12)
Some designs for zone plate lenses are shown in Figure 4, with p = 2, 4, 10, and 50. All are
designed for 300 GHz and have F = D = 10 cm, with an index of refraction equal to 1.4. A
minimum thickness of 0.1 cm has been arbitrarily chosen. In practice, it is effective to
chose the minimum thickness to make the zone plate lens central thickness resonant at the
design frequency.
If we restrict ourselves to the situation F/D >> 0.5, the first term in (12)
dominates, and we obtain the nominal lens diameter
D = [8k max fA/p]o.5 . (13a)
Third International Symposium on Space Terahertz Technology Page 351
Alternatively, we see that the number of zones in the lens is given by
k max = pD 2/8fA . (13b)
Although zone plate lens design concentrates on phase delays and ignores refraction, an
important limitation must be borne in mind, which is that the zone width must be large
enough that zones do not begin to act like waveguides, in the manner of matching layers on
a conventional dielectric lens. For this reason it is useful to determine the minimum zone
width, which is just the minimum value of Ar = rk+i — rk- This occurs at the outer radius
of the lens where we find
Ar min = (2/p).(f/D).A . (14)
For a F/D = 1 zone plate lens with p = 4 we find Ar m in = A/2, which is on the borderline
of being a problem; it is apparent that the performance of the outer portion of fast zone
plate lenses may be compromised by this effect. Detailed calculations remain to be carried
out.
IV. Zone Plate Lens Efficiency
The major issue we wish to investigate is how the wave front errors which are a
necessary consequence of the approximate nature of the zone plate lens design affect its
efficiency as an antenna. We consider a lens being used to transform a spherical wave into
a plane wave and start with a wave of radius of curvature Ri n ; the phase distribution as a
function of distance from the axis is shown in Figure 5a; in the paraxial limit this is just
that described by (1). To this we add the differential phase shift produced by the lens
which has thickness t at radius r
A^ zp i = (27r/A)-(n-l)t(r) . (15)
The lens shown in cross section in Figure 5b has p = 4 and F/D = 1. The resulting output
or aperture phase distribution is shown in Figure 5c; note that the phase delay increases
essentially quadratically as a function of radius, except at the zone boundaries. The p = 4
lens has 3 successive zone boundaries at which the phase delay decreases by 27r/4 radians,
followed by a boundary at which the phase delay increases by 27r.
Page 352 Third International Symposium on Space Terahertz Technology
In order to calculate the aperture efficiency, we assume that we have a Gaussian
feed distribution which yields an aperture field distribution with magnitude of the form
I Eap(r) | = exp Kr/w)2] , (16a)
which defines the power edge taper
Te(dB) = 8.69 (R/w)2 = 2.17 (D/w)2 . (16b)
The taper efficiency is the efficiency with which the aperture is utilized, and is defined by
ct= |// E ap -dS|2/// |E ap |2.dS • //dS , (17a)
where all integrals extend over the aperture [7]. The spillover efficiency is the fraction of
power in the feed pattern which is intercepted by the aperture, and is given by
e s = // |E ap |2.dS / // |E ap |2.dS . (17b)
aperture entire
pat tern
The aperture efficiency is the product of the two preceding contributions:
c a = ft • fs • (17c)
The integral in the numerator of (17a) includes the effects of the phase errors; any
deviation from a uniform phase distribution reduces the taper efficiency and thus the
aperture efficiency.
The efficiencies can be calculated for different input beam characteristics for a given
lens. An edge taper of approximately 10 dB yields the maximum efficiency for this type of
illumination of an unblocked antenna [7]. The variation of taper efficiency as a function of
input beam radius of curvature is shown in Figure 6a for a zone plate lens with p = 4 and
F = D = 10 cm operating at a wavelength of 0.3 cm. As expected, the maximum efficiency
occurs for Ri n = F.
The behavior of the lens efficiency as a function of p is shown in Figure 6b, for the
same lens conditions as above, fixing Rin = F. The values of the efficiencies for large p are
Third International Symposium on Space Terahertz Technology Page 353
very close to those for a perfect phase transformer with the same 10 dB edge taper:
ft = 0.9, e s = 0.9, and e a = 0.81. The efficiencies for p = 2 and 3 are quite low, but for
p = 4 we begin to approach the asymptotic behavior. Thus, the choice of p represents a
compromise between obtaining the highest efficiency and ease of fabrication together with
the requirement on minimum zone width given by (14).
V. Zone Plate Lens Measurements
We have fabricated and measured a zone plate lens designed for operation at 95
GHz, where test equipment is readily available. The lens was fabricated of Rexolite tm (n =
1.59), with p = 4, f = 12.7 cm, D = 9.53 cm, and a central thickness of 0.59 cm. For
comparison, we used a fused silica lens which was anti— reflection coated with layers of
polyethylene. This lens had the same diameter, but a slightly different focal length of
14.5 cm. Both lenses were illuminated by a scalar feed horn giving a Gaussian illumination
pattern with an edge taper of close to 10 dB. The measured patterns in one plane are
shown in Figure 7. We see that the main lobe beamwidths are very similar. The sidelobe
structure of the unzoned lens is essentially that predicted from the truncated Gaussian
illumination. The zone plate lens shows more extended error pattern which is a
consequence of the phase errors.
The gains of the two lens antennas were also measured using a compact range; their
absolute values are compromised by uncertainty in the gain of the reference horn but are
consistent with expectations. What is more reliable is the difference in gain between the
unzoned lens and the zone plate lens, which indicate that the zone plate lens has 1.0 dB
lower gain. The calculations for the p = 4 zone plate lens predict an efficiency 0.86 dB
below that of an ideal phase transformer lens. The reflection loss of the n = 1.6 zone plate
lens is 0.23 dB per surface at normal incidence. Given the possible imperfections in
matching, phase transforming, and the absorption in the unzoned lens, the measurements
and calculations are in satisfactory agreement.
VI. Conclusions
We have reviewed the theory of operation and design of zone plate lenses. Their
very small thickness makes these devices attractive for use at submillimeter wavelengths
where absorption loss of unzoned lenses can be appreciable. We have examined the
Page 354 Third International Symposium on Space Terahertz Technology
efficiency of zone plate lenses as a function of interzone phase shift, and find that for zone
boundary phase shifts < ir/2 the performance approaches that of an ideal phase
transformer. Measurements and calculations of the efficiency of a 95 GHz lens agree quite
well.
We thank John Kapitzky, Chris Koh and Ellen Moore for their contributions to this
project.
References
[1] T.S. Chu, "Geometrical Representation of Gaussian Beam Propagation," Bell Syst.
Tech. J., 45, PP- 287-299, 1966.
[2] M.N. Afsar, "Millimeter Wave Complex Refractive Index, Complex Dielectric
Permittivity and Loss Tangent Measurements of Common Polar and Non-Polar
Polymers," Proc. Tenth International Conference on Infrared and Millimeter Waves,
pp. 60-61, 1985.
[3] A.P. Sheppard, A. McSweeney, and K.H. Breeden, "Submillimeter Wave Material
Properties and Techniques: Dielectric Constant, Loss Tangent, and Transmission
Coefficients of Some Common Materials to 2000 GHz", Proc. Symposium
Submillimeter Waves, Vol. XX in Microwave Research Institute Symposia Series.
Brooklyn: New York Polytechnic Institute Press, 1970, pp. 701-705. G.W.
Chantry, SubmiUimetre Spectroscopy. New York: Academic, 1971, p. 341.
[4] G.J. Simonis, J.P. Sattler, T.L. Worchesky, and R.P. Leavitt, "Characterization of
Near— Millimeter Wave Materials by Means of Non— Dispersive Fourier Transform
Spectroscopy," Int. J. Infrared and Millimeter Waves, 5, pp. 57-72, 1984. R.H.
Giles, A.J. Gatesman, and J. Waldman, "A Study of the Far— Infrared Optical
Properties of Rexolitet m ," Int. J. Infrared and Millimeter Waves, 11, pp. 1299-1302,
1990.
[5] M. Sussman, "Elementary Diffraction Theory of Zone Plates," Amer. J. Phys., 28,
pp. 394-398, 1960. F. Sobel, F.L. Wentworth, and J.C. Wiltse, "Quasi-Optical
Surface Waveguide and Other Components for the 100- to 300 Gc Region," IEEE
Trans. Microwave Theory Tech., MTT-9 . pp. 512-518, 1961. D. N. Black and J.C.
Third International Symposium on Space Terahertz Technology Page 355
Wiltse, "Millimeter-Wave Characteristics of Phase— Correcting Fresnel Zone
Plates," IEEE Trans. Microwave Theory Tech., MTT-35 . pp. 1123-1129, 1987.
J.E. Garrett and J.C. Wiltse, "Fresnel Zone Plate Antennas at Millimeter
Wavelengths," Int. J. Infrared and Millimeter Waves, 12, pp. 195-220, 1991. This
includes a particularly complete of references on this topic.
[6] L.F. Van Buskirk and C.E. Hendrix, "The Zone Plate as a Radio— Frequency
Focusing Element," IRE Trans. Antennas Propag., AP-9 . pp. 319-320, 1961. Yu.
N. Danilov and L.A. Fedorova, "Scattering Behavior in a Zoned Reflector
Antenna," Izvestiya VUZ. Radioelektronika, 32(2) . pp. 61-65, 1989. R. Lambley,
"Fresnel Antenna," Electronics & Wireless World, 95, P- 1642, 1989. J.M. Franke
and B.D. Leighty, "Reflection Zone Plate Antenna," NASA Tech. Brief
LAR-1S5S5, 1989. M.A. Gouker and G.S. Smith, "A Millimeter-Wave Integrated
Circuit Antenna Based on the Fresnel Zone Plate," 1991 IEEE MTT-S Digest, pp.
157-160.
[7] P.F. Goldsmith, "Radiation Patterns of Circular Aperture with Gaussian
Illumination," Int. J. Infrared and Millimeter Waves, 8, pp. 771-781, 1987.
Page 356
Third International Symposium on Space Terahertz Technology
INPUT BEAM
RADIUS OF CURVATURE
1
OUTPUT BEAM
i
PHASE
TRANSFORMER
A¥ (r)
FIGURE 1 - OPERATION OF LENS AS PHASE TRANSFORMER
FOCAL
POINT
»
IS
fe
(a) UNZONED
LENS
(b) ZONED
LENS
(c) ZONE PLATE
LENS
FIGURE 2 - DIFFERENT DESIGNS OF LENSES
Third International Symposium on Space Terahertz Technology
Page 357
FOCAL POINT
FIGURE 3 - ZONE PLATE LENS DESIGN PARAMETER DEFINITIONS
GS I I 84
Page 358
Third International Symposium on Space Terahertz Technology
P = 2
P = 4
Radius CcttO
Radius Ccm)
P = 10
P = 50
u 9.1
c
Radius CcnO
Radius CcraJ
ZONE PLATE LENSES WITH F = D = 10 CM ; FREQUENCY = 300 GHz ; n = 1.4
FIGURE 4
Third International Symposium on Space Terahertz Technology
Page 359
FIGURE 5
100 GHZ ZONE PLATE LENS
FL = 10 cm DIA = 10 cm
n = 1.59 • p = A
INPUT PHASE DISTRIBUTION
a
c
<t
a
L
X
a.
a. a l.a
LENS CROSS SECTION
Radius CcnO
OUTPUT PHASE DISTRIBUTION
m
L
O
X
Q.
a*
j*
a«
sa
ja U
xa -
2* -
14
) i i i ' | ■ ■ ■ ■ ) i ' ■ < | ' ' ' ' I ' ' ' ' 1 ' ' ' ' I ' ' ' ' I ' ' '
/W^
AAA>
i
'■■■■'
_i_
_i_
' ■ ■ ■ . '
_
' i ■ ■ ■ *
■.a a. s l.a l.s a. a a. 3 a. a a.s *■» *.3 3. a
R CcnO
Page 360
Third International Symposium on Space Terahertz Technology
>•
u
z
w
1-1
u
M
Pn
w
a:
W
51
0.8 m-r
I 1 I 1 I I | I I I I I I I I I | I I I I I I I I 1 | I I I I I I I I I | I I I I I I I I l[ I I I I I I I I I | I I I I I I I I 1 | I I 1 I I I I I I | I I I I I I I I 1 | I I I I I I I 1 I .
1 ■ " ■ ■ ' "T
9 18 11
rc
INPUT BEAM RADIUS OF CURVATURE
14 IS
o
z
w
1.08
a. 98
8.88
8.78
8.68
8. SB
8.48
8.30
i — • — r
TAPER EFFICIENCY
APERTURE EFFICIENCY
_L
■ ' ■
18
P
12 14 16 18 28
FIGURE 6 - (a) EFFICIENCY AS FUNCTION OF INPUT
BEAM RADIUS OF CURVATURE
(b) EFFICIENCY VERSUS P
Third International Symposium on Space Terahertz Technology
FIGURE 7
94GHz LENS COMPARISON
Page 361
-5
-10
CD
"□
cr
-15
LU
s
o
Q_
-20
LU
>
i— i
h-
-2b
<C
_l
LU
DC
-30
-35 —
-40
- HPBW(deg): 1.98
UNZONED
LENS
K
-10 -9-8-7-6-5-4-3-2-10 1 2 3 4 5 6 7 8 9 10
ANGLE (DEGREES)
- HPBW(deg): 1.96
-10
CD
TD
DC
-15
LU
3
O
Q_
-20
LU
>
i— 1
J—
-2b
<£
_J
LU
DC
-30
-35
-40
ZONE PLATE
LENS
-10 -9-8-7-6-5-4-3-2-10 1 2 3 4 5 6 7 8 9 10
ANGLE (DEGREES)
GS1179
Page 362 Third International Symposium on Space Terahertz Technology
ONSET OF DISPERSION IN Nb MICROSTRIP TRANSMISSION LINES AT
SUBMILLIMETER WAVE FREQUENCIES
H. H. S. Javadi, W. R. McGrath, B. Bumble, H. G. LeDuc
53^-33
n *)fo Center for Space Microelectronics Technology
f ^ Jet Propulsion Laboratory ftQ Q 7 7 CC 8
California Institute of Technology N V o ~ £ * (
Pasadena, CA 91109
^^ ABSTRACT
/ '
! We have measured the dispersion in phase velocity of a Nb-SiO x -Nb
, microstrip transmission line resonator over a frequency range from 50
GHz to 800 GHz. A submicron Nb/AI-AIO x /Nb Josephson junction was used
as a voltage-controlled oscillator to excite the high order modes in the
resonator. The same junction is used as a direct detector resulting in a
series of step-like structures in the DC current-voltage characteristic at
the position of each mode frequency. The transmission line is
dispersionless up to about 500 GHz where the phase velocity begins to
decrease. This is well below the gap frequency f g = 700 GHz. Results
agree qualitatively with the expected theoretical behavior near f g . This
onset of dispersion and loss in Nb transmission lines will have a
significant impact on the design of submillimeter wave rf circuits.
Third International Symposium on Space Terahertz Technology Page 363
Superconducting transmission lines have many important
applications in high frequency analog circuits and high bit rate digital
systems, to name a few. These transmission lines are expected to be
nearly lossless and dispersion free up to frequencies near the
superconducting energy gap frequency f g = 2A / h, where A is the
superconductor energy gap parameter and h is Planck's constant. We are
interested in superconductive microstrip transmission lines as integrated
rf tuning elements for superconductor-insulator-superconductor (SIS)
quasiparticle mixers at frequencies near 630 GHz. These SIS mixers
provide near quantum-limited sensitivity throughout the millimeter wave
band [1-3], provided they are optimized with an appropriate rf embedding
circuit. The large geometrical capacitance Cj of the tunnel junction
provides a susceptance which shunts the rf signal away from the nonlinear
quasiparticle conductance channel. An open-circuited superconductive
microstrip transmission line (stub) can provide a parallel inductance to
resonate with the junction capacitance. This approach was first used
successfully with an SIS mixer operating near 36 GHz [4] and since has
been used at frequencies up to 360 GHz [5,6]. However, above 100 GHz,
uncertainties in material, transmission line, and junction parameters
generally lead to poor performance.
Page 364 Third International Symposium on Space Terahertz Technology
The phase velocity must be accurately known in order to correctly
design the microstrip stub for a particular application. At high
frequencies possible dispersion and loss are expected to degrade the
performance. It is thus important to know the frequency range at which
these effects become significant. Other groups [7,8] have measured short-
pulse propagation on superconductive coplanar transmission lines using
optical sampling techniques and indirectly determined loss and dispersion
from a Fourier transform analysis of the pulse distortion.
We have taken a different approach, using a Josephson junction as
both a sweep-oscillator and detector to sample the resonances of a
microstrip stub. This stub is connected in parallel across the junction as
shown in fig. 1. The ac Josephson effect provides a voltage-controlled
oscillator. The frequency is 2eVyh = 0.485 x Vb (u.V) GHz where Vb is the
junction bias voltage. This monochromatic signal sets up a standing wave
in the stub. The junction detects this standing wave and a signature is
developed in the DC current-voltage (l-V) curve of the junction by a
frequency downconversion process. As the bias voltage, and thus frequency
is swept, small current steps will appear in the DC l-V curve of the
junction at frequencies corresponding to the modes of the stub. The
Third International Symposium on Space Terahertz Technology Page 365
frequency spacing of the modes is given by Af = v$/2/ where v$ is the
phase velocity and / is the physical length of the stub. Thus by measuring
the voltage interval between steps, the phase velocity can readily be
obtained if it is constant over this interval. In general, this technique can
be used at high frequencies well above the energy gap frequency f g . A
similar approach using microstrip resonators with a single resonance
below f g has recently been reported [9]. Values of surface resistance and
phase velocity were determined up to about 400 GHz. We have used long
microstrip resonators with many high order modes to determine the phase
velocity as a function of frequency up to 800 GHz, which is above the gap
frequency of niobium (Nb). The onset of dispersion has clearly been
observed.
High quality Nb / AI-AlOx / Nb tunnel junctions are fabricated using
a trilayer process [10]. The junction area, defined by electron-beam
lithography, is between 0.3u2 and 1u.2. The DC sputtered Nb films are 2000 -
3300 A thick. The current density is as high as 10 4 A/cm2 and the normal
state resistance is 50Q - 70Q. SiO x serves as both a planarization layer
for the junction trilayer and the dielectric layer for the microstrip stub.
Page 366 Third International Symposium on Space Terahertz Technology
The thickness is U » 1500A and the dielectric constant is taken to be e r =
5.5 [11]. The dielectric thickness is comparable to the magnetic
penetration depth. In this case, the phase velocity is << c and hence is
strongly affected by the field penetration into the Nb. This makes v$ a
sensitive probe of the superconductive properties of Nb.
Junctions fabricated for use as SIS mixers have stubs 2ji wide by
60u, - 70u, long to provide a fundamental broad band resonance near 630
GHz. However, to study the phase velocity, junctions were fabricated with
stubs 2u. wide by 51 8u. or 1064u. long. A longer stub gives a smaller
spacing between resonant modes, thus providing a better indication of the
phase velocity in a given frequency range. Careful examination of these
stubs however, indicated jagged edges and undercutting of the SiO x .
Additional junctions were fabricated with stubs 6u. and 12u. wide to
reduce uncertainties due to these edge effects. These stubs were either
500u,, 750{i, or 1000|i long.
The fundamental resonance frequency of a 1000u, long stub is
estimated to be about 50 GHz [4]. This leads to steps in the l-V curve
which are =0.1 mV apart. The inset of fig. 2 shows one example. The
Third International Symposium on Space Terahertz Technology Page 367
subgap conductance of this junction is very large with an almost point-
contact type of appearance. While junctions with much lower subgap
conductance also showed resonant peaks, the step structure was seen
mostly in junctions similar to that shown in fig. 2. An external magnetic
field was used to enhance the steps in different voltage ranges [12]. In
order to accurately locate these small steps in current, a small amplitude,
low frequency ( ~ 200 Hz ) ac voltage was superimposed on Vt,. The
resulting ac signal was detected with a lock-in amplifier. The interval
between two adjacent peaks is then measured to determine v$ which is
plotted as a function of the frequency of the higher step. Due to the
discreteness of the data, the resolution for changes in velocity is the
frequency spacing of the modes. Figure 2 shows the results for 4 different
microstrip stubs. About 10 to 14 resonances were observed between 50
GHz and ~ 750 GHz. A smooth curve was drawn through the resulting
velocity vs frequency data. The scatter in velocity values about this curve
is typically 5%, with only a few points deviating by ~ 10%.
Curve 1 is an example of a junction with a 2|i wide stub. Resonances
of these narrow stubs were usually not observable above about 500 GHz.
This may be due to increased scattering or losses from the edge effects
Page 368 Third International Symposium on Space Terahertz Technology
mentioned earlier. Curves 2, 3, and 4 represent stubs "lOOOu, long with
widths of 6u. or 12u.. At low frequencies, a slight curvature can been seen
in the data as expected from the mode spacing (see discussion below).
Otherwise the curves are horizontal up to about 500 GHz where they begin
to bend down indicating the onset of dispersion. The gap frequency of a
pristine Nb film is f g ~ 740 GHz at OK. Thus the dispersion begins at
frequencies well below f g . This agrees with previously reported results [8]
using an optical sampling technique. While SIS mixers are predicted to
operate well up to f g [13], the resonant embedding circuit utilizing
superconductive microstrip transmission lines is limited to lower
frequencies.
Figure 3 shows the data for a 500u. long x 12ji wide stub. In this
case, resonances, and hence the phase velocity, were obtained up to =
800GHz which is above the gap frequency of Nb. At low frequencies, the
velocity is dispersionless. Around 500 GHz, the velocity begins to
decrease and reaches its lowest value near 730 GHz. For this data point,
the frequency resolution for velocity change is about 50 GHz. This
minimum is expected theoretically to occur near 770 GHz (see fig. 3) using
a gap frequency of 660 GHz as determined from the l-V curve of this
Third International Symposium on Space Terahertz Technology p a g e 359
junction at zero magnetic field (a more complete discussion of the theory,
will be given in the future [14]). In addition, we have observed a small
suppression of the energy gap with external magnetic field in these
junctions which may describe the lower frequency of the observed phase
velocity minimum. Other loss mechanisms may also play a role. The
velocity begins to increase above 730 GHz as is expected theoretically,
since the superconductor begins to behave as a normal metal for
frequencies well above f g . The solid line in fig. 3 is the theoretical
prediction for the velocity. We have used the approach followed by Kautz
[15] which employs the Mattis-Bardeen [16] theory for the electrical
conductivity. The magnetic penetration depth was adjusted to X = 900A to
fit the theory to the low frequency asymptote of the velocity. As seen
from fig. 3, theory and experiment show the same general trend. However,
based on the Mattis-Bardeen conductivity, the theory predicts a 9%
decrease in velocity at the dip whereas the experiment shows a 35% dip.
Errors in velocity resulting from uncertainties in the mode spacings can
result from end effects on the microstrip line and the possibility of a
► negative resistance loop in the l-V curve at the position of the resonance
[17], However for our geometry, the end-effect correction to the length is
<< 1%, and the worst-case negative resistance, predicted for a lightly-
Page 370 Third International Symposium on Space Terahertz Technology
damped resonance, could only cause a 5% shift in the apparant position of
the resonance. Thus neither effect can account for the large change we
observe. At low frequencies, superconductive microstrip is a slow-wave
transmission line due to the penetration of the electromagnetic field into
the superconductor over a length comparable to the thickness of the
dielectric. For frequencies well above f g , the phase velocity increases
towards the value for a normal line. The onset of dispersion just below
the gap frequency is due to the departure of the imaginary part of the pair
conductivity from a 1/f frequency dependence [15].
Some additional insight may be gained by first considering a
Josephson junction directly coupled to a lossless microstrip stub. Using
the notation in reference [18], the resonant frequencies of the stub are
solutions to the transcendental equation
■n$± -L _ » ( -*£-) ■■ (1)
'S «S
where C s = e r eo/w/td is the total capacitance of the stub and f s is the
fundamental mode of the stub. This mode occurs when the length of the
Third International Symposium on Space Terahertz Technology Page 371
stub,/, equals one wavelength and is given by f s = v^ / 21.
In fig. 4 both sides of eqn. (1) are plotted vs (f/f s ) for an arbitrary
value of Cj/C s . At low frequencies, intersections between the straight
line ( left hand side of eqn. 1) and the tangent curves are between (m-
1/2)f s and mf s where m is an integer. These solutions move progressively
toward (m-1/2)f s for m » 1. Thus at high frequencies, the solutions of
eqn. (1) (i.e.: the modes of the stub) are equidistant. If plotted in the spirit
of fig. 2, they represent a horizontal line.
We have extended eqn. (1) to the case of a lossy transmission line,
yielding
C, f 2s.n(-^- )
-K-r- t - ^~r- < 2 >
°S 's (e 2a/ + e -2a/) + ^ ( 2*1)
T s
where a is the attenuation constant. In fig. 4, the right hand side of eqn.
Page 372 Third International Symposium on Space Terahertz Technology
(2) is plotted for three arbitrary cases of loss: a/ = 0.05, 0.10, and 0.15.
As can be seen, the distance between the solutions first decreases and
then increases as losses in the microstrip line increase. In light of this,
our experimental observations can be interpreted as evidence for
increased loss in either the superconductor or the dielectric layer. Losses
due to radiation will be negligible given the cross sectional dimensions of
the stub, 0.1 5ji x 12ji, compared to relevent free space wavelengths, X Q ~
500ji [19]. It is expected that absorption of water in evaporated SiO x
films could contribute losses at microwave and millimeter wave
frequencies. Losses due to absorbed water molecules constitute broad
peaks in the frequency domain. The attenuation of a microstrip line is
linearly proportional to frequency provided the loss in the dielectric is
frequency independent. A linear dependence is much weaker than the sharp
increase in attenuation expected near the gap frequency. Moreover, since
we observe a strong dispersion near the Nb gap frequency, the
superconductors are the most probable source for the losses. The
superconducting electrodes of the microstrip line are not perfect, defect
free, bulk crystalline materials, but are polycrystalline thin films with
fine grains and possibly surface layers and interface defects. These
Third International Symposium on Space Terahertz Technology Page 373
microstructural features of real films may be responsible, in part, for the
large dispersion we observe.
In summary, we have presented evidence for the onset of dispersion
and loss at submillimeter wave frequencies in Nb-SiO x -Nb microstrip
transmission lines. This behavior is expected at frequencies approaching
the gap frequency, but the range over which Nb microstrip lines are
dispersionless was previously not well known. These results will have a
direct impact on the application of Nb microstrip lines in millimeter wave
and submillimeter wave circuits. For operation near 1 THz, higher
temperature superconductors such as NbN or NbCN will have to be
investigated.
This work was supported in part by the Jet Propulsion Laboratory,
California Institute Of Technology, under contract to the National
Aeronautics and Space Administration and the Innovative Science and
Technology Office of the Strategic Defense Initiative Organization.
Page 374 Third International Symposium on Space Terahertz Technology
REFERENCES
1. C.A. Mears, Q. Hu, P.L Richards, A. H. Worsham, D.E. Prober, and A.V.
Raisanen, Appl. Phys. Lett. 57, 2487 (1990).
2. H.H.S. Javadi, W.R. McGrath, S.R. Cypher, B. Bumble, B.D. Hunt, and H.G.
LeDuc, Digest, 15th. Int. Conf. on IR and Millimeter Waves, p. 245, Orlando,
FL (1990).
3. B.N. Ellison, P.L Schaffer, W. Schaal, D. Vail, and R.E. Miller, Int. J. IR and
mm Waves 10, 937 (1989).
4. A.V. Raisanen, W.R. McGrath, P.L. Richards, F.L. Lloyd, IEEE Trans. Microwave
Theory Techn. MTT-33, 1495 (1985).
5. Q. Hu, C.A. Mears, P.L. Richards, and F.L. Lloyd, IEEE Trans. Magn. 25, 1380
(1989).
6. W.R. McGrath, J.A. Stern, H.H.S. Javadi, S.R. Cypher, B.D. Hunt, H.G. LeDuc,
IEEE Trans. Magn. 27, 2650 (1991).
7. M.C. Nuss and K.W. Goossen, IEEE J. Quantum Electronics 25, 2596 (1989).
8. C.C. Chi, W.J. Gallagher, I.N. Duling III, D. Grischkowsky, N.J. Halas, M.B.
Ketchen, and A.W. Kleinsasser, IEEE Trans. Magn. MAG-23, 1666 (1987).
9. B. Bi, K. Wan, W. Zhang, S. Han, J.E. Lukens, IEEE Trans. Appl.
Superconductivity 1, 145 (1991).
Third International Symposium on Space Terahertz Technology Page 375
0. H.G. LeDuc, B. Bumble, S.R. Cypher, and J.A. Stern, submitted to 3rd.
International Symposium on Space Terahertz Technology, University of
Michigan, Ann Arbor, March 1992.
1. H.K. Olsson, IEEE Trans. Magn. 25, 1115 (1989).
2. I.O. Kulik, JETP Lett. 2, 84, 1965.
3. M.J. Feldman, Int. J. IR and mm Waves 8, 1287 (1987).
4. H.H.S. Javadi and W.R. McGrath, to be published.
5. R.L Kautz, J. Appl. Phys. 49, 308 (1978).
6. D.C. Mattis and J. Bardeen, Phys. Rev. 111, 412 (1958).
7. D.B. Tuckerman and J.H. Mageriein, Appl. Phys. Lett. 37, 241 (1980).
8. H.D. Jensen, A. Larsen, J. Mygind, IEEE Trans. Magn. 27, 3355 (1991).
9. T.C. Edwards, Foundations for Microstrip Circuit Design , John Wiley and
Sons, New York (1981).
Page 376 Third International Symposium on Space Terahertz Technology
FIGURE CAPTIONS
Figure 1. (a) Geometry of open-circuited microstrip stub and Josephson
junction, (b) Cross sectional view of stub and junction.
Figure 2. Phase velocity vs. frequency for four different Nb microstrip stubs
at 4.2K. The stub dimensions are 1: 2u. x 1064u.; 2: 12u. X "lOOOu,; 3: 6u. X
"lOOOu.; 4: 6u, X 1000^. Inset shows current steps in DC IV curve
associated with high order modes in the stub.
Figure 3. Phase velocity vs. frequency for a Nb stub 12ji wide X 500u, long. The
different symbols refer to data taken with different applied external
magnetic fields. The solid line is the theoretical prediction. The arrow
indicates the gap frequency as determined for the Nb junction (at zero
magnetic field) which is used in the calculation.
Figure 4. Graphical representation of both sides of the transcendental eqns (1)
and (2). X-axis is frequency normalized to f s (fundamental resonance of
the stub). Straight line is a plot of the left hand side of eqns (1) and (2)
with arbitrary slope. Right hand side of eqn (1) is represented by tangent
Third International Symposium on Space Terahertz Technology Page 377
curves while right hand side of eqn (2) is plotted for values of a/ = 0.05,
0.10, 0.15. Intersections of the straight line and the tangent curves
represent the modes of the stub. When losses are significant ( a/ > ), the
mode frequencies are determined by the closest approach of the two
curves as indicated by the arrows.
88
)-
«
«
o
E
3
<n
O
s
to
"3
o
£
J?
Dm
MICROSTRIP
LINE
JUNCTION
SiOx
LAYER
SIS
JUNCTION
MICROSTRIP
LINE
QUARTZ
SUBSTRATE
(b)
Figure 1
Javadi et al Fig. 1
rn
g
o
K
-s:
u
v.
<u
-x;
«
<u
«
a.
to
K
o
to
<3
K
•2
K
a)
00,
■p
•H
o
o
H
d)
(0
(d
1.4-
1.2--
1.0-
0.8-
0.6-
0.4-
0.2-
0.0-
1.5
•=• 1.0 ..
0.0 2.0 4.0
Voltage (mV)
+
+
+
+
+
+
+
+
t
100 200 300 400 500 600 700 800 900 1000
Frequency (GHz)
&
v.
<u
8.
to
c
©
5
.3
55
o
£
2
a;
o
o 12
0)
(0
o
w 0.8 1
. 0.6 „.
>i
+)
•H
§ 0.4 1
(0
CM
0.2 ..
0.0
t
+
+
+
+
+
+
+
+
+
100 200 300 400 500 600 700 800 900 1000
Frequency (GHz)
Javadi et al Fig. 3
s
i
c
u
<u
«
E-.
8.
K
o
S
3
a.
K
O
c
-13--
-15
Page 382 Third International Symposium on Space Terahertz Technology
The material presented below is intended as a review only. A full length paper has been
submitted for publication in IEEE/MTT (May 1992).
DOUBLE-SLOT ANTENNAS ON EXTENDED HEMISPHERICAL
DIELECTRIC LENSES MQ^-27759
Daniel F. Filipovic, Steve J. Gearhart, Brian K. Kormanyos and Gabriel M. Rebeiz
/LOW?
y
j ^ \ "^- NASA/Center for Space Terahertz Technology
j Electrical Engineering and Computer Science Department
University of Michigan
Ann Arbor, MI 48109-2122
ABSTRACT
An investigation of the coupling efficiencies to a gaussian-beam of a double-slot antenna on
a hyperhemispherical lens is presented. It is shown that both lenses couple equally well to
an appropriate gaussian beam (about 80%). The radiation patterns of both lenses with a
double-slot antenna are computed using the ray-tracing method. The experimental radiation
patterns are presented and show close agreement to the theoretically computed patterns.
Third International Symposium on Space Terahertz Technology Page 383
I. INTRODUCTION
The use of a hemispherical lens with an attached extension length can greatly improve
coupling to a gaussian-beam system. In optical theory, an extension length of r/n is used,
and this extended lens is termed a hyperhemispherical lens. This extension length was chosen
since it satisfies the sine condition, which is where first-order aberrations are removed [1].
The hyperhemispherical lens was borrowed into the millimeter- wave field [2,3,4], but it was
found that radiation patterns from these lenses were very broad and even multi-lobed in
some cases. The hyperhemispherical lens is capable of coupling well to a gaussian-beam
system. However, it couples most efficiently to a converging beam and not to a plane
wave. Recently, several researchers showed that a narrow, diffraction-limited beam could
be achieved by putting the antennas on an elliptical lens [5,6]. The same effect was also
found by taking a hyperhemispherical lens and adding a planar extension to it [7]. Figure 1
shows that the focus of this longer extension length lens superimposes exactly on the second
focus of an elliptical lens. It is known from optical theory that a plane wave converges to
the second focus of an ellipse, and therefore a lens with this extension length is simply a
close geometrical approximation to an elliptical lens. The validity of this approximation
depends on the maximum allowed phase tolerance. For high dielectric constants (see Fig.
1) and relatively low frequencies, the phase difference becomes small and the approximation
is valid. Generally, for lens diameter of 12.5mm, e larger than 4, and frequencies less than
300GHz, the approximation is very good.
Page 384 Third International Symposium on Space Terahertz Technology
II. THEORETICAL AND EXPERIMENTAL PATTERNS
The theoretical radiation patterns are computed using a ray- tracing technique [9]. First,
the feed antenna pattern into the dielectric is calculated using standard far-field methods.
Figure 2 shows the calculated radiation patterns for a double-slot antenna with L = 0.28A at >
and d = 0.16A at >. These parameters were chosen to result in a symmetric pattern inside the
dielectric and a low cross-polarization in the 45° -plane. Ray-tracing is then used to calcu^
late the electric field distribution across the aperture plane (Fig. 3). In this method, the
fields are decomposed into TE/TM components at the lens/air interface, and the appropri-
ate transmission formulas are used for each mode. The power reflected into the substrate
is neglected in this analysis. A diffraction integral over the aperture then yields the far-
field pattern from the lens. Experimental measurements were performed at 246GHz on a
13.7 mm diameter silicon lens (e=11.7) with the double-slot antenna as a feed. Different
values of extension length were achieved by adding high-resistivity silicon wafers, resulting
in 3 extension lengths: hyperhemispherical, intermediate, and elliptical (Fig. 3). Measured
patterns at the elliptical focus (Fig. 4) demonstrate a gain of 28.6dB±0.3dB with relatively
low sidelobes (-16dB). From the measured patterns, the resulting aperture efficiency (cou-
pling to a plane wave) is 73%. The theoretical patterns calculated for this position are a bit
wider than the measured patterns (Fig. 5). This discrepancy arises from the fact that rays
at a certain angle end up hitting the critical angle at the lens/air interface, resulting in no
transmission of rays after this point. This limits the aperture size and results in a wider theo-
retical pattern. Note that this discrepancy is only significant at the elliptical focus for lenses
with high dielectric constants. Measured patterns at the elliptical focus for ±10% of the
Third International Symposium on Space Terahertz Technology Page 385
246GHz design frequency (Fig. 6) result in nearly the same gain, and therefore the double-
slot antenna has good pattern bandwidth. The measured power at broadside is nearly the
same from 222GHz-270GHz, also indicating good impedance bandwidth for the double-slot
design. The measured patterns at the intermediate focus (Fig. 7) are similar to the elliptical
focus, but with a gain of 24dB±0.3dB. In this case, the critical angle is not a problem, and
there is close agreement between theory and experiment (Fig. 8). At the hyperhemispherical
focus (Fig. 9), the pattern becomes very wide with a gain of 18.1dB±0.3dB and shows a
multi-peak behaviour, as indicated by theory (Fig. 10). As will be seen later, this has no
detrimental effect on the coupling efficiency to a converging beam. The ratio of the 246GHz
measured received power at broadside for an elliptical lens and a hyperhemispherical lens
was lOdB which is the same as the difference in the measured directivities. This indicates
that no power is coupled to substrate modes that may arise in the flat wafers.
III. GAUSSIAN-BEAM COUPLING
In order to match the double-slot/extended hemisphere system to a gaussian beam, one
could compute the electric field across the aperture and match this to a gaussian beam.
Since we had already predicted the far-field amplitude and phase distributions, we chose to
compute the coupling efficiency to a gaussian beam in the far- field (see Appendix). In this
calculation, the power radiated by the slot antennas to the air-side (which is 11.5% of the
total power) is taken into account, and no lens-air interface loss is considered. The power
loss radiated to the air side could be reflected using an appropriately designed cavity at
the expense of impedance bandwidth. Figure 11 gives the gaussian-beam parameters which
Page 386 Third International Symposium on Space Terahertz Technology
yield the highest coupling efficiency, and shows that all three focus positions are capable
of coupling equally well to a gaussian beam. However, the non-elliptical foci require a
converging wavefront, whereas the elliptical focus couples directly to a Gaussian beam with
an equal phase wavefront. Note that equivalent gaussian-beam parameters in the near field
may be found through a simple inverse Fourier transform. A gaussian beam experiment was
performed at 246GHz, in which it was attempted to couple all the power coming out of a lens
into the double-slot antenna. For the elliptical focus, the lens was placed at the minimum
waist position, where the radius of curvature is infinite, indicating an equal phase wavefront.
For the hyperhemispherical position, the lens was placed closer to the lens, at a position
where there is a negative radius of curvature. The proper negative radius of curvature
and position were computed knowing the gaussian-beam parameters from Figure 11. It was
found that the ratio of powers with either focus is the same within experimental error (±4%),
indicating that both the hyperhemispherical focus and the elliptical focus will match equally
well to an appropriately designed gaussian-beam system. Similar measurements were done
on a log-periodic antenna from 90-250GHz. The results are similar to those presented in this
paper and have been submitted for publication in IRMMW (May 92).
IV. ACKNOWLEDGEMENTS
This work was supported by the NASA/Center for Space Terahertz Technology at the Uni-
versity of Michigan.
Third International Symposium on Space Terahertz Technology Page 387
APPENDIX
The field representation of a Gaussian beam is of the form: EGauss(^) = t exp - ^^ ^ exp* 7 ^/^ 2
The coupling efficiency between an antenna pattern and a gaussian beam is calculated using
the formula [12,13]:
| //[g co • F(fl, (j>f) exp-tW exp^ g /*> 2 sin fldfld<ft| 2
VGauss - jj | F ^ ^ | 2 gin mA(j) jjexp-a («/*)» sin eded(j>
where F(d, (f>) is the far-field pattern of the antenna, and e co is the co-pol unit vector. The
value 0o controls the amplitude term and 0i controls the phasing term. These values are
varied to optimize the coupling efficiency.
REFERENCES
[1] Born and Wolf, Principles of Optics , Permagon Press, New York, 1959, pp. 252-252.
[2] D.B. Rutledge, D.P. Neikirk and D.P. Kasilingam, "Integrated Circuit Antennas," Infrared
and Millimeter-Waves, Vol. 10, K.J. Button, Ed., Academic Press, New York, 1983, pp. 1-
90.
[3] D.B. Rutledge and M. Muha, "Imaging antenna arrays," IEEE Trans. Antennas Propa-
gat., Vol. AP-30, 1982, pp.535-540.
[4] J. Zmuidzinas, "Quasi-optical slot antenna SIS mixers," IEEE Trans, on Microwave
Theory Tech., accepted for publication Jan. 1992. Also presented at the 2nd Int. Symp. on
Space Terahertz Technology, CA, March 1991.
[5] A. Skalare, Th. de Graauw, and H. van de Stadt,"A Planar Dipole Array Antenna with
an Elliptical Lens," Microwave and Optical Tech. Lett., Vol. 4, No. 1, 1991, pp. 9-12. Also,
"Millimeter and Submillimeter Studies of Planar Antennas," First Int. Symp. on Space
Terahertz Technology, Ann Arbor, MI, March 1990, pp. 235-255.
[6] C.J. Adler, C.R. Brewitt-Taylor, R.J. Davis, M. Dixon, R.D. Hodges, L.D. Irving, H.D.
Rees, J. Warner, and A.R. Webb, "Microwave and Millimeter- Wave Staring Array Technol-
ogy," IEEE MTT-S Int. Microw. Symp. Digest, June 1991, pp. 1249-1252.
[7] T.H. Buttgenbach, "A Fixed Tuned Broadband Matching Structure for Submillimeter
SIS Receivers," presented at the Third Int. Symp. on Space Terahertz Technology, Ann
Arbor, MI, March 1992.
[8] A.E. Siegman, Lasers, University Science Books, New York, 1986.
[9] R.E. Collin, Antennas and Radiowave Propagation, McGraw-Hill, New York, 1985, pp.
190-199.
Page 388
e r =11.7
Third International Symposium on Space Terahertz Technology
e r =4.0
e r =2.3
Figure 1: The synthesis of an elliptical lens from a hyperhemispherical lens and planar
wafers. The extended hemisphere is a very good approximation to an elliptical lens at high
dielectric constants.
: ; V ^
i
<
•28X air ;
*
r -
^_i
.16 ?vair
u i
i
c
' ' '■■■ ' >\
j
l a =
= 0.76A m / \\
r
d,
= 0.30A m /
-
■ rl-plone _'
- 15 -pi a nc
/
-°i
V
'^
L
*
J
c
-4
\
J
J
— i
'','
CO
i
'/
\
V i
O
o
i
L
-15 L
//:
■\> i
>
' \ ~i
—J
i"
J
i
™
i
, /
j
i
-20 f-
V
/J
l
■ ■ ; ^
-90
-60 -30 30
Angle (degrees)
60 90
Figure 2: The double-clot antenna (left) and its radiation patterns into a silicon (e=11.7)
dielectric (right).
Third International Symposium on Space Terahertz Technology
Page 389
Aperture
Plane
Critical
Angle
Silicon
/ Wafers
Silicon
Wafer
Feed Antenna
Hyperhemispherical — ' | ' — Elliptical
Intermediate
Figure 3: The ray-tracing method. Note the three focus positions that are achieved by
adding high-resistivity silicon wafers.
o
4)
>
CO
0)
-10
-15
-20
-25
I i.
. aw
E-plane j
H-plane j
— — 45-plane^
-30 -20 -10 10
Angle (degrees)
20 30
>
•10
-20
-30
Gain=28.6<lB
* ri I
u
-40
E-plane
H-plane
-60 -40 -20 20 40 60
Angle (degrees)
Figure 4: Measured patterns at the elliptical focus at 246GHz. The patterns are diffraction-
limited by the size of the aperture
Page 390
Third International Symposium on Space Terahertz Technology
X3
CO
O
>
OS
-20 -10 10 20
Angle (degrees)
.5 _10 r
eo r
C I""
F
> -15 r
-20
-25
H-Exper.
- - H-Theory '
-30 -20 -10 10
Angle (degrees)
1
1
20 30
Figure 5: Comparison of theory vs. experiment for the elliptical focus. The critical angle
limits the size of the aperture, resulting in wider theoretical patterns.
m
-a
c -
a
O
>
0)
OS
-30 -20 -10 10
Angle (degrees)
20 30
Gain = 29.4dB
TV ' ' ' ~
v E-plane
i H-plane
— — 45-plane.
/ \
-10 h
> -15
-20
.1' ' . \
J
-30 -20 -10 10 20 30
Angle (degrees)
Figure 6: Measured patterns at the elliptical focus at 222GHz (left) and 270GHz (right).
Third International Symposium on Space Terahertz Technology
Page 391
s
— -■ 1 ' 1 r
Gain=25.7dB
25 I 1 L
T ] I | I
E-plane
H-plane
— — 45-plane
' ' ' '
-30 -20 -10 10 20 30
Angle (degrees)
Figure 7: Measured patterns at the intermediate focus position at 246GHz.
-5 -
CO
■rH
efl
O
4)
> -15
-20 -
-25 L
-30
1 1 1 1 1
r
I
/
\
\
\
\
1 ;
- E-Exper.
- E-Theor\
-i
-
I'-
ll
//
\
V
-
-
II
V
- !
l\l
A
-:
\
r:
•
\
u
i
-
1
\
-;
" I 1
1
1
1 * 1
1
t !
I
i
-20 -10 10
Angle (degrees)
20 30
-20 -10 10
Angle (degrees)
30
Figure 8: Comparison of theory vs. experiment for the intermediate focus.
Page 392
Third International Symposium on Space Terahertz Technology
j- Gain = 18.1dB // I
i if I ^
c -10
3
o
> -15
I An )
, .'■/'■' , .'
-20 L\S - i
M
-25
-60 -40 -20 20
Angle (degrees)
40 60
Figure 9: Measured patterns at the hyperhemispherical focus position at 246GHz.
CO
o
•10
4)
> -15
-20
-25
T <~
E-Exper. -
E-Theorv -
'. \
-30 -20 -10 10
Angle (degrees)
20
■10'
;\
• 15 U/ ;
l/V ;
-20
r ,x/
r '
-25
/
^H
H-Exper.
Theory
-30 -20 -10 10 20
Angle (degrees)
30
Figure 10: Comparison of theory vs. experiment for the hyperhemispherical focus. Notice
the predicted multi-peak behaviour.
Third International Symposium on Space Terahertz Technology
Coupling to Gaussian Beams
Page 393
Focus Position
(extension)
Gain
Gaussian Beam Parameters
Matching
Efficiency
(amplitude)
O^phase)
Elliptical
(.39 radius)
28.6dB
5.0 8
-
~ 79%
Intermediate
(.32 radius)
25.7dB
8.2°
11.3°
~ 83%
Hyper-
hemispherical
(.25 radius)
18.1dB
13.3°
13.5°
~ 81%
Gaussian Beam Electric Field: explX©/©^ 2 ] exp[j*7t*( 0/G^ 2 ]
Max. Power Elliptical
Max. Power Hyperhemispherical
= 10dB
Figure 11: Table of Gaussian beam parameters.
Page 394 Third International Symposium on Space Terahertz Technology
6 ^^ EMBEDDING IMPEDANCE APPROXIMATIONS
J6 6V8 IN THE ANALYSIS OF SIS MIXERS
A. R. Kerr and S.-K. Pan ^ ,v
National Radio Astronomy Observatory* ^93'"^**^^
Charlottesville, VA 22903
S. Withington
Cavendish Laboratory, University of Cambridge
Cambridge CB3 OHE, UK
ABSTRACT
Future nillimeter-wave radio astronomy instruments will use arrays of many
SIS receivers, either as focal plane arrays on individual radio
telescopes, or as individual receivers on the many antennas of radio
interferometers. Such applications will require broadband integrated
mixers without mechanical tuners. To produce such mixers, it will be
necessary to improve present mixer design techniques, most of which use
the three - frequency approximation to Tucker's quantum mixer theory.
This paper examines the adequacy of three approximations to Tucker's
theory: (i) the usual three -frequency approximation which assumes a
sinusoidal LO voltage at the junction, and a short-circuit at all
frequencies above the upper sideband, (ii) a five -frequency approximation
which allows two LO voltage harmonics and five small-signal sidebands, and
(iii) a quasi five -frequency approximation in which five small-signal
sidebands are allowed, but the LO voltage is assumed sinusoidal. These
are compared with a full harmonic -Newton solution of Tucker's equations,
including eight LO harmonics and their corresponding sidebands, for
realistic SIS mixer circuits.
It is shown that the accuracy of the three approximations depends
strongly on the value of uRgC for the SIS junctions used. For large a>R„C,
all three approximations approach the eight -harmonic solution. For uRgC
values in the range 0.5 to 10, the range of most practical interest, the
quasi five -frequency approximation is a considerable improvement over the
three -frequency approximation, and should be suitable for much design
work. For the realistic SIS mixers considered here, the five - frequency
approximation gives results very close to those of the eight -harmonic
solution.
Use of these approximations, where appropriate, considerably reduces
the computational effort needed to analyze an SIS mixer, and allows the
design and optimization of mixers using a personal computer.
*The National Radio Astronomy Observatory is operated by Associated
Universities, Inc. under cooperative agreement with the National Science
Foundation.
Third International Symposium on Space Terahertz Technology Page 395
INTRODUCTION
The superior sensitivity of SIS mixer receivers at millimeter
wavelengths has been clearly demonstrated in recent years [1], and SIS
receivers are now used at almost all the world's major millimeter-wave
radio astronomy observatories. SIS mixers appear for the most part to be
well described by Tucker's quantum theory [2,3], which predicts strong
non-classical behavior in a resistive mixer with an extremely sharp I-V
nonlinearity. Tucker's theory is normally applied in its three -frequency
approximation, in which form it is moderately tractable analytically and
allows optimization of designs by small computer.
At present, the most sensitive SIS mixers have one or two mechanical
waveguide tuners which allow the RF embedding impedance to be adjusted to
suit the particular junction (or array of junctions). While mechanically
tuned mixers are acceptable in applications requiring one or two
receivers, plans for future radio astronomy instruments include arrays of
many SIS receivers operating either all in the focal plane of a single
radio telescope , or individually on the many antennas of a radio
interferometer. For such applications, it is highly desirable to replace
mechanically tuned mixers with broadband integrated mixers without
mechanical tuners. To produce such mixers will require refinement of
present mixer design techniques, most of which use Tucker's three-
frequency approximation.
This paper examines the adequacy of three approximations to Tucker's
theory: (i) the usual three -frequency approximation, (ii) a five-
frequency approximation, and (iii) a quasi five -frequency approximation.
These are compared with a full harmonic -Newton solution [4,5] of Tucker's
equations up to the eighth LO harmonic for realistic SIS mixer circuits.
Use of the approximations greatly reduces the computational effort needed
to optimize a mixer design.
Page 396 Third International Symposium on Space Terahertz Technology
THE APPROXIMATIONS
In the so-called three- frequency approximation to Tucker's theory,
it is assumed, as indicated in Fig. 1(a), that the embedding impedance
seen by the junction is finite at the LO frequency w p , and at w u , w, and
w , the upper and lower sideband and intermediate frequencies. At all
higher frequencies the junction is short-circuited. This is likely to be
a good approximation for junctions with large capacitance. The three-
frequency approximation implies a sinusoidal LO voltage at the junction,
for which case Tucker gives closed- form expressions for the coefficients
Y t j of the mixer's 3 x 3 small-signal conversion matrix as functions of
the pumping parameter a - eV p /#w p and DC bias voltage V .
The quasi five- frequency approximation assumes the same sinusoidal
LO voltage waveform as above. However, while the second LO harmonic 2w p
is short-circuited at the junction, the second harmonic sidebands 2w p ± o>
are not. This is depicted in Fig. 1(b). Closed-form expressions for the
elements of the 5x5 small-signal conversion matrix are given by Tucker.
While this situation is not easy to implement practically, it is not
physically unrealizable, and is expected to be a better approximation to
a real mixer than the simple three -frequency approximation.
The full five-frequency approximation allows finite embedding
impedances at all frequencies up to 2w p + w , but requires the junction to
be short-circuited at all higher frequencies. This is depicted in Fig.
1(c). The LO voltage at the junction contains a second harmonic
component, and an iterative algorithm must be used to determine the LO
voltage waveform and thence the elements of the 5x5 small-signal
conversion matrix.
Third International Symposium on Space Terahertz Technology p a g e 397
SIMULATIONS
The three approximations are investigated using hypothetical double
sideband SIS mixers at 115 and 345 GHz, with the desired embedding
impedances at the LO harmonics and small-signal sideband frequencies. In
all the examples, the junction capacitance is tuned out at frequencies w,
and w u by the source susceptance. The embedding impedance at frequencies
w > o> p + w is either zero or that of the junction capacitance alone,
depending on the particular example. The RF source and IF load
conductances are assumed equal. These assumptions are consistent with
realistic low noise mixer designs [1] with low IF (u> « u> p ) .
The I-V curve used in the examples is that of the 4 -junction array
of Nb/Al-Al 2 3 /Nb junctions used in [1], and is shown in Fig. 2. The
theoretical equivalence between a series array of junctions and a single
junction is shown in [6] .
Five values of w p R N C are used: 0.5, 1, 2,4, and 50. In all cases,
the pumping parameter a - eVj/ftWp = 1.2, where V 1 is the amplitude of the
fundamental component of the LO voltage at the junction. The junction is.
voltage biased at the center of the first photon step below the gap
voltage; i.e., V = V gap - hu/2e.
The mixer's (equivalent input) noise temperature, (transducer)
conversion gain, input return loss, and output impedance are computed for
each case as functions of Rrf/Rn, where Rgp is the reciprocal of the RF
source conductance at the signal frequency and R N is the normal resistance
of the junction (or array of junctions) . For comparison, the results of
the full eight-harmonic analysis are shown on each graph.
Page 398 Third International Symposium on Space Terahertz Technology
RESULTS
Full five- frequency approximation
The results of the full five -frequency approximation are shown in
Figs. 3(a) -(d) for 115 GHz and 345 GHz mixers. At 115 GHz it is clear
that the only significant deviation from the eight -harmonic results is in
the gain for the lowest value of o>R N C (0.5), and that is only a small
fraction of a decibel. At 345 GHz there is almost no difference between
the full five - frequency approximation and the eight -harmonic results.
Quasi five- frequency approximation
Figures 4(a) -(d) show the results of the quasi five -frequency
analysis of the same mixers as above. At both 115 and 345 GHz, the gain
is within a decibel of that computed using eight harmonics. The mixer
noise temperatures agree closely for wR N C - 50 and are within about 10% for
wR H C - 4, but for «R N C - 0.5 they differ by as much as 40%. The input
return loss and output impedance show minor deviations from the eight-
harmonic solution.
Three- frequency approximation
In the three -frequency case, the value of wR N C has no effect. This
is because all relevant frequencies above the upper sideband are short-
circuited, and we are assuming the junction capacitance is tuned out at
the signal and image frequencies. The three -frequency case is thus
equivalent to the high-wR N C limit of any of the other cases we have
considered. The results of Figs. 3 or 4 with wR„C ■= 50 are, in fact,
indistinguishable from those of the three -frequency approximation.
Third International Symposium on Space Terahertz Technology Page 399
DISCUSSION
In an earlier paper [1] we put forward design criteria for SIS
mixers for millimeter-wave radio astronomy applications. They should have
low noise, conversion gain near unity (to avoid saturation), and a
reasonable RF input match. For wide frequency coverage, only double -
sideband operation is considered. It was found that these conditions
could be met if the mixer operated with equal RF source and IF load
conductances , and that the output impedance of the mixer was then large
(i.e., the mixer operates as a current source).
The examples here likewise assume equal RF source and IF load
conductances, and examine the mixer performance as a function of Rrf/Rh (Rrf
is the reciprocal of the RF source conductance, and R N is the normal
resistance of the junction or array of junctions). As expected, the mixer
noise temperature exhibits a broad minimum as Rrf/Rh is varied (Fig. 3).
The (transducer) gain exhibits no minimum, but increases with Rrf/Rn-
At higher frequencies, the minimum noise temperature occurs at
larger values of R^/R^. This has an important implication for SIS mixer
design: If the same mixer circuit is scaled for use at two different
frequencies (i.e., the embedding impedances are the same for the two
designs), then the normal resistance of the junction (or array) should be
smaller for the higher -frequency mixer. This is discussed in more detail
in [1].
In designing a broadband tunerless SIS mixer, the choice of the
value of wR N C is primarily a tradeoff between noise temperature and useable
frequency range; large values of C will obviously limit frequency
coverage, while too small an wR N C degrades the noise temperature as well
as being more difficult to achieve without sacrificing junction quality.
It is important, therefore, for the designer to have available a method of
mixer analysis which reflects the effect of wR N C with sufficient accuracy.
From the results in Figs. 3 and 4, it is clear that the quasi five-
Page 400 Third International Symposium on Space Terahertz Technology
frequency approximation describes the mixer performance quite well for
wR N C > 4. For 0.5 < wR N C < 4, the gain, input return loss, and output
impedance are well enough predicted by the quasi five -frequency
approximation for most mixer design work, but the mixer noise temperature
can be too high by as much as 40% . Figure 5 shows the dependence of the
minimum value of T M (with respect to Rrf/Rjj) on wR N C for each of the methods
described here .
It may seem odd that the quasi five -frequency approximation should
give quite accurate results for all the mixer parameters except the noise
temperature. This is explained by considering the origin and nature of
the noise at the output port of the mixer. This noise originates as shot
noise from the current flowing in the SIS junction. The action of the
local oscillator is, in effect, to amplitude modulate the shot noise
produced by the DC current flowing in the junction, thereby generating
correlated components at all the sideband frequencies nw p ± w , n - 0,
1 etc. The mixing action of the time-varying junction conductance
converts all these sideband components to the IF, preserving their
correlation. The relative phase of each correlated IF component depends
on the embedding impedance (including junction capacitance) at its
sideband of origin, and on the phases of the harmonics in the LO waveform
at the junction. The minimum noise occurs when these correlated IF noise
components are phased so as to cancel to the greatest degree. (It is in
this way that a classical diode mixer can, in principle, have zero noise
temperature while operating with 3 dB conversion loss and many milliamps
of rectified current.) The fact that, in low-noise operation, the SIS
mixer has several quite large but strongly correlated output noise
components which largely cancel one another, explains the sensitivity of
the noise temperature results to the small changes in LO waveform between
the quasi five -frequency and full five - frequency approximations. (A more
detailed discussion of noise conversion in mixers, and additional
references are given in [7].)
Third International Symposium on Space Terahertz Technology Page 401
REFERENCES
[1] A. R. Kerr and S.-K. Pan, "Some recent developments in the design of
SIS mixers," Int. J. Infrared Millimeter Waves, vol. 11, no. 10,
Oct. 1990.
[2] J. R. Tucker, "Quantum limited detection in tunnel junction mixers,"
IEEE J. of Quantum Electron., vol. QE-15, no. 11, pp. 1234-1258,
Nov. 1979.
[3] J. R. Tucker and M. J. Feldman, "Quantum detection at millimeter
wavelengths," .Rev. Mod. Phys . , vol. 57, no. 4, pp. 1055-1113, Oct.
1985.
[4] C.-Y. E. Tong and R. Blundell, "Simulation of superconducting
quasiparticle mixer using a five-port model," IEEE Trans. Microwave
Theory Tech., vol. MTT-38, no. 10, pp. 1391-1398, Oct. 1990.
[5] S. Withington and P. Kennedy, "Numerical procedure for simulating
the large-signal quantum behavior of superconducting tunnel -junction
circuits,"; Proc. IEE, part G, vol. 138, no. 1, pp. 70-76, Feb. 1991.
[6] M. J. Feldman and S. Rudner, "Mixing with SIS arrays," Reviews of
Infrared & Millimeter Waves, (Plenum, New York), vol. 1, p. 47-75,
1983.
[7] D. N. Held and A. R. Kerr, "Conversion loss and noise of microwave
and millimeter-wave mixers: Part 1 - Theory," IEEE Trans. Microwave
Theory Tech. , vol. MTT-26, pp. 49-55, Feb. 1978.
Page 402
Third International Symposium on Space Terahertz Technology
- 3w p
Junction
short-
circuited
- 2u p
■
USB Up +
«o
- Up LD
LSB u p -
Uq
— - u IF
■4
J-
f
Junction
short-
circuited
not shorted
shorted
not shorted
USB Up +• wo —
LSB Up - Ug — —
3u r
2u c
u p LD
— u IF
t j
Junction „
short- -- 3u
circuited p
__L_„~____
2u p + uo —
2u p - u — -
- 2u p
USB Up +■ uo — ■■
- u p LD
LSB Up - uo — -
— u IF
-a-
3-FREQ. APPRDX.
Co.)
QUASI 5-FREQ. APPRDX.
<b)
5-FREQ. APPRDX.
Fig. 1 . Embedding impedance diagram indicating which frequencies are short-circuited for the three
approximations: (a) three-frequency, (b) quasi five-frequency, and (c) full five-frequency.
200
HY430C5K11
4 junctions
4.2
K
/
I jjA
R „* = 7 2 oh
ns
|
100
\
J
.
N
5
V
10
nV
15
Fig. 2. I-V curve used in the simulations. This curve is for a four-junction array of Hypres
Nb/Al-Al 2 3 /Nb junctions at 4.2K, as used in [1].
Third International Symposium on Space Terahertz Technology
Page 403
8 Harmonics
FuH 5-freq.
CD
(/)
CO
CD
~o
C
"o
o
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Rrf/Rn
Fig. 3(a). Noise temperature and conversion gain as functions of Rrf/R n for a 1 15 GHz SIS mixer,
computed using the full five-frequency approximation (solid line) and with eight LO harmonics
(broken line). Results are shown for wR N C = 0.5, 1, 2, 4, and 50.
20
15
CO
■o
w
O 10
c
3
^> 5
3
a
c
o-
-5
8 Harmonics
Full 5-freq.
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Rrf/Rn
Fig. 3(b). Input return loss and output impedance as functions of R RF /R N for a 1 15 GHz SIS mixer,
computed using the full five-frequency approximation (solid line) and with eight LO harmonics
(broken line). The output resistance is normalized to the IF load resistance R IF . Results are shown
for wR N C = 0.5, 1, 2, 4, and 50.
Page 404
Third International Symposium on Space Terahertz Technology
8 Harmonics
Full 5-freq.
m
in
</)
m
■o
c
'a
o
-10
Fig. 3(c). Noise temperature and conversion gain as functions of Rrf/Rm for a 345 GHz SIS mixer,
computed using the full five-frequency approximation (solid line) and with eight LO harmonics
(broken lines - almost obscured by the solid lines). Results are shown for wR N C = 0.5, 1 , 2, 4, and
50.
8 Harmonics
Full. 5-freq.
or
3
O
Fig. 3(d). Input return loss and output impedance as functions of Rrf/R n for a 345 GHz SIS mixer,
computed using the full five-frequency approximation (solid line) and with eight LO harmonics
(broken lines - almost obscured by the solid lines). The output resistance is normalized to the IF
load resistance R, F . Results are shown for «R N C = 0.5, 1,2,4, and 50.
Third International Symposium on Space Terahertz Technology
Page 405
8 Harmonics
Quasi 5-freq.
CD
(A
(/)
CD
■o
C
'o
o
0.0 0.2 0.4
0.8 1.0 1.2 1.4
Rrf/Rn
-10
Fig. 4(a). Noise temperature and conversion gain as functions of R RF /R N for a 115 GHz SIS mixer,
computed using the quasi five-frequency approximation (solid line) and with eight LO harmonics
(broken line). Results are shown for wR N C = 0.5, 1 , 2, 4, and 50.
8 Harmonics
Quasi 5-freq.
0C
3
O
cc
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Rrf/Rn
Fig. 4(b). Input return loss and output impedance as functions of R„f/R n for a 115 GHz SIS mixer,
computed using the quasi five-frequency approximation (solid line) and with eight LO harmonics
(broken line). The output resistance is normalized to the IF load resistance R IF . Results are shown
for wR N C = 0.5, 1 , 2, 4, and 50.
Page 406
Third International Symposium on Space Terahertz Technology
8 Harmonics
Quasi 5-freq.
CD
CD
■o
c
o
o
Fig. 4(c). Noise temperature and conversion gain as functions of R RF /R N for a 345 GHz SIS mixer,
computed using the quasi five-frequency approximation (solid line) and with eight LO harmonics
(broken line). Results are shown for wR N C = 0.5, 1 , 2, 4, and 50.
8 Harmonics
Quasi 5-freq.
or
3
O
or
Fig. 4(d). Input return loss and output impedance as functions of Rrf/R n for a 345 GHz SIS mixer,
computed using the quasi five-frequency approximation (solid line) and with eight LO harmonics
(broken line). The output resistance is normalized to the IF load resistance R IF . Results are shown
for wR N C = 0.5, 1, 2, 4, and 50.
Third International Symposium on Space Terahertz Technology
Page 407
100
u>RnC
Fig. 5. Minimum mixer noise temperature as a function of wR N C for: (i) the full eight-harmonic and
full 5-frequency solutions (broken line), (ii) the quasi 5-frequency approximation (solid line), and (iii)
the 3-frequency approximation (horizontal lines).
Page 408 Third International Symposium on Space Terahertz Technology
SUB MICRON AREA Nb/A10 x /Nb TUNNEL JUNCTIONS FOR SUBMM MIXER
APPLICATIONS
H.G. LeDuc, B. Bumble, S.R. Cypher, A. J. Judas*, and J A.. Stern
Center for Space Microelectronics M Q Q C) ry ty „
Jet Propulsion Laboratory i\ & O m /£{ • 6 1
California Institute of Technology
Pasadena, CA 91109
*Present Address: Stanford University
Palo Alto, CA
Abstract
In this paper, we report on a fabrication process developed for submicron area tunnel
junctions. We have fabricated Nb/A10 x /Nb tunnel junctions with areas down to 0.1 nm*
using these techniques. The devices have shown excellent performance in receiver systems
up to 500 GHz and are currently in use in radio astronomy observatories at 115, 230, and
500 GHz.
The junctions are fabricated using a variant of the self-aligned-liftoff trilayer process* with
modifications for electron beam lithographic patterning of junction areas. In brief, the
technique involves patterning submicron holes in PMMA using electron beam lithography.
The negative of this pattern is formed by thermal deposition and liftoff of chromium metal
using this PMMA stencil. The chromium pattern is transferred to an underlying polyimide
film using oxygen RIE. Junctions are formed by RIE using a gas mixture containing
CQ2F2 and electrically isolated with thermally evaporated silicon monoxide. Contact
wiring and coupling/tuning structures are patterned by RIE.
Introduction
SIS tunnel junctions can be modeled as a nonlinear resistor in parallel with a shunt
capacitor. A good figure-of-merit of the high frequency performance of these devices is
Third International Symposium on Space Terahertz Technology p a o e 409
the ratio of the capacitive reactance to the real resistance (coRC). The RC product, for
SIS tunnel junctions, is determined by the tunnel barrier thickness and is independent of
the device area. The junction area is chosen to provide the best impedance match to the
mixer embedding circuit and is usually a compromise between minimizing the capacitance
while maintaining a reasonable real impedance. In the best case, the embedding circuit
can tune out the capacitance and the junction area is chosen to make the rf-resistance
match the real part of the embedding circuit impedance (approximately 50 - 100 CI). For
small RC devices, the resistance-area product is small so that achieving the appropriate
resistance using a single junction requires submicron areas. Series arrays or other novel
coupling mechanisms may relieve the constraint on submicron areas, however, designing
these elements may require a greater understanding of the high frequency characteristics
of devices and materials than is currently available. We have chosen to use single junctions
in the hope that the simplicity in understanding the high frequencies behavior of the
mixers may outweigh the complexity associated with the fabrication of submicron devices.
Since their development 2 , high quality Nb/AlOx/Nb tunnel junctions represent the only all
refractory SIS technology in use in radio astronomy receiver systems. This is primarily due
to their nearly ideal tunneling characteristics and physical robustness. In this paper we
describe techniques for fabricating submicron devices.
Tunnel Junction Fabrication
The tunnel junction fabrication process is similar to the self-aligned-liftoff process used
to fabricate refractory tunnel junctions employing optical lithography 1 ' 3 . The primary
difference arises from the need to use higher resolution lithography in the tunnel junction
patterning and to maintain this resolution throughout the fabrication process. The process
steps are shown schematically in figure 1.
Page 410
Third International Symposium on Space Terahertz Technology
Chromium
PMMA
Polyfmide
Nb
Insulator
Nb
Chromium
Polyfmide
i Nb
Insulator
2 Nb
m
Figure 1. Submicron fabrication process schematic, (a) After trilayer deposition , wafers are spin coated
with 400-600 nm of polyimide and 120 nm PMMA. (b) Chromium metal is thermally deposited, (c) Oxygen
RIE of polyimide. (d) RIE of Nb counter electrode in CCI2F2+CF4+O2 gas mixture, (e) Deposition of
SiO, lift-off, and wire electrode deposition and patterning.
a. Nb/AJO x /Nb Trilayer Deposition
The Nb/A!O x /Nb trilayer is deposited in-situ in a high vacuum system (base pressure 1.3 x
10 "7 Pa) by magnetron sputtering. The substrates are oxidized silicon or quartz and are
heat sunk to a thermal mass but not actively cooled during deposition. The large scale
features of the trilayer are formed by lift-off using AZ5214 photoresist (AZ Hoechst) and
image reversal. The Nb base and counter electrodes are approximately 160 nm and 120
nm respectively. The barrier is formed by depositing 6-10 nm of aluminum followed by an
in-situ oxidation in an argon/oxygen gas mixture in a manner similar to that described by
Morohashi et al. . During this step the total process gas pressure is maintained constant by
throttling the vacuum pump. A dc-plasma is formed during the oxidation process by
applying approximately -500V to an aluminum ring placed in the system. This plasma has
Third International Symposium on Space Terahertz Technology Page 411
been found to reduce oxidation times, but does not effect the quality of the barrier. After
the Nb counter electrode deposition, 30 nm of gold is deposited on the trilayer to act as a
contact layer.
b. Junction Patterning
The etch mask used to form the tunnel junction is patterned by electron beam (e-beam)
lithography using a JEOL JBX-5 lithography system with a minimum spot size of 8 nm.
The lithographic stencil must be robust enough to withstand Reactive Ion Etching (RIE)
and provide a means to subsequently lift-off the SiO isolation layer. The high resolution e-
beam resist, PMMA, is not suitable as the final RIE mask because it lacks the required
etch resistance. Techniques have been developed which transfer the e-beam written
pattern into polyimide while maintaining the required resolution 4 . The wafer is spin
coated with a polyimide 5 film approximately 400 - 600 nm thick. Following a hot plate
bake to drive the solvents from the polyimide, the wafer is spin coated with 120 nm of
PMMA. It is then exposed in the e-beam lithography system to form holes in the PMMA
film with the required junction dimensions. Chromium metal is thermally evaporated onto
this stencil and the PMMA is removed in acetone, leaving metal where there were holes
(polyimide is not soluble in acetone). The resulting pattern is etched in a parallel plate
RIE system using oxygen gas to remove polyimide from areas of the wafer not protected by
chromium. The RIE of polyimide is highly anisotropic, however, it is sensitive to surface
contamination such as dust or material resputtered from the electrodes of the etcher and
care must be taken to provide a clean environment for this process step. An SEM
micrograph of an etch test pattern is shown in figure 2. The square etch stencils consisting
of Cr(30nm) on Polyimide(550nm) have dimensions of 1.5, 1.0, 0.5, and 0.25 urn on a side.
The minimum area is .06p.m2 .
Page 412
Third International Symposium on Space Terahertz Technology
Figure 2. Test patterns etched in polyimide using oxygen RIE . The smallest
features are 0.06u,m2 .
c. Junction Etch
The tunnel junction is formed using RIE by first etching the gold contact layer and then
the Nb counter electrode. The gold is sputter etched using argon gas. Techniques for
aniso tropically etching Nb had to be developed. An etch profile for a submicron line
patterned in an Nb film using a standard etch process (CF4+20% O2 ,4 Pa pressure, and
0.27 W/cm^ power density ) is shown in figure 3. The isotropic component of this etch
mixture is clearly too large to be used in the fabrication of submicron devices. Anisotropy
occurs in RIE when the etch mechanism requires predominantly normal incident ion
impact energy to proceed 6 . Etching of Nb in CF4/O2 , however, occurs via a spontaneous
rather than ion assisted reaction of fluorine and fluorine radicals with Nb. We have found
a technique which achieves the required anisotropy. Etching with a gas mixture containing
CQ2F2 produces very good etch anisotropy, which may be attributed to the a nonvolatile
NbCl x product which forms on the sidewalls. Figure 4 shows the etch rate of Nb and NbN
using mixtures of CCI2F2+CF4+O2 • For these measurements, the total pressure was 4
Third International Symposium on Space Terahertz Technology
Page 413
Figure 3. Submicron Nb lines etched by RIE using CF4+O2. The large undercut of the Nb
line below the 0.4(im chromium etch stencil is evident.
Pa, the power density was 0.27 W/cm^ and the oxygen flow was constant at 2 seem, while
the CQ2F2/CF4 ratio was varied. The etch is highly anisotropic for mixtures containing
greater than 60% CQ2F2 in CQ2F2+CF4. Mixtures rich CF4 exhibited isotropic etching.
The region with approximately 20% to 50% CCI2F2 content was characterized by low etch
rates and polymer formation. Shown in figure 5 is the etch profile of Nb achieved using
62% CQ2F2 in (CCI2F2+CF4) and similar sample etched in CF4+O2. Structures etched
in the CCI2F2 gas mixture show very little undercut while CF4+O2 produced a large
undercut.
d. Electrical Isolation
Following the etch the counter electrode to form the junctions, a electrical isolation layer
of SiO is deposited with the etch mask in place . The SiO is thermally deposited from a
baffled source. To achieve good edge coverage, the samples are placed at a fixed angle
Page 414
Third International Symposium on Space Terahertz Technology
0.2
0.4
0.6
Nb
NbN
0.8
Figure 4. RIE etch rate for Nb and NbN as a function of gas composition. The
etch gas consists of 85%(x CCl2F2+(l-x)CF4)+15%02.
relative to SiO flux and rotated during the deposition. Flux angles for normal incidence to
approximately 60 degrees have been evaluated. Angles of 5-15 degrees have been found to
provide a good compromise between side wall coverage and clean lift-off. SiO film
thicknesses are typically 150-250 nanometers depending on the application. The polyimide
and SiO are removed from the junction areas using dichloromethane solvent. A short RIE
etch in oxygen is used to remove polyimide residues after the lift-off step.
e. Contact Wiring
Mixer elements are completed by depositing 250-350 nm of Nb by magnetron sputtering.
The wire layer is patterned lithographically and etched using a RIE process similar to the
one used for the junction etch. A typical current-voltage characteristic for a tunnel junction
fabricated by this process is shown in figure 6. This device is 0.25 \xxa^ in area and has a
critical current density of 7.7 kA/cm^.
Third International Symposium on Space Terahertz Technology Page 415
Summary/Conclusions
In this paper, we have described techniques developed for the fabrication of submicron
area tunnel junctions in refractory materials. The process described is applied specifically
to the fabrication of Nb/A10 x /Nb tunnel junctions, however, much of the technology has
also been used to fabricate NbN/MgO/NbN tunnel junctions 7 and is relevant to other
submicron fabrication tasks. This process extends the self-aligned lift-off process used to
fabricate refractory tunnel junctions using optical lithography. The primary new features
are the use of electron beam lithography to form a submicron pattern in PMMA and the
transfer of this pattern into chromium by lift-off. The chromium pattern is transferred into
polyimide using oxygen RIE and the resulting Cr/polyimide is used to etch the trilayer
counter electrode using a highly anisotropic RIE gas mixture containing CQ2F2.
Nb/A10 x /Nb tunnel junctions with areas down to 0.1 \im^ have been fabricated using these
techniques. Mixer elements have been fabricated using this process for both wave
guide 8 » 9 » 10 and quasi optically coupled 11 * 12 ' 13 receiver systems. In wave guide receiver
systems with operating frequencies up to 500 GHz, the capacitance associated with the
submicron area Nb/A10 x /Nb devices is small enough so that the mixer block rf-embedding
circuit provides enough tuning to achieve excellent performance (receiver noise
temperatures, Tr(DSB) = 180K at 485 GHz) 14 without integrated tuning structures. In
principle junction areas can be scaled down further, however, in order to do so the junction
relaxation times must also be scaled down so that the real part of the junction impedance in
the correct range. The junction relaxation time (RC) is determined by the insulator barrier
thickness, with thinner barriers producing smaller RCs. The limit for a given insulator
barrier is determined by the thinnest barrier that can be achieved while maintaining
suitable junction characteristics. It has been our experience with Nb/A10 x /Nb tunnel
junctions, that the I-V characteristics degrade significantly for critical current densities of
Page 416 Third International Symposium on Space Terahertz Technology
greater than 15kA/cm2 (RA= 12 Q u.m2). For junctions with this current density, a 100 Q
junction has an area of = 0.12 um^
Acknowledgements
The research described in this paper was performed by the Center for Space
Microelectronics Technology, Jet Propulsion Laboratory , California Institute of
Technology, and was jointly sponsored by the Strategic Defense Initiative Organization /
Innovative Science and Technology Office and the National Aeronautics and Space
Administration / the Office of Aeronautics and Space Technology. We would also like to
acknowledge P.D. Maker and R.E. Muller for the excellent electron beam lithography
support and technical discussions.
References:
1 A. Shoji, F. Shinoki, S. Kosaka, M. Aoyagi, and H. Hayakawa, " New Fabrication Process for
Josephson Tunnel Junctions with (Nobium Nitride, Niobium) Double-Layered Electrodes", Appl. Phys.
Lett., 41,1097, (1982).
2 M. Gurvitch, M.A. Washington, and H.A. Huggins, "High Quality Refractory Tunnel Junctions
Utilizing Thin Aluminium Layers", Appl. Phys. Lett., 42, 472 (1983).
3 S. Morohashi, F. Shinoki, A. Shoji, M. Aoyagi, and H. Hayakawa, "High Quality NblAl-AlOxINb
Josephson Junction", Appl. Phys. Lett. 46, 1179, (1985).
4 D.M. Byrne, AJ. Brouns, F.C. Case, R.C. Tiberio, B.L. Whitehead, and E.D. Wolf, "Infrared Mesh
Filters Fabricated by Electron-Beam Lithography", J. Vac. Sci. Technol., B3, 268 (1985) and references
within: M. Hatzakis, B.J. Canavello, and J.M. Shaw, IBM J. Res. Dev., 24, 452 (1980).
5 Olin Ciba-Geigy, Probimide 200 series.
6 See for instance: J.W. Coburn, Plasma Etching and Reactive Ion Etching. AVS Monograph Series,
Ed. N. Rey Whetten.
7 J.A. Stern, H.G. LeDuc, and AJ. Judas, "Fabrication and Characterization of High Current-Density,
Submicron, NbN/MgO/NbN Tunnel Junctions", this conference.
8 RES. Javadi, W.R. McGrath, S.R. Cypher, B.D. Hunt, and H.G. LeDuc, Digest 15th Int. Conf. on
IR and Millimeter Waves, p245, Orlando, FL (1990).
9 J.W. Kooi, M. Chan, T.G. Phillips, B. Bumble, and H.G. LeDuc, "A Low Noise 230 GHz Heterodyne
Third International Symposium on Space Terahertz Technology Page 417
Receiver Employing .25\un^ Area NblAlOJNb Tunnel Junctions", IEEE Microwave Theory and
Techniques Journal, to be published.
10
C.K. Walker, M. Chen, P.L. Shafer, H.G. LeDuc, J.E. Carlstrom, and T.G. Phillips, "A 492 GHz SIS
Waveguide Receiver for Submillimeter Astronomy", Int. J. of IR and Millimeter Waves, to be
published.
11 T.H. Buttgenbach, H.G. LeDuc, P.D. Maker, and T.G. Phillips, "A Fixed Tuned Broadband Matching
Structure for Submillimeter SIS Receivers", IEEE Trans. Appl. Superconductivity, to be published.
12 P.A. Stimson, RJ. Dengler, P.H. Siegel, and H.G. LeDuc, "A Planar Quasi-Optical SIS Receiver for
Array Applications", this conference.
13 J. Zmuidzinas, H.G. LeDuc, and J.A. Stem, "Slot Antenna SIS Mixers for Submillimeter Wavelengths",
this conference.
14 Private communication, C.K. Walker.
Page 418
Third International Symposium on Space Terahertz Technology
Figure 5. SEM micrographs showing etch comparrisons between CF4+O2 (left) and
CCl 2 F2+CF 4 +02(right). The RIE mask is Cr(30 nm)/Polyimide(550 nm) patterned by e-beam
lithography and oxygen RIE. The Nb film (400 nm thick) etchs anisotropically in the CCI2F2 containing
etch gas.
1mV
Figure 6. Typical tunneling current-voltage characteristic for a Nb/AlOx/Nb junction taken ar 4.2K. The
junction area is 0.25 mm^ and the critical current density is 7.7 kA/cm^ .
Third International Symposium on Space Terahertz Technology Page 419
Noise in Josephson Effect Mixers and the RSJ Model
R. Schoelkopf*, T. Phillips*, and J. Zmuidzinas* ^ " «* ^-X7 6 2
Josephson effect mixers have previously been observed to display "excess" noise r J
/ both in experiments with point contacts and in numerical simulations using the resistively
shunted junction (RSJ) model. This excess noise causes the mixer noise temperature to be
a factor of typically 20- 100 times^ the physical temperature of the device. Previously, this
excess was ascribed to conversion from unwanted sidebands of the local oscillator and
Josephson frequencies and their harmonics^. Our numerical modeling of the RSJ
equations has led to a new understanding of the excess noise, which is simply due to the
intrinsic Josephson oscillations of the device. In addition, we have extended the modeling
to include the previously ignored case of finite device capacitance (i.e. RSJ capacitance
parameter (3c ^ 0), which is more realistic for lithographically defined Josephson such as
shunted tunnel junctions or SNS bridges. For some cases, this yields an improvement of a
factor of two in noise temperature from the zero capacitance models. We will discuss the
device parameters which optimize the mixer performance for frequencies approaching the
characteristic frequency of the device, which is given by the Josephson frequency at the
IcR n voltage (v = 2eIcRn/h). These modeling results predict good conversion efficiency
and a noise temperature within a factor of a few of the physical temperature. Experiments
are in progress to determine the accuracy of this modeling using a waveguide mixer at
100 GHz with optimized, resistively shunted Nb tunnel junctions. If the modeling results
are valid, they are particularly encouraging for mixers in the submillimeter regime, given
the possibility of obtaining non-hysteretic Josephson devices with IcR n products in excess
of a millivolt, using, for instance, high-T c SNS bridges. We discuss the modifications to
the classical RSJ model which are necessary in the quantum regime (hv > kT), and
conclude the Josephson mixers may attain noise temperatures less than ten times the
quantum limit at high frequencies.
*R. Schoelkopf, T. Phillips, and J. Zmuidzinas are with California Institute of
Technology.
MY. Taur, IEEE Transactions on Electron Devices, ED-27, p. 1921, 1980.
t^K.K. Likharev, Dynamics of Josephson Junctions and Circuits, New York:
Gordon and Breach, p. 423, 1986.
Page 420 Third International Symposium on Space Terahertz Technology
FABRICATION AND CHARACTERIZATION OF HIGH CURRENT-DENSITY,
^ -y^p/L SUBMICRON, NbN/MgO/NbN TUNNEL JUNCTIONS
/hoSer/ N93-27763
Oy /. A. Stern " **
U ^ H. G. LeDuc
A. J. Judas*
Center for Space Microelectronics Technology
Jet Propulsion Laboratory
California Institute of Technology
Pasadena, California 91109
* Present Address: Stanford University
Palo Alto, California
Abstract
At near-millimeter wavelengths, heterodyne receivers based on SIS tunnel junctions are the most
sensitive available. However, in order to scale these results to submillimeter wavelengths,
certain device properties should be scaled. The tunnel-junction's current density should be
increased to reduce the RC product. The device's area should be reduced to efficiently couple
power from the antenna to the mixer. Finally, the superconductor used should have a large
energy gap to minimize RF losses. Most SIS mixers use Nb or Pb-alloy tunnel junctions; the
gap frequency for these materials is approximately 725 GHz. Above the gap frequency, these
materials exhibit losses similar to those in a normal metal. The gap frequency in NbN films is
as-large-as 1440 GHz. Therefore, we have developed a process to fabricate small area (down
to 0.13 n 2 ), high current density, NbN/MgO/NbN tunnel junctions.
Third International Symposium on Space Terahertz Technology Page 421
In this paper, we describe a process used to fabricate submicron NbN junctions. Low-leakage
current-voltage (I-V) characteristics are achieved for current densities up to 40 kA/cm 2 .
However, the quality of the I-V characteristics degrades significantly for higher current densities.
Junction areas are patterned by lifting off a Cr stencil defined by electron beam lithography.
This image is transferred to a polyimide stencil using reactive ion etching (RIE). The junctions
are then etched and isolated using a self-aligned liftoff technique. The limitations of this
technique and the quality of the resulting I-V characteristics will be discussed.
There are several device and materials parameters which must be known to properly design
mixer circuits. The optimal imbedding impedance for the mixer is determined by the
capacitance and I-V characteristic of the tunnel junction. If microstrip line circuits are used to
achieve this impedance, the magnetic penetration depth of the NbN must be known to calculate
the microstrip line impedance and propagation velocity. We have measured the junction
capacitance versus current density and microstrip line inductance using superconducting quantum
interference devices (SQUIDs). The propagation velocity was measured using long open-ended
microstrip-lines connected to Josephson tunnel junctions. The magnetic penetration depth can
be calculated from either the microstrip line inductance or the propagation velocity. The
implications of these measurements will be discussed.
Introduction
There is a growing need for sensitive submillimeter (smm) wavelength detectors for both ground
based and space based applications. Heterodyne receivers based on superconductor-insulator-
superconductor (SIS) tunnel junctions are the most sensitive at millimeter and near-millimeter
Page 422 Third International Symposium on Space Terahertz Technology
wavelengths [1][2]. At wavelengths longer than 3 mm, the noise in these detectors has
approached the quantum limit [3]. If these mixers designs could be redesigned for shorter
wavelength operation, they would fill the need for smm wavelength mixers. For good
performance to be achieved at smm wavelengths, a number of SIS tunnel junction properties
must be adjusted. We have fabricated and characterized tunnel junctions which are suitable for
smm use.
The junction capacitance is the property which has the largest impact on smm wave mixers. An
SIS tunnel junction can be modeled as a nonlinear tunneling element in parallel with a shunt
capacitance. The product of the shunt capacitance (C) and the RF resistance of the junction
(approximately the normal state resistance, RJ is one limit on the high frequency response of
a tunnel junction, because the capacitance tends to shunt RF currents away from the nonlinear
tunneling element. To reduce the R n C time constant, the tunnel junction barrier is made
thinner; this increases the junction capacitance
C=e e r A/d,
where e is the dielectric constant of free space, e r is the relative dielectric constant of the
barrier, A is the junction area and d is the barrier thickness, but it decreases the resistance more
rapidly
„ _ h 2 d-Q\p{4ird\/2m(f) /h}
e 2
■A\Jlm<\>
where h is Plank's constant, m is the mass of the quasi-particle, <t> is the barrier height and e the
charge of the quasiparticle. Because the barrier cannot be made thinner than one monolayer,
the R,jC product is ultimately limited by the barrier material properties (e r , and the lattice
Third International Symposium on Space Terahertz Technology Page 423
constant). The R n C product is independent of junction area, but the junction resistance must be
roughly 100 Q for it to couple efficiently to the antenna. Therefore, as the barrier is made
thinner, the tunnel junction's area must also be reduced. At smm wavelengths, junctions with
submicron areas are needed. Since the barrier thickness is difficult to measure, R„A products
can be used as an area-independent measure of junction relaxation time. The product of the
critical current density (J c ) and R„A is proportional to the energy gap, so the current density is
also frequently used as a measure of the relaxation time.
The embedding circuit of a mixer can be designed to resonate out the junction capacitance over
a frequency band. In waveguide mixer mounts, this can be accomplished using a back short and
E-plane tuner; however, at high frequencies these tuners do not work as well as at lower
frequencies. Additionally, quasi-optical mixers do not have these tuning elements. Monolithic
embedding circuits can also be fabricated to resonate out the junction capacitance
[4] [5] [6]. Many of these circuits use superconducting microstrip lines. To accurately
design these circuits, the propagation velocity and impedance of the microstrip line must be
known. Both of these parameters depend critically on the magnetic penetration depth of the
superconductor [7].
The energy gap of the superconductor (2A) limits the use of SIS mixers at smm wavelengths.
*
Above the gap frequency (2A/h), the superconductor has losses similar to those in a normal
metal; therefore, any embedding circuits will have losses[8]. Embedding circuits could be made
of a larger gap superconductor or a low-loss normal metal (such as gold or copper).. In this
case, the SIS mixer will perform well up to twice the gap frequency. Most mixers currently use
Page 424 Third International Symposium on Space Terahertz Technology
niobium based tunnel junctions. The gap frequency of niobium is 725 GHz. Above this
frequency, other superconductors will probably be needed.
Based on the small gap in niobium, we have developed a process for fabricating junctions with
NbN electrodes, which has a much larger gap frequency (up to 1440 GHz). The goal was to
fabricate high current-density, submicron tunnel junctions with large energy gaps. The
limitations of our process on both current density and junction area are discussed below.
Additionally, we have measured several device properties necessary to design mixer circuits
(NbN magnetic penetration depth, junction uniformity, yield and specific capacitance.)
Junction Fabrication
The tunnel junctions are fabricated by depositing the NbN/MgO/NbN trilayer over the entire
substrate [9][10]. The NbN films are deposited by reactive DC-magnetron sputtering from a
niobium target using a 5 cm diameter US inc. sputter gun in an argon and nitrogen atmosphere.
The gas flow rates are 150 and 15 seem respectively and the total pressure with the plasma off
is 14.4 mTorr. The gun is typically biased at -232 Volts and 600 mAmps. The substrate is not
heated and is approximately 6 cm from the sputter gun. Deposition rates are 45 nm/min. The
thickness of the base and counter electrode are 315, 135 nm respectively. The superconducting
transition temperature of the films is usually between 14.5 and 15.0 K.
The barrier layer is formed by depositing MgO and performing a pinhole cure in an oxygen
plasma. MgO is deposited by RF sputtering from a US inc. 5 cm diameter source in a pure
argon atmosphere. The Ar pressure is 10.0 mTorr and the power is 50 Watts. To promote
Third International Symposium on Space Terahertz Technology Page 425
uniform growth of the barrier, deposition occurs intermittently as the substrate is rotated past
the source in a circular orbit. After the barrier deposition, the sample is exposed to an oxygen-
plasma glow discharge for 1 minute at 100 mTorr. The plasma is maintained with a high purity
aluminum ring below the substrate; the ring is biased at -400 Volts and 2.0 mA. Barrier
deposition times vary between 5.5 and 6.7 min. for current densities between 40 and 10 kA/cm 2 .
The trilayer is completed by covering the entire NbN/MgO/NbN structure with 30 nm of gold.
The gold prevents the top surface of the junction from oxidizing in air. In some cases, there
is a thin (40 nm) layer of Nb between the counter electrode and the gold. The purpose of the
Nb interlayer will be discussed below.
Tunnel junctions are fabricated using a submicron, self-aligned lift-off technique. This process
is shown in figure 1 . The trilayer is patterned using RIE and a photoresist stencil. The etcher
used is a Semi Group System 1000, and the electrode area is 730 cm 2 . The gold and MgO layers
are etched using a straight argon sputter etch. The NbN layers are etched in a CF 4 -0 2 mixture.
The conditions for this and other reactive ion etches are given in table 1. After the etch, the
photoresist is stripped in organic solvents.
The submicron tunnel junction stencil is formed in a polyimide layer [11]. The etched
trilayer is coated in Ciba Geigy Probimide 286 and thinner 2:1 at 4000 rpm resulting in a 310
nm polyimide layer. The polyimide is cured at 150° C on a hot plate for 15 minutes and flood
exposed with 315 nm radiation. A KTI Chemicals Inc. 950K 2% PMMA layer is spun on top
of the polyimide at 4000 rpm and baked at 150° C. A thin aluminum layer is evaporated on top
Page 426 Third International Symposium on Space Terahertz Technology
of the PMMA to prevent charging during the exposure. The sample is exposed in a JEOL model
JBX5 electron beam lithography system and subsequently developed to reveal square holes in
the PMMA. A thin (35 nm) chromium layer is thermally evaporated and lifted off to define
the junction area. Next an AZ 5214-E stencil is patterned using image reversal [12] to reveal
an open area around the chromium dot. This stencil acts to protect the majority of the trilayer
from the junction etch. The polyimide layer is etched using an oxygen plasma. During the etch,
a polyimide (Kapton) sheet is about 3 cm above the sample to protect against particles falling
on the sample during the etch. The oxygen etch is very anisotropic and an over etch is
normally used to insure that the trilayer surface is completely clean.
The junction etch consists of an Ar sputter-etch of the gold and one of several NbN etches
[13]. For micron size junctions, the CF 4 -0 2 mixture can be used, however this etch is isotropic
and it will undercut NbN mesa. A straight CF 4 etch can be used for submicron junctions. This
etch is largely anisotropic, because a fluorinated carbon layer forms on the sidewall of the
junction as it etches; however, there is about 50 nm of undercut. A mixture of CCl 2 F2-CF 4 -0 2
produces an anisotropic etch, however it reacts at the Au-NbN interface and produces a
superconductor-normal metal-superconductor type I-V characteristic in series with the SIS I-V
characteristic. This series weak link can be avoided by inserting a niobium layer between the
gold and NbN layers and by rinsing the sample in water after the etch. The niobium buffer
removes the series weak link which seams to occur at the NbN-Au interface Regardless of the
etch used, a monitor sample is used to detect the endpoint of the junction etch. After the
junction etch, a 1 minute argon sputter etch is done to improve the adhesion of the dielectric
isolation layer.
Third International Symposium on Space Terahertz Technology
Page 427
Table 1 . Summary of Reactive Ion Etch Conditions
Material
Etch
Gases
Flow Rates
(seem)
Pressure
(mTorr)
Power
(Watts)
Etch Rate
(nm/min)
Polyimide
o 2
20
30
130
100
Gold
Ar
20
30
118
15
NbN
CF 4 ,0 2
20,2
30
118
65
NbN
CF 4
20
30
118
52
NbN
CC1 2 F 2 ,
CF 4 ,0 2
23,7,2
30
118
58
The junction is electrically isolated by thermally evaporating 150 nm of Silicon monoxide on the
sample. The wafer is rotated at a 5 to 15° angle during the evaporation to improve SiO
coverage. The photoresist protection layer is then lifted off in acetone, followed by the
polyimide in dichloromethane. The polyimide liftoff is done in an ultrasonic cleaner followed
by a mechanical scrub with a q-tip to remove the SiO flags that form on the side- walls of the
polyimide pillar. The liftoff yield, even for the smallest junctions, is excellent. Following the
polyimide liftoff, an oxygen etch is done to remove any polyimide residue which will lead to a
contact resistance.
The NbN wiring layer is deposited on top of the electrically isolated junction. To avoid breaks
at step edges, the wiring is typically 600 nm thick. The wiring is patterned and etched like the
trilayer, although if small features are desired, one of the anisotropic etches is used.
Results
Page 428
Third International Symposium on Space Terahertz Technology
The quality of tunnel junction I-V characteristics depends strongly on the critical current density.
As the current density increases, the ratio of the subgap resistance (measured at 3 mV) to the
normal state resistance decreases. Figure 2 shows a plot of this ratio as a function of critical
current density. Above 50 kA/cm 2 , the ratio R S g/Ro is reduced to two; however, for current
densities up to 40 kA/cm 2 the I-V characteristics are still reasonable. An example of a 0.42x0.42
(i 2 , 35 kA/cm 2 junction is shown in figure 3; this junction was fabricated for 626 GHz mixer
tests.
A second form of I-V degradation occurs at high current densities. For a fixed area junction,
the gap voltage decreases as the current density increases. This is caused by a non-thermal
distribution of quasi-particles near the barrier. As the tunneling current increases, the number
of quasi-particles injected into the superconducting electrode increases; this represents a local
heating of the quasiparticle bath and leads to a reduction in the gap voltage. The quasi-particle
heating can be minimized by reducing the junction area. Table 2 shows the gap for different
area junction on the same wafer; the current density is 28 kA/cm 2 .
Table 2. Energy Gap as a Function of Junction Size
size (microns)
Energy Gap (meV)
1.0
4.15
0.7
4.3
0.5
4.4
0.36
4.45
Third International Symposium on Space Terahertz Technology Page 429
Junction uniformity and yield was measured by fabricating series arrays of 100 tunnel junctions.
Typical I-V characteristics for arrays of 0.49, 0.25 and .13 ^junctions are shown in figure 4.
For these size junctions, the standard deviation of the critical currents is typically ±5.4, 6.6 and
7.7% respectively. For comparison, the standard deviation for optically defined, 4 y? junctions
is 3.9%. Typical minimum-to-maximum uniformity is ±17% for 0.25 n 2 junctions. When the
fabrication process was successful, junction yields were better than 99 %. The uniformity and
yield for this process were high enough that a monolithic array of mixers could be fabricated.
SQUIDs have been fabricated to measure the magnetic penetration depth of NbN films and the
specific capacitance of NbN/MgO/NbN junctions. The design of the SQUIDs is based on that
of Magerlein [14] and is described in detail elsewhere [15]. Briefly, the SQUID design
is a microstrip line over a ground plane with two tunnel junctions connecting the microstrip to
the ground plane. The inductance of the microstrip line, and therefore the magnetic penetration
depth, can be determined by observing the critical current of the two parallel-junctions versus
control current passing along the microstrip line. The loop of the SQUID and the junction
capacitance form an LC tank circuit. The Josephson oscillations will interact with this tank
circuit to form a current step at a voltage proportional to this resonant frequency. The junction
capacitance can be determined from the resonance voltage and the critical-current modulation
curve.
The magnetic penetration depth can also be determined using an open-ended microstrip-line
connected to a Josephson junction. The microstrip line has an impedance which is periodic with
frequency. The Josephson oscillations interact with the microstrip line circuit to yield a series
Page 430 Third International Symposium on Space Terahertz Technology
of resonances whenever the microstrip line reflects an inductance that cancels out the junction
capacitance. An example of these resonances is shown in figure 5. The propagation velocity of
the microstrip line can be calculated to be
V p =e-5V//i-2L,
where 5V is the spacing or the resonances and L is the length of the microstrip line stub. The
penetration depth can be calculated from the propagation velocity. These microstrip line stubs
are a good method of verifying the SQUID results and directly measuring the microstrip line
propagation velocity.
The specific capacitance of our junctions as a function of J c is shown in figure 6 for critical
current densities ranging from 500 to 50,000 A/cm 2 . Much of this data has been published
previously 14 ; however, we have extended the data to include the highest current densities shown
here (10 to 40 kA/cm 2 ). Also shown in figure 6 is the R„C frequency as a function of current
density. At 40 kA/cm 2 , the roll off frequency is only 140 GHz; therefore, at smm wavelengths,
the capacitance must be tuned out to achieve optimal performance.
The magnetic penetration depth of our NbN films has also been measured using SQUIDs. The
penetration depth varies significantly with film quality and is typically between 270 and 380 nm.
The propagation velocity of microstrip lines with 150 nm SiO dielectric layers is typically 0.16
to 0.18 times the speed of light in a vacuum. The penetration depth calculated from these
numbers is 300 to 365 nm, which agrees well with the SQUID results. Because the penetration
depth depends critically on film quality, SQUID and microstrip line circuits are usually added
to mixer mask-designs, so the penetration depth can be measured for each set of tunnel junctions
Third International Symposium on Space Terahertz Technology Page 431
that are fabricated. These circuits are small (2.5 mm x 2.5 mm) and 12 test dies take up
roughly one sixth of a 25 mm diameter quartz wafer.
Conclusions
We have developed a process for fabricating high critical current, submicron NbN/MgO/NbN
tunnel junctions. The yield and uniformity of this process is good and should be sufficient for
most mixer needs. However, the RJ2 frequency can only be made 140 GHz without seriously
degrading the I-V quality. Submillimeter wave mixers will either require high Q circuits to
resonate out the junction capacitance or a new barrier material with a lower dielectric constant
or barrier height. Although NbN films should have low RF losses even at smm wavelengths, the
actual losses need to be measured in order to evaluate potential high Q tuning circuits. In
addition, the effects, if any, of quasi-particle heating on mixer performance need to be
investigated.
Work Supported by NASA and SDI/IST
References
[1] C. K. Walker, M. Chen, P. L. Shafer, H. G. LeDuc, J. E. Carlstrom and T. G. Phillips,
"A 492 GHz SIS Waveguide Receiver for Submillimeter Astronomy," Int. J. of IR and
Millimeter Waves, Submitted 1992.
[2] T. H. Buttgenbach, H. G. LeDuc, P. D. Maker and T. G. Phillips, "A Fixed Tuned
Broadband Matching Structure for Submillimeter SIS Receivers," IEEE Trans. Appl.
Superconductivity, Submitted Feb. 1992.
[3] C. A. Mears, Qing Hu, P. L. Richards, A. H. Worsham, D. E. Prober and A. V.
Raisanen, "Quantum Limited Quasiparticle Mixers at 100 GHz," IEEE Trans, on Magn., vol.
27, no. 2, 1991.
Page 432 Third International Symposium on Space Terahertz Technology
[4] L. R. D'Addario, "An SIS Mixer for 90-120 GHz with Gain and Wide Bandwidth," Int.
J. ofIR and Millimeter Waves, vol. 5, no. 11, 1419-1442, 1984.
[5] A. V. Raisanen, W. R. McGrath, P. L. Richards and F. L. Lloyd, "Broad-Band Match
to a Millimeter- Wave SIS Quasi-Particle Mixer," IEEE Trans, on Microwave Theory and
Technique, vol. 4, no. 12, December 1985.
[6] S. K. Pan, A. R. Kerr, M. J. Feldman, A. W. Kleinsasser, J. Stasiak, R. L. Sandstrom
and W. J. Gallagher, "An 85-116 GHz SIS Receiver Using Inductively Shunted Edge- Junctions
," IEEE Trans, on Microwave Theory and Technique, vol. 37, no. 3, 580-592, March 1989.
[7] W. H. Chan, "The Inductance of a Superconducting Strip Transmission Line, " J. Appl.
Phys., vol.50, no. 12, Dec. 1979.
[8] R. L. Kautz, "Picosecond Pulses on Superconducting Striplines," J. Appl. Phys., vol.
49, 308-314, 1978.
[9] H. G. LeDuc, J. A. Stern, S. Thakoor and S. Khanna, "All Refractory NbN/MgO/NbN
Tunnel Junctions," IEEE Trans. Magn. vol. 23, March 1987.
[10] J. A. Stern, B. D. Hunt, H. G. LeDuc, A. Judas, W. R. McGrath, S. R. Cypher and
S. K. Khanna, "NbN/MgO/NbN SIS Tunnel Junctions for Submm Wave Mixers," IEEE Trans.
Magn., vol 25, 1989.
[11] A. W. Lichtenberger, D. M. Lea, C. Li, F. Lloyd, M. J. Feldman and R. J. Mattauch,
"Fabrication of Micron Size Nb/Al-Al 2 3 /Nb Junctions with a Trilevel Resist Liftoff Process, "
IEEE Trans on Magnetics, vol. 27, no. 2, March 1991.
[12] M. Spak, D. Mammato, S. Jain and D. Durham, "Mechanism and Lithographic
Evaluation of Image Reversal in AZ 5214 Photoresist," As Presented at: Seventh International
Technical Conference on Photopolymers, Ellenville, New York, Reprints Available from
American Hoechst Corporation, AZ Photoresist Products.
[13] H. G. LeDuc, A. Judas, S. R. Cypher, B. Bumble, B. D. Hunt and J. A. Stem,
"Submicron Area NbN/MgO/NbN Tunnel Junctions For SIS Mixer Applications," IEEE Trans,
on Magnetics, vol. 27, no. 2, March 1991.
[14] J. H. Magerlein, "Specific Capacitance of Josephson Tunnel Junctions," IEEE Trans.
Magn., vol. 17, no. 2, 286-289, 1981.
[15] J. A. Stem and H. G. LeDuc, "Characterization of NbN Films and Tunnel Junctions,"
IEEE Trans. Magn., vol 27, no. 2, March 1991.
Third International Symposium on Space Terahertz Technology Page 433
Fi gure Captions
Figure 1 Submicron, self-aligned liftoff process.
Figure 2 Subgap resistance divided by the normal state resistance as a
Junction of critical-current density
Figure 3 I-V characteristic of a .42x.42 p. 2 , 35 kA/cm 2 , tunnel junction. The
current scale is 20 pA/div. and the voltage scale 1 mV/div.
Figure 4 I-V characteristics for series arrays of 100 tunnel junctions. The
voltage scale is 50 mV/div. The area of the junctions is 0.13, 0.25,
0.49 p. 2 , and the current scale is 10 20 and 50 pA/div for figures
a, b and c respectively.
Figure 5 I-V characteristic of a 1 p 2 tunnel junction connected to a 6 p by
750 p open-ended microstrip-line stub. The current scale is 5
pA/div, and the voltage scale is 100p.V/div. The resonance spacing
is 65 p.V leading to a propagation velocity of 0.16 times the speed
of light.
Figure 6 Junction specific capacitance (a) and R n C roll-off frequency
(l/2irR„C) (b) as a function of critical-current density.
Page 434
Third International Symposium on Space Terahertz Technology
Figure 1.
[ a )
\\\\\\\\\\\\\\\\\\\\\^^^
NbN
mtzmmmm^m^.
NbN
V N \ N \ V \ N \ N \ N \ N \ N \ N \ X \ N \ N \ N \ N \ N \ N \ N \ N \ N \ N \ X \ N \ N \ N \ X
Y\\y\\\\\\Y\\\a\\\\\\\\Y
\\ v\ \\ 0\ \\ s\ n\ v\ \\ v\ \\ 0\ \\ \\ s\ \\ \\ v\ s\ \\ v\ n\ s\ v\ \
[ b )
SiO
Third International Symposium on Space Terahertz Technology
Page 435
Figure 2.
Rsg/Rn vs. J c
12.
8.0 -
0.0
4.0 L
5.0 - 10.0
20.0 50.0
100.0
200.0
kA/cm 2
Figure 3.
Page 436
Third International Symposium on Space Terahertz Technology
Figure 4.
lOmV
Third International Symposium on Space Terahertz Technology
Page 437
Figure 5.
■4
'i
i
i
Page 438
Third International Symposium on Space Terahertz Technology
Figure 6.
(a)
200.0
fF/u
Specific Capacitance vs. Current Density
0.2 0.5 1.0 2.0 5.0 10.0 20.0 50.0 100.0
kA/cm 2
(b)
RC Roll— Off Frequency vs. Current Density
160.0
2.0 5.0 10.0 20.0 50.0 100.0
kA/cm 2
Third International Symposium on Space Terahertz Technology Page 439
A Quasioptical Resonant-Tunneling-Diode Oscillator " ,
Operating Above 200 GHz fj 9 ^S^-^^T^ 4
E. R. Brown*, C. D. Parker*, K. M. Molvar*, A. R. Calawa*, and M. J. Manfra* /&$^
We have fabricated and characterized a quasioptically stabilized resonant-tunneling-
diode (RTD) oscillator having attractive performance characteristics for application as a
radiometric local oscillator. The fundamental frequency of the oscillator is tunable from
about 200 to 215 GHz, the instantaneous linewidth is between 10 and 20 kHz, and the
output power across the tuning band is about 50 H.W. The narrow linewidth and fine
tuning of the frequency are made possible by a scanning semiconfocal open cavity which
acts as the high-Q resonator for the oscillator. The cavity is compact, portable, and
insensitive to vibration and temperature variation. The total dc power consumption (RTD
plus bias supply) is only 10 mW.
The present oscillator provides the highest power obtained to date from an RTD
above 200 GHz. We attribute this partly to the use of the quasioptical resonator, but
primarily to the quality of the RTD. It is fabricated from the Ino.53 Gao.47As/AlAs
materials system, which historically has yielded the best overall resonant-tunneling
characteristics of any material system. The RTD active area is 4 (im 2 , and the room-
temperature peak current density and peak-to- valley current ratio are 2.5xl0 5 A cnr 2 and 9,
respectively. The RTD is mounted in a WR-3 standard-height rectangular waveguide and
is contacted across the waveguide by a fine wire that protrudes through a via hole in a
Si3N4 "honeycomb" overlayer. We estimate that the theoretical maximum frequency of
oscillation of this RTD is approximately 1.1 THz, and that scaled-down versions of the
same quasioptical oscillator design should operate in a fundamental mode up to frequencies
of at least 500 GHz.
This work was sponsored by NASA-OAST through the Jet Propulsion Laboratory and by the U.S.
Army Research Office.
*E. R. Brown, C. D. Parker, K. M. Molvar, A. R. Calawa, and M. J. Manfra are
with Lincoln Laboratory, Massachusetts Institute of Technology.
a-Q
(
r
Page 440 Third International Symposium on Space Terahertz Technology
Transit-Time Devices as Local Oscillators for Frequencies Above 100 GHz *)
-5JST-35 N93-27765
H. Eisele. C. Kidner, G. I. Haddad
/& O 5^3
. \ V Center for Space Terahertz Technology
~^ Department of Electrical Engineering & Computer Science
2231 EECS Building
The University of Michigan
Ann Arbor, Michigan 48 109-2122
Abstract:
Very promising preliminary experimental results have been obtained from GaAs IMP ATT diodes at F-
band frequencies (75 mW, 3.5 % @ 111.1 GHz and 20 mW, 1.4 % @ 120.6 GHz) and from GaAs
TUNNETT diodes at W-band frequencies (26 mW, 1.6 % @ 87.2 GHz and 32 mW, 2.6 % @ 93.5 GHz).
These results indicate that IMPATT, MIT ATT and TUNNETT diodes have the highest potential of deliv-
ering significant amounts of power at Terahertz frequencies. As shown recently, the noise performance of
GaAs W-band IMPATT diodes can compete with that of Gunn devices. Since TUNNETT diodes take
advantage of the quieter tunnel injection, they are expected to be especially suited for low-noise local
oscillators. This paper will focus on the two different design principles for IMPATT and TUNNETT
diodes, the material parameters involved in the design and some aspects of the present device technology.
Single-drift flat-profile GaAs D-band IMPATT diodes had oscillations up to 129 GHz with 9 mW, 0.9 %
@ 128.4 GHz. Single-drift GaAs TUNNETT diodes had oscillations up to 112.5 GHz with 16 mW and
output power levels up to 33 mW and efficiencies up to 3.4 % around 102 GHz. These results are the best
reported so far from GaAs IMPATT and TUNNETT diodes.
*) This work was supported by NASA under contract No. NAGW 1334.
Third International Symposium on Space Terahertz Technology Page 441
1. Introduction
GaAs IMPact ionization Avalanche Transit-Time (IMP ATT) diodes have long been thought to be limited
to frequencies below 60 GHz. Little has been reported regarding the operation of GaAs IMP ATT or
MITATT diodes above 100 GHz [1,2]. Experimental results of W-band IMPATT diodes (up to 320 mW,
6.0 % @ 95 GHz) [3] with excellent noise performance [4] clearly indicate that IMPATT diodes are one
prime candidate to fulfill the growing need for local oscillators above 100 GHz. TUNNE1 injection
Iransit-Time (TUNNETT) diodes were already proposed in 1958 and are considered another prime can-
didate for low-noise, medium power sources at millimeter and submillimeter frequencies. Although
pulsed oscillations were demonstrated up to 338 GHz in 1979 [5], CW power has only recently been
obtained from devices with low impact ionization carrier multiplication [6]. This significant progress is
mainly due to the fact that refined epitaxial growth techniques have become widely available. Despite the
impressive progress in oscillators with three-terminal devices at mm-wave frequencies [7] two-terminal
devices hold the highest potential in delivering significant amounts of power with clean spectra above
100 GHz.
2. Design of single-drift flat-profile IMPATT diodes
In GaAs, the first derivative of the ionization rates of electrons and holes with respect to the electric field
saturates around 500 kVcnr 1 [8-10]. Together with dead space effects in the avalanche zone [11] this sat-
uration phenomenon favors a flat-profile structure for frequencies above V-band (50 - 75 GHz). The per-
formance of a single-drift structure is the least sensitive to doping profile variations. The design of this
structure is based on the assumption that the center of the avalanche region occurs where the electron
concentration equals the hole concentration for the applied bias voltage [12] and that such a defined
avalanche region is electrically equivalent to an avalanche region of the same width / a but constant elec-
tric field and ionization rates [10]. The drift region - where ionization is to be neglected - is defined in its
length / d by the maximum in the well known transit-time function [13]
3v s
/ d = • CD
8/o
where v s is the average saturated drift velocity (4.5 x 10 6 cms 1 in GaAs for 7j = 500 K) [9,10] and/ the
operating frequency. Several structures for operating frequencies between 130 GHz and 160 GHz have
been designed and the nominal doping profile of such a p ++ nn + structure is given in Figure 1. A bias cur-
rent density Joe °f 60 kAcnr 2 extrapolated from the experimental results in W-band [3] was taken into
account.
Page 442 third International Symposium on Space Terahertz Technology
3. Design of single-drift TUNNETT diodes
The design of the TUNNETT diode structure is based on a first order large signal theory [14] and experi-
mental studies of highly doped MBE grown p ++ n + structures. The carrier generation rate due to tunneling
does not depend on the current density, but does strongly depend on the electric field. Therefore, a sharp
pulse of electrons is injected at the p ++ n + junction when the RF field reaches its maximum.
Under these assumptions for the p ++ n + nn + structure the first order large signal theory predicts a maximum
in RF output power and DC to RF conversion efficiency [13,14] for
5v s
'i + 'd = . (2)
where /j is the length of the n + region in the p ++ n + junction, / d is the length of the n region, v s is the aver-
age saturated drift velocity (4.6 x 10 6 cms 1 for 7] = 500 K) [9,15] and/ the operating frequency. Since
the design is based on considerably lower electric fields in the drift region compared to the ones in the
IMPATT diodes above, a slightly higher value for v s is appropriate. Further details of the design proce-
dure are given in References 16 and 17. The carrier concentration due to a current density /qc of
25 kAcnr 2 is taken into account in the doping profile. The nominal doping profile of this p ++ n+nn +
TUNNETT diode structure is depicted in Figure 2.
4. Device technology
The operating current density of 60 kAcnr 2 in a single-drift flat-profile GaAs D-band IMPATT diode
requires a diamond heat sink to keep the operating junction temperature below 250 °C. Therefore all
IMPATT diodes were fabricated using a well established selective etching technology for substrateless
diodes on diamond heat sinks which gives up to 600 diodes per cm 2 wafer area with high uniformity [18].
This technology implements an Al 55 Ga 45 As etch-stop layer between the substrate and the epitaxial layers
for the device. In order to get the steep transitions for doping profiles in the submicron range, all wafers
were grown by MBE. Figure 3 shows the flow chart of this technology process. The epitaxial side of the
wafer is metallized with Ti/Pt/Au for a p + ohmic contact, then selectively plated with gold to form a grat-
ing for mechanical support and glued on a ceramic carrier. In the next step the substrate is removed by
selective wet chemical etch and subsequently the etch-stop layer in another selective wet chemical etch. A
Ni/Ge/Au contact metallization is evaporated on top of the n + layer and plated with gold to ease bonding.
Contact patterns and diode mesas are defined by standard positive photoresist technology and wet chemi-
cal etching. The diodes outside the supporting grating are tested and selected for good DC characteristics
Third International Symposium on Space Terahertz Technology Page 443
and thermocompression bonded on diamond heat sinks which are embedded in plated copper blocks.
Electrical contact to the diode is provided by four metallized quartz stand-offs thermocompression
bonded onto the heat sink and tapered gold ribbons bonded on the diode and the stand-offs.
The TUNNETT diodes were designed to operate at a maximum current density of 25 kAcnr 2 and a DC
bias voltage comparable to the one of the IMP ATT diodes. This allows fabrication of TUNNETT diodes
with an integral heat sink. The wet chemical etching in the previously described technology limits the
choice of materials and the minimum diameter for the n + ohmic contact. Therefore a different process has
been developed, which likewise implements an Al 55 Ga 45 As etch-stop layer between the substrate and the
epitaxial layers for the device. Its flow chart is given in Figure 4 and further details are discussed in
Reference 19.
Before the epitaxial side of the MBE-grown wafer is metallized with Ti/Pt/Au for a p + ohmic contact,
grooves are selectively etched down to the AlGaAs etch-stop layer to divide the device layers into square
shaped islands. This reduces the stress in the device layers during annealing. Furthermore, it shapes the
Ti/Pt/Au layers and the plated gold layer of the integral heat sink thus providing additional mechanical
strength for the subsequent processing steps after the substrate has been removed. The contacts are
defined by standard lift-off technology and an additional metallization and photolithography step gives
holes on top of the n + ohmic contact through which up to 3 urn of gold is electroplated. The mesas are
formed by a wet chemical etch. After annealing the sample is diced into individual diodes and diodes with
the desired size and DC characteristic are soldered or glued to a gold plated copper block. Electrical con-
tact to the diode is provided by four metallized quartz stand-offs thermocompression bonded onto the
plated block and tapered gold ribbons bonded on the diode and the stand-offs.
5. Experimental results
RF testing is performed in full height waveguide cavities with a resonant cap on top of the diode.
IMP ATT diodes are tested both in W-band (WR-10 waveguide) and D-band (WR-6 waveguide).
TUNNETT diodes are only tested in W-band.
Figure 5 shows RF output power, DC to RF conversion efficiency and oscillation frequency of the best
IMP ATT diode in a W-band cavity as a function of the bias current. At each bias point the short plunger
of the cavity and at some bias points also the E-H-tuner were adjusted for maximum output power. As can
be seen from Figure 5, the efficiency reaches its maximum of 3.8 % at an output power of 72 mW.
An output power of 85 mW at 102.0 GHz with an efficiency of 2.5 % in WR-10 waveguide cavity and
20 mW at 120.6 GHz with an efficiency of 1.4 % in a WR-10 waveguide cavity were obtained from other
Page 444 Third International Symposium on Space Terahertz Technology
IMPATT diodes. The highest oscillation frequency of 128.4 GHz could be observed in a WR-6 wave-
guide cavity. At this frequency the output power was 9 mW and the efficiency 0.9 %. Table 1 summarizes
the experimental results obtained from these diodes. The operating junction temperature was limited up to
Tj = 550 K in order to ensure reliable long-term operation.
Frequency [GHz]
102.0
111.1
111.5
120.6
128.4
Output power [mW]
85
75
48
20
9
Efficiency [ % ]
2.5
3.5
2.3
1.4
0.9
Cavity (W/D)
W
W
W
D
D
Table 1: Experimental results of IMPATT diodes in W-band and D-band cavities.
To verify the mode of operation the DC I-V characteristics are measured at room temperature (7 = 300 K)
and an elevated temperature (7 = 370 K). As shown in Figure 6a for low bias currents and Figure 6b for
high bias currents, the breakdown of the D-band IMPATT diode is sharp and the bias voltage always
increases with increasing temperature due to the decreasing ionization rates. The breakdown voltage at
7= 300 K agrees well with breakdown voltages that were calculated from ionization rates evaluated in
Reference 9, and which are plotted in Figure 7 together with the peak electric field strength. The sharp
breakdown also proves that tunneling is significant only for electric field strengths above 1.0 MVcnr 1 .
Figure 8 shows RF output power, DC to RF conversion efficiency and oscillation frequency of two
TUNNETT diodes in W-band cavities as a function of the bias current. At each bias point the short
plunger and the E-H-tuner were adjusted for maximum output power. As can be seen from Figure 8a and
8b, neither output power nor efficiency saturate up to the highest applied bias currents. The oscillation
frequency varies only slightly and is mainly determined by the resonant cap. An output power of 33 mW
at 93.5 GHz with an efficiency of 2.65 % and an output power 31.5 mW at 107.36 GHz with an efficiency
of 3.35 % were obtained. The highest oscillation frequency of another diode was 112.5 GHz with an
output power of 16 mW and the corresponding efficiency of 2.55 %. Table 2 summarizes the
experimental results obtained from the so far best diodes. The operating junction temperature was well
below 7j = 550 K in each case.
Frequency [GHz]
87.22
93.50
102:66
107.30
112.50
Output power [mW]
27
33
33
31.5
16
Efficiency [ % ]
1.75
2.65
3.35
3.35
2.55
Table 2: Experimental results of TUNNETT diodes in W-band cavities
Third International Symposium on Space Terahertz Technology Page 445
A plot of the output power and efficiency of the W-band diodes that have been mounted and tested to date
is given in Figure 9. There appears to be a broad peak in the RF output power and DC to RF conversion
efficiency around the nominal design frequency of 100 GHz. This peak confirms that the first order
design rules accurately predict the operating frequency of the TUNNETT diodes. It also indicates that the
high field, high temperature electron average drift velocity in GaAs TUNNETT diodes is close to
4.6 x 10 6 cms -1 . The power levels and efficiencies above 100 GHz are comparable to the ones obtained
from Gunn devices in this frequency range [20-22].
To verify the mode of operation the DC I- V characteristics are measured at room temperature (T = 300 K)
and elevated temperatures (T = 470 K). The I-V curves of a 25 ujn diameter W-band TUNNETT diode
shown in Figure 10 clearly demonstrate that the injection mechanism is predominantly tunneling. For
comparison the I-V curves of a 55 urn V-band Mixed Tunneling and Avalanche Transit-Time (MIT ATT)
diode are also given in Figure 10. At room temperature the MIT ATT diode has a sharp increase in current
at about 18 V due to the onset of impact ionization [16]. The TUNNETT diode I-V curve at room temper-
ature exhibits no sign of this behavior. Tunneling as the dominant breakdown mechanism also explains
the temperature dependence of the TUNNETT diode I-V curves. Increasing the junction temperature of
the device enhances tunneling and suppresses impact ionization as can be seen in the temperature behav-
ior of the MIT ATT diode. For low bias voltages the current increases, thus indicating tunneling. The volt-
age for the sharp increase in current has a positive temperature coefficient, characteristic of impact ion-
ization as previously shown in Figure 6. For the range of applied bias voltages the current in the
TUNNETT diode always increases as a function of temperature implying that impact ionization is not
significant.
Figure 1 1 shows the measured spectra of a free running W-band IMP ATT diode oscillator with 42.8 mW
at 89.2 GHz (Figure 11a) and a free running W-band TUNNETT diode oscillator with 9.2 mW at
92.2 GHz (Figure 1 lb), and proves that the oscillations have clean spectra. The spectrum of another free
running TUNNETT diode oscillator in Figure 12 was taken using the same settings (vertical scale, scan
width and resolution bandwidth) of the spectrum analyzer as in Reference 23 for an free running InP
Gunn device oscillator and it compares favorably to the spectrum of this Gunn device oscillator.
6. Device simulation
In order to determine the capabilities of GaAs IMPATT diodes at D-band frequencies and in order to find
an explanation for the significant decrease in output power above 110 GHz, the device structures were
simulated using two IMPATT diode simulation programs, a drift-diffusion (DD) model [24] and an
energy-momentum (EM) model [25]. Table 3 shows calculated output power and efficiency at/= 95 GHz
as preliminary results for both programs. The data for the device area A D and current density Jpc are
Page 446
Third International Symposium on Space Terahertz Technology
taken from Reference 3. The energy-momentum program shows slightly higher breakdown voltages and
higher efficiency and output power. If a series resistance R s = 0.18 Q is taken into account for this diode,
the calculated output power and efficiency are much closer to the measured values. This series resistance
is comparable to the value obtained from small signal impedance measurements in forward direction at
32 MHz [9,10].
W-band IMPATT diode
Frequency: 95 GHz Area A D : 8x
lO^cm 2 Current density:
50 kAcm- 2
Model Voltage
Power Efficiency
Power Efficiency
(/? 5 = on)
(fl s = 0.18Jl)
[V]
[mW] [%]
[mW] [%] .
DD 12.2
550 11.3
320 6.5
EM 12.5
700 14.0
510 10.2
D-band IMPATT diode
Frequency: 140 GHz Area/l D : 5x
10' 6 cm 2 Current density: 60 kAcm- 2
Model Voltage
Power Efficiency
Power Efficiency
Power Efficiency
(* S = 0Q)
(/? s = o.20 n)
(R s = 0.288 £2)
[V]
[mW] [%]
[mW] [%]
[mW] [%]
DD 10.4
120 3.8
35 1.1
12 0.4
EM 10.9
215 6.5
80 2.4
18 0.5
Table 3: Calculated results for GaAs single-drift flat-profile IMPATT diodes.
The results for the D-band structure in Table 3 were calculated for no series resistance and two different
values of the series resistance. R s = 0.288 fl assumes that the series resistance is mainly due to the contact
resistances of the p + and n + layers and scales with reciprocal area, i.e. it is equivalent to R s = 0. 18 £2 of the
W-band diode. For this series resistance the output power is reduced to about one tenth of the output
Third International Symposium on Space Terahertz Technology
Page 447
power for the case of no series resistance taken into account. Since the calculated output power agrees
with the experimental value of 9 mW at 128.4 GHz, the series resistance is believed to be the main reason
for the significant rolloff in performance above 1 10 GHz. As can be seen also from Table 3, the output
power reduction is only about one third and therefore much less pronounced, if a slightly smaller
# s x ^d (1 x 10" 6 ftcm 2 ) is assumed. This demands better technology for contacts on both p + - and n + -
type GaAs.
Neither the drift-diffusion model nor the energy-momentum model consider any losses in the cavity.
These losses are due to the large transformation ratio (up to 500) from the low impedance level between
the contacts of the diode and the high impedance level of the waveguide.
The simplified large-signal model for TUNNETT diodes [ 14] which was employed in the design was also
used to determine how strongly the series resistance influences output power and efficiency of these
devices. The above mentioned drift velocity v s and the actual device dimensions (mesa height and
diameter, heat sink thickness, etc.) were used for the simulation. In Table 4 the specific contact resistance
was assumed to be p c = 1 x 10 7 Qcm 2 for the p + ohmic as well as for the n + ohmic contact. In this case
the predicted RF output power into a load of R\ = 1 Q, is 251 mW for experimentally investigated
diameters around 25 urn.
Freq
(GHz)
Drift Length
(/mi)
Drift Field
(kV/cm)
V DC
(Volts)
Vrf
(Volts)
Vdc /Vrf
Jdc
kA/cm 2
100.0
0.345
309.8
12.29
10.89
0.886
32.84
DIAM
(/im)
AREA
(/im 2 )
R,
(Ohm)
Rd
(Ohm)
Ri
(Ohm)
Vrf
(Volt)
Idc
(mA)
Pdc
(W)
Prf (Gen)
(mW)
Prf (Load)
(mW)
15
177
0.33
2.36
2.02
10.89
58
0.71
134
115
20
314
0.21
1.33
1.12
10.89
103
1.27
238
201
25
491
0.15
1.15
1.00
8.40
161
1.98
287
251
30
707
0.11
1.11
1.00
6.35
232
2.85
313
281
DIAM
(/im)
Pdc
(W)
Prf
(mW)
Rth (Cu)
(°C/W)
Rth (Di)
(°C/W)
AT (Cu)
CO
AT (Di)
(°C)
15
0.71
115
213
140
127
84
20
1.26
201
140
86
149
91
25
1.98
251
103
59
178
102
30
2.85
281
80
44
207
113
Table 4: Performance of TUNNETT diodes at 100 GHz for p c = 1 x 10 7 Qcm 2 and V RF /V DC < 0.886.
Page 448
Third International Symposium on Space Terahertz Technology
As described above for the IMP ATT diodes, present GaAs technology, however, gives a specific contact
resistance closer to p c = 7.5 x 10 7 ficm 2 . As a result, the RF output power into 1 Q decreases to 171 mW.
For the predicted results in Table 4 and 5 the maximum RF voltage was limited to 88.6 % of the applied
DC bias voltage. Since IMP ATT diodes at millimeter wave frequencies operate at an RF voltage around
or less than 50 % of the DC bias voltage, this case was also investigated for the TUNNETT diodes. As
can be seen from Table 6 the RF output power drops to 158 mW for experimentally investigated diame-
ters around 25 urn.
Freq
(GHz)
Drift Length
(/im)
Drift Field
(kV/cm)
V DC
(Volts)
v RF
(Volts)
Vdc /Vrf
Jdc
kA/cm 2
100.0
0.345
309.8
12.29
10.89
0.886
32.84
DIAM
(urn)
AREA
(/*m J )
R.
(Ohm)
R*
(Ohm)
R,
(Ohm)
Vrf
(Volt)
Idc
(mA)
Pdc
(W)
Prf (Gen)
(mW)
Prf (Load)
(mW)
15
177
1.07
2.36
1.29
10.89
58
0.71
134
73
20
314
0.62
1.62
1.00
9.15
103
1.27
200
124
25
491
0.41
1.41
1.00
7.05
161
1.98
241
171
30
707
0.30
1.30
1.00
5.61
232
2.85
276
213
DIAM
{fim)
Pdc
(W)
Prf
(mW)
Rth (Cu)
(°C/W)
Rth (Di)
(°C/W)
AT (Cu)
(°C)
AT 1 (Di)
(°C)
15
0.71
73
213
140
136
90
20
1.27
124
140
86
160
98
25
1.98
171
103
59
186
107
30
2.85
213
80
44
212
116
Table 5: Performance of TUNNETT diodes at 100 GHz for p c = 7.5 x 10 7 Qcm 2 and V^V^ < 0.886.
The calculated values of the thermal resistance R A and temperature rise AT for a copper (Cu) or diamond
(Di) heat sink are also included in Tables 4, 5 and 6. It should be noted that experimental values for the
thermal resistance always are higher than calculated. Therefore, a diode with a diameter of 30 urn on a
copper heat sink will be operated at a lower DC input power Pqq and therefore reduced RF output power
Prf to achieve an operating junction temperature below 250 "C.
Similar to the IMPATT diode simulation, the simplified large-signal TUNNETT diode simulation does
not account for any losses due to the large transformation ratio from the low impedance level between the
contacts of the diode and the high impedance level of the waveguide. Since the structure of the resonant
cap full height waveguide cavity has been optimized for IMPATT diodes its impedance transformation
losses are expected to be higher for the TUNNETT diodes.
Third International Symposium on Space Terahertz Technology
Page 449
Freq
(GHz)
Drift Length
(fim)
Drift Field
(kV/cm)
V DC
(Volts)
Vrf
(Volts)
Vdc /Vrf
Jdc
kA/cm 2
100.0
0.345
309.8
12.29
10.89
0.500
32.84
DIAM
(fim)
AREA
(/im 2 )
Rs
(Ohm)
Rd
(Ohm)
Ri
(Ohm)
Vrf
(Volt)
Idc
(mA)
Pdc
(W)
Prf (Gen)
(mW)
Prf (Load)
(mW)
15
177
1.07
4.63
3.56
6.14
58
0.71
76
58
20
314
0.62
2.61
1.99
6.14
103
1.27
134
103
25
491
0.41
1.67
1.26
6.14
161
1.98
210
158
30
707
0.30
1.30
1.00
5.61
232
2.85
276
213
DIAM
(fim)
Pdc
(W)
Prf
(mW)
Rth (Cu)
(°C/W)
Rth (Di)
(°C/W)
AT (Cu)
(•C)
AT (Di)
PC)
15
0.71
58
213
140
136
90
20
1.27
103
140
86
162
98
25
1.98
158
103
59
186
107
30
2.85
213
80
44
212
116
Table 6: Performance of TUNNETT diodes at 100 GHz for p c = 7.5 x 10" 7 Qcm 2 and V RF /V l3C < 0.50.
7. Conclusions
The experimental results clearly show that GaAs IMP ATT diodes are powerful devices not only for
frequencies below 60 GHz, but also above 100 GHz. The predicted results of two different simulation
programs agree with the experiment. These simulations also reveal that the contact technology is very
crucial for high output power and efficiency and must be improved considerably for GaAs D-band
IMPATT diodes. The results from D-band GaAs IMPATT diodes and from W-band GaAs TUNNETT
diodes are the best reported to date. Both IMPATT and TUNNETT diodes exhibit clean spectra for local
oscillator applications. The TUNNETT diodes demonstrate useful power levels and efficiencies compara-
ble to Gunn devices. Since RF output power and DC to RF conversion efficiency do not saturate up to the
highest applied DC bias currents, still higher output power levels and efficiencies can be expected from
TUNNETT diodes on diamond heat sinks in optimized cavities.
References
[1] Elta, M. E., Fettermann, H. R., Macropoulos, W. V., and Lambert, J.: "150 GHz GaAs MITATT
source", IEEE Electron Device Letters, EDL-1, 1980, pp. 115-1 16.
[2] Chang, K., Kung, J. K., Asher, P. G., Hayashibara, G. M., and Ying, R. S.: "GaAs Read-type
IMPATT diode for 130 GHz CW operation", Electronics Letters, 17, 1981, pp. 471-473.
Page 450 Third International Symposium on Space Terahertz Technology
[3] Eisele, H., and Grothe, H.: "GaAs W-band IMPATT diodes made by MBE", Proc. MIOP '89,
Sindelfingen, FRG, Feb. 28th - March 3rd 1989, Session 3A.6.
[4] Eisele, H.: "GaAs W-band IMPATT diodes for very low-noise oscillators", Electronics Letters, 26,
1990, pp. 109-110.
[5] Nishizawa, J., Motoya, K., and Okuno, Y.: "Submillimeter Wave Oscillation from GaAs TUNNETT
Diode", Proceedings of the 9th European Microwave Conference, 1979, pp. 463-467.
[6] PObl, M., Freyer, J.: "Characterization of W-Band CW TUNNETT Diode", Proceedings of the 21st
European Microwave Conference, Stuttgart, FRG, 1991, pp. 1496-1501.
[7] Kwon, Y., Pavlidis, D., Tutt, M., Ng, G. I., Lai, R., and Brock, T.: "W-Band Monolithic Oscillator
Using InAlAs/InGaAs HEMTs", Electronics Letters, 26(18), 1990, pp. 1425-1426.
[8] Rolland P. A., Friscourt M. R., Lippens D., Dalle C., and Nieruchalski, J. L.: "Millimeter Wave
Solid-State Power Sources", Proceedings of the International Workshop on Millimeter Waves,
Rome, Italy, April 2-4, 1986, pp. 125-177.
[9] Eisele, H.: "Electron properties in GaAs for the design of mm-wave IMPATTs", International
Journal of Infrared and Millimeter Waves, 4, 1991, pp. 345-354.
[10] Eisele, H.: "GaAs W-Band IMPATT diodes: The first step to higher frequencies", Microwave
Journal, 34, 1991, pp. 275-282.
[11] Okuto, Y., and Crowell, C. R., "Threshold energy effects on avalanche breakdown voltage in semi-
conductor junctions", Solid-State Electronics, 18, 1975, pp. 161-168
[12] Hulin, R.: "GroBsignalmodell von Lawinenlaufzeitdioden", Ph.D. Thesis Techn. University Braun-
schweig, Braunschweig, 1973.
[13] Harth., W., Claassen, M.: "Aktive Mikrowellendioden", Springer- Verlag, Berlin, 1981.
[14] Haddad, G. I., East, J. R., and Kidner, C.: "Tunnel Transit-Time (TUNNETT) Devices for Terahertz
Sources", Microwave and Optical Technology Letters, 4, 1991, pp. 23-29.
[15] Allam, R., and Pribetich, J.: "Temperature Dependence of Electron Saturation Velocity in GaAs",
Electronics Letters, 26, 1990, pp. 688-689.
[16] Kidner, C., Eisele, H., and Haddad, G. I.: "Tunnel Injection Transit-Time Diodes for W-Band Power
Generation", Electronics Letters, 28, 1992, pp. 511-513.
Third International Symposium on Space Terahertz Technology
Page 451
[17] Kidner, C., Eisele, H., East, J., and Haddad, G. I.: "Design, Fabrication and Evaluation of Tunnel
Transit-Time Diodes for V-Band and W-Band Power Generation", to be presented at the 1992 IEEE
MTT-S International Microwave Symposium, June 1 - June 5, 1992, Albuquerque, New Mexico.
[18] Eisele, H.: "Selective etching technology for 94 GHz GaAs IMPATT diodes on diamond heat sinks",
Solid-State Electronics, 32, 1989, pp. 253-257.
[19] Kamoua, R., Eisele, H., East, J. R., Haddad, G. I., Munns, G., Sherwin, M.: "Modeling, Design,
Fabrication, and Testing of InP Gunn Devices in the D-Band (110 GHz - 170 GHz), these
Proceedings of the 3rd International Symposium on Space Terahertz Technology, March 24-26,
1992, Ann Arbor, Michigan.
[20] Wandinger, L.: "mm-Wave InP Gunn Devices: Status and Trends", Microwave Journal., 24(3),
1981, pp. 71-78.
[21] Eddison, I. G., et al.: "Efficient fundamental frequency oscillation from mm-wave InP n + -n-n +
TEOs", Electronics Utters, 17(20), 1981, pp. 758-760.
[22] Teng, S. J. J., Goldwasser, R. E.: "High Performance Second-Harmonic Operation W-Band GaAs
Gunn Diodes", IEEE Electron Device Letters, EDL-10(9), 1989, pp. 412-414.
[23] Perrin, O., et al.: "380 GHz Receiver Front-End for the Balloon-Borne Radioastronomical
Experiment", Proceedings of the 2nd International Symposium on Space Terahertz Technology,
February 26-28, 1991, Pasadena, California, pp. 622-640.
[24] Bauhahn, P. E., and Haddad, G. I.: "IMPATT device simulation and properties", IEEE Transactions
on Electron Devices, ED-24, 1977, pp. 634-642.
[25] Mains, R. K... Haddad, G. I., and Blakey, P. A.: "Simulation of GaAs IMPATT Diodes Including
Energy and Velocity Transport Equations", IEEE Transactions on Electron Devices, ED-30, 1983,
pp. 1327-1338.
5 x 1 1 9 cm" 3
5x I0 ,8 cm" 3
Sx I0 l9 cm" 3
5x lO^cm" 3
p"
• 3 x I0 17 cm" 3
p"
n* puffer
3xlO ,8 cm' 3
n* puffer
HP
12
7.3 x lO^cm" 3
n
n"
— fc ..
_
"300 nm
_
240 nm
• 270 nm
1500 nm
240 run
40 nm
1500 nm
Fig. I: Nominal device structure of a GaAs
D-band single-drift flat-profile IMPATT
diode.
Fig. 2: Nominal device structure of a GaAs
W-band single-drift TUNNETT diode.
Page 452
Third International Symposium on Space Terahertz Technology
FLOW DIAGRAM FOR ETCH-STOP GaAs IMPATT DIODE FABRICATION PROCESS
Awimmmmmmw
pt contact (Ti/Pt/Au)
active layer ■
holder
supporting
grating (Au)
etch -stop layer
GaAs substrate
ftWl^VVVV^^VkV^^^^
1) Metal evaporation for the
p+- ohmic contact
wsssssssssssssssssssssssssssss*.
3) Thinning substrate selectively
down to the etch-stop layer
W^
photoresist
5) Etching the contact and the mesa
2) Electroplating (Au) for the supporting
grating and gluing sample to holder
n + -contact
(Au/Ge/Ni)
4) Removing the etch-stop layer
metallizing & electroplating for the
rf"- ohmic contact
supporting
grating (Au)
P- ohmic contact
n -ohmic contact
6) Annealing ohmic contacts
► Final device
Fig. 3: Flow chart for IMPATT diodes device fabrication.
Third International Symposium on Space Terahertz Technology
Page 453
Plated Au
Ti/Pt/Au
p*GaAs
n + A!GaAs
GaAs
aSSSSSSSSSBBSSBBSsS
:B55S5cg£c5gccBc£cc55cS5cccge5geo£eccBSSc££5c&S
a) Island definition, p-ohmic evaporation, and
gold plating (~ IS um)
n-ohmic
Ni/Ge/Au/Ti/Au
b) Substrate thinning, etch stop layer removal,
and n-ohmic evaporation
Ti/Au/Tl
Photoresist
c) Gold plating of ohmic contacts
d) Final diodes after annealing and mesa etch
Fig. 4: Flow chart for TUNNETT diodes device fabrication.
125
too
$?,
*•
75
-3
JB.
&
u
V
C
<u
is
o
<£
50
-2
<*4
Ct-i
W
• Power
25
- 1
■ Efficiency
+ Frequency
H in
no
109
112
50 100 150 200
Current [mA]
250
N
as
0)
a
Fig. 5: Output power, efficiency and oscillation
frequency as a function of bias current
for a GaAs single-drift flat-profile
IMPATT diode.
Fig. 7: Breakdown voltage V hT and peak electric
field £ max of an abrupt p + n-junction.
■ * : measured
: calculated.
Page 454
Third International Symposium on Space Terahertz Technology
IF
(jiA)
SO. 00
10.00
/dlv
7=300K -»
-SO. 00
T= 300 K -»
«- r=3
/
I'"
■4
1 _ Tn 370 K
.
'
-1.200
VF
forward I ravaraa -
1.200/div ( V)
9.600
IF
(DA)
50.00
10.00
/dlv
T=300K
-50.001
■
1
/
/
//
r=300K -♦//«- r
i
«- Ts 370 K
!
-1.200
VF
t- forward I raversa -
7= 370 K
1.200/dlv ( V)
9.600
(a) (b)
Fig. 6: Current-voltage characteristics of a GaAs single-drift flat-profile IMPATT diode at room
temperature (300 K) and an elevated temperature (370 K).
J.
40
30
20
10
+ Frequency
■ Efficiency
• Power
c
V
'5
w
94
93
50 75 100 125
Current [mA]
150
N
X
c
V
a
a"
I
b
40
30
20
10
1
-\ —
r
/•
-3
* /
HI
ji J
N
Efficiency
•"' /
-1
■'
ji
• Power
■ Efficiency
i
+ Frequency •
i
108
107
106
30 50 70 90
Current [mA]
110
x
a
<u
3
c*
9)
£
(a) (b)
Fig. 8: Output power, efficiency and oscillation frequency as a function of bias current for two W-band
GaAs single-drift TUNNETT diodes.
Third International Symposium on Space Terahertz Technology
Page 455
I
u
I
100
frequency [GHz]
120
i
9)
1
Fig. 9: Output power and efficiency of GaAs
single-drift TUNNETT diodes in
W-band.
20 mA
200 pA
20|iA
2|iA
/ 7
y-+
TUNNETT A
y 1
MFTATT J
/
..-/J
470K— -V/— JOOK
/y~*
Fig. 10: Reverse bias current-voltage character-
istics for pure tunnel injection
(TUNNETT) and mixed tunnel injection
and impact ionization (MITATT) at
room temperature (300 K) and elevated
temperatures (470 K and 400 K,
respectively).
l*wl ranuBcv
-1006M ca B9.158 71GHZ
-10.8DBH »» 99.1 6 7I6HZ
VAH/DIV
200KHZ
M^rf^ -
1008/ RT 79-140
wrnot » m
onuir ArrouTin mm
10KHZ
LUTIO
WIDTH
(a)
Fig. 11: Spectrum of a W-band IMP ATT diode
free running oscillator, power level
42.8 mW, center frequency 89.16 GHz,
vertical scale 10 dB/div, horizontal scale
200 kHz/div, BW 10 kHz.
TEX
Z755P
leva.
-1008M
^lOW"^^
92.210 92GHZ
92.210 916HZ
VM/OZV
20OKHZ
^Wf^r^ri
1006/
EXT
75-140
10KHZ
FT
Mnufriw
OimjtT
ATTBtUTItM
W«
MMKIDTH
(b)
Fig. 1 1 : Spectrum of a W-band TUNNETT diode
free running oscillator, power level
9.2 mW, center frequency 92.21 GHz,
vertical scale 10 dB/div, horizontal scale
200 kHz/div, BW 10 kHz.
Page 456
Third International Symposium on Space Terahertz Technology
LEVEL
ref -iODBM
hkh -16.4DBM
TEK
2755P
FREQUENCY
cen 94.068GHZ
hkr 94.06BGHZ
SPAN/DIV
500KHZ
DBH
-10
-20
-30
-40
-00
-60
-70
-BO
-90
10DB/
VERTICAL
DISPLAY
EXT 75-140
RF FREQ
ATTENUATION RANGE
3KHZ 100KHZ
VIDEO RESOLUTION
FILTER BANDHIDTH
Fig. 12: Spectrum of a W-band TUNNETT diode free running oscillator, power level 8.8 mW, center
frequency 94.07 GHz, vertical scale 10 dB/div, horizontal scale 500 kHz/div, BW 100 kHz.
Third International Symposium on Space Terahertz Technology Page 457
Negative Differential Resistance (NDR) 5f0%33
Frequency Conversion with Gain /^>OSSj
R. J. Hwu, R. W. Aim, and S. C. Lee
Department of Electrical Engineering
University of Utah
Abstract—The dependence of the I-V characteristic of the negative differential resistance
(NDR) devices on the power level and frequency of the rf input signal has been
theoretically analyzed with a modified large- and small-signal nonlinear circuit analysis
program [1,2]. The NDR devices we used in this work include both the tunnel diode
(without the antisymmetry in the I-V characteristic) and resonant-tunneling devices (with
the antisymmetry in the I-V characteristic). Absolute negative conductance can be found
from a zero-biased resonant tunneling device when the applied pump power is within a
small range. This study verifies the work of Sollner et al. [3]. Variable negative
conductances at the fundamental and harmonic frequencies can also be obtained from both
the unbiased and biased tunnel diodes. The magnitude of the negative conductances can be
adjusted by varying the pump amplitude— a very useful circuit property. However, the
voltage range over which the negative conductance occurs moves towards the more positive
side of the voltage axis with increasing frequency. Furthermore, the range of the pumping
amplitude to obtain negative conductance varies with the parasitics (resistance and
capacitance) of the device. The theoretical observation of the dependence of the I-V
characteristic of the NDR devices on the power and frequency of the applied pump signal is
supported by the experimental results. In addition, novel functions of a NDR device such
as self-oscillating frequency multiplier and mixer with gain have been experimentally
demonstrated. The unbiased oscillator have also been successfully realized with a NDR c •>
Page 458 Third International Symposium on Space Terahertz Technology
device with an antisymmetrical I-V characteristic. Finally, the applications of these device
functions will be discussed. % -
INTRODUCTION
There have been increased interest in the study of resonant tunneling devices due to
the fact that the characteristics of these devices can be engineered to have properties for
very high-speed applications. In particular, their ability to exhibit negative differential
resistance (NDR) regions lead to their potential use as gain elements in circuits and offers a
new opportunity for circuit design. The presence of peaks and valleys in the I-V curve
combined with the overall antisymmetry of the I-V curve about the origin [i.e., I(V) = -I(-
V)], also offers the potential for efficient odd-harmonic generation with an unbiased
resonant tunneling device [4,5]. The key lies in pumping the device so that the peak
amplitude of the voltage across the device occurs above the resonant current peaks. This
will produce more than three local maxima in the device current waveform over one cycle,
corresponding to third or higher odd harmonic generation. The resonant tunneling
frequency multiplier, therefore, has several distinct advantages over existing resistive
multipliers, which are usually based on Schottky barrier diodes. The antisymmetrical
response provides the potential for efficient odd harmonic frequency multiplication with an
unbiased resonant tunneling device due to cancellation of the even harmonics, therefore
greatly simplifying the circuit design. The maximum harmonic generation efficiency of a
resonant tunneling device is significantly higher than the 1/n^ (n is the harmonic number)
value that applies to standard resistive multipliers because of its negative resistance [4,5]
(i.e., nonmonotonically increasing function I-V characteristic). The resonant tunneling
device also has the ability to act as an efficient mixer due to the rapid variation of the
dynamic conductance with voltage near the NDR region of the I-V curve. The resonant
tunneling mixer has the potential to displace the Schottky diode in many microwave and
millimeter-wave applications. The most intriguing aspect of the resonant tunneling
Third International Symposium on Space Terahertz Technology Page 459
frequency multiplier and mixer is its intrinsic capability to achieve conversion gain
(efficiency > 1).
LARGE- AND SMALL-SIGNAL NONLINEAR-CIRCUIT ANALYSIS
A large- and small-signal analysis program has been developed to analyze the
behavior of dc and microwave negative conductance of a NDR device. The analysis
technique was developed by T. Kerr [6] and the computer program was implemented for
analyzing ideal Schottky barrier diodes by Siegel et. al. [1]. The analysis program has
been modified to take into account the negative resistance of the NDR device [2]. The I-V
characteristics measured from the NDR devices as can be seen in Fig. 1 have been used in
the nonlinear-circuit analysis. Since the devices were mounted on a 50 Q. microstrip line
for the measurements in this work, the embedding impedance of 50 Q. at every harmonic
frequency has been used. A simple experiment has been carried out to verify that the
embedding impedance at higher harmonic frequencies is equal to 50 Q [7].
Tunnel Diode— Without Antisymmetry in the I-V Characteristic
The differential conductance of a tunnel diode biased at zero voltage and in the
positive differential resistance (PDR) region (close to the current peak) under different rf
pumping conditions has been studied in this work. No negative conductance has been
observed at dc for a tunnel diode biased at zero voltage and in the PDR region. However,
negative conductances at the fundamental and different harmonic frequencies have been
observed from a tunnel diode biased at zero voltage and in the PDR region. The magnitude
of the negative conductance varies with the pump amplitude (see Fig. 2). The pump
amplitude region required to achieve the negative conductance moves toward the more
positive side of the power axis with increasing pumping frequency (also see Fig. 2). This
can be easily explained using the equivalent circuit model of a tunnel diode (see Fig. 1 (c)).
Since the impedance of the parallel circuit section decreases with the increasing frequency,
more voltage will be distributed on the series resistance and less voltage on the parallel
circuit section for the higher pumping frequency. The negative conductance observed at the
Page 460 Third International Symposium on Space Terahertz Technology
fundamental and different harmonic frequencies can be used as the basis for harmonic
oscillators. It should be pointed out that the magnitude of the negative conductance at the
fundamental is much higher than those at other harmonic frequencies with a runnel diode
biased in the PDR region, close to the current peak (see Fig. 2 (b) and (c)). At this bias
point, near the region of greatest curvature, the Fourier series of the conductance
waveform has a predominant coefficient at the oscillation frequency. In addition, the
negative conductances at the odd harmonic frequencies are higher than those at the even
harmonic frequencies for the tunnel diode biased in the PDR region, close to the current
peak due to the antisymmetrical conductance-voltage (G-V) characteristic at this bias point.
The power levels at which the maximum negative conductances at the fundamental and
second harmonic frequency occur are smaller than those for a unbiased tunnel diode.
The differential conductance of a tunnel diode biased at the center of the NDR
region has also been studied. From the results shown in Fig. 3 (a), an absolute negative
conductance to dc has been obtained from a tunnel diode biased in the NDR region when
the applied pump amplitude is within a small range. That is, the conductance of the
resonant tunneling device will be negative at any frequency when the pump amplitude is
within this small range. The value of the absolute negative conductance is approximately
the same as that found in the NDR region. The magnitude of the negative conductances
changes with the pumping power level (see Fig. 3). That the absolute negative
conductance occurs for the tunnel diode biased in the NDR region asserts that oscillation
can occur at any frequency if the pumping power is within the region that the negative
conductance occurs. The magnitude of the negative conductance at the second harmonic is
higher than those at dc, the fundamental and third harmonic frequencies due to the
symmetrical G-V characteristic at this bias point (see Fig. 3 (b)). The variable absolute
negative conductance observed can be used as the basis for oscillators and harmonic
oscillators up to the cut-off frequency of the diode. The self-oscillation capability of a
tunnel diode biased in the NDR region can, therefore, finds applications as biased self-
Third International Sytnposium on Space Terahertz Technology Page 461
oscillating mixers and frequency multipliers. The self-oscillating frequency multiplier and
mixer discussed here do not require a large-signal rf pump.
It should be noted that the self-oscillation at the fundamental generates its own
harmonics using the nonlinearity of the NDR device; this will be referred to as the self-
oscillating frequency multiplier. While the harmonic oscillator refers to the case that the
NDR device oscillates at a particular harmonic frequency using the negative conductance at
that harmonic frequency. It should be pointed out that the conversion gain (efficiency > 1)
can be achieved from the biased self-oscillating frequency multiplier and mixer.
Resonant Tunneling Device- With Antisymmetry in the I-V Characteristic
From the nonlinear-circuit analysis results, an absolute negative conductance to dc
can be found from a resonant tunneling device at zero bias when the applied pump power is
within a small range (see Figs. 4 (a) and (b)). The value of the negative conductance is
approximately the same as that found in the NDR region. As can be seen from the results
in Figs. 4 (a) and (b), the magnitude of the negative conductance can be adjusted by
varying the pump amplitude. This study verifies the work of Sollner et al. [3]. However,
the voltage range over which the negative conductance occurs is strongly dependent on the
pumping frequency. This region moves towards the more positive side of the power axis
when the pumping frequency increases (see Figs. 4 (a) and (b)). The reason for this can,
again, be explained with a equivalent circuit model (discussed for the tunnel diode in
previous section).
We found that the range of the applied pump power to obtain absolute negative
conductance varies with the parasitics (series resistance and capacitance) of the device. As
can be seen from Fig. 5, the pumping power region over which the negative conductance
occurs move towards the more positive side of the power axis with increasing capacitance
and series resistance of the device. This again can be seen from the equivalent circuit
model of a NDR diode. When the series resistance and/or capacitance of the device
increases, more voltage drops across the series resistance and, therefore, less power is
Page 462 Third International Symposium on Space Terahertz Technology
developed across the parallel circuit section of the equivalent circuit. It should be pointed
out that the dependence of the conductance on the parasitics of the device is similar to that
on the rf input frequency.
The conductance of the resonant tunneling diode will be negative at any frequency
when the pump amplitude is within a small range. Figures 4 (c) and (d) show the
differential conductance at the second harmonic frequency versus pumping power and
pump amplitude, respectively. The magnitude of the negative conductance at the second
harmonic frequency is larger than the negative conductances at dc and at fundamental
frequency. This is due to the symmetrical G-V characteristic of the resonant tunneling
device. The variable absolute negative conductance observed can be used for oscillators up
to the cut-off frequency of the diode.
From these studies, one can expect to find absolute negative resistance whenever a
material with negative differential conductance and an I-V curve that is antisymmetrical is
driven with a pump of the right amplitude and frequency. The resonant tunneling device
can also perform the same functions such as the self -oscillating frequency multiplier and
mixer discussed for a tunnel diode if it is biased in the NDR region. It should be pointed
out that the biased tunnel and resonant tunneling device (in the PDR region) requires less
pumping power to achieve self-oscillation than the unbiased device. The biased self-
oscillation tunnel diode and resonant tunneling frequency multipliers and mixers have the
intrinsic capability of conversion gain. For the unbiased oscillator operations, it should be
noted that little in the way of negative conductance or dynamic range has been sacrificed
with this operation, and the advantage of operating with zero DC bias voltage has been
gained. It should be further noted that the increase in negative conductance at a specific
frequency (depending upon the operating point) could simplify frequency selection for
oscillator designs.
EXPERIMENTAL RESULTS
Third International Symposium on Space Terahertz Technology ^ge 463
During the experimental measurements, it was observed that the dc I-V
characteristic of the NDR device is very strongly dependent on the power level of the rf
input signal. The dc I-V characteristics of the NDR devices (with and without the
antisymmetrical I-V characteristics) measured at different rf input power levels are shown
in Figs. 6 and 7, respectively. The dc I-V characteristics of the tunnel diode measured at
different rf input frequencies are shown in Fig. 8. Based on these results, the dc I-V
characteristics of the NDR device changes dramatically with the increasing input power
level and frequency. The dependence of the I-V characteristics of the NDR device on the
frequency of the rf input signal can easily be seen from the equivalent circuit model of the
NDR device. The frequency dependence of the impedance across the parallel circuit section
of the equivalent circuit results in the power dependence of the I-V characteristics. The
power dependence of the negative conductance complicates the dependence of the negative
conductance on the frequency. The nonlinear circuit analysis was used to theoretically
verify this observation. The I-V curves measured at different pumping power levels as
shown in Figs. 6 and 7 and the I-V curves measured at different pumping frequencies as
shown in Fig. 8 compare favorably to the simulation results.
Based upon this study, the design, operation, and performance of the NDR
frequency multiplier, self-oscillating frequency multiplier and mixer and harmonic oscillator
can be complicated. For a constant rf input frequency, the biasing and pumping conditions
and output power of the self oscillation of the NDR device vary with the rf input power
level. In addition, the onset of self oscillation of a NDR device biased in the NDR region
also depends upon the rf input power level. For example, the self oscillation can be
suppressed by changing the rf input power level (tuning the I-V characteristic) in the
frequency multiplication operation of a NDR device. However, this will cause the
conversion efficiency of the device to change as well since the nonlinearity is not the same.
In addition, the self-oscillation with a constant rf input frequency may disappear for a given
dc bias depending upon the rf input power level for the self-oscillating frequency
Page 464 Third International Symposium on Space Terahertz Technology
multiplier, mixer and harmonic oscillator applications. Determination of the biasing and
operating conditions and performance of the NDR frequency multiplier, self-oscillating
frequency multiplier and mixer, and harmonic oscillator, therefore, requires complete
information of the I-V characteristics at different rf input frequencies and power levels.
This can be accomplished by extensive simulations of the NDR device under different
pumping conditions using the modified large- and small-signal nonlinear circuit analysis
program as mentioned above.
A NDR biased in the NDR region can be used for the self-oscillating frequency
multiplier and mixer. Both the self-oscillating frequency multiplier and mixer have been
successfully demonstrated using a tunnel diode biased in the NDR region. The results
from a self-oscillating frequency multiplier can be seen in Fig. 9. The highest tripling
efficiency has been obtained at the center of the NDR region while the highest doubling
efficiency has been obtained at the edges of the NDR region of the tunnel diode as
expected. This is due to the I-V characteristic being antisymmetrical when biased at the
center of the NDR region, and the I-V characteristic being almost symmetrical when biased
close to the current peak. It should be pointed out that the circuit used does not allow
independent tuning of the harmonics.
APPLICATIONS
The wide use of resonant tunnel devices is limited, to a considerable extent, by the
low level of their output power. Power combining techniques are employed to increase the
output power of the resonant tunneling devices. The device-grid array approach is a
potentially attractive way to spatially combine the output power of large numbers of
resonant tunneling devices. In this approach, a grid is monolithically integrated with
thousands of devices thereby overcoming the power limitations of a single device since the
power is distributed among the many devices making possible watt- level CW output power
throughout the microwave and millimeter- wave region [8,9]. This kind of array can find
applications as a high frequency, high power solid-state rf power source. All the
Third International Symposium on Space Terahertz Technology Page 465
interconnections of the high- and low-frequency leads of each port of each device
(especially, three-terminal devices) present an extremely difficult problem for the
development of such arrays. The demonstration of novel unbiased oscillators is most
useful for the development of monolithic planar wafer-scale device arrays since no dc bias
lines are required, which greatly simplify the grid design.
In addition, the pumping power for a device grid is significantly higher than that for
a single device (proportional to the number of the devices). Therefore, it is important to
minimize the amount of input power required to pump each individual device. Based upon
the theoretical and experimental work which have been performed in this study, the
possibility of biasing a NDR diode to minimize the amount of power required to pump each
individual diode into the desired operation point has been verified. The bias lines can be
easily employed in the design of diode grid (two terminal device grid) to provide dc bias
and minimize the required pumping power [9]. In addition, frequency multiplication and
mixing with gain can be obtained from these biased NDR diodes as discussed in this paper.
CONCLUSION
This work employs a modified large- and small- signal nonlinear-circuit analysis [2]
to verify the previous work of Sollner et al. using a simple mathematical model [3]. The
absolute negative conductance can be obtained from an unbiased resonant tunneling device
when the applied pump power is within a small region. The variable absolute negative
conductance can be used as the basis for oscillators up to the cutoff frequency of the
device. Furthermore, a NDR device biased in the NDR region can be used as the basis for
the self-oscillating frequency multiplier and mixer. The biased self-oscillating frequency
multiplier and mixer can achieve conversion gain (efficiency > 1). These functions have
been experimentally demonstrated in this work. The advantage of a unbiased oscillator
using a resonant tunneling device comes from the fact that the negative conductance can be
adjusted by varying the pump amplitude-a very useful circuit property. In addition, the
negative conductance is larger at even harmonic frequencies which could simplify
Page 466 Third International Symposium on Space Terahertz Technology
frequency selection of an oscillator design based upon this effect. The advantage of
operating with zero dc bias voltage is also gained. Through this study, the power
dependence of the negative conductance of a NDR device on the rf input signal has been
observed. The biasing and pumping conditions and performance of the frequency
multiplier, self-oscillating frequency multiplier and mixer, and harmonic oscillator requires
complete information of the I-V characteristics of a NDR device at different input
frequencies and power levels. This information can be obtained using the modified large-
and small-signal nonlinear-circuit analysis as discussed in this paper.
References
[1] H. Siegel, A. R. Kerr, and W. Hwang, "Topics in the Optimization of MM Wave
Mixers," NASA Technical Paper #2287, 1987.
[2] R. J. Hwu and N. C. Luhmann, Jr., "Quantum Well Diode Frequency Multiplier
Study," Second International Symposium on Space Terahertz Technology,
Proceedings, pp. 226-237, 1991.
[3] T. C. L. G. Sollner, E. R. Brown, and W. D. Goodhue, Picosecond Electronics
and Optoelectronics II, Editors: F. J. Leonberger, C. H. Lee, F. Capasso, and H.
Morkoc, Springer-Series in Electronics and Photonics, Vol. 24, pp. 102-108,
1987.
[4] P. D. Batelaan and M. A. Frerking, 13th Int'l Conf. on Infrared and Millimeter
Waves, Proceedings, 1988.
[5] T. C. L. G. Sollner, E. R. Brown, W. D. Goodhue, and C. A. Correa, J. Appl.
Phys., Vol. 64, P. 4248, 1988.
[6] D. N. Held and A. R. Kerr, "Conversion Loss and Noise of Microwave and
Millimeter- Wave Mixers: Part 1 - Theory," IEEE Trans, on Microwave Theory and
Tech., MTT-26, PP. 49-55, 1978.
[7] P. P. Huang, "I-V Characterization of Negative Resistance Device by Microwave
Reflection Coefficients," M.S. thesis, UCLA, pp. 49-51, 1989.
[ 8 ] D. B. Rutledge and S. E. Schwarz, "Planar Multimode Detector Arrays for Infrared
and Millimeter Waves Applications," IEEE J. Quantum Electronics, QE-17, pp.
407-414, 1981.
[9] R. J. Hwu, C. F. Jou, N. C. Luhmann, Jr., M. Kim, W. W. Lam, Z. B. Popovic,
D. B. Rutledge, "Array Concepts for Solid-State and Vacuum Microelectronics
Third International Symposium on Space Terahertz Technology Page 467
Millimeter- Wave Generation," IEEE Trans, on Electron Device, ED-36, No. 11,
pp. 2645-2650, 1989.
Figure Captions
Fig. 1 The I-V curves of a (a), tunnel diode and a (b). resonant tunneling device used in
the nonlinear circuit analysis, (c). The equivalent circuit of a NDR device.
Fig. 2 The differential conductance of a tunnel diode biased in the PDR region (close to the
current peak) at the (a), dc, (b). fundamental, and (c). third harmonic frequency
from the nonlinear circuit analysis.
Fig. 3 The differential conductance of a tunnel diode biased at the center of the NDR
region at the (a), dc, and (b). second harmonic frequency from the nonlinear circuit
analysis.
Fig. 4 The differential conductance at dc versus (a), pump power level, and (b). pump
amplitude of a zero-biased resonant tunneling device. The differential conductance
at the second harmonic frequency versus (c). pump power level and (d). pump
amplitude of a zero-biased resonant tunneling device from the nonlinear circuit
analysis.
Fig. 5 The differential conductance at dc of a zero-biased resonant tunneling device with
different (a), capacitance values of 1 pF and 1 fF with three rf input frequencies of
0.7, 12, and 90 GHz, and (b). series resistance values of 12.5 Q, 625 Q, and 1.25
kQ with an rf input frequency of 0.7 GHz from the nonlinear circuit analysis.
Fig. 6 The measured I-V curves of a tunnel diode with different rf input power levels at
two rf input frequencies of (a). 0.7 GHz, and (b). 2.5 GHz.
Fig. 7 The measured I-V curves of a NDR device (with an antisymrnetrical I-V
characteristic) with two different rf input power levels at an rf input frequency of
10 MHz.
Fig. 8 The measured I-V curves of a tunnel diode with different rf input frequencies at an
rf input power level of 2 mW.
Fig. 9 The ratio of the output power to the fundamental power of a self-oscillating
frequency multiplier using the tunnel diode of Fig. 6 versus different bias points in
the NDR region with a rf input frequency of 0.7 GHz.
Page 468
Third International Symposium on Space Terahertz Technology
0.05
0.04
0.03
0.02-
0.01 -
0.00-
-0.01
-0.02
-O.03
-0.04 -
-0.05
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
V (V)
Fig. 1 (a)
<
E
-i — |— h — | — i — i — i — | — i — | — i — | — i — i — r— , — i — | — i — | — t — | — i-
.6-0.5-0.4-0.3-0.2-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Fig. 1 (b)
V (V)
R.
G(V)
$
Fig. 1 (c)
Third International Symposium on Space Terahertz Technology
Page 469
0.10
100
Fig. 2 (a)
0.00
-0.01
-0.02 -
-0.03
-0.04
-0.OS
-0.06
Fig. 2 (b)
Power (mW)
Power (mW)
-a Gdc - 0.7 GHz
» Gdc - 2.5 GHz
-» Gdc - 5.S GHz
-o Gdc - 8.5 GHz
Grf1 - 0.7 GHz
Grfl - 2.5 GHz
Grfl - S.5 GHz
Grfl - 8.5 GHz
-• Grf3 -0.7 GHz
-a Grf3 - 2.5 GHz
-* Grf3 - 5.5 GHz
-o Grf3 - 8.5 GHz
Fig. 2 (c)
Power (mW)
Page 470
Third International Symposium on Space Terahertz Technology
0.020
tn
o
•D
o
0.018-
0.016
0.014-
0.012-
0.010-
0.008 -
0.006
0.004 -
0.002 -
0.000 4
-0.002
100
Power (mW)
Gdc - 0.7 GHz
Gdc - 2.5 GHz
Gdc - 5.5 GHz
Gdc - 8.5 GHz
Gdc - 20 GHz
Fig. 3 (a)
tn
O
Power (mW)
Grf2 - 0.7 GHz
Grf2 - 2.5 GHz
Grf2 - 5.5 GHz
Grf2 - 8.5 GHz
Fig. 3 (b)
Third International Symposium on Space Terahertz Technology
Page 471
Pumping Power (mW)
Fig. 4 (a)
Fig. 4(b)
0.6 GHz
93 GHz
ITHz
1.5 THz
2.2 THz
12.00-
11.00-
10.00-
9.00-
8.00-
7.00
6.00-
5.00-
4.00
3.00
2.00
1.00
0.00 -*
-1.00-
-£00
.1
«sS££§SF
a~2=fi^o— q^s=SSq<
Pumping Power (mW)
Voltage (V)
Fig. 4 (c)
Fig. 4 (d)
Page 472
Third International Symposium on Space Terahertz Technology
E
u
O
E
u
(3
Fig. 5 (a)
10.00'
9.00 H
8.00
7.00-
6.00-
5.00
4.00-
3.00-
2.00-
1.00
0.00
-1.00
;1
Pump Power (mW)
1 ' ■ i
1
■ i i i i
-o 90 GHz, lpF
-• 90 GHz, 10 fF
■^ 12 GHz, 1 pF
12 GHz, 10 fF
0.7 GHz, 1 pF
0.7 GHz, 10 fF
-• Gdc-Rs=12.5n
-a Gdc - Rs=625 £2
-a Gdc-Rs=1.25KQ
1
Power (mW)
Fig. 5 (b)
Third International Symposium on Space Terahertz Technology
Page 473
0.003
0.002 -
< 0.001 -
c
o
0.000 -
-0.001
0.002 i ■ T*-> 1 ' 1 ■ 1 ' 1 ' 1 ' 1 "-
-0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Voltage (V)
Fig. 6 (a)
0.003
0.002
< 0.001 -
c
4)
o
0.000 -
-0.001 -
-0.002 H r
Voltage (V)
■a — 5nW
-* — 2.96 mW
49.8 mW
4^W
-* — 0.56 mW
16.7 mW
Fig. 6 (b)
Page 474
Third International Symposium on Space Terahertz Technology
<
G
*
L.
3
o
10 mW
OmW
— i — i — ■ — i — ■ — i — i — i — ■ — i — ■ — i — i — i — i — i — ■ — i — r-
.5 -1.2 -0.9 -0.6 -0.3 -0.0 0.3 0.6 0.9 1.2 1.5
Voltage (V)
Fie. 7
Third International Symposium on Space Terahertz Technology
Page 475
15.00
<
E
c
©
3
o
10.00
5,00-
0,00-
-5.00 -
-10.00-
-15.00 | ■ i • i — i 1 1 | i | i i
-0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
-o I-.2 GHz
-* — 1-3.62 GHz
-o — 1-8.45 GHz
-® — l-t1.95GHz
Voltage (V)
Fig. 8
Page 476
Third International Symposium on Space Terahertz Technology
-o
0.1 0.15 0.16 0.19 0.21 0.22 0.28
Bias (V)
P2/P1
♦ P3/P1
Fig. 9
Third International Symposium on Space Terahertz Technology p a g e 477
Modeling, Design, Fabrication, and Testing of InP Gunn /&03-OO
Devices in the D-band (110 GHz - 170 GHz) l f- / ^
R. Kamoua 2 , H. Eisele, J. R. East, G. I. Haddad,
G. Munns, and M. Sherwin
Solid-State Electronics Laboratory
Department of Electrical Engineering and Computer Science
The University of Michigan, Ann Arbor, MI 48109-2122
Abstract
The development of fundamental Gunn sources for D-band frequencies requires improve-
ments of doping profiles, processing technology, and circuit design. We have developed a
' technology for fabricating InP Gunn diodes using an InGaAs etch-stop layer between the
InP substrate and the device layers. The epitaxial layers were grown by CBE. During
device processing, the substrate is completely removed. Substrateless devices with an n +
InGaAs cap layer are expected to have reduced contact and series resistances, and skin
effect losses. This technology gives better uniformity and control of the device geometry
across the processed chip.
InP Gunn devices with a 1.7 fim long active region (doping : 9 x 10 15 cm -3 ) have been
mounted on copper heat sinks. Two tapered leads were then bonded to the diode and to
four quartz standoffs. As a preliminary result, an output power of 13 mW at 82 GHz was
obtained. Based on these RF measurements, we determine appropriate material parameters
to be used in the Ensemble Monte Carlo model. Subsequently, we use this model to design
and evaluate the performance of InP Gunn Devices for D-band frequencies. Using the same
technology, we are currently processing Gunn devices with a 1/j.m long active region for
operation at higher frequencies.
'This work was supported by the Center for Space Terahertz Technology under Contract No. NAGW-
1334
2 R. Kamoua is currently with the Department of Electrical Engineering, State University of New York
at Stony Brook, Stony Brook, NY 11794-2350
Page 478 Third International Symposium on Space Terahertz Technology
1 Introduction
Gunn devices are widely used as local pump oscillators in the W-band frequency region.
At these frequencies, the technology is well developed. In particular at 94 GHz, GaAs and
InP Gunn devices are available with very good performance. At frequencies above 100
GHz, the situation is quite different. Although there is a strong demand for sources at
these frequencies, fundamental Gunn devices are generally not available. There are two
reasons. First, most of the research effort has been focused at 94 GHz. Second, extending
the fundamental oscillation frequency of Gunn devices into the submillimeter region is
exceedingly difficult because the Gunn effect is being pushed to its high frequency limit.
This paper discusses a systematic approach toward the modeling, design, fabrication,
and testing of InP Gunn devices in the D-band region. Compared with GaAs, InP material
parameters are more favorable for operating Gunn devices in the D-band. The approach
taken in this work is both experimental and theoretical. Experimentally, the conventional
processing technology is improved by incorporating etch-stop layers in the wafer design.
The RF results obtained from devices fabricated using this technology are then used to
improve the accuracy of the theoretical model. Subsequently, the improved model is used
to design the optimum Gunn structure for the desired frequency of operation.
2 Simulation Model
The self consistent Ensemble Monte Carlo model is used to estimate the performance of
InP Gunn devices at high frequencies. This model is an extension of the one particle
Monte Carlo technique [1]. In order to describe the transport process in a structure with
nonuniform doping and with time varying fields, an ensemble of electrons needs to be
simulated simultaneously. The electric field has to be updated regularly since it is evolving .
as the electrons redistribute in the structure. The analysis is carried out assuming the
device behavior is mainly one dimensional which is justified for two terminal devices. A
diode structure of length L is divided into cells of equal length (Ax = 50 A). Any attributes
of the electrons are averaged over each cell and assigned to the midway position of the cell.
The cell size should be smaller than the smallest Debye length in the structure which occurs
at the highly doped regions.
The simulation algorithm consists of monitoring the evolution in real space and momen-
tum space of an ensemble of electrons. The simulation time is partitioned into time steps
(At = 5xl0 -15 sec) each terminated by a call to a Poisson solver in order to update the
field. In each time step every electron is submitted to successive free flights terminated by
a scattering process which is selected using a random number generator. Electrons cross-
ing cell boundaries are temporarily stopped at that boundary and then resumed with the
electric field in the new cell. An analogous procedure is followed when it is time to update
the electric field and the electron is in the middle of a free flight. In this case the remaining
flight time is stored, and the flight is resumed when all other electrons are simulated for one
time step, the carrier density is calculated, and the electric field is updated. The predicted
performance of a particular Gunn structure is estimated by applying an RF voltage across
the device and simulating the current response over many RF periods (about 10). The
resulting particle current density is Fourier analyzed and the fundamental component is
Third International Symposium on Space Terahertz Technology
Page 479
N +
3 x 10 17 cm- 3
InP
9 x 10 15 cm -3
InP
3 x 10 17 cm- 3
InP
O.ltim 1.7/xm 0.2pm
Figure 1: Dimensions and doping profile of Gunn structures.
used to determine the device admittance. The RF output power is estimated by consider-
ing the Gunn device in a resonant circuit represented by a load resistance and a resonating
inductance. A series resistance is included in the equivalent circuit that takes into account
effects of contact resistances, any substrate resistances, and skin effect losses.
The accuracy of the Monte Carlo model is strongly dependent on the accuracy of the
material parameters used. Unfortunately, one finds a wide range of values that are being
used in the literature. In particular, some of the material parameters that are important
to the Gunn effect have the following range of values ([2, 3, 4, 5]):
r - L valley separation (eV)
L valley effective mass ratio ( j^- )
V - L coupling constant (xlO 9 eV.cm -1 )
r — X coupling constant (xlO 9 eV.cm -1 )
0.4
«=>
0.832,
0.26
<^>
0.4,
0.1
<=3>
2.5,
0.43
<=>
1.0.
There is more than an order of magnitude uncertainty in the T to L intervally coupling
constant. In the next section, the material parameters that yield the best agreement with
the RF measured results will be identified.
3 Method for Extracting Accurate InP Material Parame-
ters for a 3- Valley Nonparabolic EMC Model
The appropriate material parameters are determined by comparing measurements at high
frequencies with results predicted by the model. The structure considered for comparison
is shown in Figure 1. It has 1.7 /xm long active region doped at 9xl0 15 cm -3 , a 0.1/xm
17 cm" 3 .
cathode region doped at 3x10 cm and a 0.2 /im anode region doped at 3x10
An InP wafer with this structure has been processed. The relevant fabrication technol-
ogy will be described later. Diodes with various sizes have been mounted on copper heat
sinks. Tapered ribbons were then used to bond a diode to four quartz standoffs. A 40 /xm
diode was tested in a W-band resonant cavity with the following results:
Page 480 Third International Symposium on Space Terahertz Technology
Bias voltage 4.0 V to 4.5 V,
Oscillation frequency 82 GHz,
output power 13 mW,
conversion efficiency 1 %,
DC current 350 mA.
The structure shown in Figure 1 is simulated using the model described above. The
DC bias is 4.5 V, the RF voltage is 1.0 V, and the operating temperature is assumed
to be 400 K. The material parameters are considered to be accurate if oscillations at 82
GHz are predicted with a performance comparable to experimental results. The starting
combination of parameters is listed in Table 1 and is referred to as the initial parameter
set. No oscillations occurred with this set of values for frequencies ranging from 75 GHz to
120 GHz. It appears that the T - L intervalley energy separation of 0.832 eV is too large.
As a result, the following modifications to the initial set are considered:
•
Case 1. The same parameters as the initial set are used except 17 — L valley separation
= 0.6 eV. No oscillations occurred at 82 GHz.
• Case 2. The same parameters as the initial set are used except T — L valley separation
= 0.4 eV. No oscillations occurred at 82 GHz.
• Case 3. The same parameters as the initial set are used except T — L energy separation
= 0.4 eV and T — X energy separation = 0.7 eV. No oscillations occurred at 82 GHz.
•
•
Case 4. The same parameters as the initial set are used except L valley effective mass
ratio = 0.4 and X valley effective mass ratio = 0.4. No oscillations occurred at 82
GHz.
Case 5. The same parameters as the initial set are used except T — L coupling constant
= 1.0 x 10 9 eV.cm -1 . No oscillations occurred at 82 GHz.
• Case 6. The same material parameters as the initial set are used except T — L energy
separation = 0.5 eV, T - X energy separation = 0.8 eV, T — L coupling constant =
1.0 XlO 9 eV.cm -1 and T - X coupling constant = 1.0 xlO 9 eV.cm -1 . Very weak
oscillations were obtained. The efficiency was 0.032 % and the output power was 0.22
mW.
• Case 7. The same parameters as the initial set are used except T — L energy separation
= 0.45 eV, T — X energy separation = 0.775 eV, L valley effective mass ratio = 0.4 eV,
X valley effective mass ratio = 0.4 eV, T — L coupling constant = 1.0 xlO 9 eV.cm" 1 ,
T — X coupling constant = 1.0 xlO 9 eV.cm -1 and acoustic deformation potential
= 5 eV. Oscillations were obtained at 82 GHz even though the DC bias was 4.0 V
instead of 4.5 V. With 1.0 V RF voltage, the predicted output power for a 40 fj,m
diode was 7.2 mW with 0.52 % efficiency.
The last parameter set appears to be promising and was considered in more detail. At an
RF voltage of 1.5 V, the predicted efficiency was 1.1 % and the predicted output power
Third International Symposium on Space Terahertz Technology
Page 481
Energy Separation (eV)
r-L
0.832
r-x
1.5
Effective Mass (g)
r
0.082
L
0.26
X
0.325
Nonparabolicity factor ([eV] -1 )
r
0.83
L
0.23
X
0.38
Intervalley Coupling Constant
(K^eV.cm- 1 )
r-L
0.506
r-x
0.498
L-X
0.468
L-L
0.575
x-x
0.28
Acoustic Deformation Potential (eV)
r
7
L
7
X
7
LO Phonon Energy (eV)
r
0.043
L
0.0423
X
0.0416
Static Dielectric Constant
12.61
Optical Dielectric Constant
9.61
Table 1: InP initial material parameter set.
Page 482
Third International Symposium on Space Terahertz Technology
n:
1.5 -
0.5 -
60
I
1 1 "
-
/
/ • Vrf = 05 V \
J a Vrf = 10 V
/ ♦ Vrf - 15 V
y -
*r i
i i
80
100
120
140
Frequency [GHz]
Figure 2: Efficiency vs. frequency for a 1.7 /zm long InP Gunn device doped at 9xl0 15
cm" 3 . Diameter = 40 /im, Vdc = 4.0 V, T = 450 K.
was 14.9 mW for a 40 /xm diode. The necessary load resistance for matching the diode
was 6.5 SI. The simulated DC current density has a value of 2.7xl0 4 A. cm -2 which results
in a current of 340 mA. The simulated device performance is in good agreement with the
measured RF results. In summary, we found it necessary to decrease the intervalley energy
separation and increase the electron effective mass in the upper valleys and the intervalley
deformation potentials in order to obtain oscillations at 82 GHz. The actual device was
operated at higher temperature than room temperature. This suggests that the lower values
for the intervalley separation and the higher values for the deformation potentials might be
caused by temperature effects on the band-structure. The last parameter set will be used
to analyze various InP Gunn structures. In the next section, the 1.7 ^m structure will be
considered in more detail.
4 Simulation of 1.7 ^m InP Gunn Devices
The results presented in this section correspond to a DC bias voltage of 4.0 V and an
operating temperature of 450 K. For comparison purposes, the load resistance was adjusted
so that the device area is 40 /im. Figure 2 shows a comparison of the conversion efficiency as
a function of frequency for three RF voltage amplitudes: 0.5 V, 1.0 V and 1.5 V. The peak
efficiency for a 1.5 V RF voltage is 1.86 % at 100 GHz. The corresponding comparison of
the output power is shown in Figure 3. A peak power of 23 mW at 100 GHz was obtained
for the case where the RF voltage is 1.5 V.
In general, the operating temperature is 80 K to 90 K above the room temperature.
Therefore, it is possible to extract more power by either increasing the device area or the
DC bias or a combination of both. Increasing the device area is limited by thermal effects
Third International Symposium on Space Terahertz Technology
Page 483
25
20 -
15 -
10 -
5 -
60
I
, , ,
1
/ • Vrf = 05 V \
/ o Vrf = 1.0 V \
-
1
/ ♦ Wf - 15 V
\ -
- ,„. -.1
i i
80
100
120
140
Frequency [GHz]
Figure 3: Power vs. frequency for a 1.7 /im long InP Gunn device doped at 9xl0 15 cm -3 .
Diameter = 40 /im, Vdc = 4.0 V, T = 450 K.
and by the minimum load provided by the resonant circuit. On the other hand, increasing
the DC bias voltage is limited by thermal effects and breakdown due to the large electric
field at the anode. For example, increasing the diode diameter from 40 ^m to 60 \xxa. results
in 50 mW output power at 100 GHz with an RF voltage of 1.5 V. The required matching
load is 2 Q compared to 3.75 $7 for the smaller device and the temperature increase is 120
K compared to 86 K.
5 Simulation of 1.0 /im InP Gunn Devices for Operation
in D-band Frequency Region
The 1.7 fj,m device considered in the previous section had an optimum operating frequency
around 100 GHz. For higher frequencies in D-band, structures with shorter active regions
need to be considered. In this section, simulation results of a 1 /jm long InP device are
presented.
5.1 Flat Doping Profile
Two flat doping profiles in the active region lxlO 16 cm -3 and 2xl0 16 cm -3 are considered.
Figure 4 compares the corresponding efficiency as a function of frequency for a DC bias of
4.0 V, an RF voltage of 0.5 V, and an operating temperature of 450 K. The structure with
2xl0 16 cm* 3 doping results in a higher efficiency at all frequencies. The output power into
a 2 ft load is shown in figure 5. Near 160 GHz, the predicted output power from the highly
doped structure is more than 5 times higher than the output power from the low doped
structure. However, for a meaningful comparison, thermal effects and current density levels
Page 484
Third International Symposium on Space Terahertz Technology
0.7
0.0
120 140 160 180 200 220
Frequency [GHz]
Figure 4: Efficiency versus frequency for a 1.0 ^m long InP Gunn device. Vdc = 4.5 V, Vrf
= 0.5 V,^ = 2fl,T = 450 K.
need to be considered.
The DC current density in the structure with a lower doping is about 3.95 xlO 4 A. cm -2
at 160 GHz whereas it is about 6.7xl0 4 A. cm -2 for the other structure. The device with
the higher doping has a very large current density which results in an operating temperature
approaching the limit for a 30 /xm diameter device. For smaller diodes, bonding becomes
very difficult. Therefore, there is a need for improving the efficiency while keeping the
current density from increasing rapidly. It will be shown in the next section that a graded
doping profile in the active region provides such an improvement.
5.2 Graded Doping Profile
This section examines methods of improving the efficiency of InP Gunn devices and opti-
mizing the design for operation around 160 GHz. In addition to the structure with a flat
doping of lxlO 16 cm -3 , three linearly graded doping profiles are considered:
Ni (xlO 16 cm -3 )
N 2 (xlO^cm" 3 )
Structure 1
1.0
1.0
Structure 2
0.8
1.5
Structure 3
0.8
3.0
Structure 4
0.8
4.0
In the above table, Ni is the doping density in the active region at the cathode side and
N2 is the corresponding doping at the anode side. The results presented in this section
correspond to a bias of 4.0 V, an RF voltage of 1.0 V, an operating temperature of 450 K,
Third International Symposium on Space Terahertz Technology
Page 485
220
Frequency [GHz]
Figure 5: Power versus frequency for a 1.0 /im long InP Gunn device. Vdc
0.5 V, R L = 2 ft, T = 450 K.
4.5 V, Vrf =
and a matching load of 2 ft.
Figure 6 shows the efficiency versus frequency for the three structures. The efficiency
is increasing as the doping profile becomes steeper. At 180 GHz, structure 3 results in
2 % efficiency which is twice the efficiency obtained from structure 1. A comparison of the
output power is shown in Figure 7. The optimum frequency for power generation is 160
GHz where structure 3 resulted in 73.4 mW compared with 8.6 mW for structure 1. The
DC current density in structure 3 is 5.18xl0 4 A. cm -2 compared with 3.84xl0 4 A. cm -2
in structure 1 at 160 GHz. This increase in the current density is much smaller than
the corresponding increase in the 2xl0 16 cm -3 doped structure. Structure 4 has higher
efficiencies than structure 3 but the current is also higher. For diodes with a mesa diameter
of 30 fim or larger, structure 3 is likely to be the optimum design for operation at 160 GHz.
The graded doping profile results in a higher electric field near the cathode and a lower
field near the anode due to electron diffusion toward the lower doped region. This change in
the field has two consequences: first, a higher cathode field results in a larger fraction of the
electrons transferring to the upper valleys, second, a lower anode field permits application of
a larger DC bias without breakdown. The electric field in structure 1 peaks at 125 kV.cm -1
near the anode side whereas the field in structure 4 is less than 100 kV.cm -1 at the anode.
In addition to improving the efficiency and reducing the field at the anode, a graded
doping profile provides a modest increase in the current density. The graded doping profile
can provide the same performance as the flat doped profile with a smaller current density.
The reason is that the higher fraction of the electron population transferring to the upper
valleys in the graded structure results in a reduction of the average velocity. This is not
the case for the structure with a flat doping where an increase in the carrier density does
not alter the distribution of the field across the structure. Figure 8 compares the current
Page 486
Third International Symposium on Space Terahertz Technology
2.5
1.5 -
0.5
Frequency [GHz]
200
Figure 6: Efficiency versus frequency for a 1.0 //m InP Gunn device. Vdc = 4.0 V, Vrf '=
1.0 V, RL = 2 fi, T = 450 K.
CD
80
70
60
50
40
30
20
10
-
1 1 1
-
-
/• flat 10 x 10" on" 1 \
/ ON2 = 15«10* cm'' \
-
-/
♦ N2 - 3 x 10" cm" 1 \
-
- .
i i i
\ -
120 140 160 180
Frequency [GHz]
200
Figure 7: Power versus frequency for a 1.0 //m InP Gunn device. Vdc = 4.0 V, Vrf = 1.0
V, RL = 2 n, T = 450 K.
Third International Symposium on Space Terahertz Technology
Page 487
'$2
•3
7.5
6.5
5.5
4.5
3.5
2.5
- 4.5
3.5
125
250 275
Time (ps)
625
o
3
Figure 8: Comparison of the current density in structure 1 and structure 3 at 160 GHz.
Vdc = 4.0 V, Vrf = 1.0 V, T = 450 K.
waveforms in one RF period as obtained from structure 1 and structure 3. The current
density in structure 3 resembles a pulse 180 degrees out of phase with the RF voltage
whereas in structure 1, it is more sinusoidal and is not perfectly out of phase. This shows
that structure 3 is more efficient and the space charge layers are more developed.
To verify that the highly doped side of the active region should be at the anode side for
best performance, structure 3 was considered with the opposite polarity. Figure 9 shows a
comparison of the output power as a function of frequency. The structure with the doping
decreasing toward the anode contribute to much smaller output power levels when compared
with structure 3. An examination of the electric field revealed a peak electric field near the
anode close to 200 kV.cm -1 compared with 100 kV.cm -1 in structure 3. In addition the
current density is higher, at 160 GHz it has a average value of 6.9 XlO 4 A. cm -2 compared
with 5.2xl0 4 A. cm -2 in structure 3.
6 InP GUNN DEVICE TECHNOLOGY
Among the many techniques used for fabricating Gunn devices, the most common are based
on the Integral Heat Sink process (IHS) [7, 8] or the flip-chip process [9]. In the IHS process,
the heat sink is formed as an integral part of the diode. The wafer front side is metallized
and plated with copper, silver, or gold to a thickness of several thousands of an inch. The
substrate is chemically or mechanically thinned to a thickness of 10 fim to 15 /xm . Next,
ohmic contacts are metallized and standard photoresist techniques are used to define the
mesas. Individual chips are then mounted in standard packages. In the flip-chip process
the mesas are defined on the epitaxial side. The mesa chip is then flipped and mounted on
a heat sink pedestal. Finally, the substrate is thinned to about 100 /zm .
Page 488
Third International Symposium on Space Terahertz Technology
100 c
cu
0.1
200
Frequency [GHz]
Figure 9: Comparison of the power versus frequency for structure 3 with different bias
polarities . Vdc = 4.0 V, Vrf = 1.0 V, RL = 2 Q, T = 450 K.
These two processing techniques have been used successfully for fabricating Gunn de-
vices in the W-band. At higher frequencies, it is necessary to reduce further the substrate
thickness to minimize the series resistance. In addition, the smaller size of the mesas
presents new challenges to the bonding procedure. In this chapter a new fabrication tech-
nology allowing the complete removal of the substrate is developed. This process is an
extension of the integral heat sink technique with the additional step of plating the top
contact to facilitate bonding.
Figure 10 shows the epitaxial layers of an unprocessed InP wafer. The different layers,
starting from the n + doped substrate consist of
• an n + InP substrate,
• a 0.5 /xm n + InGaAs layer doped at 2.0xl0 18 cm -3 ,
• a 0.6 /xm n + InP contact layer doped at 2.0xl0 18 cm -3 ,
• a 1.0 /xm n InP active region doped doped at l.OxlO 16 cm -3 ,
• a 0.2 /xm n + InP contact layer doped at 2.0xl0 18 cm -3 , and
• a 0.1 /xm n + InGaAs cap layer doped at 2.0xl0 18 cm' 3 .
Prior to processing the wafer, the doping profile in the active region is characterized through
C-V measurements. To perform these measurements, the top n + InGaAs and InP layers
are chemically etched from a small sample so that Schottky contacts can be formed. The
processing sequence for Gunn device fabrication is shown in figure 11 and described below.
Third International Symposium on Space Terahertz Technology Page 489
6.1 Island Definition and Integral Heat Sink Formation
The first step consists of defining square islands approximately 400 //m x 400 /an in size
separated on all sides by 100 /xm wide trenches. These trenches are etched down to the
InGaAs etch-stop layer. An n-ohmic contact (Ni/Ge/Au/Ti/Au) is evaporated over the
whole surface and then gold is plated to a thickness of 25 /xm to form the integral heat
sink. The top InGaAs cap layer reduces the contact resistance [10] of the ohmic contact
because InGaAs has a lower bandgap than GaAs. The isolation provided by the trenches is
helpful in reducing the cracking of the semiconductor epilayers during the annealing process.
Cracks occur because the gold heat sink and the InP semiconductor have different thermal
expansion coefficients. Figure 11(a) shows a cross section of the sample at the end of the
gold plating.
6.2 Substrate Thinning and Top Contact Definition
The InGaAs layer, referred to as an etch-stop layer, permits the complete removal of the
substrate by chemical etching. The chemical solution HCl:H20 (4:1) selectively etches InP
and does not etch InGaAs. Once the substrate is removed, the InGaAs etch-stop layer
is etched away using HzPO4.H2O2.H2O (1:1:8) which does not attack the InP n + region.
Standard lift-off techniques are used to define circular diodes with sizes varying from 30 to
65 /im in diameter which are then metallized to form n-ohmic contacts. Figure 11(b) shows
a cross section of the sample after the substrate thinning and the heat sink formation.
6.3 Gold Plating of Top Contacts
In the standard IHS technique, the next step would be to etch the mesa. However problems
in bonding have been encountered due to the thin ohmic contact. A thick ohmic contact is
obtained by plating gold on top of the evaporated ohmic contacts. A conductive metal layer
is needed to electroplate uniformly over all the contacts. A Ti/Au/Ti layer is evaporated
over the whole surface, then an alignment over the ohmic contacts is used to open holes in
a thick photoresist (3 /jm ). The photoresist is removed from a small region at the edge of
the sample. The exposed Ti layer is also removed in Buffered HF and the sample is plated
for a thickness of 2 fim - 3 fim. Figure 11(c) shows a cross section of the sample after the
plating step.
6.4 Mesa Definition and Annealing
The final step before mounting individual chips consists of etching the mesas and annealing
the ohmic contacts. The final structure is shown in figure 11(d).
Page 490
Third International Symposium on Space Terahertz Technology
InGaAs cap layer
) l l l|l lll l l l|l ll |l l l|l l l|lll| l ll|l ll )llll ll lj ll l|ll l | l l l |lip
nlnP
InGaAs etch -stop -
layer
InP Substrate
Figure 10: InP Gunn structure
7 Conclusions
A new method has been developed for estimating the material parameters used in the
Monte Carlo model. By comparing simulation and experimental results in the W-band,
we obtained more accurate material parameters. Lower values for the intervalley energy
separation and higher values for the deformation potentials than stated in the literature
were used. A possible explanation for these trends is the high operating temperature of the
Gunn device which perturbs the band-structure.
Using these parameters, it was shown that it is possible to operate fundamental mode
InP Gunn devices in the D-band. The performance of a flat doped structure can be con-
siderably improved by employing a graded doping profile in the active region. Specifically,
a linearly graded doping increasing from the cathode toward the anode improves the con-
version efficiency, the output power, reduces the electric field at the anode, and results in a
smaller current density compared with a flat profile. A structure with a doping decreasing
toward the anode is not desirable because it increases the electric field at the anode and
does not reduce the dead zone. As a result, the device breaks down at lower voltages and
the performance is degraded.
A processing technology for GaAs and InP Gunn devices has been developed based on
the integral heat sink processing technique. An Etch-stop layer between the substrate and
the epilayers was included in the wafer design in order to completely remove the substrate
and obtain better uniformity across the chip. InGaAs cap layers were used to reduce the
contact resistance. A process was developed for plating the top contacts with gold to
facilitate bonding.
Third International Symposium on Space Terahertz Technology
Page 491
Plated Au
n-ohmic
n + InGaAs
IbP Substrate
a) Island definition, n-ohmic evaporation, and
gold plating (25 Jim)
n-ohmic
Ni/Ge/Au/Ti/Au
b) Substrate thinning, etch stop layer removal,
and n-ohmic evaporation
Figure 11: Processing sequence for InP Gunn fabrication
Page 492
Third International Symposium on Space Terahertz Technology
Ti/Au/Ti
Photoresist
c) Gold plating of ohmic contacts
d) Final diodes after annealing and mesa etch
Figure 11: Cont. Processing sequence for InP Gunn fabrication
Third International Symposium on Space Terahertz Technology p a g e 493
References
[1] W. Fawcett, A. D. Boardman and S. Swain, "Monte Carlo Determination of Electron
Transport Properties in Gallium Arsenide," J. Phys. Chem. Solids, 30, 1969, pp. 643.
[2] K. Brennan, K. Hess, J. Y. Tang, and G. J. Iafrate, "Transient Electronic Transport
in InP Under the Condition of High-Energy Electron Injection," IEEE Trans, on
Electron Dev., ED-30, 12, pp. 1750-1753, Dec. 1983.
[3] D. C. Herbert, W. Fawcett, and C. Hilsum, "High Field Transport in Indium Phos-
phide," J. Phys. C: Solid State Phys., Vol. 9, pp. 3969-3975, 1976.
[4] G. H. Glover, "Study of Electron energy Relaxation Times in GaAs and InP," J. Appl.
Phys., 44, No. 3, pp. 1295-1301, March 1973.
[5] T. J. Maloney, and J. Frey, "Transient and Steady-State Electron Transport Proper-
ties of GaAs and InP," J. Appl. Phys., 48, No. 2, pp. 781-787, Feb. 1977.
[6] M. V. Fischetti, "Monte Carlo Simulation of Transport in Technologically Significant
Semiconductors of the Diamond and Zinc- Blende Structures- Part I: Homogeneous
Transport," IEEE Trans, on Electron Dev., ED-38, No. 3, pp. 634-649, March 1991.
[7] R. A. Zettler, and A. M. Cowley, "Batch fabrication of Integral-Heat Sink IMPATT
Diodes," Electronics Letters, Vol. 5, No. 26, pp. 693-694, Dec. 1969.
[8] S. Y. Narayan, J. P. Paczkowski, "Integral Heat Sink Transferred Electron Oscilla-
tors," RCA Review, Vol. 33, pp. 752-765, Dec. 1972.
[9] A. Paolella, R. L. Ross, and J. Ondria, "Advanced mm- Wave Sources by Automated
MBE," Microwave Journal, p. 149, April 1986.
[10] J. M. Woodall, J. L. Freeouf, G. D. Pettit, T. Jackson, and P. Kirchner, "Ohmic Con-
tacts to n-GaAs using Graded Band Gap Layers of Ga\_ x In x As Grown by Molecular
Beam Epitaxy," J. Vac. Sci. TechnoL, 19, No. 3, Sept./Oct. 1981, p. 626.
Page 494 Third International Symposium on Space Terahertz Technology
A Recent Advances in Superconducting- Mixer
JJ 3. -27 ? 68
Stimulations
S. Withington and P.R. Kennedy
Cavendish Laboratory,
University of Cambridge,
England.
March 22, 1992
1 Introduction
Over the last few years, considerable progress has been made in the development of tech-
niques for fabricating high-quality superconducting circuits, and this success, together with
major advances in the theoretical understanding of quantum detection and mixing at mil-
limetre and submillimetre wavelengths [1], has made the development of CAD techniques
for superconducting nonlinear circuits an important new enterprise. For example, arrays of
quasioptical mixers are now being manufactured, where the antennas, matching networks,
filters and superconducting tunnel junctions are all fabricated by depositing niobium and
a variety of oxides on a single quartz substrate. There are no adjustable tuning elements
on these integrated circuits, and therefore, one must be able to predict their electrical
behaviour precisely. This requirement, together with a general interest in the generic be-
haviour of devices such as direct detectors and harmonic mixers, has lead us to develop
a range of CAD tools for simulating the large-signal, small-signal, and noise behaviour of
superconducting tunnel junction circuits.
2 Large-signal analysis
To model the behaviour of a quasiparticle mixer, it is first necessary to simulate the large-
signal steady-state dynamics of the local-oscillator circuit. Once the large-signal operating
point is known, it is then possible to perturb, either numerically or analytically, the under-
lying system of equations to gain information about the linear relationships between signal
and noise variables.
The main problem is how does one calculate the periodic current that flows through a
tunnel junction when a periodic voltage is applied? For semiconductor devices this calcula-
tion is almost always carried out in the time domain, and fast Fourier transforms are used
to interface the terminal waveforms to the frequency- domain description of the embedding
circuit.
Classical resistive mixer diodes are relatively easy to simulate because the induced
current is an instantaneous function of the terminal voltage. Quantum mixer diodes on
the other hand are difficult to simulate because the tunnelling current depends on the
voltage that was across the junction at very long times in the past. In the time domain
Third International Symposium on Space Terahertz Technology Page 495
the current is calculated through an integral which is similar to the convolution integral
of linear systems theory, and the tunnel junction is characterised by a response function
■which oscillates at the gap frequency with an envelope which decays inversely with time
at large times. To evaluate the tunnelling current it is necessary to sample the terminal
voltage at a rate greater than the gap frequency , and to integrate beyond a limit which is
inversely related to the voltage width of the dc nonlinearity. Time-domain simulations are
useful for studying the switching behaviour of tunnel junctions, but they are inappropriate
for studying the steady-state behaviour of RF circuits.
When a sinusoidal potential is applied to a superconducting tunnel junction, the wave-
functions associated with the quasiparticle states on one side of the barrier are coherently
phase modulated. The spectrum of the phase factor is a comb of delta functions whose
coefficients are the elements of a Bessel-function sequence. The Bessel functions have the
same argument, determined by the voltage drive level, and consecutive orders ranging from
some large positive integer to the same negative integer. The trick is to recognize that when
a periodic potential is applied, the spectrum of the overall phase factor is the convolution
of the spectra associated with the individual harmonic contributions. Once the spectrum
of the overall phase factor is known, it is possible to calculate the harmonic phasors of the
tunnelling current from the dc I-V curve and its Hilbert transform [2].
The above procedure describes a way of calculating the periodic current that flows in a
tunnel junction when a periodic potential is applied. In a real circuit the tunnel junction
is embedded in a linear network and the problem of determining the various voltages and
currents is complex. Applying the method of harmonic balance [3] to a generic circuit
comprising a tunnel junction and a Thevenin voltage source, leads to a system of coupled
nonlinear algebraic equations. Mathematically, the problem then consists of finding the
roots of these equations; electrically, the problem is equivalent to searching for a waveform
that simultaneously satisfies the circuit equations at every harmonic frequency. In nonlinear
CAD terminology, the scheme is a frequency- domain spectral-balance method, however,
unlike other versions, the spectral decomposition is based on device physics, rather than on
expanding the terminal behaviour in a set of basis functions.
The set of algebraic equations that results from applying the method of harmonic bal-
ance to a tunnel-junction circuit must, in general, be solved numerically. By repeatedly
analyzing, in different ways, a large pseudo-random set of tunnel-junction circuits, we have
investigated the speeds and stabilities of a range of iterative root-finding techniques. A
comparison of the techniques is shown in Fig. 1, where we have plotted the percentage
of circuits that converge, and the mean number of iterations taken, as a function of the
damping factor. The damping factor is a coefficient between and 1 which determines the
degree to which the result of an iteration influences the next guess. A small value improves
stability at the expense of reducing the rate of convergence. The solid and dashed lines in
Fig. 1 correspond to two different quality characteristics. It should be appreciated that the
plots represent a total of around 20,000 circuit simulations.
The tunnel junction is a nonlinear admittance in the sense that it is most easy to
calculate the current in terms of the terminal voltage. However, fixed-point voltage-update
methods [4] are inappropriate for analyzing tunnel- junction circuits, especially in a common-
user environment, because they fail to converge when the large- signal harmonic admittances
of the tunnel junction are much greater than those of the embedding circuit. This problem
is clearly demonstrated in Fig. 1 where it is seen that the routine will only converge if
the system is heavily damped. A slightly more sophisticated way of finding the roots is
Page 496
Third International Symposium on Space Terahertz Technology
-a
v
ho
u
V
>
S
o
u
s
u
Fixed Point
80
HI 1 1 1 1 1 III 1 ILL
60
— —
40
^r x —
20
r\ -:
1 1 1 1 1 1 1 rm 1 1 1 1
Secant
80
III
|...
|i.i|iLL
60
—
40
—
20
—
—
771
Llll
inlirr
.2 .4 .6 .8
.2 .4 .6 .8
Harmonic Newton
iiijmjmjjii
-rr 1 1 1 1 1 1 1 1 1 1 1 i-r
.2 .4 .6 .8
L_l 1 I I I I f 1 I I I I 1 I 1_|
.2 .4 .6 .8
Jlllllll II II IIL|
200 E-
150
100
50
OhIihIiiiIuj:
Jll I I II I I III ML
.2 .4 .6 .8
Damping factor y
.2 .4 .6 .8
Figure 1: Comparison of various techniques for calculating the large-signal quantum be-
haviour of superconducting tunnel-junction circuits.
to use a multi-dimensional variant of the secant method [5]. This method is similar to
the fixed-point method, in the sense that it is only necessary to calculate the tunnelling
current once per iteration, however, because coarse derivative information is included one
might expect the routine to behave more reasonably. Somewhat surprisingly, the routine is
significantly worse despite the additional information. The problem is caused by the fact
that, effectively, only the terms on the leading diagonal of the Jacobian matrix are included,
and coupling between harmonics relies on the current calculations. As long as the current
at a given harmonic is most strongly influenced by the voltage at the same harmonic then
the routine will work well. In a highly nonlinear tunnel-junction circuit, however, there
is strong coupling between harmonics and the routine is inadequate. Fig. 1 shows the
behaviour of a harmonic-Newton [6] [7] scheme where the full Jacobian matrix is used. It is
possible to calculate the Jacobian matrix analytically, however, we prefer to calculate the
Jacobian matrix using finite differences. Harmonic Newton results in the least number of
failures; in fact, it finds a solution for 75 % of the circuits studied, and to a large extent,
the stability of the method is independent of the quality of the junction being investigated.
The fact that the convergence parameter has little effect on this fraction, together with the
almost reciprocal dependence of the mean number of iterations, shows that if it is possible
for the method to find a solution then it will eventually do so. Reducing the damping factor
Third International Symposium on Space Terahertz Technology Page 497
simply reduces the size of the voltage steps taken at each iteration, however, these steps
are usually in the correct direction.
We have now performed a very large number of real circuit simulations, and despite
the fact that 25 % of the randomly generated circuits failed to converge, we have never
come across a real circuit that has not converged. We have investigated this problem in
some detail, and we have found that many of the circuits that do not converge are behaving
in a non-periodic manner. This behaviour usually requires that the embedding circuit
impedances are very much larger than the normal-state resistance.
An alternative approach to finding a root in many dimensions is to recast the problem
into a multidimensional optimization. To do this change, the error function is used to
construct a scalar quantity that has a global minimum at the required root. One of the
attractions of optimization is that uninteresting variables, such as the local oscillator drive
level, can be eliminated from the analysis by making the variable part of the objective
function. In general, we have found that optimization methods are slow and should not be
used unless there is a particular reason to do so. Unfortunately, there is insufficient time
to discuss this more advanced topic in this short paper.
The results of a typical large-signal analysis are shown in Fig. 2. The sequence of plots
shows how the pumped dc I-V curve of a typical Nb-AlOx-Nb tunnel junction evolves as
the wCR product is changed. As the capacitance decreases the subgap current increases,
and non-classical negative differential resistance is induced on a number of photon steps.
Notice that large capacitances are required before the characteristic relaxes to its constant
sinusoidal- voltage-driven form. Also shown, for comparison, is an analysis where the har-
monic feedback is turned off. Curiously, it seems as if internal harmonic pumping can
enhance the negative differential resistance induced on high-order photon steps — for this
reason it is possible, in certain circumstances, for the small-signal behaviour of a mixer to
be very sensitive to harmonic impedance levels.
3 Small-signal analysis
Once the large-signal behaviour of a mixer has been established, it is possible to calcu-
late the small-signal and noise performance. The admittance and noise-current correlation
matrices are determined, in the usual way, through quantum-mechanical generalizations of
commonly used classical concepts. We then use a selection of linear transforms to reduce
the admittance parameters to two-port impedance and scattering parameters, and the cur-
rent correlation matrix to a noise-temperature matrix from which the standard two-port
noise parameters can be deduced [8] [9]. It transpires that the whole scheme, both signal
and noise, can be very elegantly normalized to the gap voltage and gap current of the
tunnel barrier. The advantages of our generalized approach are that one does not have to
specify before hand which ports are to be used for the input and output, and one can easily
calculate the two-port small-signal and noise parameters which can then be loaded into pro-
prietary microwave circuit simulators for further analysis. For example, we are interested in
designing mixers that have the first stage of low-noise IF amplification in the mixer block.
A further advantage of our scheme is that the noise performance is described in terms of
correlated travelling noise waves, and this approach is an elegant way of considering a mixer
as an integral part of a quasioptical system along which noise waves propagate.
Page 498
Third International Symposium on Space Terahertz Technology
IS
No harmonics.
Rs = 6.0 , u>CR = 0.0
2
J T r 1 1 | 1
1 1 1 I 1 1 1 1 1 1
1 1 1 1
! Rs = 2.0
uCR =1.0
/ -
l.S
-
-
1
-
-
-
IiimIi
, , 1 , , , , 1 , ,
, 1.!
J I I I I I I I I ' I I I I 1 I
Rs = 2.0 , u/CR = 0.0
Figure 2: The dc I-V curve of a pumped Nb-AlOx-Nb tunnel junction for different values
of ujCR product. The first plot does not include internal harmonic pumping.
4 Mixer simulations
To date we have studied the large-signal, small-signal, and noise behaviour of mixers by
adopting the design procedure suggested by Kerr [10]. That is to say, the mixer is operated
in a double-sideband mode, and the source and load impedances are assumed to be real. The
ratio of the source and load impedances, which in practice is determined by the geometry
of the mount, is set at some fixed value. In general, we use a value of unity as a higher
value tends to degrade the input return loss of the mixer. It is interesting to note, however,
that it may be possible to choose the ratio so as to minimize the sensitivity of the gain
to variations in the source resistance. The free parameter, as far as the design process
is concerned, is the normal-state resistance. Although we assume that the capacitance of
the tunnel junction is tuned out at the fundamental, we assume that the impedance at
the harmonics is given by the capacitance of the tunnel junction alone. The effects of
junction capacitance are considered in companion paper [11], here we simply demonstrate
the procedure by plotting, in Fig. 3, the transducer gain, noise temperature, input return
loss, and normalized output impedance of a typical Nb-AlOx-Nb mixer as a function of the
normalized source resistance; the various curves are for different normalized frequencies (
normalized to the gap frequency). It is interesting to note that the overall performance is
Third International Symposium on Space Terahertz Tech7iology
Page 499
s
•3
bo
CQ
CO
CO
-10
V
V
a.
S
V
o
a
CO
co
CO
1.5 2
_ 20
05
w
O
J 10 -
s
a
0)
3
O.
a
-
-10
_i r t - i [ 1 1 1 i | 1 1
1 r | f i i r
" °- 2 _ .101
__
^^0.8 N.
^s. ^_
~i 1 1 1 1 1 1 1 1 1 1 1
i\ 1 1 1 1 1"
1.5
Normalized source resistance
Figure 3: The transducer gain, noise temperature, input return loss, and normalized output
impedance of a typical Nb-AlOx-Nb mixer as a function of the normalized source resistance.
very poor for low values of source resistance, and this is probably the single most important
reason why mixer performances improve significantly when integrated tuning elements are
used.
A useful normalized expression can easily be derived for the source resistance at which
unity gain, good input match, and minimum noise temperature can be achieved. The
expression is
|^ = 0.5Vp° 92 = 78
f(GHz)
V„ (mV)
-0.92
(1)
and it applies for frequencies between 0.2 and 0.8 of the gap frequency. The exponent is
slightly different from that given by Kerr and Pan, because we have taken into account the
fact, that the optimum bias point does not remain in the middle of the first photon step
below the gap for frequencies greater than about 0.5 of the gap frequency.
It is well known that if one plots the conversion gain, at a given frequency, as a function
of the uiC R product, at some point the conversion gain becomes depressed. This can be
regarded as the frequency at which harmonic effects become significant; or equivalently, the
frequency at which the five-port, rather than the three-port model should be used. Using
the above value for the source resistance, we have investigated this behaviour and generated
Page 500
Third International Symposium on Space Terahertz Technology
wCRn = 0.5
10000
a
■a
t
w
O
1000
* \
^^^
\
\
_^^~
S r — -
_^00*~
^•"^
r»*
^t*~~
^***~
^0* *'
.
+***
^^
^00**^ •
^-"^.
,'' ^^--^
^^-^
^^^- *■
-"' ' ^s^*Z^"
r ^O-* **Z*>* <^ *^ ^""^
i^"*" ^^ 00r ' ^*** 0r
r*^^**^
*"
1.0
2.0
9.0
100
Frequency / (GHz)
1000
Figure 4: Critical current density against frequency for different values of u/CR product.
The dotted line shows the optimum current density as a function of frequency.
the following expression for the optimum wCR product
uCR
\opt
= Vf° 75 = 61
V g (mV.
If (GHz)}
0.75
(2)
and this in turn generates the following expression for the optimum value of the critical
current density: J c (Acm- 2 ) - OAf(GHz) 175 . We have assumed I c R n = 1.8ml' and a
specific capacitance of 45fFfim~ 2 . Once again the above requirement is less severe than
that published by Kerr. Finally, in Fig. 4 we plot the optimum critical current density as
a function of frequency, and we show lines of constant u)C R product. Above the dotted
line, harmonic effects are important, whereas below the dotted line harmonic effects can be
ignored.
References
[1] J.R. Tucker and M.J. Feldman, "Quantum detection at millimetre wavelengths," Rev.
Mod. Phys., vol. 57, pp. 1055-1113, 1985.
Third International Symposium on Space Terahertz Technology Page 501
[2] S. Withington and E.L. Kollberg, "Spectral-domain analysis of harmonic effects in
superconducting quasi-particle mixers,"IEEE Trans. Microwave Theory Tech., vol.
MTT-37, pp. 231-238, 1989.
[3] K.S. Kundert and A. Sangiovanni- Vincent elli, "Simulation of nonlinear circuits in the
frequency domain.'TEEE Trans. Computer-Aided Design, vol. CAD-5, pp. 521-535,
Oct. 1986.
(4] R.G. Hicks and P.Kahn, "Numerical analysis of nonlinear solid-state device excitation
in microwave circuits," IEEE Trans. Microwave Theory Tech., vol. MTT-30, pp. 251-
259, Mar. 1982.
[5] C. Camacho-Penalosa, "Numerical steady-state analysis of nonlinear microwave cir-
cuits with periodic excitation,"IEEE Trans. Microwave Theory Tech., vol. MTT-31,
pp. 724-730, Sep. 1983.
(6] C.-Y.E. Tong and R. Blundell, "Simulation of superconducting quasiparticle mixer us-
ing a five-port model," IEEE Trans. Microwave Theory Tech., vol. MTT-38, pp. 1391-
1398, Oct. 1990.
[7] S. Withington and P. Kennedy, "Numerical procedure for simulating the large-signal
quantum behaviour of superconducting tunnel-junction circuits," Proc. IEE, part G,
vol. 138, pp. 70-76, Feb. 1991.
[8] S. Withington, "Scattered noise waves in microwave and millimetre- wave net-
works, "Microwave Journal, vol. 32, pp. 169-178, 1989.
[9] L.R. D'Addario, "Noise parameters of SIS mixers," IEEE Trans. Microwave Theory
Tech., vol. MTT-36, pp. 1196-1206, July 1988.
[10] A.R. Kerr and S.-K. Pan, "Some recent developments in the design of SIS mixers,"
Infrared and Millimetre Waves, vol. 11, pp. 1169-1187, 1990.
[11] A.R. Kerr, S.-K. Pan, and S. Withington, "Embedding impedance approximations in
the analysis of SIS mixers," Proceedings Space Terahertz Conference, 1992.
Page 502 Third International Symposium on Space Terahertz Technology
SUBMILLIMETER WAVE DETECTION WITH
SUPERCONDUCTING TUNNEL DIODES
Michael J. Wengler M Q 3 - %7 '7 6 9
University of Rochester
ABSTRACT
Superconductor-Insulator-Superconductor (SIS) diodes are the detector elements in the
most sensitive heterodyne receivers available from 100 to 500 GHz. SIS mixers are the front end
of radio astronomical systems around the world. SIS mixer technology is being extended to 1 THz
and higher frequencies for eventual use on spacebome astronomical experiments.
Here is a short review of submillimeter SIS mixers. The role of impedance matching in the
proper design of an SIS mixer is described. A variety of methods for achieving good impedance
match at submillimeter frequencies are presented. The experimental state of the submillimeter SIS
jnixer art is described and summarized.
1 . INTRODUCTION
Twelve years ago, the first descriptions of mixing on a superconducting tunnel diode called
a Superconductor-Insulator-Superconductor (SIS) were published [1,2]. At about the same time,
a comprehensive theoretical investigation of the SIS made clear that SIS's respond to photons at
millimeter and submillimeter wavelengths [3]. The SIS has become the instrument of choice for
millimeter spectroscopic radio astronomy, finding use on radio telescopes around the world [4-13].
Since the discovery of the SIS mixer, the technology for fabricating SIS circuits has improved
immensely so that nearly ideal niobium devices with sub-micron feature sizes can be fabricated
[14-20]. Recently, the high performance of SIS mixers has been extended to submillimeter
wavelength radio astronomy [8, 21-23, 20, 24].
This paper is a shortened version of a review published elsewhere [25]. The full review
includes an in-depth introduction to the field of SIS mixing. An excellent and comprehensive
earlier review of both theory and experiment is that of Tucker and Feldman [26]. Reviews of the
state of the art of SIS mixers have appeared regularly [27-30]. Other reviews include both SIS
mixers and competing receiver technologies used in radio astronomy [31, 32].
Third International Symposium on Space Terahertz Technology Page 503
2 . RF ADMITTANCE OF THE SIS
The SIS mixer has two important if input admittances. The first one is Ylo» the admittance
the SIS presents to the LO. The second is Yin, the admittance the SIS mixer presents at the signal
frequency. Because the signal power must always be much less than the absorbed LO power for a
mixer to avoid saturation, it is useful to think of Ylo as the "large signal admittance" of the SIS,
and Yin as the "small signal admittance."
For a submillimeter SIS mixer, Ylo is generally quite close to Gn, the dc normal state
admittance of the SIS. Ylo is nearly constant over the range of useful dc and LO biasing
conditions. Ylo is completely independent of the signal, image, LO, and harmonic source
admittances to which the SIS is coupled. In fact, Ylo is the admittance presented by the SIS seen
as a passive absorber of radiation.
Yin, on the other hand, is a highly variable quantity. It does change with all of the
parameters mentioned above. It is possible for Yin to have a negative real part, while Ylo mus t
always have a positive real part. For this reason, Yin m ust be interpreted as the input admittance
of the SIS viewed as an active device powered by the LO.
2.1. SIS Parasitic Capacitance
In submillimeter SIS mixers, a dominant part of the SIS's rf admittance is due to its
parasitic capacitance. The capacitance per unit area of the SIS junctions used in submillimeter
mixers is nearly a constant depending only on the materials from which the junctions are made.
Numbers commonly used in design are 50 fF/(jim) 2 for lead alloy SIS's, and 50 to 70 fF/(|i.m) 2
for niobium SIS's with aluminum-oxide insulating layers.
Many mixer designers aim to minimize the SIS capacitance. In fact, many of the best SIS
mixers have been built with low capacitance junctions. The lowest SIS receiver noise temperature
at 230 GHz are achieved with 0.25 (|im) 2 .Nb junctions with C = 1 .7 x G^/lnf [13]. Open-
structure SIS mixers have work fairly well at their low frequency ends where C < 1.0 x GN/27rf
[8, 33].
However, there is some evidence that C = 3 x G^/2Kf provides more benefit than harm in a
well designed SIS mixer. The reasons for this, and the value of the optimum capacitance are
Page 504 Third International Symposium on Space Terahertz Technology
discussed in these references [1 1, 34]. The larger capacitance seems especially important when
series arrays of SIS's are used [35].
Whether or not a non-zero capacitance is helpful becomes a less important question as the
operation frequency of SIS mixers is raised. All suggestions of a non-zero capacitance still have
that optimum falling as f" 1 or f" 2 . As f is raised it becomes increasingly difficult to fabricate SIS's
with C less than any of the proposed optimum values.
2 .2 . Optimum Signal Source Admittance: Match to Yio
The signal appears to come from a source admittance Ys. The value of Ys is one of the
major design available to an SIS mixer designer. For submillimeter SIS mixers, the proper choice
is Ys = Y* . This choice of Ys minimizes the SIS mixer noise temperature. This result is
predicted from photodiode mixer theory [25], and also by detailed Tucker theory calculations [36].
It is paradoxical that it is not the actual small signal admittance Yin to which the mixer
structure must supply a match, but the large signal LO admittance, Ylo In fact, a choice of
Ys = YjL will maximize the SIS mixer's gain instead of its noise. Because IF amplifiers have very
low noise levels, it is much more important to minimize mixer noise than to maximize mixer gain in
submillimeter SIS mixers.
3 . IMPEDANCE MATCHING TECHNIQUES
There are two tasks which must be accomplished to match the signal source admittance Ys
to the SIS's input admittance Ylo- Fu " st > the values of Ylo typical of SIS's are much larger than
the source admittances presented by most antennas and waveguide structures. Second, the
parasitic capacitance of the SIS must be tuned out. These two tasks can be accomplished in a few
different ways. In some cases, a single tuning structure can do both of these things
simultaneously.
3.1. Integrated tuning
It is a relatively simple matter to integrate tuning structures with the SIS diode in its
photolithographic stage of fabrication. There are at least two layers of low loss superconductor
required in SIS fabrication, which allows for various "two-wire" tuning elements. With the
addition of a thick (around 2,000 A) insulating layer, stripline tuning structures can be fabricated.
Third International Symposium on Space Terahertz Technology
Page 505
I
T i'
a )
b )
Figure 1 [38]. Two tuning structures integrated with SIS junctions, a) The sickle-
shaped piece is the bottom layer of superconductor, the straight piece the top.
Where the two layers overlap, they are actually separated by a thick (~2,000 A)
layer of insulator, except in the dashed square, where they come much closer to
form the SIS. b) A tuning structure based entirely on microstrip.
Direct measurements of the efficacy of tuning elements in the 100-500 GHz range have been made
using Fourier Transform Spectrometer (FTS) measurements [37].
The two tuning structures shown in fig. 1 are compact and simple designs proposed by
Kerr, Pan, and Feldman [38]. Their primary role is to provide an inductance at the SIS which
tunes out its parasitic capacitance. The structure shown in b) is particularly appropriate for higher --
rf frequencies. In both a) and b) structures are shown which are A/4 in length. Here, A refers to
the wavelength associated with rf radiation propagating along the integrated tuning structure. The
structures labeled with length A/4 are designed to present a short circuit to the rf at their left ends,
but are open circuited at dc and fjp. In a), the tuning structure shown to be / long is essentially a
single-turn inductor. In b), the tuning structure shown to be / long is a short length of high
impedance microstrip. Since it is rf-shorted at one end, it presents an inductive susceptance across
the SIS. The reactances associated with these structures are very small at the IF and so have no
effect on the IF or dc properties of the mixer.
Using SIS circuits designed with integrated tuning, it is possible to build SIS mixers with
no mechanical tuning which have excellent responsivity over an entire waveguide band at
millimeter wavelengths [11, 39]. These two mixers are quite different in their design, both should
be reviewed for a good appreciation of the range of integrated tuning circuitry which is possible.
Page 506
Third International Symposium on Space Terahertz Technology
Figure 2 [40]. A spiral antenna SIS mixer with transmission line tuning a) The
radiauon coupling structure is a 3.4 mm diameter spiral antenna, b) Instead of
SfXJ S at *? e °T^ lead i ° f the spiral (1) ' a mi <™stripline (3) is formed
and the SIS (2) is placed at its end.
3 .2 . Microstrip transmission line transformers
The inductive structures shown above address only the problem of tuning out SIS parasitic
capacitance. Using microstrip transmission lines, it is possible to tune out the SIS's capacitance
and to lower Ylo to a more convenient value.
The simplest circuit which does this is shown in fig. 2. The large-scale structure shown in
a) is a spiral dipole antenna. Radiation coupled by this antenna to the antenna leads at its center has
a source admittance of about (1 14 Q)-l. Rather than placing the SIS at the center of the antenna,
the SIS is placed at the end of a microstrip transmission line as shown in b). One of the poles of
the antenna is used as the ground plane for this transmission line. The transmission line has a
fairly low impedance and a length chosen so that the SIS's high value of Y L0 is transformed down
to about (1 14 Q)-l where it is connected to the center of the antenna. In fact, the transmission line
transformer is a little longer than a standard quarter-wave transformer so that it also tunes out the
SIS's parasitic capacitance.
Third International Symposium on Space Terahertz Technology Page 507
This simple technique can be extended in useful ways. The junction need not be placed at
the end of the transmission line structure [41]. A symmetric structure with junctions and
transmission lines on both antenna leads adds flexibility to design and will avoid mismatch due to
the transition from a balanced mode on the antenna to an unbalanced mode on the microstrip [41].
A multi-step (Chebyshev) transformer can be used for excellent broadband performance [42]. The
transformation to low impedance junctions improves the dynamic range of the mixer [41]. A
tapered transmission line transformer can provide a broadband match [23].
4. SIS MIXER RESULTS
SIS mixers of some variety have been built. Niobium technology is the best current choice
for submillimeter SIS's. Both waveguide and quasioptical coupling structures are useful at these
high frequencies. Integrated tuning structures of increasing complexity are being used to improve
mixer performance.
4.1. SIS Junction Fabrication
Many of the best SIS mixers use lead alloy SIS junctions. However, niobium and niobium
alloy junctions have such great advantages that even the fabricators of lead alloy SIS's are
developing niobium technology. Niobium junctions have been fabricated and used in high
performance SIS mixers [15, 18, 19, 43, 13, 23]. Niobium-nitride alloy SIS's [17, 44] are
interesting as they will operate at higher temperatures, and possibly higher frequencies than pure .
niobium SIS's.
4 .2 . Radiation coupling structures
A sensitive SIS mixer requires an SIS junction which is well coupled to its input radiation.
The coupling structure chosen effects this in two important ways. 1) The coupling structure
determines Ys, the source admittance from which the radiation appears to come. 2) The coupling
structure determines the beam pattern of the SIS mixer. A good beam pattern is essential for high
efficiency coupling between the mixer and the radiation it is intended to detect. Table 1 shows the
reported results of a variety of SIS mixers and receivers along with the coupling structure that each
one uses.
Page 508
Third International Symposium on Space Terahertz Technology
Table 1. Summary of some of the best reported SIS receiver results. The bold
lines separate the results into frequency ranges. Within a frequency range, the
results are listed in order of increasing Trecdsb-
Rf
(GHz)
Trec
DSB
(K)
Tmtx
DSB
(K)
Gain
DSB
(dB)
SIS material, size,
configuration
Coupling
(waveguide or
antenna shape)
Reference
230
48
38
-2.6
0.25 sq. micron Nb
waveguide
[13]
230
48
Nb
waveguide
[11]
230
80
60
-2
PbBi alloy
waveguide
[7]
240
100
Nb
waveguide
[47]
228
114
PblnAu submicron
waveguide
[6]
230
116
PblnAu submicron
spiral
[8]
241
153
85
-7.5
PblnAu submicron
waveguide
[5]
228
163
PblnAu submicron
waveguide
[6]
230
200
-8
PblnAu submicron
waveguide
[10]
220
250
25
-9
SIN! (NOT an SIS)
waveguide
[48]
345
150
PblnAu submicron
waveguide
[21]
345
200
Nb
spiral
[42]
345
215
PblnAu submicron
spiral
[8]
342
214
3 PbBi in series
quasiopuc
[24]
312
275
-9.5
Nb/PblnAu
waveguide
[22]
492
171
114
-9
0.25 sq. micron Nb
waveguide
[20]
426
220
Nb
spiral
[42]
490
420
240
-10
Nb
twin slot
[23]
525
470
PblnAu submicron
spiral
[8]
492
500
Nb
spiral
[42]
761
1100
PblnAu submicron
spiral
[8]
4.3. Waveguide SIS mixers
Until 1985, all SIS mixers used waveguide coupling structures. The rf input side of the
mixer is almost always coupled to a scalar feedhorn [45, 46] which provides a very high quality
Gaussian beam pattern. The waveguide behind the mixer usually contains a movable back short
Which can help tune out the parasitic reactances of the SIS and its mounting structure. In some
designs, a second waveguide tuner can be included to increase the range of tuning which is
possible. In other designs, no mechanical tuning is necessary.
A full height rectangular waveguide has dimensions a little smaller than 1/2 x 1/4 Xo where
Ao is the free-space wavelength at frf. At 500 GHz, this waveguide is quite small, 300 x 150 um.
It is hard to say what the upper frequency limit is for a waveguide design. SIS's in waveguides
Third International Symposium on Space Terahertz Technology
Page 509
EPUNETtWAVEOUOE
TTUNgFORMER SECTKM
FOUTPUT
fzzzzz&L
^TTTzm
CORRUQATED-
FEEOHORNSECTON
MICROMETER DRIVE
(80TPI)-
TO SECOND MICROMETER ORIVE
Figure 3 [6]. The two-tuner waveguide design which has been implemented for
230, 345, and 492 GHz band mixers.
are the most sensitive available mixer at 500 GHz Walker, 1991 #134]. It seems clear that
waveguide designs will be attempted to at least 800 GHz.
The good features of waveguide designs are summarized. They provide excellent control
over the rf source impedance seen by the SIS, either through fixed or mechanically tuned ""
structures. They have excellent beam patterns capable of high efficiency coupling to radio
telescopes and other signal sources.
4.3. 1 . Ellison's Waveguide Mixer
Some of the lowest noise SIS receivers for the 200 to 500 GHz frequency range [13, 21,
20] have been built at the California Institute of Technology using a mixer designed by Ellison [6].
This design, shown in fig. 3, includes two movable waveguide shorts, one behind the junction
(back-short), and one in a waveguide tee in the E-plane wall of the waveguide near the junction 03-
plane tuner). The SIS is mounted across the E-plane of the waveguide. One side of the SIS is
shorted directly to the mixer block. The other side is coupled through an rf-choke structure to the
IF and dc-bias circuits. In later designs, a symmetric mount is designed so that the ground
connection is made at the end of an rf-choke.
This design has been verified and improved through the use of microwave scale modeling
[49]. These models showed that the rf impedance seen by the SIS can be tuned over more than
Page 510 Third International Symposium on Space Terahertz Technology
half the area of a Smith chart at almost all frequencies in the waveguide band. This modeling also
demonstrated the critical importance of the placement of the SIS in the waveguide. The SIS must
be centered in the waveguide to avoid extra reactance due to coupling to a higher order waveguide
mode.
In all of the mixers based on Ellison's design, a great amount of attention has been paid to
the IF matching circuit This circuit is designed for two criteria: 1) transform the 50 Q IF amplifier
to a higher impedance, 100 to 200 Q, at the SIS and 2) provide a low impedance at the SIS at
frequencies outside the desired IF band. The first criterion improves the coupled gain of the SIS
mixer since its IF output impedance is very high. The second criterion protects the SIS from IF
saturation [50, 51].
4.3.2. Round Waveguide Mixer
An early attempt at designs that would be appropriate for high frequency scaling is
Woody' s round waveguide mixer [4]. Mixers based on this design have been built for 115 GHz
[4], 230 GHz [5], and 345 GHz [22]. This design is relatively easy to scale to high frequencies.
The round waveguide is fabricated by drilling the proper sized hole in a block. The rf choke is
built into the mixer block. The SIS junction has no rf choking or any other frequency dependent
structures on it, so the chip can be made very small. While the sensitivity of the round waveguide
mixers is good, the greater complexity of the other designs discussed here is clearly justified by
their lower noise temperatures.
4.3.3. IRAM Mixer
SIS mixers are built at the Institut de Radio Astronomie Millimetrique (IRAM) for use on
European radio telescopes. The IRAM mixer block uses a reduced height rectangular waveguide
with the SIS junction mounted across its E-plane. The reduced height waveguide provides a lower
rf input impedance than full height waveguide, and eliminates resonant coupling to higher order
modes which trouble Ellison's design. There is a single mechanical tuner, a backshort in the
waveguide behind the SIS. The rf choke is fabricated integrally with the SIS and is symmetric
around the junction: it is the basis for the rf choke structure now used in Ellison designed mixers.
IRAM mixers have been built for 100 GHz [19] and 230 GHz [7, 47]. An SIS mixer is now being
Third International Symposium on Space Terahertz Technology
Page 511
Figure 4 [1 1]. The NRAO-401 170-260 GHz mixer. The center channel contains
microstrip circuitry including the SIS junctions with inductive tuning. The channel
on the left is a waveguide carrying the signal and LO into the mixer. The top and
bottom channels on the right are waveguides containing movable shorts for tuning.
Stripline to waveguide coupling structures are shown in solid black.
developed for 345 GHz [47]. The simplicity and success of the IRAM design make it a good
candidate for higher frequency operation.
4.4. Waveguide-Substrate Mixers
A number of mixer designs use a mechanically simple waveguide design to couple input
radiation to a microstrip or coplanar transmission line built on a dielectric substrate. Critical if
circuit components in these mixers are fabricated photolithographically. It is easier to fabricate
small components this way than to machine them into a waveguide block. Because of this, the
approach is very promising for higher frequency mixers. These designs have been fabricated up to
230 GHz with excellent results. They should perform equally well at higher frequencies in the
future. A few of these designs are presented below.
4.4.1. NRAO-401
The National Radio Astronomy Observatory makes a wide variety of receivers for radio
astronomy in the US. Their NRAO-401 SIS mixer is shown in fig. 4 [11]. The rf is coupled from
an input waveguide into stripline circuit through a broadband transition very similar to commercial
waveguide to coaxial transitions available at much lower frequencies. Two adjustable waveguide
tuning elements are also coupled to the circuit through similar probes to provide series and parallel
reactances to the SIS. This mixer's sensitivity in a 230 GHz DSB receiver is the best currendy
Page 512 Third International Symposium on Space Terahertz Technology
available. A similar mixer is part of a 40 K DSB receiver at 100 GHz [9]. Both of these mixers
can be tuned for SSB operation with 20 dB image rejection.
4.4.2. NRAO Tunerless Mixer
Another NRAO SIS mixer achieves 40-80 K DSB receiver temperatures over the entire
WR-10 waveguide band, 75-1 10 GHz, without any mechanical tuning [11]. As in the NRAO-
401, the waveguide to stripline transition couples radiation from the input waveguide to the
substrate circuit. The stripline radiation makes a second transition to coplanar waveguide before it
reaches the SIS and its integrated tuners.
4.4.3. Yale Waveguide-Microstrip Mixer
In this design, the input waveguide is coupled to a microstrip circuit with a 4-section
Chebyshev single ridge transformer [39]. Nearly full WR-10 band coverage is achieved with no
mechanical tuning. Circuits with microstrip inductive compensation of the SIS capacitance were
investigated, but the lowest noise temperatures for this mixer were achieved without inductive
tuning. Because of its simple mechanical design, this mixer is a good candidate for scaling to
higher frequencies.
4.5. Quasioptic SIS mixers
An interesting alternative to waveguide coupling is shown in fig. 5. The SIS is fabricated
integrally with a planar antenna. The SIS- antenna is placed on a hyperhemispherical lens. The
lens focuses the input radiation into the center of the SIS-antenna which earns this scheme the
name "quasioptic." A comprehensive introduction to the properties of antennas on dielectrics is
given by Rutledge et al. [52]. The superconducting films from which the SIS-antenna is made are
very good conductors, so the fabricated antennas will have low resistive losses despite the
submillimeter frequencies involved. This low-loss property may not hold for fjf > 2A/h, about
700 GHz for niobium technology.
The planar antenna designs have the following positive features. Many of these mixers can
be used over a few octaves of frf. The spiral and bowtie planar antennas are frequency independent
both in their beam partem and in their antenna impedance. The SIS-antennas are fabricated
photolithographically so no waveguide or feedhorn must be machined. Large substrates can be
used. If an SIS is to be placed across a waveguide, the SIS and its substrate must be much smaller
Third International Symposium on Space Terahertz Technology
Page 513
cm
H
SIS in
Bowtie
Quartz
Hyperhemisphere
Teflon Lens
Quartz
Substrate
Figure 5 [33]. The bowtie dipole antenna is formed as an extension of the leads of
the SIS diode. The radiation from this antenna is focussed into a fairly narrow
beam by a hyperhemispherical lens placed up against the substrate. Further
focussing is provided by a plastic lens
than an rf wavelength. An SIS-antenna, however, should actually be fabricated on a substrate
which is bigger than a wavelength.
4.5. 1 . Spiral and B owtie Antennas
The bowtie mixer in fig. 5 was improved on by using a spiral-shaped dipole antenna [8]
like the one shown in fig. 2a. The mixer with spiral antenna had noise temperatures about half
those of the bowtie mixer over the same large frequency range. This mixer has been used at the
Caltech Submillimeter Observatory for radio astronomy at 1 15, 230, 345 and 492 GHz. Its beam
couples to the telescope about as well as the beam from a waveguide receiver, resolving a difficulty
with the bowtie design.
The spiral antenna has a much cleaner beam pattern than the bowtie, with sidelobe and
pedestal structure at the -20 dB level. The cleaner beam is primarily due to the better radiation
pattern of the spiral, but it is also helped by the use of a mirror less than 1 mm behind the SIS-
antenna. The lower noise of the spiral SIS is attributable to its better beam pattern, as is its
excellent coupling efficiency to the CSO radio telescope.
The submillimeter spiral SIS mixer has very recently been improved remarkably at the
expense of its multi-octave bandwidth [42]. The design is similar to that shown in fig. 2b, but the
Page 514
Third International Symposium on Space Terahertz Technology
SIS
junction
I
A/4
radial
stubs
^^"" l ' ' ' ' "i^TTTT
A'^^ li muni ' - '™
4000A SiO
Insulating
Film
Tapered
Microstrip
Line
RF
choke
Ground
Plane
Slot
Antennas
Figure 6. The planar antenna-SIS mixer of Zmuidzinas and LeDuc [23]. The entire
structure is built on a niobium ground plane which is etched away to create the two
slot antennas.
transformer is a more complicated two section Chebyshev design. The result is 75% or better
coupling between the SIS and the spiral antenna in a one-octave rf band centered at 350 GHz.
Receiver results using this mixer are included in Table 1. The fall-off in performance at 492 GHz
is due to the low center frequency of this particular mixer. It should be a simple matter to adjust
the design so that Trecdsb ^ 250 K is achieved at 492 GHz.
4.5.2. Twin-slot antenna
Fig. 6 shows a different planar antenna design used in the mixer developed by Zmuidzinas
and LeDuc [23]. This receiver is reported in further detail elsewhere in this conference publication.
This design couples a single SIS junction to two planar slot dipole antennas as shown in fig. 6.
Unlike the bowtie and spiral, the twin-slot is a resonant antenna, but it is reasonable to expect the rf
bandwidth of this mixer to be similar to that of waveguide designs: about one-half octave.
The slot antennas have a low antenna impedance. They are coupled to the SIS through
tapered transmission lines which reduce the impedance even further. Finally, the radiation
impedance is reduced even further since the two slots appear in parallel across the SIS. The source
admittance of radiation presented to the SIS at the design frequency is Y$ = (4 Q)" 1 . As a result,
an SIS with a large junction area, 2.3 (|im) 2 , can be used. The result is Trecdsb = 420 K at
Third International Symposium on Space Terahertz Technology
Page 515
60-
50 -
■
Quasioptic
■
o
Waveguide
(/)
o
c
o
o
■
u
sz
40-
CL
o
o
■ "
CD
10
O
30-
o
o
1
>
0)
?0 -
B
■
o
CD
cr
10 -
-
o
o
o
o
— ■ 1 "-
o
1 ' 1
200 300 400 500 600
Frequency (GHz)
700
800
Figure 7. The noise of the receivers summarized in Table 1. Plotted is
TRECDSB/(hf/2ke). This normalizes the receiver noise to its quantum limit value.
492 GHz (the conference publication on this receiver should be reviewed for the most up to date
information. It is reasonable to expect that the noise of this kind of receiver will drop quickly as
more is learned about its design.
4.6. Overall performance summary
A summary of the best reported SIS receiver results is shown in fig. 7 and in Table 1. The
most common results reported are Trecdsb, so those are tabulated. DSB values of gain or mixer
temperature are included in the table when they are quoted in the literature. Generally, SSB values
can be estimated by doubling Trecdsb. doubling Tmixdsb. and subtracting 3 dB from the DSB
Gain. The receiver noises in fig. 7 are normalized to the quantum limit of Nrec = 1 for an ideal
photodiode mixer. The noise values achieved are many times this limiting value. However, these
noise values represent a very great improvement over values available just a few years ago. It is
reasonable to expect that they will continue to fall in the future.
Page 516 Third International Symposium on Space Terahertz Technology
4 . 7. Unusual Mixers for Higher Frequencies
This section is concluded with the mention of some unusual modes of superconducting
mixer operation. None of these methods have been developed to the point that they are useful on a
radio telescope. However, each method addresses problems with submillimeter SIS mixers in
ways which may be useful in higher frequency superconducting mixers.
4.7.1. Harmonic mixin g
A mixer can be designed so that it is sensitive to input signals at fs = n fLO ± fiF where n is
an integer. If n > 2, the mixer is said to be subharmonically pumped. Belitsky et al. investigated
the gain of their spiral-transformer mixer with subharmonic pumps [53]. The mixer gain fell off
slowly as n increased from 1 to 3. The advantages of subharmonic pumping include 1) it is easier
to get powerful oscillators at lower LO frequencies and 2) the large separation between fLO and fs
makes it easier to couple both of these to the SIS with high efficiency.
4.7.2. SIN junction mixer
If one of the superconducting sides of an SIS is replaced by a non-superconducting metal
film, the resulting diode is a Superconductor-Insulator-Normal metal junction, or SIN. The SIN
does not have as sharp a non-linearity in its IV as the SIS. But it also doesn't have any Josephson
currents, which can be a problem in high frequency SIS mixers.
An SIN mixer at 230 GHz was built and tested by Blundell and Gundlach [48]. Compared
to a similar SIS mixer, its gain was about 7 dB lower, but its mixer noise was quite low.
Considering the very low noise IF amplifiers now available (Tip ~ 3 K at 1 .5 GHz), this is a very
promising result. While it does not give any advantage over an SIS at 230 GHz, its lack of
Josephson currents might make the SIN mixer useful at submillimeter wavelengths. Theoretical
work on the SIN mixer operated both as a fundamental mixer [36, 54] and as a subharmonically
pumped mixer [55] suggest it should be capable of TrecdsB < 100 K up to 660 GHz.
4.7.3. Josephson effect mixing
While the Josephson effect is known to produce excess noise in submillimeter mixers, it is
also capable of enhancing the gain of these mixers [33, 56]. While the current results for
Josephson mixing show the noise increasing faster than the gain, the mixers tested are not
Third International Symposium on Space Terahertz Technology Page 517
designed to take advantage of the Josephson currents. Much work remains to be done on this type
of mixer before it is known if it will be useful.
5 . SUMMARY AND CONCLUSIONS
SIS mixers are used extensively for 100 to 500 GHz radio astronomical receivers. Many
useful techniques have been developed for using SIS's at submillimeter wavelengths. The
waveguide-feedhorn and the planar-antenna-lens (quasioptical) techniques provide excellent
radiation beam patterns for SIS mixers. Integrated tuning elements including inductive
components and transmission-line-transformers allow impedance matching of the SIS to both of
these radiation coupling structures. SIS mixers using these techniques have been reviewed here.
6. ACKNOWLEDGEMENTS
The author thanks the many researchers who sent letters and reprints which helped him to
write this review. The development of this article was supported by National Science Foundation
Grant ECS-8857868.
7. REFERENCES
[1] P. L. Richards, T. M. Shen, R. E. Harris and F. L. Lloyd, "Quasiparticle heterodyne
mixing in SIS tunnel junctions," Appl. Phys. Lett., vol. 34, 1 March, 1979, pp. 345-347.
[2] G. J. Dolan, T. G. Phillips and D. P. Woody, "Low-noise 1 15-GHz mixing in
superconducting oxide-barrier tunnel junctions," Appl. Phys. Lett., vol. 34, 1 March,
1979, pp. 347-349.
[3] J. R. Tucker, "Quantum limited detection in tunnel junction mixers," IEEE J. Quantum
Electron., vol. QE-15, November, 1979, pp. 1234-1258.
[4] D. P. Woody, R. E. Miller and M. J. Wengler, "85-1 15 GHz receivers for radio
astronomy," IEEE Trans. Microwaves Theory and Techniques, vol. MTT-33, February,
1985, pp. 90-95.
[5] E. C. Sutton, "A superconducting tunnel junction receiver for 230 GHz," IEEE Trans.
Microwaves Theory and Techniques, vol. MTT-31, July, 1983, pp. 589-592.
[6] B. N. Ellison and R. E. Miller, "A low noise 230 GHz receiver," Int. J. ofIR and MM
Waves, vol. 8, 1987, pp. 608-625.
Page 518 Third International Symposium on Space Terahertz Technology
[7] R. Blundell, M. Carter and K. H. Gundlach, "A low noise SIS receiver covering the
frequency range 215-250 GHz," Intl. J. ofIR and MM Waves, vol. 9, April, 1988, pp.
361-370.
[8] T. H. Buttgenbach, R. E. Miller, M. J. Wengler, D. M. Watson and T. G. Phillips, "A
broad-band low-noise SIS receiver for submillimeter astronomy," IEEE Transactions on
Microwave Theory and Techniques, vol. 36, December, 1988, pp. 1720-1726.
[9] S.-K. Pan, A. R. Kerr, M. J. Feldman, A. W. Kleinsasser, J. W. Stasiak, R. L.
Sandstrom and W. J. Gallagher, "An 85-1 16 GHz SIS receiver using inductively shunted
edge-junctions," IEEE Trans. Microwaves Theory and Techniques, vol. 37, March, 1989,
pp. 580-592.
[10] D. P. Woody, C. J. Giovanine and R. E. Miller, "Dual channel 1 15 and 230 GHz SIS
receivers in operation at the Owens Valley Radio Observatory," IEEE Trans. Magn., vol.
25, March, 1989, pp. 1366-1370.
[11] A. R. Kerr and S.-K. Pan, "Some recent developments in the design of SIS mixers," Int'l
J. ofIR and MM Waves, vol. 11, October, 1990, pp. 1169-1187.
[12] H. Ogawa, A. Mizuno, H. Hoko, H. Ishikawa and Y. Fukui, "A 1 10 GHz SIS receiver
for radio astronomy," Intl. J. ofIR and MM Waves, vol. 11, June, 1990, pp. 717-726.
[13] J. W. Kooi, M. Chan, T. G. Phillips, B. Bumble and H. G. Leduc, "A low noise 230
GHz heterodyne receiver employing .25," vol. 1991,
[14] H. G. LeDuc, S. K. Khanna, J. A. Stern and S. Thakoor, "All refractory NbN/MgO/NbN
tunnel-junctions," IEEE Trans. Magn., vol. 23, March, 1987, pp. 863-865.
[15] A. W. Lichtenberger, C. P. McClay, R. J. Mattauch, M. J. Feldman, S.-K. Pan and A. R.
Kerr, "Fabrication of Nb/Al-A^OyNb junctions with extremely low leakage current,"
IEEE Trans. Magn., vol. 25, March, 1989, pp. 1247-1250.
[16] W. C. Danchi, E. C. Sutton, P. A. Jaminet and R. H. Ono, "Nb edge junction process for
submillimeter wave SIS mixers," IEEE Trans. Magn., vol. 25, March, 1989, pp. 1064-
1067.
[17] J. A. Stern, B. D. Hunt, H. G. LeDuc, A. Judas, W. R. McGrath, S. R. Cypher and S.
K. Khanna, "NbN/MgO/NbN SIS tunnel junctions for submm wave mixers," IEEE Trans.
Magn., vol. 25, March, 1989, pp. 1054-1059.
[18] A. H. Worsham, D. E. Prober, J. H. Kang, J. X. Przybysz and M. J. Rooks, "High
quality sub-micron Nb Trilayer tunnel junctions for a 100 GHz SIS receiver," IEEE Trans.
Magn., vol. 27, March, 1991, pp. 3165-3167.
[19] T. Lehnert, C. Grassl, K. H. Gundlach and J. Blondel, "Nb-Aloxide-Nb junctions for 3-
mm SIS receivers," ISEC, vol. 1991,
[20] C. K. Walker, J. W. Kooi, M. Chan, H. G. LeDuc, J. E. Carlstrom and T. G. Phillips,
"A 492 GHz SIS waveguide receiver for submillimeter astronomy," Intl. J. ofIR and MM
Waves, vol. submitted, 1991,
Third International Symposium on Space Terahertz Technology Page 519
[21] B. N. Ellison, P. L. Schaffer, W. Schaal, D. Vail and R. E. Miller, "A 345 GHz receiver
for radio astronomy," Int. J. ofIR and MM Waves, vol. 10, 1989, pp. 937-947.
[22] E. C. Sutton, W. C. Danchi, P. A. Jaminet and R. H. Ono, "A superconducting tunnel
junction receiver for 345 GHz," Intl. J. ofIR and MM Waves, vol. 11, February, 1990,
pp. 133-149.
[23] J. Zmuidzinas and H. G. LeDuc, "Quasi-optical slot antenna SIS mixers," IEEE Trans.
Microwaves Theory and Techniques, vol. Submitted, 1991,
[24] H. Rothermel, D. Billon-Pierron and K. H. Gundlach, "An open structure SIS mixer for
350 GHz," Digest ofthel6th Int. Conf. on Infrared and Millimeter Waves, Lausanne, vol.
August, 1991,
[25] M. J. Wengler, "Submillimeter wave detection with superconducting tunnel diodes," Proc.
IEEE, vol. Accepted, 1992,
[26] J. R. Tucker and M. J. Feldman, "Quantum detection at millimeter wavelengths," Rev.
Mod. Phys., vol. 57, 1985, pp. 1055-1113.
[27] T. G. Phillips, "Submillimeter and far-infrared detectors," Astro. Lett, and
Communications, vol. 26, 1988, pp. 293-304.
[28] P. L. Richards and Q. Hu, "Superconducting components for infrared and millimeter-wave
receivers," Proc. IEEE, vol. 77, August, 1989, pp. 1233-1246.
[29] R. Blundell and D. Winkler, "The superconductor insulator superconductor mixer receiver
- a review," in NATO Applied Research Workshop on Superconducting Electronics and
2nd Workshop on Josephson Devices. Capri, Italy: 1990.
[30] D. Winkler, Z. Ivanov and T. Claeson, "Superconducting detectors for mm and sub-mm
waves," in 2nd Nordic Symposium on Superconductivity. R0ros, Norway: World
Scientific, 1991.
[31] J. W. Archer, "Low-noise heterodyne receivers for near-millimeter wave radio
astronomy," Proc. IEEE, vol. 73, 1985, pp. 109-130.
[32] J. M. Payne, "Millimeter and Submillimeter Wavelength Radio Astronomy," Proc. IEEE,
vol. 77, 1989, pp. 993-1017.
[33] M. J. Wengler, D. P. Woody, R. E. Miller and T. G. Phillips, "A low noise receiver for
millimeter and submillimeter wavelengths," Intl. J. ofIR and MM Waves, vol. 6, August,
1985, pp. 697-706.
[34] S. Withington and E. L. Kollberg, "Spectral-domain analysis of harmonic effects in
superconducting quasiparticle mixers," IEEE Trans. Microwaves Theory and Techniques,
vol. 37, January, 1989, pp. 231-238.
[35] D.-G. Crete, W. R. McGrath, P. L. Richards and F. L. Lloyd, "Performance of arrays of
SIS junctions in heterodyne mixers," IEEE Trans. Microwaves Theory and Techniques,
vol. MTT-35, April, 1987, pp. 435-440.
[36] M. J. Wengler and D. P. Woody, "Quantum noise in heterodyne detection," IEEE J. of
Quantum Electron., vol. QE-23, May, 1987, pp. 613-622.
Page 520 Third International Symposium on Space Terahertz Technology
[37] Q. Hu, C. A. Mears, P. L. Richards and F. L. Lloyd, "Measurement of integrated tuning
elements for SIS mixers with a Fourier transform spectrometer," Intl. J. ofIR and MM
Waves, vol. 9, April, 1988, pp. 303-320.
[38] A. R. Kerr, S.-K. Pan and M. J. Feldman, "Integrated tuning elements for SIS mixers,"
Intl. J. ofIR and MM Waves, vol. 9, February, 1988, pp. 203-212.
[39] D. Winkler, N. G. Ugras, A. H. Worsham, D. E. Prober, N. R. Erickson and P. F.
Goldsmith, "A full-band waveguide SIS receiver with integrated tuning for 75-110 GHz,"
IEEE Trans. Magn., vol. 27, March, 1991, pp. 2634-2637.
[40] V. Y. Belitsky, M. A. Tarasov, S. A. Kovtonjuk, L. V. Filippenko and O. V.
Kaplunenko, "Low noise completely quasioptical SIS receiver for radioastronomy at 1 15
GHz," 21st European Microwave Conference, Stuttgart, vol. September, 1991,
[41] J. A. Carpenter, A. D. Smith, E. R. Arambula, L. P. S. Lee , T. Nelson and L. Yujiri,
"100 GHz SIS mixer with improved rf matching," IEEE Trans. Magn., vol. 1991,
[42] T. H. Buttgenbach, H. G. LeDuc, P. D. Maker and T. G. Phillips, "A Fixed Tuned
Broadband Matching Structure For Submillimeter SIS Receivers," IEEE Trans. Appl.
Super conduct.., vol. submitted, 1992,
[43] A. B. Ermakov, V. P. Koshelets, I. L. Serpuchenko, L. F. Filippenko, S. V. Shitov and
A. N. Vystavkin, "SNAP structures with Nb-AlO-Nb junctions for mm wave receivers,"
IEEE Trans. Magn., vol. 25, March, 1989, pp. 1060-1064.
[44] W. R. McGrath, J. A. Stern, H. H. S. Javadi, S. R. Cypher, B. D. Hunt and H. G.
LeDuc, "Performance of NbN Superconductive tunnel junctions as SIS mixers at 205
GHz,"/£££ Trans. Magn., vol. 27, March, 1991, pp. 2650-2653.
[45] B. M. Thomas, "Design of corrugated conical horns," IEEE Trans. Antennas Propagat.,
vol. AP-26, March, 1978, pp. 367-372.
[46] P. F. Goldsmith, "Quasi-optical techniques at millimeter and submillimeter wavelengths,"
in Infrared and Millimeter Waves, K. J. Button, Editor. 1982, Academic Press: New York,
pp. 277-343.
[47] M. C. Carter, S. Navarro, A. Karpov, D. Billon-Pierron, T. Lehnert and K. H. Gundlach,
"Receiver development with Nb/Al-Oxide/Nb SIS mixers in the frequency ranges of (201-
270) and (320-370) GHz," Intl. J. ofIR and MM Waves, vol. in press, 1991,
[48] R. Blundell and K. H. Gundlach, "A quasiparticle SIN mixer for the 230 GHz frequency
range," Intl. J. ofIR and MM Waves, vol. 8, December, 1987, pp. 1573-1579.
[49] T. H. Buttgenbach, T. D. Groesbeck and B. N. Ellison, "A scale mixer model for SIS
waveguide receivers," Intl. J. ofIR and MM Waves, vol. 11, January, 1990, pp. 1-20.
[50] M. J. Feldman and S. Rudner, "Mixing with SIS arrays," in Reviews of Infrared and
Millimeter Waves, K. J. Button, Editor. 1983, Plenum: pp. 47-75.
[51] A. D. Smith and P. L. Richards, "Analytic solutions to superconductor insulator
superconductor quantum mixer theory," J. Appl. Phys., vol. 53, May, 1982, pp. 3806-
3812.
Third International Symposium on Space Terahertz Technology Page 521
[52] D. B. Rutledge, D. P. Neikirk and D. P. Kasilingam, "Integrated-circuit antennas," in
Infrared and Millimeter Waves, K. J. Button, Editor. 1983, Academic Press: New York,
pp. 1-90.
[53] V. Y. Belitsky, I. L. Serpuchenko, M. A. Tarasov and A. N. Vystavkin,
"Subharmonically pumped SIS mixer," in 19th European microwave conference. London,
UK: 1989.
[54] C. E. Tong, L. M. Chernin and R. Blundell, "Harmonic mixing in a superconducting
tunnel junction,"/. Appl. Phys., vol. 68, 15 October, 1990, pp. 4192-4198.
[55] L. M. Chernin and R. Blundell, "Harmonic mixing in a superconductor-insulator-normal
metal tunnel junction receiver," /. Appl. Phys., vol. 69, 15 February, 1991, pp. 2682-
2684.
[56] M. J. Wengler, N. Dubash, G. Pance and R. E. Miller, "Josephson effect gain and noise
in SIS mixers," IEEETrans. Microwaves Theory and Techniques, vol. 40, in press, 1992,
Page 522 Third International Symposium on Space Terahertz Technology
. trzrft Evaluation of integrated tuning elements with SIS devices.
N93-27770
\p
M.M.T.M. Dierichs 1 , C.E. Honingh 2 , R.A. Panhuyzen 2 , BJ. Feenstra 1 , A. Skalare 2 ^, JJ.
Wijnbergen 2 , H. v.d. Stadt 2 , Th. de Graauw 2 .
1: Dept. of Applied Physics and Materials Science Centre, University of Groningen,
Nijenborgh 4, 9747 AG Groningen, The Netherlands.
2: Space Research Organization of the Netherlands, Groningen,
Landleven 12, 9747 AD Groningen, The Netherlands.
3: Dept, of Applied Electron Physics, Chalmers University of Technology,
GSteborg, Sweden.
Abstract.
The resonance of integrated tuning stubs in combination with SIS detectors is measured
and modeled. The predicted resonances are compared with measurements of stubs
integrated with Nb/Al 2 3 /Nb junctions in a log-periodic antenna using a Michelson
interferometer. Different stub lengths were made on different substrates (on 200 nm
thick quartz and on a 7 jrni thick silicon membrane) and the results show a fairly good
Third International Symposium on Space Terahertz Technology Page 523
^agreement with the model calculations. Quartz substrates showed resonances up to 580
GHz, silicon membrane stub resonances reach as high as 480 GHz. An observed
resonance at 560 GHz is probably a substrate effect from the membrane. The gap
frequency for all the samples is 650 GHz and no resonances are detected above this
frequency. Up to the maximum detected frequency dispersion is found to be negligible.
I Introduction.
SIS mixers with Nb/Al 2 3 /Nb junctions are very sensitive submm detectors. Recent
progress in SIS mixer development is due to the ability to manufacture smaller junctions
down to sub-micron dimensions 1,2 . Instead of continuing to put more effort into the
fabrication of smaller junctions and thus reducing the junction capacitance, it is also
possible to implement integrated tuning elements, which are fairly easy to fabricate and
result in a high sensitivity and broad bandwith. It has been shown that junctions with
integrated tuning used in submm-wave mixers give good results 3 ' 4 .
The first published stub measurements used the self-pumped steps in the I-V
characteristic to measure the resonance of the stub 5 . A more accurate and complete
evaluation can be performed with a wide-band Michelson interferometer as first shown
by Hu et. al. 6 .
In this paper we first describe our design criteria for niobium stubs in combination with
niobium junctions. Next, we describe how they can be analyzed on a log-periodic antenna
with two 1 Mm 2 junctions in series. Each junction has its own stub. Stubs for 100 GHz
pa e 524 Third International Symposium on Space Terahertz Technology
and 350 GHz have been designed. The first type is expected to have multiple resonances
from which the dispersion in niobium can be calculated. These antennas are made on
200 /xm thick quartz substrates and on 7 /im silicon membranes. The results are used to
separate stub and antenna resonances and to estimate the dispersion.
The organization of this paper is as follows: the theoretical background will be
introduced in Sec.II, the fabrication results are presented in Sec.III, the experimental
details are described in Sec.IV, the comparison between theory and experiment is
discussed in Sec.V, and the conclusion will be drawn in Sec. VI.
II Model calculation.
To tune out the geometric capacitance an inductive tuning element was used. An
example of the devices studied is shown in Fig.l. Two junctions in series, placed in the
center of a log-periodic antenna, were used. To each junction a stripline type inductor
is attached. The total arrangement can be modelled with the circuit shown in Fig.2. For
completeness the connecting strip between the two junctions is included as an inductor
Lieads- I n practice this inductance is negligible in evaluating the frequency response.
Using integrated tuning, the junction impedance can not simply be described as a pure
resistor with a parallel capacitor. Instead, it must be described as a capacitor in parallel
with a complex admittance with a conductive part (G Q ) and a susceptive part (B Q ). Since
the experiment works in the small signal limit and the Josephson effect is suppressed by
a magnetic field, the junction admittance can be described as follows 7,8 :
Third International Symposium on Space Terahertz Technology Page 525
G °- ■&5- lz - ir '* i T ) "'-'"i-*?" <l>
These equations show that the resonant frequency depends on the bias voltage and the
photon-energy. In these equations is >iS>/e the energy of the photon step, V is the bias
voltage, and
where 1,^ is the Kramers-Kronig transform, which can be calculated from the dc-IV
curve.
The inductance per unit length of the stub can be calculated as follows 9 :
L <- &■"*♦ — hr * — h~ ] <«
tanh(-ri) tanh(-r^)
where w is the width of the stub, k is the fringing factor 10 , t x , tj and t d are the thicknesses
of the ground plane, the stub, and the dielectric layer respectively, and X. is the
penetration depth of the niobium layers. The capacitance per unit length of the stub is
given by:
c = jt-» •« 'Jul '^'
Knowing the capacitance and the inductance of the stub,^the impedance Z„ and the
phase velocity v follow from the definitions:
Page 526 Third International Symposium on Space Terahertz Technology
Zo =
v = X (7)
The impedance of a transmission line with small loss and dispersion follows from:
Z= 2 + 2 (8)
sin 2 (-^) j.tan(-^l)
Where a is the loss per unit length and 1 is the length of the stub. The RF coupling
coefficient C^p defined as the fraction of the available power dissipated in the junction
is given by:
Crf= 1 '
**-£
Ya+Yj
(9)
where Y A = 1/R A is the simplified admittance of the antenna (1/120 n), and Yj is the
admittance of the right hand side of Fig.2.
If both areas and stub lengths are equal then the resonant frequency can be
approximated from:
<•>• C: + i_ + i_ = o
J <o- x s <■»• X q
(10)
In our equations we do not take into account the behaviour of the log-periodic antenna,
this is rather complex and not known in all detail. Therefore it will be very difficult to
identify the loss from the observed resonances, because it could be loss in niobium, bad
coupling to the antenna or a combination of both.
Third International Symposium on Space Terahertz Technology
Page 527
*ig.l. Photo of a log-periodic antenna with two junctions in series. Each junction has
its own stub.
I ^junctic
5
stub ,-J-, * quantum L K quantum
L leads
1 stuDJU * quantum L ^ quantum
Fig.2 Electrical equivalent of two junctions in series with integrated tuning elements.
page 528 Third International Symposium on Space Terahertz Technology
III Fabrication of devices.
The detector is positioned in the center of a broad-banded log-periodic antenna. We
used 2 junctions in series of each 1 /im 2 and a current density of 12000 A/cm 2 . On top
of the junctions the wiring layer was defined with a stub for each junction (Fig.l.). The
dielectric layer between the ground plane (antenna) and the stub was 250 nm thick
sputtered SiO z . The junctions have been fabricated with the Selective Niobium Over-Etch
Process (SNOEP) 11 .
Antennas have been fabricated on 200 /xm thick quartz substrates and on 7 Mm thick
silicon membranes (Fig.3.). The membranes have been etched in ethylenediamine-
pyrocatechol-water (EPW) 12 . The junctions on the membranes were fabricated after the
etching of the membranes. With the obtained thickness, the membrane is transparent
which simplifies the alignment of the antenna on the membrane.
Two different stub lengths were fabricated for different purposes. Firstly, short stubs
(around 120 /ira) were designed to resonate at 350 GHz. A single resonance simplifies
the comparison with the model and it can easily be implemented in the waveguide mixer
chip design. Secondly, long stubs (around 500 /im) were designed to have a fundamental
resonance around 100 GHz and multiple resonances at higher frequencies. From the
frequencies of the resonances in principle the dispersion and loss in niobium can be
estimated.
Third International Symposium on Space Terahertz Technology
Page 529
Fig.3. Photo of the antenna fabricated on a 7 nm thick silicon membrane.
IV Measurement set-up
For measuring the response of the detector, we used a Michelson interferometer with
a Hg arc lamp as source 13 (Fig.4.). The operating frequency range was determined by a
50 urn thick kapton film beam-splitter. The mechanical traveling distance was 50 mm
resulting in a resolution of 4 GHz. Both single sided and double sided interferograms
were measured. An example of the resulting spectra with multiple resonances is plotted
in Fig.5.
The antenna was mounted in a liquid helium dewar with dc-bias connections. Since the
G-7
Page 530
Third International Symposium on Space Terahertz Technology
Michelson
interferometer
S = source (Hg-arc)
Ml = mirror 1
(moving)
M2 = mirror 2
(static)
B = beam splitter
4.2 K Dewar
"ig.4. Schematic of the Michelson interferometer.
Michelson was not under vacuum during the measurement, the water absorption lines
at 380 GHz, 448 GHz, 557 GHz and 752 GHz were visible when sufficient resolution was
used. All antennas fabricated on a quartz substrate were glued to a quartz hyper-
hemispherical lens. The lens optimizes the optical coupling and results in a better
sensitivity. We have also performed measurements with the log-periodic antenna on a
thin membrane. In that case no lens was used, resulting in a much lower signal. By using
longer integration times we could improve the signal to noise ratio to an acceptable
level.
Third International Symposium on Space Terahertz Technology
Page 531
7
6
-
~ 5
a
~ 4
\ I
>%
! »
*» .
a
2
|
1
\ \ A
(
> 100 200 300 400 S00 600 700
800 900 1000
Frequency (GHz)
7 ig.5 Spec
trogram of a device
with multiple resonances.
V Results
A: Short stubs on 200 /im quartz substrates.
Antennas with stub lenghts around 120 pm were investigated to determine the specific
capacitance of the junction and the penetration depth of the niobium layers. Resonant
frequencies were measured from two different batches. The resonant frequency is
depending on the bias voltage due to the behaviour of the quantum impedance (eq.l and
eq.2). Calculated and measured results are shown in Fig.6. Best agreement between
Page 532
Third International Symposium on Space Terahertz Technology
theory and experiment was obtained with the assumption of a specific capacitance of 55
fF/Mm 2 and a penetration depth of 100 nm. These values were further used in
calculations of the long stubs. Differences between calculations and measurements are
due to the noise in the spectrogram which complicates the determination of the resonant
frequencies. No multiple resonance is observed.
~ theory theory ♦ measure • meainre
6A1 9A2 6A1 9A2
AdSJ
400
-
"n
3
380
/ * \
its \
/ ' * \
/ / \ \
/ ' * \
/ ' * \.
>
360
I
Li.
340
320
j^^O
300
1 1 1 1 1
0.00
0.50 1.00 1.50 2.00 2.50 3.00
Bias Voltage (mV)
Fie.6
Calculi
ited
and measured resonant frequencies as a function of bias vc
tltage
for two different junction batches.
B; Long stubs on 200 \xm quarz substrates.
Next, the resonances of a 527 ^m long stub on a quartz substrate were measured. The
results of the measurements at different bias voltages are plotted in Fig.7 and are
Third International Symposium on Space Terahertz Technology
Page 533
compared with model calculations. At the first two resonances the measurements agree
fairly well with the model both below as well as above the gap voltage. The two higher
resonances have a larger frequency shift close to the gap than the model predicts. More
measurements with different lengths are planned for a more detailed evaluation. Stub
resonances up to 580 GHz are observed, while no antenna resonances are visible. We
do not see any dispersion in the resonances of the stubs (Fig.5).
calc
calc
calc
calc
a meat • meat + meat ♦ meaf
500
410
«♦♦♦♦♦♦♦♦♦♦
^♦♦W'W*'*"
Fig.7 Calculated and measured resonant frequencies for an antenna with a 527 nm
long stub on a quartz substrate as a function of bias voltage.
Page 534
Third International Symposium on Space Terahertz Technology
C: Long stubs on 7 pm silicon membranes.
The resonances of a 527 j*m long stub on a 7 nm thick silicon membrane were measured.
A double sided interferogram was used to decrease the noise in the spectrogram. The
resolution (8 GHz) is lower because of the decreased scan length. The effect of the
quantum susceptance is not clearly observed because of the loss in resolution and signal.
We measured stub resonances at 110, 200 and 325 GHz. Incidentally resonances at 165,
310 and 540 GHz occur which are probably substrate resonances. The highest stub
resonance appears to be at 450 GHz which is lower than the measurements on quartz.
The lowest three resonances are compared with the model calculations in Fig.8. It is not
clear why above 330 GHz no well defined resonances occur except for the substrate
resonance at 540 GHz.
M
x
>>
O
a
e
3
9
o
u
a.
400
330
260
190
120
cal —
• me
+
me
cal ' cal ^ me
+ ¥-++*^
++
• +*"$
•
—
A J^
A*
1
i
+KMA
i i i i
SO
c
>
1
2 3 4 5
Biaa voltage (mV)
6
7
1
I
Fig.8 Calculated and measured resonance frequencies for a 527 urn long stub
on a silicon membrane as a function of bias voltage.
Third International Symposium on Space Terahertz Technology Pa i« 535
VI Conclusions
The theoretical model and the experimental results for the short stubs lead to a
penetration depth of 100 nm and a specific capacitance of 55 fF/nm 2 . This is
independent of the measured batch. For long stubs the model predicts a different
behaviour at higher resonances than is measured. The measured frequency shift at the
higher resonances is larger than the model predicts. Both below and above the gap
voltage the resonances agree fairly well with the model. Resonances up to 580 GHz are
observed. For antennas fabricated on 7 nm thick silicon membranes resonances up to 480
GHz are observed. Also possible substrate resonances are measured. No resonances
above 600 GHz are observed, which is close to the gap frequency of niobium (650 GHz).
Acknowledgements
We thank T.M. Klapwijk for his stimulating discussions, H.G. Golstein and G. de Groot
for their help with the Michelson interferometer, MJ. de Boer for etching the
membranes, G. de Lange for his support with the model calculation and H.H.A.
Schaeffer for the mechanical support. This work is supported by the Stichting Technische
Wetenschappen (STW) and the Stichting voor Fundamenteel Onderzoek der Materie
(FOM), which are part of the Nederlandse Organisatie voor Wetenschappelijk
Page 536 Third International Symposium on Space Terahertz Technology
Onderzoek (NWO). We also acknowledge the financial support of the European Space
Research (ESA) through contract 7898/88/NL/PB(SC).
References
1. W.R. McGrath, H.H.S. Javadi, S.R. Cypher, B. Bumble, B.D. Hunt, and H.G. LeDuc,
Proceedings of Second International Symposium on Space TeraHertz Technology, 423
(1991)
2. J.W. Kooi, M. Chan, T.G. Phillips, B. Bumble, and H.G. LeDuc, Proceedings of
Second International Symposium on Space TeraHertz Technology, 459 (1991)
3. A.W. Lichtenberger, D.M. Lea, A.C. Hicks, J.D. Prince, R. Densing, D. Peterson, and
B.S. Deaver, Proceedings of Second International Symposium on Space TeraHertz
Technology, 439 (1991)
4: J. Zmuidzinas, and H.G. LeDuc, Proceedings of Second International Symposium on
Space TeraHertz Technology, 481 (1991)
5. A. V. Raisanen, W.R. McGrath, P.L. Richards, and F.L. Lloyd, IEEE Trans Microwave
Theory Tech. 33, 1495 (1985)
6. Q. Hu, C.A. Mears, P.L, Richards, and F.L. Lloyd, Int. J. of Infrared and MM Waves,
9, 303 (1988)
7. J.R. Tucker, and MJ. Feldman, Rev. Mod. Phys. 57, 1055 (1985)
8. Q. Hu, C.A. Mears, PL. Richards, and F.L. Lloyd, Phys. Rev. B, 42,10250 (1990)
9. J.C. Swihart, J. Appl. Phys. 32, 461 (1961)
Third International Symposium on Space Terahertz Technology Page 537
10. W.H. Chang, J. Appl. Phys. 50, 8129 (1979)
1 1. M.M.T.M. Dierichs, R.A. Panhuyzen, C.E. Honingh, M J. de Boer, and T.M. Klapwijk
(unpublished results)
12. K.E. Petersen, Proc. of the IEEE 70, 420 (1982)
13. P.L. Richards, in Spectroscopic Technique for Far Infrared Submillimeter and MiUmeter
Waves, (North-Holland, Amsterdam, 1967)
Page 538 Third International Symposium on Space Terahertz Technology
Syf-33
/£o6&J SOURCE CONDUCTANCE SCALING FOR HIGH FREQUENCY
« SUPERCONDUCTING QUASIPARTICLE RECEIVERS
N93-27771
Qing Ke 1 and M. J. Feldman 2
Physics Department 1 and Department of Electrical Engineering 2
University of Rochester, Rochester, NY 14627
/
ABSTRACT
It has been suggested that the optimum source conductance G s for the supercon-
ductor-insulator-superconductor (SIS) quasiparticle mixer should have a 1/f dependence.
This would imply that the critical current density of SIS junctions used for mixing should
increase as frequency squared, a stringent constraint on the design of submillimeter SIS
mixers, rather than in simple proportion to frequency as previously believed. We have
used Tucker's quantum theory of mixing for extensive numerical calculations to
determine G s for an optimized SIS receiver. We find that G s is very roughly independent
of frequency (except for the best junctions at low frequency), and discuss the implications
our results for the design of submillimeter SIS mixers.
INTRODUCTION
Superconductor-insulator-superconductor (SIS) quasiparticle mixers [1] are now
firmly established as the most sensitive receiving devices in the vicinity of 100 to 200
GHz. Their behavior is well described by Tucker's quantum theory of mixing [2]. There
are now many publications which show excellent agreement between the theory's
predictions of a mixer's conversion properties and experimental results, especially at 100
GHz, and the theory also appears to be successful in predicting the noise temperature of
the most sensitive SIS mixers.
Given the success of the Tucker theory at 100 GHz, it is desirable to know the pre-
dicted performance of SIS mixers at higher frequencies, where there are fewer
experimental results but many experiments underway. A large step in this direction was
taken by Kerr and Pan [3], who developed a "design procedure" for SIS mixers, really a
Third International Symposium on Space Terahertz Technology Page 539
set of rules for scaling a successful and reasonably understood low-frequency SIS mixer
design to higher frequency. Their argument was carried further and ratified in Ref. [4].
Kerr and Pan concluded that the critical current density of SIS junctions used for mixing
should increase as frequency squared, rather than in simple proportion to frequency as
previously believed. This result presents a stringent constraint on the design of
submillimeter SIS mixers, implying that high frequency SIS mixers are much more
difficult to realize than had previously been appreciated. This widely quoted conclusion
certainly is influencing the design of the current generation of submillimeter SIS mixers.
Kerr and Pan based their analysis on the "coRnC = 4 rule": the best SIS mixer
performance appears to be obtained when the characteristic parameter ©RnC is near 4,
where w is the LO frequency and Rn is the normal state resistance and C the capacitance
of the SIS junction. As first advanced in Ref. [5] and more recently discussed in Ref.
[6], all SIS mixer experiments exhibiting infinite available gain have coRnC ^ 4, while
coRnC < 1 has always resulted in considerable conversion loss. (To our knowledge this
correlation still holds to date.) Presumably, good mixer conversion requires the reduction
of harmonic conversion effects by the relatively large capacitance. Indeed, computer
simulations show that harmonic conversion becomes significant for coRnC < 4 [7]. On
the other hand, unnecessarily large capacitance entails greater difficulty in tuning and
narrower bandwidth.
The damping time RnC of an SIS junction varies in inverse proportion to its critical
current density, j c - Therefore j c must increase proportional to frequency to maintain a
constant coRnC, and this alone requires an inconveniently large j c for submillimeter SIS
mixers. However, Kerr and Pan rightly note that while the coRnC = 4 rule may be valid
for 100 GHz SIS mixers, there is no reason to expect that the optimum ©RnC is
independent of frequency. In particular, their calculations indicate that the quantity GsRn.
the mixer source conductance normalized to Rn, "should have a 1/f dependence for
mixers in the quantum-limited regime." This immediately implies that j c should increase
as frequency squared.
CALCULATIONS
It is not feasible to optimize an SIS mixer by maximizing the calculated conversion
gain. There is no unique optimum bias point: the quantum theory of mixing predicts
infinite gain for high quality SIS junctions over a wide range of parameter values. Such
high gain is unrealistic and undesirable. Kerr and Pan avoid this difficulty by positing a
set of requirements, including unity gain and moderately well matched input (VSWR £
2), for optimum SIS design. We take a different approach.
Page 540
Third International Symposium on Space Terahertz Technology
We use the quantum mixer theory for extensive numerical calculations, to determine
the minimum value of the SSB (single sideband) noise temperature Tr of an SIS receiver ,
subject to reasonable experimental constraints. Thus our calculation involves a trade-off
between minimizing the mixer noise temperature and maximizing the mixer conversion
gain, which is mediated by the noise temperature of the IF amplifier Tip. Full details will
appear elsewhere. For our current purpose we make the following approximations: We
consider DSB (double sideband) operation in the three-frequency low-IF approximation,
which should be a fairly good representation of most well-designed experimental mixers.
We do not include any interference from the Josephson effect, although this is likely to be
a problem for experiments at the higher frequencies. In addition, we ignore all
reactances. Taken together, these approximations are equivalent to assuming 1) that the
geometrical capacitance of the SIS junction is large enough to both short out the LO
harmonics and their sidebands and to eliminate Josephson interference, 2) that the
capacitance is itself resonated by a relatively broadband external tuning circuit, so that the
intrinsic junction nonlinearity is presented with a resistive embedding impedance at all
relevant frequencies, and 3) that the quantum susceptance has no significant effect. This
third assumption is controversial. It has recently been argued that the quantum
susceptance is a central element of the behavior of SIS mixers [8]. Nevertheless, we
believe that this nonlinear reactance has little effect on the performance of an optimized
SIS receiver, though it may affect the optimum bias point. This question will be
addressed in further research.
Fig. 1. Three synthetic normalized I- V characteristics used for these calculations.
Third International Symposium on Space Terahertz Technology
Page 541
The equations employed in the calculation of Tr are taken from Ref. [1] and will not
be reproduced here. For convenience we assume zero physical temperature; the only
serious effect of this is to ignore the thermal noise from the IF termination which is
reflected from the mixer back into the IF amplifier. For real SIS receivers this can be an
important contribution to the total noise. We require a reasonable input match: in
particular we require that both the signal reflection gain and also the signal-to-image
conversion gain be < 1/4 (which corresponds to VSWR < 3). We find that this constraint
completely eliminates every instance of high conversion gain. What remains is a distinct
solution with stable moderate realistic conversion gain and low mixer noise. Moreover
we find that our quantitative results are extremely insensitive to the level of returned
signal or image power allowed. These topics are discussed at length in Ref. [9].
Frequency (GHz), for V g = 3 mV
50 100 500 1000
2
3
8.
E
o
Z
>
u
■ 1 1 i
■ A
' /'
1000
-
• /'
'■/' i
1 ' :
/•' >*
500
r'
r s
pi
hi
100
/ ' •'
/ ' •'
50
f"
t X
Ji i
/'/
Dull *~S* /
10
Medium •• /
5
/ /
Sharp .••"' /
. . . . i /
2k •;
1
1
1.05 0.1 0.5 1 2
Normalized Frequency co/cOg
Fig. 2. The SSB noise temperature of a DSB SIS receiver optimized at each frequency,
calculated for the three I-V curves of Fig. 1, Gl= 0.3/Rn, Tip = 3 K, and V g = 3 mV.
Page 542
Third International Symposium on Space Terahertz Technology
We have performed these calculations for each frequency for a wide range of
parameters, but only a few of the results can be presented here. The illustrations given in
this paper use the synthetic SIS junction I-V curves depicted in Fig. 1. The "sharp" curve
corresponds to the best experimental SIS I-V curves, the "medium" curve corresponds to
a good quality junction, and the "dull" curve corresponds to a moderate quality junction.
We normalize voltages to the energy gap voltage V g , conductances to the normal state
resistance Rn, and frequencies to the energy gap frequency <o g = eV g //?.
Frequency (GHz), for Vg = 3 mV
50 100 500 1000
0.05 0.1 0.5 1 2
Normalized Frequency co/Wg
Fig. 3. The normalized source conductance G s Rn required to optimize
the receiver of Fig. 2, calculated for the three I-V curves of Fig. 1.
Third International Symposium on Space Terahertz Technology Page 543
RESULTS
Figure 2 shows the minimum theoretical SSB noise temperature of a DSB SIS
receiver with IF load conductance Gl = 0.3/Rn, Tip = 3 K, and V g = 3 mV, for the three
I-V curves of Fig. 1. Figure 3 shows the optimum value of the normalized source
conductance G s Rn required to achieve the minimum Tr. At lower frequencies (below the
vertical rise in each curve) the mixer is biased on higher number photon steps and G s is
relatively constant as expected for classical behavior. On the first photon step, however,
the behavior of G s is quite different. At the lowest frequencies on the first step G s is
strongly dependent on the I-V curve quality; for high quality junctions the optimum G s is
quite large. As the frequency increases, the optimum G s gradually changes to approach a
value = 0.7, for all three I-V curves at frequencies near 2co g .
Figure 3 clearly shows that the optimum G s does not have a 1/f dependence. To
emphasize this point, in Fig. 4 we plot the quantity G s co vs. co for the data of Fig. 3. The
1/f dependence predicted by Kerr and Pan [3] would give horizontal lines in Fig. 4, and
horizontal lines are nowhere seen. Rather, the optimum G s for the sharp curve is given
by the empirical formula G s = 1/2 + 0.25/co for bias points on the first photon step. This
behavior is quite widespread. For instance, Fig. 5 shows the the optimum G s computed
for SIS receivers with various values of Tip, for the sharp I-V curve. The same empirical
formula also works well when we consider different values of load conductance, I-V
curves with considerable leakage current, etc.
In order to better understand the behavior of the optimum G s , in Fig. 6 we compare
it with all of the important "input" conductances in our calculations. It is seen that even
though G s is determined by a trade-off between the gain and the shot noise, the optimum
Gs is quite close to that which minimizes the shot noise, G s hot> Dut far from that which
maximizes the gain, G s '. This surprising result can be explained by examination of the
equations of the SIS mixer. On one hand, the dependence of the conversion gain upon
Gs is given by a simple impedance matching formula which has its minimum at G s = IG S 'I;
a fairly large mismatch therefore results in only a small decrease in gain. On the other
hand, the mixer noise is minimized by the exact cancellation of the correlated components
of the shot noise at the IF and the signal and image frequencies, which occurs at G s =
Gshot- ^ Gs strays from this value the shot noise grows rapidly. The optimum G s is also
far from the signal input conductance, Gs, but never more than a factor of three lest the
signal reflection gain become too large.
Page 544
Third International Symposium on Space Terahertz Technology
O
Frequency (GHz), for V g = 3 mV
500 1000 1500
0.5 1.0 1.5
Normalized Frequency co/o) g
Fig. 4. The data of Fig. 3 are multiplied by w and replotted (in normalized
units), and are compared to an empirical formula.
O
2.0
1.5
1.0
0.5
Frequency (GHz), for V g = 3 mV
500 1000 1500
0.0
1 ' ' r
h K
'
0.0
0.5 1.0 1.5
Normalized Frequency ca/co g
Fig. 5. The optimum source conductance G s of an SIS receiver whose IF
amplifier noise temperature Tip = 10 K, 3 K, and K, respectively, using the
"sharp" I-V curve of Fig. 1, G L = 0.3/R N , and V g = 3 mV.
Third International Symposium on Space Terahertz Technology
Page 545
VI
u
"O
c
o
U
Frequency (GHz), for V g = 3 mV
500
1000
— • -G°
:."*•"•»
■
1
0.5 1 1.5
Normalized Frequency oo/cog
Fig. 6. The optimum source conductance G s of an SIS receiver using the
"medium" I-V curve of Fig. 1, G L = 0.3/R N , Ti F = 3 K, and Vg = 3 mV,
compared to various "input" conductances: Glo and Gs are the input con-
ductances of the mixer at the LO and the signal frequencies, respectively,
G S hot is the value of G s which would minimize the shot noise of the mixer,
and G s ' is the value of G s which would maximize the gain of the mixer.
Note in Fig. 6 that G s hot> and thus the optimum G s , follows closely the input
conductance at the LO frequency, Glo- This is exactly as predicted by the simple
photodiode theory of SIS mixing [10], which reproduces the the equations of the
quantum theory of mixing in the limit of small LO voltage amplitude (small a). It is
surprising that G S hot follows Glo so closely for the relatively large a of our simulations.
In any case this enables us to explain the empirical formula G s = 1/2 + 0.25/co. In the
limit of small a, Glo is the slope of the chord connecting the photon point LjcCVo " h®/e)
to the photon point I<jc(Vo + frw/e) on the unpumped dc I-V curve. Therefore, using the
preferred value for the optimum dc bias voltage Vo = 0.9 for the sharp I-V curve, this
gives Glo = 1/2 + 0.35/co in the small a limit. G s follows but is slightly less than Glo
(Fig. 6) and so is very well approximated by the empirical formula.
Page 546
Third International Symposium on Space Terahertz Technology
DISCUSSION
The results presented here are for particular parameter values, but they are quite
general and representative of our more extensive calculations. In disagreement with Ref.
[3], we find that G s Rn is very roughly independent of frequency (except for the best
junctions at low frequency). This means that there is no reason to suppose that it is
advantageous to increase j c as frequency squared in the design of high frequency SIS
mixers.
Why do our results differ from Ref. [3]? It is likely that this disagreement arises
because in Ref. [3] the gain was fixed to unity, whereas we find that the mixer gain of an
optimized SIS receiver falls off roughly as 1/co for bias points on the first photon step
(Fig. 7). Note in Fig. 7 that we find conversion gain as high as 8 dB in the vicinity of
100 GHz, but in our solution the mixer is operating far from instability [9] with low noise
and quite low returned signal and image power.
CO
T3
O
c
o
>
c
o
U
10r
Frequency (GHz), for V g = 3 mV
50 100 500 1000
0.05 0.1 0.5 1 2
Normalized Frequency co/co g
Fig. 7. The IF conversion gain corresponding to the three curves of Fig. 2.
Third International Symposium on Space Terahertz Technology Page 547
Nevertheless, we agree with Ref. [3] that the "uRnC = 4 rule" should be modified
for submillimeter SIS mixers. Harmonic conversion effects should become much less
important as the frequency is increased, because the SIS junction presents a much weaker
nonlinearity for harmonic frequencies above co g , especially so for frequencies above 2co g .
This implies that the beneficial effects of the capacitance are reduced as the frequency is
increased, and smaller values of ©RnC can be tolerated. Since it is more difficult to re-
sonate the capacitance at high frequency, smaller values of coR^C are desirable.
However, small area and high critical current SIS junctions are difficult to fabricate, and
usually entail undesirable consequences such as inferior junction quality, poorer yield,
etc. Therefore, the choice of wRnC for submillimeter SIS mixers will at best be an
informed compromise.
Acknowledgment: Parts of this work were performed under funding from the Air Force
Office of Scientific Research and from the National Science Foundation.
[1] J.R. Tucker and M.J. Feldman, "Quantum detection at millimeter wavelengths," Rev.
Mod. Phys., vol. 57, pp. 1055-1 113, Oct. 1985.
[2] J.R. Tucker, "Quantum limited detection in tunnel junction mixers," IEEE J. Quantum
Electron., vol. QE-15, pp. 1234-58, Nov. 1979.
[3] A.R. Kerr and S.-K. Pan, "Some recent developments in the design of SIS mixers," in
Proceedings of the First International Symposium on Space Terahertz Technology , 1990,
pp. 363-376; published in Int. J. Infrared Millimeter Waves, vol. 1 1, 1 169 (1990).
[4] R. Blundell and D. Winkler, "The superconductor-insulator-superconductor mixer receiver
- a review," in NATO Applied Research Workshop on Superconducting Electronics and
2nd Workshop on Josephson Devices, 1990, Capri, Italy.
[5] M.J. Feldman and S. Rudner, "Mixing with SIS arrays," Reviews of Infrared & Millimeter
Waves, edited by K.J. Button (Plenum, New York), vol.1, pp.47-75, 1983.
[6] M.J. Feldman, "Theoretical considerations for THz SIS mixers," Int. J. Infrared
Millimeter Waves, vol. 8, pp. 1287-1292, Oct. 1987.
[7] S. Withington and E.L. Kollberg, "Spectral-domain analysis of harmonic effects in
superconducting quasiparticle mixers," IEEE Trans. Microwave Theory Tech., vol. MTT-
37, pp. 231-238, Jan. 1989.
[8] C.A. Mears, Qing Hu, and P.L. Richards, "The effect of the quantum susceptance on the
gain of superconducting quasiparticle mixers," IEEE Trans. Magnetics, vol. MAG-27, pp.
3384-3387, March 1991.
[9] Qing Ke and M.J. Feldman, "Reflected power effects and high gain in the quantum theory
of mixing," submitted to IEEE Trans. Microwave Theory Tech.
[10] M.J. Wengler, "Submillimeter wave detection with superconducting tunnel diodes,"
submitted to Proc. IEEE; see also M.J. Wengler and D.P. Woody, "Optimizing double-
sideband SIS quasiparticle mixers," IEEE Trans. Magnetics, vol. MAG-27, pp. 3388-
3390, March 1991.
Page 548 Third International Symposium on Space Terahertz Technology
/Losbo N93-27772
V
RESONANT TUNNELING DIODES AS
SOURCES FOR MILLIMETER AND
SUBMILLIMETER WAVELENGTHS
O. Vanbesien, R. Bouregba, P. Mounaix
and D. Lippens
Centre Hyperfrequences et Semiconducteurs U.A. CNRS N° 287
Universite de Lille - 59655 Villeneuve d'Ascq Cedex - France
L. Palmateer + , J.C. Pernot, G. Beaudin*
and P. Encrenaz
Ecole Normale Superieure 24, rue Lhomond 75231 PARIS
* Observatoire de Meudon 92195 Meudon Principal Cedex
E. Bockenhoff + + , J. Nagle, P. Bois, F. Che voir
and B. Vinter
Laboratoire Central de Recherche Thomson CSF
Domaine de Corbeville 91404 Orsay Cedex - France.
ABSTRACT
High-quality Resonant Tunneling Diodes have been fabricated
and tested as sources for millimeter and submillimeter
wavelengths. The devices have shown excellent I-V characteristics
with peak-to-valley current ratios as high as 6:1 and current
densities in the range of 50-150 kA/cm2 at 300 K. Used as local
oscillators, the diodes are capable of state of the art output power
delivered by AlGaAs-based tunneling devices. As harmonic
multipliers, a frequency of 320 GHz has been achieved by
quintupling the fundamental oscillation of a klystron source.
+ Now at IBM, Yorktown Heights + + at Mercedes Benz, Stuttgart
Third International Symposium on Space Terahertz Technology Page 549
1. INTRODUCTION
Resonant Tunneling Diodes (RTD's) exhibit very strong non linearity with
short time response which make them attractive in non linear applications for
millimeter and submillimeter wavelengths [1]. RTD's have already demonstrated
their potential for a variety of high speed/high frequency applications [2]-[5]. In
this paper we report on the effort of a group of laboratories in France on these
novel devices with special emphasis on local oscillators and harmonic multipliers.
The fabrication procedures in a whisker contacted technology and in a microwave
compatible technology suitable for monolithic integration are outlined in section
2. The DC and AC characterizations are reported in section 3 whereas the
oscillator and multiplier results using the devices are described in section 4.
2. TECHNOLOGICAL PROCESS
The two types of epitaxial structures grown by molecular beam epitaxy are
given in Figure 1(a) and (b). Both samples noted A and B had 17 A - thick AlAs
barriers and access regions with a stepped doping profile from l-2xl0 1 '7 cm-3 to 2-
3x1018 cm-3. They differ mainly owing to the strained Gao.9Ino.1As layers so that
structure B resembles a triple well resonant tunneling structure. By placing a
GalnAs well just prior the growth of the double barrier heterostructure (DBH) it
is expected that the peak-to-valley current ratio (PVCR's) should be enhanced
because the negative differential resistance effect involves the anticrossing of two
confined states [6] [7]. In addition by placing a GalnAs well rather than a GaAs
one and by a proper choice of the well width (Lw = 40 A), the ground state can be
lowered in energy while keeping the excited state practically unchanged. The
associated benefits are a reduction of the peak voltage and higher PVCR's [8].
Also note that structure A and B are grown on n+ and S-I substrates respectively.
Page 550
Third International Symposium on Space Terahertz Technology
GaAs
2 10 13 cm -3
490 nm
GaAs
2 10 17 cm- 3
50 nm
GaAs
undoDed (UD}
5 nm
AlAs
UD
1.7 nm
GaAs
UD
4.5 nm
AlAs
UD
1.7 nm
GaAs
UD
5 nm
GaAs
2 10 17 cm- 3
50 nm
GaAs
2 10 18 cm -3
500 nm
(a)
n + substrate
GaAs
3 10 18 cm" 3
500 nm
GaAs
10 17 cm- 3
10 nm
GaAs
undoped (UD)
5 nm
In i Gag As
UD
5 nm
GaAs
UD
0.5 nm
AlAs
UD
1.7 nm
GaAs
UD
0.5 nm
IniGa 9 As
UD
4 nm
Symmetrical layers
(b)
Figure 1 .Growth sequence for the epilayer on n + substrate (sample A), (a)
Sample B grown on semi -insulating subtrate. (b)
The epilayers on n + substrate were processed using a whisker contacted
technology including patterning of Ni/GeAu layers into matrix of 3.5 um diameter
dots on the epitaxial side of the wafer and uniform deposition on the back side
followed by alloying of these layers to make ohmic contacts. Mesa isolation was
performed by chlorine ion beam assisted etching as shown in figure 2, using the
patterned metal as a mask. As last stages some of the samples were thinned to a
thickness of about 120 urn and polyimide was used to surround the diodes in order
to aid whisker contact.
Third International Symposium on Space Terahertz Technology
Page 551
Figure 2 : SEM Photos of diodes formed by RIE.
For the epilayers on S-I substrate, the diodes were fabricated in a
microwave-compatible two-step mesa technology [9]. In that case, the devices
were connected to low-loss transmission lines in such a way that they can be
characterized at the wafer level. Such vertically integrated devices require a
means of connecting the contact on the top of the mesa to the pad of the
transmission line. We thus developed two versions : (i) a dielectric assisted cross-
over and (ii) an air bridge interconnection. A scanning electron micrograph of two
representative devices are shown in figures 3 and 4. In figure 3 a coplanar probe
configuration is apparent. Also clearly shown is the deposited strap which crosses
over the mesa edges covered with Si3 N4 layer appearing in dark. In the second
Version, the dielectric cross-over is replaced by an air bridge yielding a reduced
parasitic capacitance. Figure 4 illustrates the technology employed with the
Page 552
Third International Symposium on Space Terahertz Technology
mushroom shaped metallization which enables one to connect small anode
fingers.
AuGeNi Ohmic Contacts
Overlay Metallization
Figure 3 :Schematic cross section and SEM photo of the RTD fabricated in a
planar technology.
Figure 4 : SEM of a diode fabricated using air bridge techniques.
Third International Symposium on Space Terahertz Technology
Page 553
3. DC AND AC CHARACTERIZATION
Figure 5a shows a typical current voltage I-V characteristic for a GaAs/AlAs
device on n+ substrate at 300K. The device exhibits a peak current of ~ 16 mA at
~ 1.8 V which corresponds to a peak current density of— 160 kA/cm2 for a 3,5 um
diameter diode. For larger size of the diodes, heating of the samples prevents us
from achieving these densities. A typical DC characteristic for a device on SI
substrate is displayed in Figure 5b. The device exhibits excellent characteristics
with PVCR's as high as 6:1 along with simultaneously peak current density of
50 kA/cm2 which compare favorably to the best published results [10] [11] Note
also the high degree of symmetry in the I-V curve which is a good indicator of
quality interfaces.
201-
16
< 12
E 8
§
1-4
u -8
-12
-16
-20
T-300K
A
."
-2-1 1
Voltage V
8
T-300K
i
6
k i
< 4
-
\ /
E
/ /
2
/ -L/
/
,. /-, /
c -2
■ r\ /
^
/ /
w
/ /
3-4
' V
-6
V
-8
b
_l L_ 1 1
-U5 -.8 .8
Voltage V
1J6
Figure 5 .Typical DC characteristics for a GaAs/AlAs device on n+ substrate (a)
and for a GalnAs/GaAs/AlAs pseudomorphic device on SI substrate (b).
Page 554
Third International Symposium on Space Terahertz Technology
Following our previous work [12], on-wafer reflection gain measurements
were performed between 50 MHz and 40 GHz using cascade RF probes and an
85107 A HP network analyser. Shown in Figure 6 is the one port measurement of
a vertically integrated sample. The active area is 20 um2. The diode is biased in
the NDR region. Note that ho de-embedding was used at this stage to correct for
parasitics. For frequency evaluation, we used the equivalent circuit which
consists of a single capacitor Cd with a parallel negative resistance Rd- -hese
intrinsic lumped elements are completed by the parasitic capacitance C p , the
inductance Lp attributable to the bonding and Rs the overall series resistance. A
good fit was obtained for Cd = 36 fF, Rd = - 172 Q, R s = 9 Q, L p = 60 pH and
C p = 13 fF (air-bridge technology). With this set of data derived from experiment,
the cut off frequency for NDR is in excess of 100 GHz. This frequency is limited by
the high impedance level needed to satisfy the stability criteria.
$
c
A
su z
REF 2.5 Units
900.0 iriJnita/
30.09 Q -32.191 Q
V
MARKER
40.0,
START 0.050000000 GHz
STOP 40.000000000 GHz
Figuer 6 .One port measurement of the impedance. The bias is adjusted in the NDR
region to satisfy the stability criteria.
Third International Symposium on Space Terahertz Technology
Page 555
4. OSCILLATOR AND MULTIPLIER RESULTS
The wafers on n+ substrate were sawed into chips of 100x100 um2 and
mounted in a test waveguide for measuring the oscillator power at 35 and 110
GHz. The power levels measured at 300 K with a bolometer were 36 pW at 38.6
GHz and 12 pW at 110 GHz. Referring to the oscillator results from the published
literature on AlGaAs based RTD's [13] given in figure 7 the output power are
state of the art results.
100-
\
10
EXPERIMENT
1s
0.1
E ■ Sollneretal
- A our work
j i ' mm ' i ' Him
1
10 100
Frequency GHz
mill
1000
Figure 7 : Experimental powers for a 4 pm diameter diode from reference [13] and
results from the present work for a 3.5 pm diameter diode.
For harmonic multiplication, the samples were mounted in a commercially
available multiplier mount. The measurement set up has a quasi-optical scheme
and was initially developed to study reactive species of astrophysical interest [14].
The diodes were driven by a klystron at 64 GHz and, in the output path, high
filters enables one to spectrally analyze the power delivered by the diode. The
receiver is a helium cooled InSb detector. Figure 8 shows the power response at
3rd harmonic (192 GHz) and 5th harmonic (320 GHz). For comparison in terms of
available power commercial Schottky diodes were also tested under the same
__
Page 556
Third International Symposium on Space Terahertz Technology
experimental conditions. It is interesting to note that equivalent performances
were obtained for both types of devices by increasing the multiplication order to
frequency quintupling.
£
a
UJ
r-
O
UJ
fc3
Q
IU
O
<
r-
o
>
500 T
400 -
300 -
200 -
100 -
SCHOTTKY fifth hirm.
OBRTD third harm.
♦
♦
♦
D8RTD fifth harm.
••;/
-+-
±^
▲
▲
16 14 12 10 8 6 4 2
POWER SOURCE ATTENUATION (dB)
Figure 8 : Measured voltage of InSb detector against input power delivered by a
klystron at 64 GHz.
Third International Symposium on Space Terahertz Technology Page 557
In the multiplier experiment the devices were unbiased and driven in the
NDR region to take advantage of multiple extrema in the current waveform. This
requirement can be unfavorable especially for high threshold voltage devices
when input power is limited [14]. From this viewpoint, it is clear that
pseudomorphic structures with a buried well may overcome partly this difficulty.
From Figure 3 it is apparent that a drastic decrease in the peak voltage 0.8 V
instead of 1.8 V has been achieved by comparing structures A and B. In addition it
can be noted that the PVCR's were enhanced. This suggests the superiority of
these new tunneling devices for multiplication in view of the large harmonic
content in the current and of the reduction of the amount of input power required
to pump the diode.
5. CONCLUSION
High performance resonant tunneling diodes were sucessfully fabricated in
a whisker contacted and in a planar technology. The RF capabilities of the diodes
were demonstrated either by direct measurement of their small-signal impedance
or by using them for oscillator and multiplier.
ACKNOWLEDGEMENTS
This work was supported by the Ministere de la Recherche et de la
Technologic Technical assistance of M. Bogey, J.L. Destombes and A. Lescluse
with the Laboratoire de Spectroscopic Hertzienne of the Lille University is highly
appreciated.
Page 558 Third International Symposium on Space Terahertz Technology
REFERENCES
[1] T.C.L.G- Sollner, E.R. Brown and H.Q. Le, "Microwave and Millemeter-
wave Resonant Tunneling Devices". Physics of Quantum Electron Devices
edited by F. Capasso Springer- Verlag.
[2] E.R. Brown, J.R. Soderstorm, CD. Parker, L.J. Mahoney, K.M. Kolvar and
T.C. Mc uill, "Oscillations up to 712 GHz in InAs/AlSb resonant tunneling
diodes". Appl. Phys. Lett. 58, pp. 2291-2293, May 1991.
[3] A. Rydberg and H. Gronquist "Quantum well high efficiency millimeter-
wave frequency tripler". Electronics Letters, Vol. 25, pp. 348-349, 1989.
[4] P.D. Batelaan and M.A. Frerking, "A quantum well frequency multiplier
with millimeter wave output". Proc. 4th Conf. Infrared Physics Zurich,
pp. 527-529, 1988.
[5] R. Bouregba, D. Lippens, L. Palmateer, E. Bockenhoff, M. Bogey, J.L.
Destombes and A. Lecluse, "Frequency multiplication using resonant
tunneling diode with output at submillimeter wavelengths". Electronics
Lett. Vol.26, pp. 1804-1905, October 1990.
[6] D. Thomas, F. Chevoir, E. Barbier, Y. Guldner and J.P. Vieren, "Magneto
tunneling of charge build up in double barrier diodes". Proc. of 4th
International Conference on Superlattices Microstructures and
Microdevices 5, pp. 219-224, 1989. .
[7] P. Mounaix, O. Vanbesien and D. Lippens, "Effect of cathode spacer layer on
the current-voltage characteristics of resonant tunneling diodes". Appl.
Phys. Lett. 57, pp 1517-1519, October 1990.
Third International Symposium on Space Terahertz Technology Page 559
[8] T.P.E. Broeckaert, W. Lee and C.G. Fonstad "Pseudomorphic
Ino.53Gao.47As/AlAs/InAs resonant tunneling diodes with peak-to-valley
current ratios of 30 at room temperature.
[9] D. Lippens, E. Barbier and P. Mounaix, "Fabrication of High-Performance
Al x Gai- x As/In y Gai-yAs/GaAs Resonant Tunneling Diodes using a
Microwave-compatible Technology. IEEE Electron Dev. Lett. Vol. 12, pp.
114-116, March 1991.
[10] R.M. Kapre, A. Madhukar, and S. Guha "Highly strained
GaAs/InGaAs/AlAs resonant tunneling diodes with simultaneously high
peak current densities and peak-to-valley current ratios". Appl. Phys. Lett.
58, pp. 2255-2257, May 1991.
[11] H. Brugger, U. Meiners, C. Wolk, R. Deufel, A. Morten, M. Rossmanith,
K.V. Klitzing and R. Sauer "Pseudomorphic Two Dimensional Electron-Gas-
Emitter Resonant Tunneling Devices" Microelectronics Engineering
Elsevier, 15, pp. 663-666, 1991.
[12] P. Mounaix, P. Bedu, D. Lippens and E. Barbier "Measurement of negative
differential conductance up to 40 GHz for vertically integrated resonant
tunneling diodes". Electronics Letters Vol. 27, pp. 1358-1359, July 1991.
[13] E.R. Brown, T.C.L.G. Sollner, C.D. Parker, W.D. Goodhue and C. Chen.
"Oscillations up to 420 GHz in GaAs/AlAs resonant tunneling diodes" Appl.
Phys. Lett. 55, pp. 1777-1779, October 1989.
[14] J.L. Destombes, C. Demuynck and M. Bogey "Millimeter-wave and
submillimeter-wave spectroscopy of molecular ions". Phil. Trans. R. Soc. A
324, pp. 147-162,1988.
Page 560 Third International Symposium on Space Terahertz Technology
/£*&/ /1 J N93-27773
- \?
Simulation of Electron Transport
in
Quantum Well Devices
D. R. Miller, K. K. GuIIapalli, V. R. Reddy, and D. P. Neikirk
Department of Electrical and Computer Engineering
Microelectronics Research Center
The University of Texas at Austin
1.0 Introduction
Double barrier resonant tunneling diodes (DBRTD) have received much attention as
possible terahertz devices. Experimentally, DBRTD 's have shown detection capabilities at
sub-millimeter wavelengths 1 . When used as oscillators, small amounts of power have also
been measured in sub-millimeter range 2 . Despite these impressive experimental results, the
specific of the device physics (i.e., how the electrons propagate through the structure) are
only qualitatively understood. Therefore, better transport models are warranted if this
technology is to mature.
Near the heterostructure double barrier region, it is generally accepted that quantum
mechanical transport, via tunneling and reflections, dominate the electron dynamics.
However, most DBRTDs in use today are designed with extended spacer regions. These
spacer regions serve the function of increasing the real part of the overall device impedance
while simultaneously reducing the imaginary part, thereby incorporating DBRTDs in
millimeter wave circuits a far easier task. Since the spacer regions are sufficiendy removed
from the heterojunctions, electron propagation should be govern by semiclassical and not
quantum mechanical considerations. Here, semiclassical refers to transport which is
adequately describe using some form of the semiclassical Boltzmann equation.
Third International Symposium on Space Terahertz Technology Page 561
Past simulation models of DBRTD structures have evolved from simple
Schrodinger equation solutions of a free electron in a double barrier potential to more
complicated methods involving multiband, multivalley Schrodinger solutions 3 ' 4 or single
valley kinetic equations that utilizes quantum Wigner functions 5,6 ' 7 . These methods are
expected to work reasonably well for ideal DBRTD structures with parabolic bands in
which only quantum mechanical reflections and tunneling are important. However,
because the Schrodinger or single valley Wigner models do not include realistic phonon
scattering or band structure effects (i.e., multiple valleys, rton-parabolicity, multiple bands,
etc.) these models do not adequately address the transport through the semiclassical region.
Alternatively, the semiclassical Boltzmann equation provides an adequate description of the
semiclassical region, but fails completely near the DBRTD heterostructure region.
To model a DBRTD structure with two distinct transport regions, two options are
available. The first option incorporates a composite scheme by which each region is
modeled with an equation suitable for that region. The two solutions are then matched at a
quantum / classical interface to obtain a self consistent solution throughout the device. We
find combinations of a free particle Schrodinger equation for the quantum region coupled
with either the drift/diffusion 8 or Monte Carlo 9 formalism for the semiclassical region quite
useful. However, agreement between experiment and theory is still lacking since the
simple Schrodinger equation is only an approximate solution of the electron transport
within the heterostructure region.
The second option is to model the entire device with one self consistent formalism
that, in principle, can account for all the important device physics for each region. The
kinetic equation based upon the Lattice Wigner function is a promising candidate for such a
task. In this paper, we will use the Lattice Wigner function to explain important transport
issues associated with DBRTD device behavior.
2.0 The Lattice Wigner Function
The lattice Wigner function 10 - 11 we employ is based on the discrete spectrum
composed of Wannier and Bloch crystal representations, making it different from other
Wigner function methods. Because of the choice of representations, band structure effects
are explicitly included in the kinetic equation. Thus, issues such as T to X tunneling, non-
parabolicity of the conduction bands, or effective mass variations across the heterojunction,
Page 562
Third International Symposium on Space Terahertz Technology
can be examined in detail. Phonon scattering is also included through the standard
semiclassical Boltzmann collision term.
In a multiple barrier heterostructure device with multiple non-parabolic conduction
band valleys and no interband coupling, it is possible to write separate, but coupled, kinetic
equations for each valley. The Wigner function for the i* valley, f { , is found from the
solution of
_ 9f i w (R,k) [ I 3e.(k) Bf^flU) eE Bf^R.k) r afflU) ^
3t h 3k 3R h 9k I, 3t J,
Col
#Valley»#Bamers
-X £ Br+Br+B^
j=l m=l
(1)
where the barrier scattering matrices, Bo, Bi, and B2 are given by
Bl 2 £ 3R [* h f [V^y** 2 j + V^^K 2
3R 2
i Ie ^- k ,r V]
VN L R
'Pulie(ni)| ** + ~ j Ypulie(m)l** ~
(2)
(3)
(4)
8im A k .
The indices, j and q, indicate the j th conduction band valley and the q th barrier. The
coefficients, B^M, are the potential energy terms that account for one heterostructure
barrier. In equations 1 through 4, n is the band index, k and k ' are crystal momenta, R
and R' are lattice vectors, Nl is the number of lattice sites, and E is the electric field.
V'uise is the potential energy diagonal matrix element of the mft 1 pulse function that localizes
u *i *i
the barrier. The change in the i" 1 valley effective mass (m A ), velocity (v A ), and the offset
energy (AEO are given by
m
A "^Barrier
Barrier /
m
Bulk
i _ 1 ae^Ck) 1 ae^Ck)
V A =
dk
3k
AE i (k) = e^ raier (k)-ef ulk (k)
(5)
(6)
(7)
Third International Symposium on Space Terahertz Technology Page 563
where £^ amer (ki) and e n u (kj) are the minimum conduction band energies of the i* valley
for the barrier and bulk materials, respectively.
As noted earlier, equation 1 accounts for both effective mass variations across the
heterostructure interface and intervalley coupling. The intervalley coupling is possible
since the i m valley distribution function f { is coupled to the j 01 valley distribution function
w
f through the barrier terms. Effective mass variation effects are due to the crystal
momentum dependence of the barrier, and result in the spatial derivatives of the distribution
function in the B^ q and Bj q terms. Mass variations are also responsible for raising and/or
lowering the barrier height, as seen in the expression for Bj q . This is because the energy
band of the bulk material may rise at a different rate than the energy band of the barrier
material for a given momentum change. For semiconductors with non-parabolic valleys,
additional barrier scattering terms arise due to the higher derivatives associated with the
Wigner-Moyal expansion. However, in this paper we will assume that the effective masses
and the non-parabolicity factors are approximately the same for the bulk and barrier
material. Therefore, Bi, B2, and all other higher barrier scattering terms will be set to zero.
Including some sort of phonon collision processes into a Wigner function
calculation of an DBRTD is not new. Most calculations have approximated the influence
on carrier transport from phonon collisions using the relaxation time approximation 7 - 12 ' 13 .
In addition, these calculations have restricted phonon scattering events to one valley.
However, because of the high electric fields within the DBRTD structure, intervalley
scattering is required. Thus, we will assume that the total scattering matrix, Sjotal. is given
by
S^(k\k) = S^ p (k , ,k) + SiL Mtic (k I ,k) + Si,L vllUey (k , ,k) (9)
where Spop, S Acoustic, and Simervalley refer to the polar optical, acoustical, and intervalley
phonon scattering matrices, respectively. Assuming non-degenerate statistics, we can write
the collision term in equation 1 for the i* valley Wigner function as
fa&Bdo] = 1 1 j J ^ L(k)k . )f w (R)kl) _ s ^ (k . >k) ^ (R)k J
V tit J Col W Lpl [ k' J (g)
where Ny is the total number of valleys. The material parameters as well as the functional
form for each scattering matrix are identical to those used in Monte Carlo device
simulations and are given in reference 14.
Page 564 Third International Symposium on Space Terahertz Technology
The Wigner distribution function for each valley is defined in the three dimensional
crystal momentum space. If the structure and the electric field are homogeneous in two of
the spatial variables (independent of R y and R z ), equation 1 reduces to a four dimensional
integral/differential equation. Since solutions to the four dimensional problems are
exceedingly difficult, further simplification is required. We do this by characterizing the
distribution function that is transverse to the electric fields by a Maxwellian defined at some
transverse temperature Tt. Furthermore Tt is assumed to vary with longitudinal position.
Thus, for electric fields in the (100) direction with quantum well barriers grown in the
(100) plane, the total distribution function is approximated by the product of a transverse
and a longitudinal distribution functions
f i w (R x ,k x ,k y ,k z ) = f i R "(R x ,k x )exp(-E;/k b T t (Rj) (9)
where k x , k y , and k z are defined with respect to the valley minimum. The transverse
energy for non-parabolic valleys , Et, is assumed to be of the form
r"H r"
2m' 2m, 1
e; = t — \ y ; do)
where aj is the non-parabolicity factor for the i th valley. Note that although the
longitudinal and transverse distribution functions are not coupled through the electric field
or the barrier scattering matrix, they are coupled by the phonon scattering matrix. It is
because of the this coupling that the transverse temperature significantly impacts the device
physics, as discussed in the next section.
3.0 Simulation Results
The first obvious question that can be addressed with the lattice Wigner function is
how close to the quantum well an electron is when quantum mechanical reflections and/or
tunneling affect the electron's behavior. One measure of this distance is obtained by
comparing the size of the barrier matrix (Bo) with the size of the phonon matrix (STotal) at
each lattice site. However, such a direct comparison may provide an overestimate of the
extent of the quantum region. During one phonon scattering interval, an electron may
Third International Symposium on Space Terahertz Technology
Page 565
propagate through many lattice sites due to either a high electric field or a high initial
velocity. Therefore, the effective barrier scattering strength is an average over all lattice
sites occupied during the mean free flight time between collisions. In other words, the
barriers' influence on the electrons is best determined by the matrix
-I (Dlfflol
b Tou1 =-— x f ' Br(R x (t),k x (t))dt
: i ~ r o j=i °
(11)
integrated over the classical trajectory. In equation 1 1, Bo is the matrix defined by equation
2, and R x (t) and k x (t) are given by the semiclassical equations of motion.
l/Tn^OOkV/cm)
l/xj;0kV/cm)
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 ■ ■ 1 1 1 1 1 1 1 1 1 1 1
M 2( 4( 6C 8( 10( li( 14(
Position (monolayer!
Figure 1. Effective quantum well scattering strength versus
position from the center of a 6/19/6 GaAs/AlAs quantum
well.
It is difficult to evaluate the above equation for the general case. However, Figure
1 shows the effective quantum well scattering strengths determined by taking the matrix
norm of Bjotal for two simplified conditions. The structure is a 6/19/6 GaAs/AlAs
quantum well. The first solid line assumes the case Of an extremely slow electron in which
the electron transverses only one lattice site between phonon collisions. The second line
assumes that the mean free path was 32 monolayers, corresponding to a distance of 90.6
A. In both cases the electric field was set to zero, which significantly simplified the
integration along the classical trajectory. Also plotted are the momentum relaxation times
corresponding to different electric fields. These values represent the size of the phonon
Page 566 Third International Symposium on Space Terahertz Technology
scattering matrix (Sjotal) determined using standard Monte Carlo simulations of
homogeneous material 9 . Thus, when comparing the relative strengths of the quantum well
and the phonon scattering matrices, one should compare the fast electron case with the high
field momentum relaxation time since these two conditions usually exist together.
Similarly, the slow electron case corresponds to the zero electric field relaxation time.
Two points are worth emphasizing from Figure 1. The first is that slow electrons
are influenced by the quantum well with greater strength and at greater distances when
compared to fast electrons. The second point, however, is the one which we want to
emphasize. DBRTD structures are usually operated at extremely high electric fields (200
kV/cm) and under ballistic conditions. Therefore, as can be seen from the fast electron
curve of Figure 1, the region where the barrier potential has the greatest impact on electron
dynamics is within 65 monolayers of the center of the well. Beyond that region, phonon
scattering is more important. This fact supports the hypothesis that much of the DBRTD is
governed by semiclassical considerations for devices that extend 400 to 800 monolayers
from the quantum well.
The Wigner function has been used extensively to account for the quantum
transport within the heterostructure region of a DBRTD device. However, the question
arises as too its ability to properly account for electron transport within the semiclassical
region. To show that this is possible, we have used the Lattice Wigner Function to
calculate the velocity versus electric field characteristics for homogeneous GaAs. For
homogeneous samples, equation 1 reduces to
= _^ af^jo_ j_| K | ey ( s ^ (k>k . )f w (R>kl) _ s ^ (k . >k)f w (R>k) j| (12)
To solve this equation, the first step is to integrate over the transverse crystal momentum
directions. The resulting one dimensional phonon scattering matrix is then combined with
the matrix generated by the electric field term to yield the equation
= T^jJ?-f R '- j - k "» (13)
where T is the combined phonon/electric field matrix and fj Rx J. kx is the vectorized
longitudinal distribution function. The valid solution exists when the minimum eigenvalue
of T is equal to zero 15 . If the minimum eigenvalue is not zero, the electric field and the
valley transverse temperatures, T t , are adjusted so that homogeneous equation can be
Third International Symposium on Space Terahertz Technology
Page 567
satisfied. In affect, we are using the transverse valley temperatures as fitting parameters to
achieve a reasonable velocity versus field curve. Once the zero eigenvalue is calculated, the
eigenvector for that eigenvalue can be determined, from which the velocity is obtained.
Note that pure state tunneling models using only the Schrodinger equation do not produce
reasonable velocity-field curves for GaAs.
The above procedure yields good agreement with the Monte Carlo results, as
indicated in Figure 2. Here, a velocity field curve generated by equations 12 and 13 is
compared against a three valley model calculated using standard Monte Carlo techniques.
Furthermore, the temperatures required to obtain this level of agreement are comparable to
the transverse temperatures calculated from the homogeneous Monte Carlo simulations.
However, at low electric fields there is a larger discrepancy between the two alternative
approaches. This difference is probably due to the fact that the transverse distribution
function is not precisely Maxwellian. However, this error is fairly small, demonstrating
that the Lattice Wigner function is capable of simulating both the quantum transport region
and the classical transport regions of a device.
4.00e+7 i
o
3.50e+7 -
3.00e+7 ■:
. Lattice
b
2.50e+7 ■:
<**
2.00e+7 -
/ • •^
o
o
1.50e+7 -
>
1.00e+7 ■:
5.00e+6 "
/ Monte Carlo
0.00e+0
i i i i
5000 10000 15000
Electric Field (V/cm)
Figure 2. Velocity field calculation for a 3 valley spherical
non-parabolic Wigner function model. Also shown is the
corresponding result from a Monte Carlo simulation. The
scattering parameters used for each calculation are identical,
and are found in reference 14.
Thus far, we have applied the Lattice Wigner function to determine generic
properties of bulk materials and of resonant tunneling transport. This formalism's real
Page 568
Third International Symposium on Space Terahertz Technology
usefulness can be demonstrated by applying it to an actual DBRTD device. Shown in
Figure 3 is a typical GaAs/AlAs DBRTD. The quantum well consists of 6 monolayer
AlAs barriers separated by 19 monolayer GaAs well. As can be seen in Figure 3, a
moderately doped, extended spacer region is added to the right hand side of the quantum
well. As will be shown latter, it is the semiclassical transport through the extended spacer
layer that can have a serious impact on device behavior.
t — ' i ' i — « — r— « — r
-200 -100 100 200 300 400
Position (monolayers)
Figure 3. 6/19/6 GaAs/AlAs DBRTD used for the device
simulations. The x axis origin is defined as the center of the
quantum well.
The equilibrium electron concentration, found by integrating the Lattice Wigner
function over momentum space, is shown in Figure 4. However, under bias conditions,
obtaining the electron concentration via the Lattice Wigner function is much more
complicated. The basic problem is in determining the transverse temperature profile of the
spacer region. We have previously shown using a composite Schrodinger/Monte Carlo
model 9 , that for typical bias voltages, electrons are quickly scattered into the upper satellite
values once they emerge from the quantum well. This intervalley scattering between
equivalent and non-equivalent conduction band valleys quickly heats up the carrier
distribution functions in the transverse direction. The effect of the carrier heating on the
charge densities can be considerable.
Third International Symposium on Space Terahertz Technology
Page 569
10
19.
10 1S i
10
14.
Electron Density
'Impurity Density
-200 -100 100 200 300 400
Position (monolayers)
Figure 4. Equilibrium electron concentration for the device
shown in Figure 3.
10
19.
10
14
Non constant T
T=300 degrees
-i — | — i — i — i — i ■ i — i— i — r-
200 -100 100 200 300 400
Position (monolayers)
Figure 5. Electron concentrations obtained from the Lattice
Wigner formalism for two different transverse temperature
profiles. The dc bias is at 0.7 V.
Page 570
Third International Symposium on Space Terahertz Technology
Figure 5 illustrates the effect of transverse temperature on the simulated electron
concentration for this device biased at 0.7 V. The dashed curve shown in this figure was
generated under the assumption of constant transverse temperature within the spacer
region. As seen from this curve, the electron density actually decreases past the quantum
well region, indicating an increase in the overall electron velocities. The increase in the
electron velocities is a result of a non-physical assumption. Because the transverse
temperature in this space charge region is kept artificially low, the electron population is
also artificially cooled, keeping all electrons in the fast T valley. Therefore, unrealistically
high velocities would be predicted under this constant temperature assumption. This can
lead to unrealistic high frequency performance predictions, since fast electrons generally
result in improved frequency response.
Also shown in Figure 5 are the results of a Lattice Wigner function calculation
using a transverse temperature profile obtained through the Schrodinger/Monte Carlo
model. As seen from this figure, the electron concentration is increased significantly over
the constant temperature model, indicating a slowing down of the electron population
within the spacer region. The reduced velocity is a result of the increased phonon
scattering within the T valley, as well as some T to L intervalley transfer. This slowing of
the electrons can have an significant impact on the device performance, as discussed below.
8000
t — ■ — i — ■ — i — ' — r— • — r
■200 -100 100 200 300
Position (monolayers)
400
Figure 6. Transverse temperature profile for the device
given in Figure 3. The dc bias is at 2.0 volts. The
temperature was obtained from a Schrodinger/Monte Carlo
transport model discussed in reference 9.
Third International Symposium on Space Terahertz Technology Page 571
The bias of 0.7 V is well below the voltage at which peak current is expected in this
device. However, even under these conditions, the Monte Carlo simulated T transverse
temperature peaks at 2500 Kelvin. At even higher biases, the transverse temperature
becomes extremely high. Figure 6 shows the transverse temperature profiles extracted
through a three valley Schrodinger/Monte Carlo model for a device biased at 2.0 V . As
seen from this figure, the T temperature can exceed 7000 degrees Kelvin. Even the upper
satellite valleys are exceptionally hot Thus, we would expect that most of the electrons for
this bias voltage are in the upper satellite valley, which is indeed the case.
4.0 Impact on High Frequency Behavior
Much of the discussion concerning the ultimate frequency limitations of DBRTD
devices has centered on the frequency limitations imposed by the quantum well itself. It
has been projected that the quantum well is capable of operating at terahertz frequencies 16 .
However, the preceding discussion illustrates that the semiclassical spacer regions can have
a dramatic impact on the behavior of DBRTD devices. Because of the efficient phonon
scattering processes, the actual number of free carriers within the spacer layer is much
higher than predicted by either a pure state Schrodinger solution, a one valley Wigner
solution, or a multiple valley constant temperature Lattice Wigner solution. This free
charge contributes to a positive resistance which is equivalent in every way to the space
charge resistance found in transit time diodes 17 . The magnitude of the space charge
resistance is fundamentally determined by the total number of free carriers within the
region. Therefore, any analysis which does not realistically describe electron densities
cannot be used to project the high frequency performance of a real device.
Understanding space charge resistance is important because it is generally felt that
in order to improve the output power of a DBRTD device, one must dramatically increase
the current density. In reality, this procedure will be useful only up the point where
deleterious space charge resistance effects become dominant. This can be seen even under
static conditions. For example, considered the measured dc-IV curve of an InGaAs/AlAs
DBRTD structure, shown in Figure 7. The total spacer region for this device was 1250A.
With available simulation tools, it is possible to model comparable structures with
artificially increased current densities. Thus, five static simulations where performed using
a composite quantum injector/drift/diffusion model, as described in reference 18. For each
Page 572
Third International Symposium on Space Terahertz Technology
simulation, the DBRTD spacer regions where kept constant. The only difference between
each simulation is an assumed increase in the current density of the device, starting with the
dc-IV curve given in Figure 7. The results of the simulations are shown in Figure 8. As
the current density is magnified, the presence of the free carriers within the spacer regions
can cause a major portion of the negative resistance regime to become positive. It is
obvious that this positive resistance would prevent useful device operation, despite the fact
that the difference between the peak and valley current is exceptionally high.
E
u
a-
m
<
>.
°35
c
o
a
c
s
O
Figure 7. The experimentally measured dc-IV curve for an
InGaAs/AlAs DBRTD. The structure is similar to Figure 3
except that the total spacer layer to the right of the quantum
well is 1250A.
The current density where space charge resistance becomes important depends on
two factors. The first is the length of the spacer regions. As the length is decreased, the
total number of carriers within the region is also reduced. The price paid for this reduction
is a corresponding decrease in the overall device impedance. Furthermore, it is doubtful
that the spacer region can be eliminated beyond a certain point, since there is always a
depletion region formed due to the high electric fields near the quantum well. This is even
true if the heavily doped contacts are immediately adjacent to the quantum well.
Third International Symposium on Space Terahertz Technology
Page 573
e"
2S0-;
u
jdftOX
cr
»
200 -
<
MOX
*-"
>.
150 ■;
«
30X
c
9
Q
100 ■;
**
20X
c
a>
h*
■
3
50-
10X
u
0-i
1
2 3
4
5 6
Voltage
Figure 8. The simulated effect of the space charge resistance
on the dc-/V curve. These curves were generated using the
Schrodinger/Drift Diffusion composite model in which the
current given by Figure 7 was scaled upward 10, 20, 30,
40, and 50 times.
The second factor dictating when the space charge resistance becomes important is
determined by the velocity of the carriers. Higher velocities result in lower electron
concentrations, since total current must be conserved. It is in predicting these
concentrations and velocities that previous quantum mechanical models have failed, thus
failing to predict the importance of the spacer regions in overall low and high frequency
device behavior. Thus, in order to accurately project the ultimate performance of DBRTDs,
it is critical to use a quantum kinetic formalism such as the Lattice Wigner function. Future
results using this model should lead to a determination of the behavior of DBRTDs at
terahertz frequencies.
This work has been supported by the Texas Advance Research Program.
Page 574 Third International Symposium on Space Terahertz Technology
5.0 References
1 . T. C. L. G. Sollner, W. D. Goodhue, P. E. Tannenwald, C. D. Parker, and D. D.
Peck, Applied Physics Letters, 43, 588, (1983).
2. E. R. Brown, J. R. Soderstrom, C. D. Parker, L. J. Mahoney, K. M. Molvar, and
T. C. McGill, Applied Physics Letters, 58, 2291, (1991).
3. D. Z. Y. Ting, M. K. Jackson, D. H. Chow, J. R. Soderstrom, D. A. Collins, and
T. C. McGill, Solid State Electronics, 32, 1513, (1989).
4. E. T. Yu, M. K. Jackson, and T. C. McGill, Applied Physics Letters, 55, 744,
(1989).
5. W. R. Frensley, Physical Review B, 36, 1570, (1987).
6. K. L. Jensen and F. A. Buot, Journal of Applied Physics, 65, 5248, (1989).
7. N. C. Kluksdahl, A. M. Kriman, D. K. Ferry, Physical Review B, 39, 7720,
(1989).
8. D. R. Miller, V. K. Reddy, R. L. Rogers, and D. P. Neikirk, SPIE Proceedings
on High-Speed Electronics and Device Scaling, 1288, 167, March 18-19 1990.
9. K. K. Gullapalli, D. R. Miller, and D. P. Neikirk, 1991 International Electron
Devices Meeting, Washington D. C, December 8-11, pg. 51 1, 1991.
10. F. A. Buot, Physical Review A, 33, 2544 (1986).
11. D. R. Miller and D. P. Neikirk, Applied Physics Letters, 58, 2803, (1991)
12. K. L. Jensen and F. A. Buot, Journal of Applied Physics, 67, 7602, (1990)
13. W. Frensley, Proceedings of the International Symposium on Nanostructure
Physics and Fabrication, College Station, March 13-15, pg. 231, 1989.
14. C. Jacoboni and P. Lugli, The Monte Carlo Method for Semiconductor Device
Simulation . Springer- Verlag Wien, New York, 1989.
15. J. P. Aubert, J. C. Vaissiere, and J. P. Nougier, Journal of Applied Physics, 56,
1128,(1984)
16. W. R. Frensley, Applied Physics Letters, 51, 448, 1987.
17. S. M. Sze, Physics of Semiconductor Devices . 2nd Edition, Chapter 10, John
Wiley, New York, 1981.
18. D. R. Miller, V. P. Kesan, R. L Rogers, C. M. Maziar, D. P. Neikirk, The 13th
International Conference on Infrared and Millimeter Wave, 1988.
19 E. Wigner, Physical Review, 40, 749, (1932)
/'
Third International Symposium on Space Terahertz Technology Page 575
Parallel Arrays of Josephson Junctions for ^f£>-33
Submillimeter Local Oscillators i, ~ ,-
Aleksandar Pance* and Michael J. Wengler p ^
Department of Electrical Engineering INJQQ 9 *y ^ ry A
University of Rochester M & tJ m 7L i i t rt
Rochester, NY 14627
Abstract
In this paper we discuss the influence of the DC biasing circuit on operation of parallel
biased quasioptical Josephson junction oscillator arrays. Because of nonuniform distribution of
the DC biasing current along the length of the bias lines, there is a nonuniform distribution of
magnetic flux in superconducting loops connecting every two junctions of the array. These DC
self-field effects determine the states of the array. We present analysis and time-domain numerical
simulations of these states for four biasing configurations. We find conditions for the in-phase
states with maximum power output. We compare arrays with small and large inductances and
determine the low inductance limit for nearly-in-phase array operation. We show how arrays can
be steered in H-plane using the externally applied DC magnetic field.
~^""~-—-~- Introduction
The Josephson junction is a natural choice for submillimeter local oscillator since it is a
"voltage controlled oscillator" with typical voltage scales of mV and an oscillation frequency fj =
483 GHz per mV of dc bias. The existence of Josephson radiation into the terahertz range has
been demonstrated at Cornell [1]. A major disadvantage of the Josephson junction is its very low
output power. With DC voltage bias of 1 mV at 483 GHz, a junction which could accept 100 ^A
will put out less than 100 nW of RF power. Therefore, practical local oscillators must use arrays
of many junctions oscillating in phase. Submillimeter Josephson oscillator arrays with usable
power levels have been made at Stony Brook [2] and NIST [3].
We have proposed to build a large 2-D active grid array of parallel biased Josephson
junctions [4]. In our design, every junction drives a single antenna and the power from the whole
array is quasioptically combined. By biasing all junctions in parallel, we assure that all of them
* 1991 Link Energy Foundation Fellow.
Page 576 Third International Symposium on Space Terahertz Technology
radiate at exactly the same frequency. For maximum radiated power, all junctions must also be in
phase.
The DC biasing circuit of the 2-D quasioptical Josephson array plays a very important role
in phase-locking of Josephson junctions. In a two-dimensional array the DC biasing current is
supplied at the ends (Fig. 1). Because of that, the DC current is nonuniformly distributed along
the length of the biasing line. This current induces the nonuniform DC magnetic flux in
superconducting loops between every two neighboring junctions. Because of the superconducting
quantum interference effects [5], these self-induced fluxes determine the phase differences between
the neighboring junctions, and therefore the states of the array. These effects will be referred to as
the DC self-field effects. It is clear that, depending on the particular bias circuit, the in-phase state
can only be a special, rather than common state of the parallel 2-D Josephson junction array.
If the rows of the 2-D parallel Josephson array are biased independent from each other, the
DC self-field effects are, to the first order, limited to each row, and the whole 2-D array can be
looked at as a collection of 1-D parallel arrays. We will therefore investigate these DC self -field
effects in linear parallel arrays.
N-junction linear parallel array
The most general biasing scheme for the linear parallel array is presented in Fig. 2. We use
the RSJC model of the Josephson junction that consists of ideal Josephson junction, shunt
resistance and parasitic capacitance (Fig. 2). The ideal Josephson junction is described by
relationships between its current I, voltage V and phase difference <J) of the superconducting
quantum mechanical wave function between two sides of the junction
dd> 2e
I = Ic sin(<{)), f = f V
where Ic is the junction critical current. Assuming that all junctions are identical, the circuit from
Fig. 2 can be described with the following system of equations:
Third International Symposium on Space Terahertz Technology Page 577
ii(t) = ^<!>2(*H>i(t))+^-- — - y <Pex
ij(x) = f(<t> j+ ,(t)-2 ♦ J <tH> j . 1 (t)) + i , j = (2, N-l)
A. 2
W = - t<^H»n-iW) + v + T + T 1 *«
A. 2 2 A,
(La)
where <()j is the superconducting wave function phase difference across the j* junction, ij is the
total current through the j* junction, Yj, yl and Yr are biasing currents and <p e x is the normalized
externally applied DC magnetic flux,
ij(T) = p<t)(T) + <|>(t)+sin(<|)(x))
Yj=iuj + iDj»
"/L =1 Lin + ^out ' YR =1 Rin + ^R out
(Lb)
with capacitance and inductance parameters P and X, respectively, given by
.2
RT , L , ^o ^ h
*o=
■j 2tiL 2e
c
(Lc)
where Lj is the zero-bias Josephson junction inductance and Oq = 2.07 X 10* 15 Wb is the flux
quantum. In equations (La-c), time is given in units of t£, all currents in units of lc and
normalized junction voltages, that are just time derivatives of junction phases <t>j, in units of IcR.
In the case of 2-junction array, the in-phase solution has been reported by Ben- Jakob et al
[6]. Here we present solutions for several N-junction arrays with different biasing configurations:
the LL ("left"- "left"), LR ("left"-"right"), UD ("up"-"down") and CB ("central bias") biased array
(Fig. 3). Although the LL and LR bias are directly applicable to parallel biased two-dimensional
arrays, the other two configurations are used in other array architectures, such as series-parallel
combinations, etc.
Page 578 Third International Symposium on Space Terahertz Technology
In-phase states
The general solution of eq. (1) for the junction phases <()j is
<|) j (T) = ajT + f J <x) + <(> j (0)
CO = (<t>j(T))
f/T+T) = f /T), T= — , (ifx))=0
0)
(2)
where "< >" denotes time average, co is the normalized DC voltage across junctions, fj's arc some
general, periodic functions with zero time-average and <{>j(0) are constants. For the in-phase
solution, the following condition must hold for every two neighboring junctions
<t» j+1 (x)-<}) j (x) = 2 7tm j
where mj must be an integer. Note that mj represents the number of fluxons in the j* loop.
Condition (3) is fulfilled if
f j(t) = f (t)
j+1 CO = 4>j(t + tp
X > =m i T (4)
where Tj is the time delay between the phases of the two neighboring junctions. Substituting (3)
into the system (l.a) and equating all the currents ij leads solutions
m j = J a+ <P ex
(LL bias)
(5.a)
m j = - y ) a + cp ex
(LR bias)
(5.b)
m j =( Pex
(UD bias)
(5.c)
mj = ([y]-j)a+(p ex
(CB bias)
(5.d)
where "" in eq. 5.d denotes integer division, and parameter 'a' is given by
Third International Symposium on Space Terahertz Technology Page 579
X
J DC
a =
2tcN (6)
where ioc is me tota ^ DC biasing current. In order for mj to be an integer, which is the
precondition for the in-phase solution, it is necessary that both (p ex and a be integers:
<Pex= k q>
Xi DC .
a= =k
2tiN (7)
where k<p and k are integers. The only exception is the LR array with odd number of junctions N,
where "a" must be an even integer (2 k). The arrays will be in phase for all currents i^ that satisfy
, 2tcN
'k= k — ~
* (8)
Note that these in-phase states are achieved without external locking mechanisms.
Numerical simulations
System (1) is solved numerically using the 4th order Runge-Kutta method [7]. Figure 4
shows the I-V and dV/dl-I curves of the 4 junction LL biased array with X,=20 and P=0.5.
According to eq. (8), the in-phase states appear for current bias ik= 1.256 k (ik'=0.314 k for bias
current normalized to the number of junctions, N, as in the Figure 4). The in-phase states are
visible as voltage maximums in the I-V curve and sharp and deep minimums in dV/dl-I curve, for
k=4 to 7. Similar structure has been observed experimentally by Clarke et al [8].
Other states
The dV/dl-I curve of Figure 4 reveals considerable periodic structure between the in-phase
states. Under certain conditions, that will be specified below, these "other" states, for current bias
iDC * ik» correspond to the general solution of eq. (2) that satisfies the following:
ViW-fjM-e/T)
<t> j+1 (T) * *f x + T j )
1 J Vj (9)
Page 580 Third International Symposium on Space Terahertz Technology
where ej(x) is an error term and Hj does not need to be an integer. Furthermore, |0j is found from
the same equations as mj (5), except that a and cp ex are no more restricted to integer numbers. In
other words, all states of the parallel array are described with phases at neighboring junctions
shifted in time by an amount determined by the DC biasing current and external magnetic field (eq.
5). It is convenient to define the relative normalized time shift 8j between the waveforms of
functions fj+i and fj
6 j =Tj mod T = u,j mod 1
(10)
where "mod" is the modulus function, so that < 8j < 1. It has been shown by perturbation
analysis [6] that in the case of 2-junction array solution (9) holds in the neighborhood of the in-
phase state (ioc = *k + Aioc) and it has been suggested that it will hold for any state between the
in-phase states, for the case of weak coupling (k »1).
Figure 5.a shows the circled part of the dV/dl-I curve of Fig. 4. Points labeled "1" and "4"
correspond to the in-phase states with k=4 and k=5, respectively. The voltage waveforms on
individual junctions for these two states are shown in Figures 6.a and 7.a, respectively. Point "2"
of Fig. 5.a correspond to DC biasing current ioc = 5.65, so eq. (5.a) gives Hi=4.5, H2=5 and
|I3=5.5 for the number of fluxons in each loop. From eq. (10) we find that relative time shifts
should be 0i=O.5 between the voltages of the junctions 2 and 1, 82=0 for junctions 3 and 2 and
03=0.5 between junctions 4 and 3. Numerical simulations shown in Figure 6.b confirm this
prediction.
Point "3" of Fig. 5.a correspond to ioc = 5.42, and again from equations (5.a) and (10)
we obtain 9i=0.333, 82=0.666 and 83=0. The voltage (and phase) at junction 2 is time shifted by
third of a period from that of junction 1 , voltage at junction 3 is shifted by two-thirds from that of
junction 2, so that it is in phase with junction 1. Finally, junction 4 is in phase with junctions 1
and 3. This situation is shown in Figure 6.c. All other states can be determined in a similar
fashion.
Radiated power
As a measure of how good an array performs as an oscillator for a particular bias, we
calculate the available radiation power. We are interested in power array would radiate broadside
in the far-field. We define m* harmonic power on unit (1Q) resistance as
p(m) =(£Vj( m )) 2
Third International Symposium on Space Terahertz Technology Page 581
where Vj( m ) is m* harmonic voltage on j^ 1 junction. This power is given in units of (Ic R) 2 . We
assume that the resistance R of the RS JC model (Fig. 2) includes both the radiation resistance and
losses. The actual radiated power will at best be the power P( m ) on resistance -j.
Figure 5.b shows the normalized first harmonic power radiated in the broadside direction
for different states of the array. The maximum power is obtained only in the in-phase states
(points "1" and "4"). Significant amount of power is also obtained in states where most of the
array works in phase, as is the case with state "2".
Array properties
Several important properties of arrays of Fig. 3 can be derived from equations (5-10):
1 ) In the absence of external magnetic field:
- the UD array will be in phase for any DC bias; the in-phase state is a natural one
for this array. Maximum power will be radiated at every operating frequency (Fig.
5.b).
- LR and CB arrays are symmetrical around the middle of the array; mN-j = -mj,
which means that the junctions j and N-j are always in phase.
2) The LR array is equivalent to the LL array with equivalent external magnetic field
N
<Pex = <Pex - y a.
3) The larger the inductance parameter X , the more in-phase states will be found in the
given current bias span (eq. 8), and corresponding DC voltage and operating frequencies
span. Similarly, the larger the array (N), the more identifiable "other" states will be found
in the dV/dl-I curve (Fig. 5.a).
Magnetically steerable array
When an array is biased in the in-phase state (irx: = ik) the normalized relative "time shift
between every two neighboring junctions is the same and proportional to the external DC magnetic
field:
e j =e = q> ex modl, Vj
(11)
Page 582 Third International Symposium on Space Terahertz Technology
This situation is shown in Figure 7. In Fig. 7.a the LL array is biased at the in-phase state (point
"4", Fig. 5.a) with no external magnetic field. When an external magnetic field equal to a quarter
of the flux quantum is applied, the time shift between the voltages of every two neighbors is equal
to a quarter of period.
The time shift 0T translates into the linear phase shift in the frequency domain 2tc0.
Assuming that every junction drives one antenna, the quasioptical Josephson array becomes a
phased array [9]. The angular position ceo of the main beam in the H-plane far-field radiation
pattern of the linear Josephson phased array becomes
,,2710 9
cos(ao) = -- ="d^o
2 it , a x
_d x
*o (12)
where dx is the spacing between the antennas and Xo is the free space wavelength . The broadside
radiation corresponds to oco = 90° (0 = 0). Equations (11,12) suggest that by changing the
externally applied DC magnetic field cp e x it is possible to steer the Josephson array radiation pattern
in the H-plane. As stated earlier, for the LL, LR and CB biased array, this is only true if the array
is biased in the in-phase state. Since the UD array is always in the in-phase state, it can be steered
using DC magnetic field at any bias.
Limitations
The expressions (9,10) derived for the "other" states will hold only in certain range of array
parameters and bias conditions. We have derived expressions (5) for the in-phase states starting
from (1) and assuming that all currents ij are equal These expressions always hold for the in-
phase states. The same expressions (5) are found for "other" states if we solve (1) with an
assumption that DC currents <ij> are approximately the same. The only part of the DC junction
current that is different at every junction, due to DC self-field effects, is the supercurrent
<sin(<j)j)>. This part will be negligible if either the biasing current per junction is -^» 1 or if
there is non-vanishing capacitance (P >1).
In our account of DC self-field effects we assumed the noiseless environment and the
identical junctions. Therefore, the stability of in-phase and "other" states to noise and variations in
junction and array parameters remains to be further investigated.
Third International Symposium on Space Terahertz Technology Page 583
Strongly coupled arrays (X < 1)
When the inductance parameter X is small, it is evident from eq. (8) that the first in-phase
state appears for very large DC biasing current, which translates into large DC voltage and
operating frequency much above the critical frequency co c = (2e,#) Ic R. Depending on the
capacitance parameter f$ and shunt resistance R, the operating frequency range is at best of the
order of several co c . Therefore, the arrays with small inductances are operated in "other" states
throughout the operating range. According to eq. (5), these states should be characterized with
small time shifts between the junction phases/voltages. This is obvious, because in the limit of X -
> the whole array operates as a single junction.
Figure 8.a presents simulations for 4 junction LL biased array with small inductance (X =
0.628). The normalized harmonic radiated power is shown in the wide range of bias currents.
The bias points a=0.25 and a=0.5 with no radiated first harmonic power correspond to the states
where half of the junctions are in-phase and the other half out-of-phase, according to (5.a). As
seen from the Figure 8.b, the maximum first harmonic power is below that of the UD array, with
all junctions in-phase. Figure 9 compares the first harmonic power of 4- and 5-junction array with
Same parameters. The 5-junction array shows additional minimums in radiated power
corresponding to states a=0.125 k. These minimums occur whenever there is one or more loops
of array occupied by odd number of half-flux-quantums.
It is clear that in order for the array with small inductance to approach the performance of
the UD array in the wide operating range, the condition for the inductance parameter must be
X«tj. More precisely, if the total time shift across the array is required to be less than a quarter of
a period, the condition is:
X iDCmax^j^y ( 13 )
where lDCmax is the DC bias at the end of the operating range. So, if we wanted a 4-junction LL
biased array to approximately match the performance of UD array in Fig. 8.b, the inductance
parameter should have been X = t™ instead of ■?. Such small inductances are rather unrealistic,
specially because the X parameter is proportional to the critical current Ic (eq. 1) which should be as
large as possible for large output power.
As a final illustration, Figure 10 shows the influence of the Josephson junction capacitance.
Page 584 Third International Symposium on Space Terahertz Technology
The capacitance does not influence the occurrence or existence of in-phase and "other" states.
However, it has a severe impact on radiated power. Even at not very big capacitance (|3 = 3) the
first harmonic power is decreased at least an order of magnitude and the operating range is reduced
below 2 cflfc.compared to the case of very small or no capacitance.
Conclusion
We have discussed how the DC biasing circuit determines the operation of linear parallel
quasioptical Josephson junction arrays. We have shown that the maximum radiated power from
the array can be achieved only at certain operating points, corresponding to the in-phase states. We
have found that other states can be described by time-shifted phases and voltages of individual
junctions, where the time-shift is determined from the DC biasing conditions. We have shown
how the array can be steered from when in the in-phase state by application of DC magnetic field
perpendicular to the array.
When the inductance parameter X is large, there will be numerous in-phase bias points in
the desired operating range. However, the stability of these states to noise and variations in
junction parameters needs to be further investigated. When the inductance is relatively small, the
radiated power will continuously change across the wide operating range, with several points
where almost no power is radiated.
If one dimensional quasioptical arrays are designed, the UD biased array is a definite
choice, because it is in the in-phase state at every bias point. The operation of this array need to be
further analyzed when junction parameters are not identical. The extension of our considerations to
2-D arrays is straightforward, as long as rows of junctions are separately biased.
Acknowledgment
This work is supported by the Air Force Office of Scientific Research grant AFOSR-90-
0233.
References
1. Robertazzi, R.P. and R.A. Buhrman, Josephson terahertz local oscillator. IEEE
Trans. Magn., 1989. 25: p. 1384-1387.
2. Wan, K., A.K. Jain and J.E. Lukens, Submillimeter wave generation using
Josephson junction arrays. Appl. Phys. Lett., 1989. 54: p. 1805-1807.
Third International Symposium on Space Terahertz Technology Page 585
3. Benz, S.P. and C.J. Burroughs, Coherent emission from two-dimensional
Josephson junction arrays. Appl. Phys. Lett., 1991. 58(19): p. 2162-2164.
4. Wengler, M.J., A. Pance, B. Liu and R.E. Miller, Quasioptical Josephson
Oscillator. DEEE Trans. Magn., 1991. 27(2): p. 2708-2711.
5. Van Duzer, T. and C.W. Turner, Principles of Supeconductive Devices and
Circuits. 1981, New York: Elsevier.
6. Ben-Jacob, E. and Y. Imry, Dynamics of the DC-SQUID. J. Appl. Phys., 1981.
52(11): p. 6806-6815.
7. Press, W.H., B.P. Flannery, S.A. Teukolsky and W.T. Vetterling, Numerical
Recipes in C: The Art of Scientific Computing. 1988, Cambridge, Mass.: Cambridge
University Press.
8. Clarke, J. and T.A. Fulton, Origin of Low-Voltage Structure and Asymmetry in the
I-V Characteristics of Multiply-Connected Superconducting Junctions. J. Appl. Phys.,
1969. 40(1.1): p. 4470-4476.
9. Steinberg, B.D., Principles of Aperture and Array System Design. 1976, New
York: John Wiley & Sons, Inc.
-. '5-'
Page 586
Third International Symposium on Space Terahertz Technology
Idc
+
Idc
+
X )( X X )(
Idc
Idc
+
X X
X X
Idc
X X X X X
Idc
Figure 1. Separate row bias for parallel 2-D Josephson junction array.
Third International Symposium on Space Terahertz Technology
Page 587
out
Figure 2. General biasing scheme for one-dimensional parallel Josephson junction array.
The RSJC model used is shown below.
Page 588
Third International Symposium on Space Terahertz Technology
+
Idc
Idc XN X 2X lXL-Lbias
a)
+
L-R bias
+
X X- -
N N-l
UIdc
\
72- %
'2 i
X CB bias
^Idc
b)
c)
d)
Figure 3. Four common biasing configurations of 1-D array.
Third International Symposium on Space Terahertz Technology
Page 589
>
1-
k=4
h
k=5
k=4
1 2
Voltage [IcR]
k=5 k=6
k=7
1 2
Current [N Ic]
Figure 4. Dynamic resistance and I-V curve of 4 junction LL biased array with X = 20 and
j} = 0.5. The in-phase states are seen as small steps in the I-V curve and sharp minimums
in dV/dl-I curve , labeled k=4 to k=7. The area inside a circle is shown enlarged in Fig. 5.
Third International Symposium on Space Terahertz Technology
5q
<? 4.5-
& a'-
— 3.5^
O
a<
o
•^■«
c
o
DC Voltage [Ic R]
a)
J-i
2.5 -i
2-.
1.5-i
1 J
0.5-
/
UD array
0.8
0.9
| i 1 1 \ m\ 1 1 1 i ■"] r— i— r— i [—
1 1.1 1.2 1.3
i i i
1.4
Dc Voltage [Ic R]
b)
Figure 5. Four junction array, LL bias, X = 20, p = 0.5. a) Enlarged portion of the dV/dl-
I curve (Fig. 4) with in-phase states labeled "1" and "4" and two "other" states "2" and "3".
The waveforms of individual junction voltages for these states are shown in Fig. 6 and 7.
b) First harmonic power that correspond to states in a). The power is maximum in the in-
phase states and equal to that of the equivalent UD array.
Third International Symposium on Space Terahertz Technology
Page 591
240
time [Lj/R]
a)
240
time [Lj/R]
250
b)
240
time [Lj/R]
c)
Figure 6. The waveforms of individual junction voltages for the states of Fig. 5. a) In-
phase state, b) Junctions 1 &4 in phase, junctions 2 & 3 in phase , but out of phase with
(1&4). No power radiated, c) Junctions 1,3 & 4 in phase, junction 2 "leads" a third of a
period. Half the maximum power is radiated.
Page 592
Third International Symposium on Space Terahertz Technology
Z.3-
©A
2-
ST
a 1 - 5 -
>
0.5-
#4
#2
(pex=0
o :
#3 -
■-• #1
1 1 '
230
240
time [Lj/R]
250
a)
240
time [Lj/R]
250
b)
Figure 7. The waveforms of individual junction voltages for the in-phase state labeled "4"
in Fig. 5. a) No DC magnetic field supplied, b) Quarter of the flux quantum in every
loop supplied by the external magnetic field. Voltage waveforms uniformly shifted by
quarter of a period. The main beam of the radiation is steered from the broadside direction.
Third International Symposium on Space Terahertz Technology
Page 593
LL bias
<
0)
o
o
'2
o
0.000001
I I I I
0.3
a [FiO]
0.6
12 3 4 5 6
Vdc [IcR]
Figure 8. a) Hannonic power radiated by the 4-junction LL-biased array with X = n/5 and
(} = 0.03. Minimums correspond to odd numbeT of half-flux-quantums in some of the
loops, b) Comparison between the LL and UD array with same parameters. Varying
amount of power is radiated in the very broad operating range, but never a maximum
possible power, as in the case of UD array.
Page 594
Third International Symposium on Space Terahertz Technology
4.5 -r
cs
4-
<
^^*\
.
3.5 r
HH
■
"W
■
•— '
V
I*
<u
_
o
2.5^
D.
.
2-
e
-
o
1.5-
ed
XI
1-
•*-»
.
C/5
-
0.5^
4 junctions
5 junctions
A
! \
o
i i i i i i i i i i
Vdc [IcR]
Figure 9. Comparison between the 4-junction and 5-junction small inductance arrays; \ =
7i/5, P = 0.03. As the number of junctions increase, more maximums and minimums
appear throughout the operating range.
10rr
<
u
beta=3.00
- •• - beta=0.03
2 2.5
DC Voltage [Ic R]
i i i i i i
3.5
Figure 10. Four junction LL biased array, X = jc/10: Influence of the Josephson junction
capacitance on power output. As the capacitance is increased, the maximum power and the
operating range rapidly decreases.
Third International Symposium on Space Terahertz Technology Page 595
N 9 3 - 2fr ^ 5
MONOLITHIC MILLIMETER-WAVE DIODE GRID FREQUENCY
MULTIPLIER ARRAYS
Hong-Xia L. Liu, X-H. Qin, L. B. Sjogren, W. Wu, E. Chung,
C. W. Domier, N. C. Luhmann, Jr.,
Center for High Frequency Electronics
Department of Electrical Engineering
University of California
Los Angeles, California 90024-1594
ABSTRACT
Monolithic diode frequency multiplier arrays, including barrier-N-N + (BNN) dou-
bler, multi-quantum-barrier- varactor (MQBV) tripler, Schottky-quantum-barrier-varactor
(SQBV) tripler, and resonant-tunneling-diode (RTD) tripler arrays, have been suc-
cessfully fabricated with yields between 85% and 99%. Frequency doubling and/or
tripling have been observed for all the arrays. Output powers of 2.4-2.6 W (7?=10-
18%) at 66 GHz with the BNN doubler and 3.8-10 W ( 77=1.7-4%) at 99 GHz with
the SQBV tripler have been achieved.
INTRODUCTION
Quasi-optical spatial power combining techniques have, in recent years, been ex-
tensively investigated for millimeter and submillimeter power generation [1] [2] [3].
Our research efforts have been focused on monolithically integrating thousands of
solid state devices to generate Watt level harmonics in the millimeter region. In
addition to arrays of familiar devices such as the BNN and RTD, several new de-
vice concepts (MQBV and SQBV [4] [5]) have been developed in the course of this
work which promise to significantly improve the performance of frequency multipliers.
Arrays of all of these devices have been successfully fabricated and tested. Several
exciting results have been obtained. Improvements both in device design and in the
matching system are underway to further optimize array performance.
Page 596
Third International Symposium on Space Terahertz Technology
FABRICATION AND RESULTS
(a) BNN frequency doubler array
A four-mask process based on the self-aligned aluminum Schottky diode process
employed by C. Zah [6] is utilized to fabricate the BNN doubler array. Figure 1
shows an individual BNN device after fabrication. The array was then mounted on a
quarter wavelength thick quartz plate. No bias is required due to the adjustment of
the build-in voltage resulting from the 6 doped layer (see Fig.2). The test system for
the BNN frequency doubler is shown in Fig.3 except that the output filter was not
used. Cutoff waveguides for the fundamental have been used to prevent contamination
of the detected signal due to the pump signal. A variety of tests were employed
to conclusively verify that the received signal was actually frequency doubled. All
waveguide components including attenuators have been calibrated using at least two
methods. The input and receiving horns have also been calibrated and compared
with the theoretical values. Figure 4 displays the measured RF results. An output
power of 2.4-2.6 W and a maximum efficiency of 10-18% have been achieved. The
calculated cutoff frequency based on the measured low frequency parameters of this
array is 280 GHz, which results in a maximum theoretical conversion efficiency of
~20%.
5|im
Figure 1. The fabricated individual BNN diode.
f*t t *t t M"t > *y*? w ?*f w W**t^
Semi-insulating GaAs substrate
2000A Al
; 200A GaAs undoped
* 50A n GaAs 2x10 cm'
1200A n GaAs 6xlo"cni
. it •>
1.3jim n GaAs 4x10 cm
Figure 2. The profile of the BNN diode.
Third International Symposium on Space Terahertz Technology
Page 597
2 5
-e Output Power- — »•-•
Efficiency
1-7
$ 2
_i i i i 1 i i i i | i i i i | i i i i | i i i
'- ° o © /
't>
■©
10
% 1-5
o
t 1
%
6 °- 5
o
ft. > X5
, 1 . .
8 g
6 1
4 9
2 w
n
5 10 15 20
Input Power (W)
25
30
Figure 3. The frequency multiplier setup.
Figure 4. Measured frequency doubling output power
and efficiency at 66 GHz as the function of
the input power.
(b) Frequency tripler arrays
MQBV, SQBV and RTD arrays have been successfully fabricated using a back-
to-back processing method [7]. As a result, all these arrays are suitable for odd
harmonic generation due to the resulting symmetric C-V characteristics. Figure 5
displays the array layout. The period of each cell is 400 /ttn. There are 2250 devices
on the MQBV array, 1300 devices on the SQBV array, and 500 devices on the RTD
array, respectively. These arrays are tested using the system shown in Fig. 3. Since
the input is identical for both the doubler and tripler arrays, the same input filter
and tuning slabs have been used. The output filter for the frequency tripler is ah
inductive metal grid array [8], and quartz tuning slabs with a thickness of a quarter
wavelength at the tripled output have been used for the output impedence matching.
Figure 6 shows the measured output power and efficiency of the SQBV array at an
output frequency of 99 GHz. An output power of 3.8-10 W and an efficiency of 1.7-4%
have been achieved. Due to excessive pumping in the initial tests, the performance of
the MQBV array was degraded significantly (f c dropped from 550 GHz to 100 GHz).
However, an output power of 0.1 W and an efficiency of 0.4% have been obtained for
the degraded array which is in good agreement with the theoretical prediction (0.5%).
Finally, a frequency tripling signal has also been observed with the RTD array. Tests
are underway to measure the output power and efficiency as the function of the input
power.
Page 598
Third International Symposium on Space Terahertz Technology
< MQBVr
' 400|i
I
if"
urn
4
3.5
3
— Output power
(W) Efficiency (%)
1.8
1.6
1.4
*
4
1 | 1 1 ! 1 | 1 1 1 1 | 1. 1 1 1
2.5
/
^/
2
1.2
1
1.5
1
i X
-
0.8
0.5
I
/
0.6
50 100
150 200 250 300
Input
power (W)
Figure 5. The MQBV and SQB V arrays layout.
Figure 6. Measured frequency tripling output power
and efficiency at 99 GHz as the function of
the input power.
CONCLUSIONS
BNN, MQBV, SQBV, and RTD frequency multiplier arrays have been successfully
fabricated with yields between 85 and 99%. All of these arrays have yielded frequency
multiplication. An output power of 2.4-2.6 W with maximum efficiency of 10-18%
has been achieved at 66 GHz with the BNN doubler array; an output power of 3.8-10
W with a maximum efficiency of 1.7-4% has been achieved at 99 GHz with the SQBV
array.
ACKNOWLEDGEMENTS
This work was supported by the US Army Research Office and the US Department
of Energy. The authors wish to thank Dr. J. Maserjian and P. Smith of Jet Proposion
Lab. for providing the processing facilities and the MBE wafers. The authors also
wish to acknowledge the generous assistance of Dr. R. Bhat and Dr. L. Florez
of Bellcore, Professor M. Spencer of Howard University, and Dr. A. Miura of the
Yokogawa Electric Corp. in providing MBE and MOCVD wafers for these studies.
Third International Symposium on Space Terahertz Technology Page 599
REFERENCE
1. C.F. Jou, W.W. Lam, H.Z. Chen, K.S. Stolt, N.C. Luhmann,Jr., and D.B. Rutledge,
"Millimeter Wave Diode-grid Frequency Doubler," IEEE Trans, on Microwave Theory
and Techniques, 36, No. 11,1 988.
2. W.W. Lam, C.F. Jou, N.C. Luhmannjr., and D.B. Rutledge, "Millimeter- W ave Diode-
grid Phase Shifters," IEEE Transactions on Microwave Theory and Techniques 36, No.
5, pp. 902, 1988.
3. Z.B. Popovic, R.M. Weikle, M. Kim, K.A. Potter, and D.B. Rutledge, "Bar Grid
Oscillators," IEEE Transactions on Microwave Theory and Techniques , 38 No. 3,
pp. 225, 1990.
4. Hong-Xia L. King, L.B. Sjogren, and N.C. Luhmann,Jr., "New Concepts for High
Frequency and High Power Frequency Multipliers and Their Impact on Quasi- Optical
Monolithic Array Design", International Journal of Infared and Millimeter Waves,
13, pp.251, 1992.
5. Hong-Xia L. King, N.C. Luhmannjr., X-H. Qin, L.B. Sjogren, W. Wu, D.B. Rutledge,
J. Maserjian, U. Lieneweg, C. Zah, and R. Bhat, "Millimeter Wave Quasi- optical
Active Arrays", Proc. and Conference on Space Terahertz Technology, pp. 293-305,
Feb. 1991.
6. C. Zah, D.P. Kasilingam, J.S. Smith, D.B. Rutledge, T. Wang, and S.E. Schwartz, "Millimeter-
wave Monolithic Schottky Diode Imaging Arrays", Int. J. of Infrared and Millimeter
Waves, 6, pp. 981-997, 1985.
7. R.J. Hwu, C.F. Jou, N.C. Luhmann, Jr., M. Kim, W.W. Lam, Z.B. Popovic, D.B.
Rutledge, "Array Concepts for Solid-State and Vacuum Microelectronics Millimeter-
Wave Generation," IEEE Trans, on Elec. Dev., 36, No. 11, pp. 2645-2650, 1989. '
8. Hong-Xia L. Liu, L.B. Sjogren, and N.C. Luhmann,Jr.,"Grid Bandpass Filters for
Quasi- Optical Frequency Multiplier Array Application", submitted for publication in
Microwave and Optical Technology Letters, 1992.
Page 600 Third International Symposium on Space Terahertz Technology
PLANAR GaAs DIODES FOR THz FREQUENCY MIXING APPLICATIONS
William L. Bishop, Thomas W. Crowe, Robert J. Mattauch, and Hasan Dossal
Semiconductor Device Laboratory
Department of Electrical Engineering _ ^ yy ry (»
Thornton Hall f| 9 3 " * • f
University of Virginia
Charlottesville, VA 22903-2442
I. Introduction
For many scientific applications in the terahertz frequency range, heterodyne reception
is the only technique which exhibits the necessary combination of high spectral resolution,
large instantaneous bandwidth and excellent sensitivity. A key component in these receivers
is the non-linear resistive mixer element. In general, the mixer element should have high
intrinsic speed, the sharpest possible non-linearity, low parasitic element values, low intrinsic
noise and impedance levels which can be easily matched to the RF circuit. However, no
single device exhibits all of these properties and some tradeoffs are necessary [1].
The GaAs Schottky barrier diode is the most widely used mixer element at
submillimeter wavelengths. These diodes are commonly used in the temperature range from
300 K to 10 K and have demonstrated excellent performance from below 100 GHz to over
3 THz [2,3]. The closest competitor for Schottky diodes is the SIS element which has
demonstrated record sensitivities at millimeter and long-submillimeter wavelengths [4,5,6].
However, SIS devices are not yet competitive at terahertz frequencies and present
superconductor mixer elements require cryogenic cooling which increases the cost and size
of the receiver system.
Third International Symposium on Space Terahertz Technology Page 601
Schottky barrier diodes for terahertz applications are typically fabricated as a micron
to sub-micron circular anode metallization on GaAs which is contacted with a sharp wire
(whisker). This structure has the benefits of the simplicity of the fabrication of the diode chip,
the minimal shunt capacitance of the whisker contact and the ability of the whisker wire to
couple energy to the diode. However, whisker-contacted diodes are costly to assemble and
difficult to qualify for space applications. Also, complex receiver systems which require many
diodes are difficult to assemble [7,8].
The objective of this paper is to discuss the advantages of planar Schottky diodes for
high frequency receiver applications and to summarize the problems of advancing the planar
technology to the terahertz frequency range. Section II will discuss the structure, fabrication
and performance of state-of-the-art planar Schottky diodes. In Section m the problems of
designing and fabricating planar diodes for terahertz frequency operation are discussed along
with a number of viable solutions. Section IV summarizes the need for futher research and
cooperation between diode designers and RF engineers.
II. Planar Mixer Diodes
Planar Schottky barrier diodes have been developed by numerous laboratories over the
past ten years [9,10,11,12]. This effort has resulted in many benefits. Not only has the
troublesome and somewhat fragile whisker contact been eliminated, but receivers which
require two or more individual diode chips, such as balanced mixers, are much easier to
assemble. Single chips with two or more diodes in a fixed configuration, such as an
antiparallel diode pair, are easy to fabricate and the extension of the diode contact pads to
form a planar antenna has been demonstrated. Future work should allow additional receiver
components such as filters, oscillators and amplifiers to be integrated with the diode.
Page 602 Third International Symposium on Space Terahertz Technology
The surface channel planar diode, shown in Figs. 1 and 2, has been developed for use
at both millimeter and submillimeter wavelengths [9,13]. The chip substrate is semi-insulating
GaAs. The epitaxial GaAs structure consists of a thin n-type layer on top of a thick, heavily
doped n+ buffer layer. The anode is formed on the n-type GaAs with Si0 2 providing
passivation and insulation. An ohmic cathode pad is formed on one end of the chip in close
proximity to the anode. The anode is connected to a bonding pad by means of a narrow
finger. A trench is formed beneath the finger and completely across the width of the chip to
isolate the anode contact pad from the cathode. The isolation trench can be etched deeply into
the semi-insulating substrate and the wall of this trench can be positioned very close to the
anode. These two features combine with the inherent air-bridge to reduce the shunt
capacitance between the contact pads and the shunt capacitance from the contact finger to the
conductive GaAs of the cathode. This structure produces lower shunt capacitance than other
designs which rely on mesa or proton isolation.
The major fabrication steps of the surface channel structure are illustrated in Fig. 3.
Starting with the GaAs wafer (1), a layer of silicon dioxide is deposited using chemical vapor
deposition from silane and oxygen (2). The ohmic contact region is patterned, the Si0 2 and
n-GaAs are removed and the ohmic contact metallization is deposited and alloyed (3). An
opening for the anode is patterned and etched into the Si0 2 , leaving a thin layer of oxide to
protect the GaAs until the anode metallization can be deposited. The remaining oxide in the
anode well is removed with buffered hydrofluoric acid and platinum and gold are
electroplated to form the diode and fill the oxide well (4). A thin layer of chromium and gold
is deposited over the entire wafer by sputtering. Photoresist is applied and patterned and gold
is plated into the opening to form the anode contact pad and finger. The resist is removed and
the sputter deposited gold and chromium surrounding the anode contact pad and finger are
Third International Symposium on Space Terahertz Technology
Page 603
^^^ Ohmic Contact
Surface Channel
Air Bridge Finger
Anode Contact
Pad
Anode
(beneath finger)
Semi — Insulating
GaAs Substrate
Figure 1. Surface Channel Planar Diode Structure
Figure 2. SEM Photographs of a Surface Channel Planar Diode
Page 604
Third International Symposium on Space Terahertz Technology
1. n / n+ / S.I. GaAs
n GaAs
n
n +
S.I.
GaAs
2. Deposit Silicon Dioxide
\/////////7////////////////////////A
n +
S.I. GaAs
3. Forn Dhnic Contact
'////////////////////////,
D.C.
n
n +
S.I. GaAs
4, Form Anode
'/////////////////////Wy
D.C.
n
n +
S.I. GaAs
5. Forn Anode Pad/Finger
Anode
Pad V y///// //A
|SiD 2 |
Dhnic
Contact
'///////////////7/////M.
Anode Pad/Finger
' r inqs
7 777?
n +
D.C.
S.I. GaAs
6. Etch Surface Channel
Anode
Dhnic
Contact
V7
Pad
S.I.
GaAs
Anode Pad/Finqer
V////////777\
Air
n +
i
D.C.
S.I. GaAs
Figure 3. Surface Channel Diode Fabrication Sequence
Third International Symposium on Space Terahertz Technology
Page 605
etched away (5). Finally, the surface channel is patterned with photoresist and the Si0 2 and
GaAs are etched to form the isolating trench (6).
This fabrication sequence offers several advantages compared to other configurations:
(1) expensive and troublesome proton bombardment is not required, (2) planarization is
unnecessary, and (3) the wafer surface is nearly flat for the critical steps of anode formation
and anode-to-finger alignment.
SEM photographs of two surface channel diode chips are shown in Figs. 4 and 5. The
SC2T1 single anode chip is about 125 x 375 x 75 microns. This device has a total
capacitance of about 14 fF, zero-bias junction capacitance of 2.5 fF and series resistance of
12-15 Q. This gives a figure-of-merit cutoff frequency of 4.2 THz for the junction. The
SC2T1 has been tested in a room temperature mixer at 345 GHz with a mixer noise
temperature of 1,370 K (DSB) and a conversion loss of 9.5 dB (SSB) [14]. This is
comparable to the best whisker-contacted diode results. The SC1T4 chip is an antiparallel
diode pair for subharmonic pumping. It is only 80 x 180 x 50 microns. These chips have a
Figure 4. SC2T1 Planar Diode Chip Figure 5. SC1T4 Planar Antiparallel
Diode Pair
Page 606
Third International Symposium on Space Terahertz Technology
total capacitance of about 16 fF, zero-bias junction capacitance of 3 fF per anode and series
resistance of 7-9 Q. This diode has been successfully used in a room temperature mixer at
205 GHz with a mixer noise temperature of 800 K (DSB) and a conversion loss of 4.4 dB
(DSB) using an LO of approximately 100 GHz [15]. This result is better than has been
previously reported for antiparallel subharmonic mixers of either planar or whisker-contacted
design.
A dual anode planar diode chip for balanced mixer operation is shown in Fig. 6. This
chip was developed in collaboration with Aerojet General, Electronic Systems Division under
the direction of Robert Haas. This configuration allows individual DC bias of each diode.
This device has excellent DC electrical characteristics and is being evaluated in a 100 GHz
mixer.
Figure 6. Planar Balanced
Mixer Diode Chip
Third International Symposium on Space Terahertz Technology Page 607
HI. Planar Diodes for THz Frequency Applications
The surface channel diode structure must be optimized for terahertz operation. These
improvements reflect the fundamental need to reduce the RsC j0 product, minimize shunt
capacitance, and to efficiently couple energy into the diode. These optimization issues are
addressed in the following subsections:
A. Reduction of Anode Diameter
Theory and experimental results with whisker-contacted diodes have shown that very
small anodes combined with higher active layer doping are necessary for good performance
in the THz range [16]. Whisker-contacted diode chips have been fabricated at UVa with
anodes as small as 0.25 microns using direct write electron beam lithography and reactive ion
etching [17]. Planar diodes have been fabricated at UVa with 0.5 micron diameter, anodes
using optical lithography and reactive ion etching. We are also investigating a novel
Electroplate Window Shrink (EWS) technique. In this method, circular openings are etched
through a thin (0. 1 micron) metal layer which overlies silicon dioxide, using UV lithography
and wet or dry etch methods. Metal is then electroplated onto this thin conductive layer.
Since the plating proceeds laterally as well as vertically, the diameter of the openings is
reduced. These reduced-diameter windows are then used as a non-eroding mask to RIE etch
the silicon dioxide. Etched wells less than 0.2 microns in diameter have been formed in this
manner.
It should be realized that the main issue is not just the fundamental task of forming
small anode wells, but also the problems of uniformity and control of anode size. The UVa
anode formation process depends on leaving a thin layer of Si0 2 of known thickness in the
bottom of the anode wells after RIE. This protective layer is removed by etching with
Page 608 Third International Symposium on Space Terahertz Technology
■ buffered hydofluoric acid just prior to anode formation. Underetching of this remaining oxide
results in open circuits or high resistance. Overetching can result in high Cj and in some
cases, excessive diode noise [18]. Unlike whisker-contacted diode chips which can be etched
and plated on a chip-by-chip basis, all planar diode anodes on a wafer are formed
simultaneously. This obviously places very tight limits on dielectric thickness, thickness
uniformity and etch rate calibration.
For these reasons, it would be most helpful to have a very thin RE etch stop layer
to protect the GaAs. This etch stop layer would relax the requirements for oxide thickness
and uniformity and allow reasonable overetching during RIE without the risk of damage or
contamination of the junction area. Schemes which utilize mutiple layers of different
dielectrics could, in principle, satisfy this need. A very thin layer (100-500 A) of silicon
dioxide could first be deposited onto the GaAs. This would be followed with a thicker layer
of another dielectric, such as silicon nitride, polyimide or boron-doped silicon dioxide. This
thick layer could be patterned and selectively etched (possibly with a dielectric or metal
mask) so that the underlying thin layer of oxide acts as an RIE etch stop. Research in this
important fabrication area will provide improved control of anode diameter and the reliable
production of sub-half micron planar diode anodes.
B. Optimization of Chip Geometry
The dimensions and layout of planar diode chips must be optimized for terahertz
frequency applications. The volume of the chip must be reduced to minimize the field
disturbing effect of high dielectric constant GaAs and to allow the devices to fit into the
smaller waveguides which are required at higher frequencies. The geometry of the planar
diode must be improved to minimize shunt capacitance.
Third International Symposium on Space Terahertz Technology
Page 609
Shunt capacitance in the planar diode structure can be separated into two primary
components: capacitance from the anode contact finger and pad-to-pad capacitance through
the high dielectric constant substrate. Finger capacitance will be reduced by several means.
The width of the contact finger can be reduced from the current value of about 2.5 microns
to 1 micron. Improved mask design, alignment and surface channel etch control will allow
the surface channel wall to be etched as close as possible to the anode. A thick (1 micron)
dielectric, perhaps a polyimide, would further reduce finger capacitance.
Pad-to-pad shunt capacitance can be reduced by decreasing pad area, increasing pad
separation, increasing the surface channel depth and/or reducing substrate thickness. Our
present technology produces chips which are 50 microns thick with pads which are 30 x 60
microns and a surface channel depth of 10 microns. For the lowest possible pad-to-pad
capacitance, the GaAs substrate can be removed. This has been demonstrated in a procedure
that replaces the GaAs with quartz, as shown in Fig. 7 [19]. The quartz substrate can be
permanent or it can be removed once the chip is bonded to a circuit as shown in Fig. 8.
Figure 7. Surface Channel Diode Chip Figure 8. Surface Channel Diode Chip
with Quartz Substrate with Quartz Substrate
Removed After Bonding
Page siO Third International Symposium on Space Terahertz Technology
The effect of finger length on planar diode performance is an important issue,
particularly for waveguide mixers. Longer fingers result in reduced pad-to-pad capacitance
but increased finger inductance. A new mask set has been fabricated which will provide
small area, antiparallel planar diodes with finger lengths from 10 microns to 50 microns in
10 micron steps on the same wafer. This mask was designed in collaboration with Peter
Seigel of JPL and the devices will be RF tested at JPL in a waveguide mixer at frequencies
as high as 600 GHz.
Very short contact fingers are required in integrated antenna designs. Surface channel
formation is very difficult when the contact finger is under 10 microns in length. Research
is underway to characterize a combination of chlorine-based reactive ion etching and wet
chemical etching processes to form the surface channel isolation trench with these short
contact fingers. The new mask sets for both the small area antiparallel chips and the log
periodic antenna designs include levels for this new process.
C. Minimization of Ohmic Contact Resistance
Ohmic contact resistance contributes to diode series resistance and thus reduces cutoff
frequency. As contact pad dimensions shrink, ohmic contact resistance increases. This is of
particular importance in the case of integrated antenna devices where the pad geometry is
dictated by the antenna design. Specific contact resistance can be improved by using a very
highly doped buffer layer and through the use of a more advanced ohmic contact technology.
For example, ohmic contacts to an n 4 " 1 " InGaAs layer are reported to have specific contact
resistivity as low as 10" 7 Sl-cm 2 , a factor of 50 to 100 better than our present ohmic contacts.
Third International Symposium on Space Terahertz Technology Page 611
This would be most beneficial for planar THz antenna structures which require small pad
geometries near the anode.
D. Integration of Antenna Structures
The problem of efficient energy coupling to the planar diode is exacerbated at higher
frequencies where the wavelength begins to approach the size of the chip. For whisker-
contacted diodes, the whisker itself is used as the antenna element and mixers with a long
whisker (4X) positioned parallel to the axis of a corner cube have demonstrated excellent
performance at frequencies as high as 4 THz [3].
Another approach for planar diodes is to integrate an antenna, in the form of a bow-tie
or log periodic shape onto the chip [20]. The fabrication is straightforward, with the antenna
being an extension of the anode and cathode pads and the radiation can be coupled to the
antenna through the substrate (GaAs or quartz). An integrated bowtie antenna-diode is shown
in Fig. 9. It is 700 x 1000 x 50 microns thick with a 0.5 micron anode and an 8 micron
finger length. Preliminary RF testing with unoptimized coupling produced video response of
10 V/W.
Optimization of the integrated antenna will require close interaction between diode
designers and RF engineers. As a first step towards this goal, a mask set for the fabrication
of log periodic antenna-diodes has been designed in cooperation with Gabriel Rebeiz of the
University of Michigan and devices will be fabricated in the near future. With proper diode
design and good coupling of energy- to the antenna and the diode, it is hoped that RF
performance will exceed that of the best whisker-contacted diodes.
Page 612
Third International Symposium on Space Terahertz Technology
Figure 9. Integrated Bowtie Antenna-Diode
IV. Discussion
Development of the planar mixer diode was driven by the need for a rugged device
which is inherently simple and easy to assemble in a mixer. However, the tradeoffs for this
structural ruggedness and simplicity are a more complex and expensive fabrication procedure,
and a more complex chip geometry with larger shunt capacitance. The RF circuit must be
redesigned to efficiently couple energy to the diode. In spite of these changes, planar GaAs
Schottky barrier diodes have demonstrated performance in the millimeter wavelength range
equal to or better than that of the best whisker-contacted diodes.
Successful operation of planar diodes at THz frequencies will require several
improvements in the diode chip including reduced anode diameter, improved control of anode
diameter, smaller chip dimensions to reduce shunt capacitance, and reduced ohmic contact
resistance. These concerns are being addressed through research of novel structures and
fabrication methods. Successful application of planar diodes in the THz frequency range will
also require optimization of the embedding circuitry and improved methods of coupling
energy to the diode. Research is underway to apply novel antenna designs to this problem and
to begin to test high performance planar diodes in waveguide assemblies and to test integrated
Third International Symposium on Space Terahertz Technology Page 613
antennas in open structure mixers. The success of this effort will be hastened by very close
interaction and cooperation between diode designers and RF engineers.
Acknowledgements
The authors wish to express their sincere appreciation to Peter Siegel of the
Jet Propulsion Laboratory, Israel Galin and Robert Haas of Aerojet General and Gabriel
Rebeiz of the University of Michigan for many helpful discussions regarding the design of
planar diodes. This work was supported by the National Science Foundation (ECS-8720850),
the. U.S. Army and the Jet Propulsion Laboratory (958202).
References
[1] T.W. Crowe, R.J. Mattauch, H.P. Roeser, W.L. Bishop, W.C.B. Peatman, "GaAs
Schottky Diodes for THz Mixing Application," Invited paper accepted for IEEE Proa,
Special Issue on Terahertz Technology, to appear in 1992.
[2] C.R. Predmore, A.R. Raisanen, N.R. Erikson, P.F. Goldsmith, and J.L.R. Marrero, "A
Broad-Band, Ultra-Low-Noise Schottky Diode Mixer Receiver for 80-1 15 GHz," IEEE
Trans. Microwave Theory Tech., Vol. MTT-32, pp. 498-506, May 1984.
[3] H.P. Roser, R. Wattenbach, E.J. Durwen, and G.V. Schultz, "A High Resolution
Spectrometer for 100 urn to 1000 um and Detection of CO (J=7-6), CO (J=6-5) and
13 CO (J=3-2)," Astron. Astrophys., 165, 287-299, 1986.
[4] S.K. Pan, A.R. Kerr, M.J. Feldman, A. Kleinsasser, J. Stasiak, R.L. Sandstrom, and
W.J. Gallagher," An 85-116 GHz SIS Receiver Using Inductively Shunted Edge-
Junctions," IEEE Trans. Microwave Theory Tech., Vol. MTT-37, pp. 580-592, March
1989.
[5] A. W. Lichtenberger, D.M. Lea, and A.C. Hicks, "Nb-based SIS Mixer Elements for
Millimeter and Submillimeter Wavelengths," 2nd Int'l. Symp. on Space Terahertz
Tech., pp. 439-458, Feb. 1991.
[6] J. Zmuidzinas and H.G. LeDuc, "Quasi-Optical Slot Antenna SIS Mixers," 2nd Int'l.
Symp. on Space Terehertz Tech., pp. 481-490, Feb, 1991.
Page 614 Third International Symposium on Space Terahertz Technology
[7] J.W. Waters, "A Proposal of the Earth Observing System, Microwave Limb Sounder,"
Jet Propulsion Laboratory, California Institute of Tech, July 1988.
[8] M.A. Frerking, "The Submillimeter Mission (SMMM) Heterodyne Instrument," 2nd
Int'l. Symp. Space Terahertz Tech., pp. 17-31, Feb. 1991.
[9] W.L. Bishop, K. Mckinney, R.J. Mattauch, T.W. Crowe, and G. Green, "A Novel
Whiskerless Schottky Diode for Millimeter and Submillimeter Wave Applications,"
Proc. 1987 IEEE MTT-S Int'l Symp, pp.607-610, June 1987.
[10] J.W. Archer, R.A. Batchelor, and C.J. Smith, "Low-Parasitic, Planar Schottky Diodes
for Millimeter- Wave Integrated Circuits," IEEE Trans. Microwave Theory Tech., Vol.
MTT-38, No. 1, pp. 15-25, Jan. 1990.
[11] N.J. Cronin, and VJ. Law, "Planar Millimeter-Wave Diode Mixer," IEEE Trans, on
Microwave Theory Tech., Vol. MTT-33, No. 9, pp. 827-830, Sept. 1985.
[12] J.A. Calviello, S. Nussbaum, and P.R. Bie, "High Performance GaAs Beam-Lead
Mixer Diodes for Millimeter and Submillimeter Applications," Proc. of Intl. Electron
Device Meeting, Dec. 7-9, 1981.
[13] W.L. Bishop, K.A. McLeod, R.J. Mattauch, "Whiskerless Schottky Diode," U.S.
Patent 5,041,881, Aug. 20, 1991.
[14] T. Newman, W.L. Bishop, K.T. Ng, and S. Weinreb, "A Novel Planar Diode Mixer
for Submillimeter- Wave Applications," IEEE Trans. Microwave Theory Tech., Vol.
39, No. 12, pp. 1964-1971, Dec. 1991.
[15] P.H. Seigel, R.J. Dengler, I. Mehdi, J.E. Oswald, W.L. Bishop, T.W. Crowe and R.J.
Mattauch, "Measurements on a 215 GHz Subharmonically Pumped Waveguide Mixer
Using Planar Back-to-Back Air Bridge Schottky Diodes," submitted for publication
to IEEE Microwave and Guided Wave Letters, Oct. 1991.
[16] T.W. Crowe and R.J. Mattauch, "Analysis and Optimization of Millimeter-and-
Submillimeter- Wavelength Mixer Diodes," IEEE Trans. Microwave Theory Tech.,
Vol. MTT-35, Vol. 2, pp. 159-168, Feb. 1987.
[17] W.C.B Peatman, P.A.D. Wood, D. Poterfield, T.W. Crowe and M.J. Rooks, "A
Quarter-Micron GaAs Schottky Barrier Diode with High Video Responsivity at 118
Microns," submitted to the Appl. Physics Lett., Feb. 1992.
[18] E.M. Winkler, "A Study of the Effect of Reactive Ion Etching on the Noise
Characteristics of Schottky Diodes," Master of Science Thesis, University of Virginia,
Charlottesville, VA, August 1991.
Third International Symposium on Space Terahertz Technology P a S e 615
[19] W.L. Bishop, E.R. Meiburg, RJ. Mattauch, T.W. Crowe, and L. Poli, "A Micron-
Thickness, Planar Schottky Diode Chip For Terahertz Applications with Theoretical
Minimum Parasitic Capacitance," Proc. 1990 IEEE MTT-S Int'l. Symp., pp. 1 SOS-
BOS, May, 1990.
[20] P.H. Siegel, "A Planar Log-Periodic Mixtenna for Millimeter and Submillimeter
Wavelengths," Proc. 1986 IEEE MTT-S Int'l. Symp., pp. 649-652, 1986.
Page 616 Third International Symposium on Space Terahertz Technology
5st-2>2>
/ 1 Q&bh
Planar Doped Barrier Subharmonic Mixers*
T. H. Lee. J. R. East and G. I. Haddad
/ Center for Space Terahertz Technology
\ ^ The University of Michigan
^ Ann Arbor, Michigan
Abstract
The fPDB (Planar Doped Barrierf* diode is a device consisting of a p + doping spike
' between two' intrinsic layers and n + ohmic contacts. This device has the advantages of
controllable barrier height, diode capacitance and forward to reverse current ratio. A
symmetrically designed PDB has an anti-symmetric current vs. voltage characteristic and is ideal
for use as millimeter wave subharmonic mixers. We have fabricated such devices with barrier
heights of 0.3, 0.5 and 0.7 volts from GaAs and InGaAs using a multijunction honeycomb
structure with junction diameters between one and ten microns. Initial RF measurements are
encouraging. The 0.7 volt barrier height 4 micron GaAs devices were tested as subharmonic
mixers at 202 Ghz with an IF frequency of 1 GHz and had 18 dB of conversion loss. The
estimated mismatch loss was 7 dB and was due to higher diode capacitance. The LO frequency
was 100.5 GHz and the pump power was 8 mW.
* This work was supported by NASA under Grant No. NAG W- 13 34
Third International Symposium on Space Terahertz Technology Page 617
I. Introduction
Planar Doped Barrier devices were first proposed in 1980 by Malik etal[l]. The structure
can be understood as a planar doped p* spike sandwiched between two lightly doped regions and
heavily doped n-type ohmic contacts. Such an n*-i-p+-i-n + device has a triangular potential barrier
which is adjustable by the parameters of epi-layer growth such as sheet charge doping density
and intrinsic layer dimensions. The devices are unipolar and the charge transport over the
potential barrier can be modeled by the thermionic emission theory. The current vs. voltage
characteristic is similar to that of Schottky diodes. Accordingly PDB's can be used as a Schottky
barrier diode replacement with the additional advantage of barrier height control.
The major applications of PDB devices are as mixers and detectors. If the device structure
is designed symmetrically with the p + doping spike in the middle of the intrinsic region, the PDB
diode has an anti-symmetrical I-V characteristics and is ideal for millimeter-wave subharmonic
mixer applications. A summary of the theory and design techniques for symmetric PDB diodes
as subharmonic mixers is given by Lee et al[2]. Several subharmonic PDB diodes operating at
microwave frequencies have been reported[3,4,5]. Giittich et al. measured a D-band silicon PDB
subharmonic mixer with a minimum conversion loss of 10.8 dB[6], The structure can also be
designed to have lightly doped regions of different thickness. This will provide a current vs.
voltage characteristic that is useful in low barrier detectors applications. Dale et a/[7,8]. reported
a PDB single-ended mixer had a noise figure of approximately 6 dB at 9.4 GHz and required
only 280 |nW of local oscillator power. When used as video detectors, Anand et al[9]. showed
zero bias PDBs for low level detection had a burnout limit comparable to high barrier Schottky
diodes and were less sensitive to the electrostatic discharge.
PDB devices suffer from several problems. The space charge resistance is relatively high.
By making the /-layer as short as possible we are able to reduce the space charge resistance.
However, the /-layer width has to be larger than the Debye length of n + region to avoid the
charge redistribution at i-n* interface. For a subharmonic PDB diode the ideality factor is at least
Page si 8 Third International Symposium on Space Terahertz Technology
two. The conversion ability is thus degraded, A careful design is necessary to obtain an optimum
performance.
In this paper we present a group of MBE grown GaAs wafers with a 250A i-layer
thickness and the p + spike doping densities of 1.5, 2.0 and 2.5xl0 12 cm" 2 for mixer operation. The
barrier height was designed for 0.3, 0.5 and 0.7 volts respectively. The impurity concentration
of the intrinsic region is nominally less than 10 14 cm' 3 . Also presented are a group of InGaAs
wafers grown via gas MBE at the University of Michigan. These wafers are of the same
specifications as the GaAs ones except with longer n + layers. A preliminary RF result is reported.
II. Fabrication of Whisker-contact Subharmonic PDB Diodes
Two material systems have been chosen: GaAs and InGaAs. The use of GaAs has some
advantages. GaAs is a more mature fabrication technology and tight parameter control of MBE
grown GaAs is relatively easy. InGaAs, however, has a higher electron mobility and much lower
contact resistivity than GaAs. For terahertz operation InGaAs might be a better material system.
In this paper, we present diodes from both material systems.
A typical wafer structure for the subharmonic GaAs PDB is shown in Fig. 1. An optimum
structure for small diameter whisker-contact PDB diodes with low series resistance and parasitic
capacitance is achieved by completely removing the substrate and forming mesas. An etch stop
layer of AlGaAs between the substrate and epi-layer was included for this substrate removal. A
good selectivity between the substrate and etch stop layer is possible. In the InGaAs system
hydrochloric acid was used to remove the InP substrate without affecting the InGaAs epi-layer.
The fabrication of whisker-contact PDB diodes consists of seven major steps as shown
in Fig. 2:
(1) diode definition and metallization,
(2) mesa etch,
1
Third International Symposium on Space Terahertz Technology Page 619
(3) Si0 2 passivation,
(4) opening contact holes,
(5) front side protection,
(6) backside thinning,
(7) final metallization and plating.
More details on the process steps are given in the next section.
The first step is diode definition and metallization. An image reversal photolithography
process was characterized to attain desired diode patterns for ohmic metal lift-off. A positive
photoresist was spun uniformly on the wafers, and then soft-baked at 105°C and UV exposed in
the conventional way. Next a reversal bake and flood exposure altered the polarity of the
solubility of photoresist in the images and fields, so that a negative image could be obtained after
development. The exposure and bake parameters are optimized to obtain an undercut profile
desirable for the lift-off.
Metallization was then performed by depositing layers of thin film metals on the
photoresist patterned samples. Ni/Ge/Au/Ti/Au were evaporated in sequence. The metal covered
wafers were soaked in the acetone for lift-off. The resulting dot-like patterns also served as a
self-aligned mask for mesa etch.
The next step is the mesa etch. To avoid serious undercut we used a dry plasma etch
instead of a wet chemical etch. Mesa etching was accomplished by a reactive-ion etching system.
The GaAs wafers were etched in a low pressure chamber filled with BC1 3 and Ar gaseous
plasmas of 11:9 ratio to obtain highly anisotropic sidewalls, while InGaAs wafers were etched
by the mix of methane, hydrogen, and argon. The effective area of diodes was determined by the
area of the p* doping spike. With plasma etching the precise control of device area is possible.
The desired etch depth is into the first few hundred A of the bottom n + layer. Sometimes a
subsequent slow wet etch is used to remove a thin damaged layer.
eY
Page 620 Third International Symposium on Space Terahertz Technology
The third step is Si0 2 deposition. After the wafers were mesa etched, a dielectric film was
deposited for step coverage. A plasma-enhanced chemical vapor deposition system was used. A
low temperature process was adopted to avoid heating the devices. A silane/oxygen plasma at
200°C for 90 minutes was used for optimum step coverage. This low temperature silicon dioxide
layer was amorphous with low dielectric strength. A dielectric covered sample is shown in Fig.
3.
The fourth step is opening contact holes. At this stage, wafers were covered by PECVD
Si0 2 everywhere. The mesa mask was used to open a hole where previously ohmic metals were
deposited. The image reversal process as described in the first step with a careful alignment was
required. The silicon dioxide on the top of ohmic metals was RIE-etched using gaseous plasmas
of CHF 3 and CF 4 . It was difficult to tell, under the microscope, whether ohmic metal has been
reached. By probing the adjacent diodes and measuring the electrical properties, we made certain
the dielectric had been removed completely and could move to next step. Fig. 4 shows the diode
at this stage.
The fifth step is front side protection. The wafer prior to substrate removal has to be
protected and supported on its front side. Wafers after backside thinning were less than one (im
thick and simply too fragile to work on. A remedy is to spin a thin photoresist on the front side
and then plate thick metal for support. The thin PR and plated metal could be striped off in the
last step.
The sixth step is backside thinning. Selective etching was used for the substrate removal.
In the GaAs/AlGaAs system, NH 4 OH:H 2 2 (1:24) was used. Hydrogen peroxide helps oxidizing
GaAs whose oxide can be removed by ammonium hydroxide. The selectivity is a function of the
percentage of Al composition in AlGaAs, and has to be high enough to stop etching. The subtle
change of sample color implied the remaining thickness of the GaAs substrate. After the stop
layer was reached, the thin AlGaAs film was then removed by hydrofluoric acid which would
not attack GaAs. In the InGaAs system, hydrochloric acid quickly removed the InP substrate but
Third International Symposium on Space Terahertz Technology Page 621
did not attack the InGaAs epi-layer. These thinned samples were ready for the final metallization.
The final step is the metallization and plating. Backside metallization was performed after
substrate thinning. The same ohmic metals as for the front side contact were evaporated. If
samples have been exposed in the air, the accumulated surface oxide should be removed before
metallization. After metallization gold, for its good conductivity, was plated on the back for
mechanical support. The process was done after we removed the front side protection by acetone.
The samples were about 20 |im thick.
The finished device was a vertical diode with ohmic metal on the top and bottom. We
have been successful in fabricating 1 |im, 2 jim, 4 ^im and 10 (im diameter diodes. Fig. 5 shows
the multijunction honeycomb structure for 4 fim diodes. All the diodes were evenly arranged in
125x125 urn squares, cut by diamond saws and bonded in a fixture. They are now ready for RF
testing in a whisker mount.
III. Planar Doped Barrier Subharmonic Mixer Performance
The subharmonic mixer diodes were first characterized by their DC performance. Fig. 6
shows the I-V characteristics of a 4 (im diameter GaAs PDB diode of 0.7 volt barrier height. The
low frequency capacitance was about 20 fF. The effective depletion length is approximately 700
A. Fig. 7 shows the I-V curves for InGaAs 2 jim diameter devices with barrier heights of 0.3,
0,5 and 0.7 volts respectively. The capacitance of InGaAs 10 (im diodes was 170 fF, which
scaled down to a value of 6.9 fF for 2 (im ones. The ideality factor was determined by the slope
of I-V between 10 ]iA and 100 ^lA. The resulting value for PDB diodes with 0.7 volt barrier
height was 2.34 for GaAs and 2.39 for InGaAs.
The RF measurement results for 4 (im GaAs diodes are shown in Table 1 . The diode was
tested as a subharmonic mixer with a pump frequency of 100.5 GHz, a signal frequency of 202
GHz and an IF frequency of 1 GHz. The mixer was a whisker-contact diode mount which was
Page 622 Third International Symposium on Space Terahertz Technology
designed to match a device of 5 fF capacitance at about 200 GHz. The tested 4 jim diode
capacitance was 20 fF. With a pump power of 8 milliwatts, the mixer conversion loss was 18 dB.
The mismatch was estimated at 7 dB, which implied a diode conversion loss of 11 dB.
Measurements on the smaller devices and InGaAs devices are in progress.
V. Conclusion
The properties and fabrication techniques for planar doped barrier subharmonic mixer
diodes have been described. The fabrication process produces mesa diodes for whisker contact
mounts with diameters between one and ten microns. The diodes are based on the GaAs or
InGaAs system and have barrier heights of 0.3, 0.5 and 0.7 volts. Initial subharmonic mixer
results for a 4 Jim diameter GaAs diode with 0.7 volt barrier height gave a conversion loss of
18 dB at 202 GHz with a local oscillator power of 8 milliwatts and an expected mismatch loss
of 7 dB.
Third International Symposium on Space Terahertz Technology Page 623
Reference
[1] Malik, R. J., T. R. Aucoin, R. L. Ross, K. Board, C. E. C. Wood and L. F. Eastman,
"Planar-Doped Barriers in GaAs by Molecular Beam Epitaxy," Electron. Lett., vol. 16,
pp. 836-838, Oct. 1980.
[2] Lee, T. H., J. R. East and G. I. Haddad, "Planar Doped Barrier Devices for Subharmonic
Mixers," Microwave and Optical Technology Letters, vol. 4, No. 1, pp.53-60, Jan. 1991.
[3] Malik, R. J., and S. Dixson, "A Subharmonic Mixer Using a Planar Doped Barrier Diode
with Symmetric Conductance," IEEE Electron Device Lett., vol. EDL-3, No. 7, pp. 205-
207, Jul. 1982.
[4] Dixon, S., and R. J. Malik, "Subharmonic Planar Doped Barrier Mixer Conversion Loss
Characteristics," IEEE Trans. Microwave Theory Tech., vol. MTT-31, pp. 155-158, Feb.
1983.
[5] Chen, J., and D. Wong, "W-Band Beam Lead Planar Doped Barrier Subharmonic Mixer,"
IEEE Microwave Symp Technical Digest, pp. 178-180, 1985.
[6] Guttich, Ulrich, K. M. Strohm and F. Schaffler, "D-Band Subharmonic Mixer with Silicon
Planar Doped Barrier Diodes," IEEE Trans. Microwave Theory Tech., vol MTT-39, pp.
366-368, Feb. 1991.
[7] Dale, I., A. Condle, S. Neylon and M. Kearney, "Planar Doped Barrier Mixer and
Detector Diodes as Alternatives to Schottky Diodes for Both Microwave and Millimeter
Wave Applications," IEEE Microwave Symp. Technical Digest, pp. 467-470, 1989.
[8] Kearney, M. J., M. J. Kelly, R. A. Davies, T. M. Kerr, P. K. Rees, A. Condie and I. Dale,
"Asymmetric Planar Doped Barrier Diodes for Mixer and Detector Applications,"
Electron. Lett., vol. 25, pp. 1454-1456, Oct. 1989.
[9] Anand, Y., J. Hillson, A. Torabi and J. R. East, "Millimeter Wave Planar Doped Barrier
Detector Diodes," Proceedings of 2nd Int. Symp. on Space Terahertz Technology, pp. 340-
352, 1991.
Page 624
Third International Symposium on Space Terahertz Technology
Device profile ofGaAs subharmonic PDB diodes
GaAs substrate
nwwwwmmff
Figure (1) The device profile of GaAs subharmonic PDB diodes. The width of the intrinsic
layer is 250 A and that of the n + is 2500 A on either side of the p + spike. The sheet
charge doping density of the p + spike is 1.5, 2.0 and 2.5xl0 12 cm' 2 for 0.3, 0.5 and
0.7 volt barrier height respectively.
Third International Symposium on Space Terahertz Technology
Page 625
(1) Diode definition and metallization
Ohmic metal layer
if Device layer
M««M««MMM«8MM«M««1M«M^^ A [Q aAs
GaAs substrate
(2) Mesa etch
raMWflrow?
Ohmic metal layer
/ Device layer
s^s« m "'"«^ AIGaAs
GaAs substrate
(5) Front side protection
GaAs substrate
Support metal
Ph itoresist
AIGaAs
(6) Backside thinning
;;:;B|;||;;;ig|:^ig^ t metal
-AHioAs-
Ph itoresisl
(3) Si02 Passivation
PECVD silicon dioxide
/ \ $ Devic e layer
W :.:""\if
AIGaAs
GaAs substrate
(4) Opening contact holes
PECVD silicon dioxide
ff Devic e layer
AIGaAs
GaAs substrate
i. ............ ... . j .
(7) Final metallization and plating
Si02
-Tbp'dKnirdSfmcr
IF— 1
4. j
i. Device layer
- — ■* . /
Ohn tic contact
Figure (2) The fabrication sequence for GaAs subharmonic mixer diodes.
Page 626
Third International Symposium on Space Terahertz Technology
r<- ,/^m^.
WdS 10KV X
U-'rii WD 14
Figure (3) Diodes are covered by PECVD silicon dioxide everywhere. This picture was take:;
after fabrication step (3) as described in figure (2).
Third International Symposium on Space Terahertz Technology
Page 627
:i
1
Figure (4) Contact holes ore opened on the ohfrric metals,
fabrication step (4) as described in figure (2).
This picture was taken ar^.cr
Page 628
Third International Symposium on Space Terahertz Technology
Figure (5) Finished diodes of honeycomb structure are in 125x125 |a.m squares. The dark area
is covered by the dielectric layer.
Third International Symposium on Space Terahertz Technology
Page 629
The I-V Curve of a GaAs 4 Microm Diode
l.OE-02-
7SE-03-
5.0E-OS-
§3
<U 25E-03- "
O.OE+00-
1
-25E-03-
-S.OE-03-
■75E-03- "
-1 .0£-02^ 1 H 1 1
■2.0 -IS -1.0 -OS 0.0 OS 10 15 2.0
Voltage ( volts )
Figure (6) The I-V curve of a 4 ^m GaAs subharmonic PDB diode.
I-V Curves for InGaAs 2 Microm Diodes
l.OE-02
■l.OE-02-
Voltage ( volts )
Figure (7) The I-V curves for 2 |um InGaAs diodes of 0.3, 0.5 and 0.7 volt barrier height,
respectively.
Page 630
Third International Symposium on Space Terahertz Technology
Mixer Diode Performance (GaAs)
Diode diameter
4 Jim
p + spike doping
2.5xl0 12 cm' 2
Total / layer length
525 A
Estimated barrier height
0.7 V
Ideality factor
2.39
Diode capacitance (measured)
20 fF
Contact resistance
(by p=5xl0" 7 Q-cnV 2 )
3.9 Q.
RF
202 GHz
IF
1.0 GHz
LO
100.5 GHz, 8 mW
Mixer conversion loss
18 dB
Mixer waveguide impedence
150 Q.
Series inductance
0.01 nH
Mismatch loss
7dB
Diode conversion loss
11 dB
Table (1) The performance of a 4 |im GaAs PDB diode as a subharmonic mixer.
Third International Symposium on Space Terahertz Technology Page 631
New Approach to the Design of Schottky Barrier
Diodes for THz Mixers N 9 3 "3g#J£!g8
A. Jelenski + , A. Griib *, V. Krozer *, H.L. Hartnagel * * ,.
* Institut fur Hochfrequenztechnik, Technische Hochschule Darmstadt, Merckstr. 25, D-
6100 Darmstadt, Germany, Phone: +49 6151 162162, Fax: +49 6151 164367
+ Institut fur Hochfrequenztechnik, Technische Universitat Hannover, Appelstr. 9A, D-
3000 Hannover, Germany, on leave from the Institute of Electronic Materials Tech-
nology, Warsaw, Poland Phone: +48 22 354416, Fax: +48 39120764
Abstract
a Near-ideal GaAs Schottky barrier diodes especially designed for mixing applications in
the THz frequency range are presented.
A diode fabrication process for submicron diodes with near-ideal electrical and noise
characteristics is described. This process is based on the electrolytic pulse etching of
GaAs in combination with an in-situ platinum plating for the formation of the Schottky
contacts. Schottky barrier diodes with a diameter of 1 \irrx fabricated by the process have
already shown excellent results in a 650 GHz waveguide mixer at room temperature. A
conversion loss of 7.5 dB and a mixer noise temperature of less than 2000 K have been
obtained at an intermediate frequency of 4 GHz.
The optimization of the diode structure and the technology was possible due to the
development of a generalized Schottky barrier diode model which is valid also at high
current densities. The common diode design and optimization is discussed on the basis
of the classical theory. However, the conventional formulas are valid only in a limited
forward bias range corresponding to currents much smaller than the operating currents
under submillimeter mixing conditions. The generalized new model takes into account not
only the phenomena occurring at the junction such as current dependent recombination
and drift /diffusion velocities, but also mobility and electron temperature variations in the
undepleted epi-layer.
Calculated diode I/V and noise characteristics are in excellent agreement with the
measured values. Thus, the model offers the possibility of optimizing the diode structure
and predicting the diode performance under mixing conditions at THz frequencies, r "j/
Page 632 Third International Symposium on Space Terahertz Technology
Introduction
Schottky-barrier diodes have been recently applied for heterodyne receivers in the fre-
quency range up to a few THz, the interesting point being how their realization and
technology can be improved to obtain still better results at higher frequencies. Much ef-
fort has been done to reduce the feature sizes, but further reduction beyond approximately
0.5fim seems to be unpractical because of contacting difficulties. Since maximizing of the
conversion efficiency requires a resistive mixing, the junction resistance at the operating
point has to be much smaller than its capacitive reactance to avoid losses related with
parametric downconversion.
Therefore, for the highest frequencies despite the decreasing diameter, the DC current
at the operating point is always of the order of 0.5 mA, leading to an increased current
density and therefore the bias approaches the so called flat-band voltage. The expressions
for the I/V and the C/V characteristics of Schottky diodes are usually determined by
the following well known formulas:
7 = / JOt exp^J giving Rj = — (1)
C A J q ' Nd - ° 3 ° (2)
Cl - A \J2{V D -V T -V)- I Vr__ V_ (2)
V v D V D
When the difference between the bias and the flat-band voltage is smaller than 3 Vj =
3kT/q ~ 80 mV at room temperature, eq. 1 and eq. 2 become invalid [1].
To avoid discrepancies between calculated and measured results some authors were
forced to make unphysical assumptions like assuming that the fiat-band voltage Vq equals
the barrier height <f>b [2] and neglecting the Vj/Vq term in eq. 2 [3]. But these assumptions
still do not solve the problem which results from the fact that under the so called flat -band
condition, when the depleted layer disappears, the capacitance C, becomes infinite, while
the junction resistance R, (eq. 1) is still finite and greater than zero. This phenomenon
results from the approximations made in the derivation of eq. 1 and 2.
The examination of the I/V characteristics of the fabricated diodes shows that the
dominating carrier transport mechanism is the thermionic field emission in the reverse
bias and the thermionic emission and thermionic emission/ diffusion for forward biases.
In the frame of the diffusion theory, which determines the diode behavior near fiat-band,
eq. 1 is just an approximation of the solution given by the Dawson integral valid for
V < Vrj — Wj. A better approximation of this solution similar to that given already by
Third International Symposium on Space Terahertz Technology
Page 633
Schottky [4] is
2sinb
V nV T )
giving Rj = — — tanh
\ nV T )
(3)
In this bias range the depletion approximation utilized in the derivation of eq. 2 is
also not valid any more and the space charge must be taken into account. This leads to
a novel expression for the junction capacitance:
V D -V\
1 - exp
C> =
*-%-*
Vd
1 -exp
Vp-V "
(4)
These expressions should be used if one wants to calculate the mixer performance
when the diode is operating near flat-band conditions. They can be easily inserted into
the generalised program, results of which are presented in this contribution. The actual
program takes into account the current dependent recombination velocity, the field depen-
dent mobility and electron heating at high forward bias. This program can therefore be
efficiently used to calculate mixer or frequency multiplier performances enabling a more
realistic analysis and optimization similar to the analysis given in [5].
The noise generating mechanisms in Schottky-barrier diodes are also well understood
and are indicated in fig.l. The basic noise generating mechanisms are: a) shot noise in
the junction with tj = 2g/A/ and b) thermal noise in the series resistance with e] =
4kT R t Af.
O (►
a
K£>
e* e,
R
<X>^.
Figure 1: The equivalent circuit of the Schottky-barrier diode.
The well known expression first derived by Weisskopf in 1943 [6] gives the diode
noise temperature for frequencies much lower than the junction cut-off frequency w ; =
Page 634 Third International Symposium on Space Terahertz Technology
U(RiCj).
2 R, + i?j if, + iij
For higher forward biases and diodes with low doped epi-layers, electrons are heated
by the high electric field in the undepleted epi-layer and the excess noise temperature T^
due to the increased electron temperature has to be added to the thermal noise [7, 8].
t\ = 4kT K h I 2 R s Af (6)
In some diodes an excess noise can be observed probably due to some trapping effects
at the interface. This phenomenon has been described by [8, 9] and the noise source i t
should be added to ij.
«-"*&l+fcF (7)
where Nt is the concentration of presumed traps and r is the time constant of this process.
The general expression to be compared with the experiment is then
T n = nT ° Rj + Ct + T Rs (1 + K k I 2 ) (8)
2 R s + Rj R s + Rj Rs + Rj
where Ct and Kh are constants. If a good fitting is obtained with these 2 constants
it means that the above simple model well describes noise generation mechanisms in
measured diodes.
Diode fabrication
The first major requirement for GaAs Schottky diodes for mixing applications in the
submillimeter frequency range is a small Schottky contact area in order to achieve junc-
tion capacitances in the low fF or even subfF region. Secondly, a homogeneous me-
tal/semiconductor contact free of interfacial layers is required in order to achieve near-
ideal electrical and noise performance. Since practical submm Schottky diodes have a
metal/semiconductor contact area of less than 1 jum 2 , the GaAs surface treatment and
Schottky metal deposition techniques are much more important than for other semi-
conductor devices. Therefore, the fabrication of low-noise Schottky diodes requires an
especially optimized device technology in order to avoid any damage to the GaAs surface.
The diodes presented in this paper have been fabricated by applying a novel GaAs etching
technique which is called anodic pulse etching [10, 1-1]. Since the initial fabrication steps
such as Si02~deposition, ohmic backside contact and Sz'02-structuring are the same as
commonly used, the main subject of the fabrication section will be the description of the
anodic pulse etching in combination with the electrolytic Pt deposition for the formation
of near-ideal small-area Schottky contacts.
Third International Symposium on Space Terahertz Technology Page 635
For the fabrication of the diodes high-quality MBE-grown layers have been used [12].
Besides the diode diameter, thickness and doping concentration of these layers are the
important process parameters for optimum device performance. Epitaxial layers having
doping concentrations of 2 • 10 16 cm' 3 , 8 • 10 16 crn~ 3 and 2 • 10 17 cm' 3 with an original
epi-layer thickness of 200 nm have been used. In addition, one epi-layer with a graded
doping profile has been used. This layer has a surface doping concentration of 2- 10 16 cm~ 3
which is increased exponentially to 6 • 10 18 cm~ 3 within 90 nm. By controlled etching of
the GaAs surface, hence any surface doping concentration can be achieved.
After thinning the n + -substrate to a thickness of 50-70 fim by mechanical lapping
and polishing in order to reduce the substrate resistance, a 500 nm thick Si02 layer is
deposited onto the epitaxial side by e-beam evaporation. The Si02 is necessary in order
to avoid As outdiffusion during the following formation of the ohmic backside contact
and finally separates the single Schottky diodes and serves as a mechanical guide for the
whisker contact. The ohmic backside contact is formed by evaporation of Ni/AuGe/Ni
followed by a rapid thermal annealing step in ^-atmosphere. Subsequently, the honey-
comb diode structure is transferred to a photoresist layer on the S1O2 by conventional
UV-lithography. The structured photoresist serves as an etch mask for the reactive ion
etching of the SiO-i- The applied RIE-process with CHF3/O2 assures highly anisotropic
etching of the Si02. Thus, honeycomb structures with smallest hole diameters of 0.8 fim
have been defined. The RIE-process is followed by a short dip in buffered HF in order to
remove any possible Si02 residues.
♦
The next step is the formation of the small-area Schottky contacts which is of course
most important since it defines the quality of the Schottky contact. Because of the
introduced GaAs surface damage due to the Si02 e-beam evaporation and the plasma
etching, it is necessary to remove some ten nanometers of GaAs before the deposition of
the Schottky metal.
The etching step usually is performed by wet chemical etching [13, 14] or anodic
oxidation of the GaAs surface with subsequent dissolution of the anodic oxide in a Pt
electrolyte [15]. Since wet chemical etching is isotropic it leads to an enlargement of
the contact area and thus to larger junction capacitances. Furthermore, the etched GaAs
surface is in contact with air, leading to the formation of a thin interfacial oxide layer. The
anodic oxidation process allows the in-situ Pt plating avoiding these interfacial layers.
The drawbacks of this technique are the rather isotropic etching and, as well as for the wet
chemical process, the poor control of the etched depth. Nevertheless, the anodic oxidation
process has become the standard technique for the fabrication of GaAs Schottky diodes
for submm applications.
Page 636 Third International Symposium on Space Terahertz Technology
The anodic pulse etching technique avoids the above stated problems [10, 11]. The
principle of this technique is outlined below. The GaAs surface is brought into contact
with a Pt electrolyte. The electrolyte/GaAs junction which behaves like a Schottky
junction is driven into an Avalanche breakdown by the application of short voltage pulses.
During the impact ionization in the space charge region electron-hole pairs are generated.
The holes drift to the electrolyte/GaAs interface where they are essential for the anodic
dissolution of GaAs. The short pulse width of 300 ns therefore enables an excellent control
of the removed GaAs thickness by the number of applied voltage pulses. The short pulses
are also essential for the anisotropic etching since saturation effects due to diffusion limited
transport of reaction species are avoided. Since the solution for the anodic pulse etching
is the same which is used for the electrolytic Pt deposition, the in-situ metallization is
possible. The most important aspects of this technique are summarized below.
1. anisotropic etching ^ suitable for fabrication of submicron structures
2. excellent control and reproducibility => suitable for process-oriented modelling
3. in-situ metallization => suitable for fabrication of near-ideal Schottky contacts be-
cause surface damage and interfacial layers are avoided
By application of the anodic pulse etching technique, 100 nm of epitaxial GaAs have
been removed, followed by the in-situ Pt deposition of 150 nm and a final electrolytic
150 nm thick Au deposition. After the formation of the Schottky junctions the samples
are cut into single diode chips of 100 • 100 /im 2 .
I/V-, C/V— and noise characteristics at 1.5 GHz were recorded by whisker contac-
ting the diode chips soldered to a BeCu whisker post. The whisker consists of a 15 nm
AuNi wire with an electrochemically etched tip. The whisker is soldered to another whis-
ker post, which is mechanically advanced to the diode chip by a micromanipulator for
contacting.
Experimental Results
Several Schottky diodes having different diameters, doping concentrations and doping pro-
files have been fabricated. All diodes show'very good I/V characteristics in agreement
with parameters predicted by the thermionic-field emission for reverse bias and ideality
factors close to unity for forward bias. Fig.3 shows the measured and calculated values of
the forward I/V characteristic of a Schottky diode with 1/im diameter and 2 • 10 17 cm~ 3
doping concentration. The higher value of the ideality factor are due to interfacial states
and to thermionic-field emission for higher doping concentrations, as it can be inferred
Third International Symposium on Space Terahertz Technology Page 637
from fig. 4 in which ideality factors of several diodes manufactured at the Technical Uni-
versity of Darmstadt and at the University of Virginia are presented as a function of
doping concentration.
In fig. 5 and 6 diode noise temperatures are presented. The diode examined in fig. 5 does
not exhibit any trap noise, however due to the low doping concentration, hot electron noise
becomes appreciable at a relatively low current density, its characteristic being fairly well
described by eq. 8 with negligible trap noise. The opposite situation can be inferred from
fig. 6. Noise originating from a mechanism described by eq.7 makes a major contribution
and again the agreement between the measured and calculated values is excellent.
Noise measurements performed at 3 distinct frequencies give an estimation for the
value of r (r ~ 0.2ns) corresponding to very shallow traps at the interface.
Noise characteristics of a diode with a graded doping concentration (fig. 7) in the epi-
layer exhibits the noise temperature similar to the diodes with moderate doping in the
region of the diode operation and lower than that for the 1/xm diode, but the increase of
the noise temperature at higher current densities as is the case for highly doped diodes.
The Ifim diode has already shown excellent results in a 650GHz waveguide mixer
at room temperature. A noise temperature smaller than 2000-ftT and a corresponding
conversion loss of l.bdB have been obtained at an intermediate frequency of 4GHz [16].
These results are comparable to others obtained with smaller diodes. It is expected that
the lower noise temperature in the operating current region of the 0.8/im diode with a
graded doping profile in the epi-layer will enable to even improve these results.
Page 638
Third International Symposium on Space Terahertz Technology
Figure 2: SEM photograph of the fabricated Schottky diode chip with 0.8/jm diodes
cr
0)
o
CD
O
10
10
10
-3
-4
10
10
0.8
0.9
1.0
measured
calculated
1.1
1.2
Diode voltage [V]
Figure 3: Measured and calculated values of the forward I/V characteristic for a diode
with l.Ofim diode diameter and 2 • 10 17 cm~ 3 doping conctration.
Third International Symposium on Space Terahertz Technology
Page 639
O
CO
Doping concentration N D [cm" ]
Figure 4: The ideality factor of Schottky-barrier diodes as a function of the doping con-
centration for diodes fabricated by the Technical University of Darmstadt (DA)
and the University of Virginia (VI)
10000
3
«
a
£
V
o
c
100
i i — i i i i 1 1 1 i i i i i i i i i —
N d =2* 10' 6 cm" 3 , *=3.0Aim. R,=Sn.
77=1.08
10
■
100 1000
current density [A/sqmm]
i i i. i i i
'
10000
Figure 5: Comparison between measured and calculated (from eq. 8) noise characteristics
of a low doped Schottky diode. The crosses indicated the measured data, the
solid line stands for T n + T/, and the dotted line is T n +T k + T t .
Page 640
Third International Symposium on Space Terahertz Technology
10000
3
o
£ 1000
£
V
o
c
I I I I I I
I I I I I I I 1 1
; N d =2 , 10 ,7 cm- J . $ = 1 . 1 M .-n, P, = 1 50.
77 = 1.18
i i i i i i 1 1
100 4
10
-I— I — I I I 1 1 1 I I I '
100 1000
current density [A/sqmm]
1C000
Figure 6: Comparison between measured and calculated (from eq. 8) noise characteristics
of a highly doped Schottky diode. The crosses indicated the measured data,
the solid line stands for T n + 7\ and the dotted line is T„ + 7\ + TV
2000
^
^ 1500
-•— >
CO
<5 1000 •
E
CD
CD
O
500 •
1um (constant doping)
0.8um (graded doping)
i ■ i i i «■ i
10
100
1000
10000
Diode current density [A/mm 2 !
Figure 7: Noise characteristics of a 0.8/im diode with graded junction in comparison to
the lfim diode with constant (2 • 10 17 cro -3 ) doping concentration
Third International Symposium on Space Terahertz Technology Page 641
References
[1] E. Rhoderick and R. Williams, "Metal-Semiconductor Contacts,". Monographs in
Electrical and Electron. Eng., No. 19, Oxford Science Publ., 2 ed., (1988).
[2] M. McColl, "Conversion loss limitations in Schottky barrier mixers," IEEE Trans.
Microwave Theory & Techniques, Vol. MTT-25, (1977), S. 54-59.
[3] T. W. Crowe and R. J. Mattauch, "Conversion loss in GaAs Schottky-barrier mixer
diodes," IEEE Trans. Microwave Theory & Techniques, Vol. MTT-34, (1986), Nr. 7,
S. 753-759.
[4] W. Schottky and E. Spenke, "Zur quanitativen Durchfuhrung der Raumladungs-
und Randschichttheorie der Kristallgleichrichter," Mitt. Zentralabt. Femmeldetech.
Siemens & Halske AG, Vol. 18, (1939), Nr. 3, S. 225-291.
[5] G. Hegazi, A. Jelenski and S. Yngvesson, "Limitations of microwave and millimeter-
wave mixers due to excess noise," in Proc. of 1985 IEEE MTT-S Int. Microwave
Syrup. Dig., (St. Louis, USA), S. 431-434, (June 4th-6th 1985).
[6] Lawson, "Thereshold Signals," MIT Rad. Lab. Series, Vol. 24, (1946), S. 111.
[7] N. Keen and H. Zirath, "Hot-electron noise generation in Gallium Arsenide Schottky
-barrier diodes," Electronics Letters, Vol. 19, (1983), Nr. 20, S. 853-854.
[8] A. Jelenski, E. Kollberg and H. Zirath, "Broad-band noise mechanisms and noise
measurements of metal-semiconductor junctions," IEEE Trans. Microwave Theory
& Techniques, Vol. MTT-34, (1986), Nr. 11, S. 1193-1201.
[9] M. VanVliet and J. R. Fawcett, "Fluctuations due to electronic transitions and trans-
port in solids,". New York Academic Press, (1965).
[10] A. Griib, K. Fricke and H. Hartnagel, "Highly controllable etching of epitaxial GaAs
layers by the pulse etching method," J. of Electrochemical Society, Vol. 138, (1991),
Nr. 3, S. 856-857.
[11] A. Grub, R. Richter and H. Hartnagel, "Electrolytic processes for etching and me-
tal deposition towards nanometre quantum structures," Electronics Letters, Vol. 27,
(1991), Nr. 4, S. 306-307.
Page 642 Third International Symposium on Space Terahertz Technology
[12] H. Grothe and J. Freyer, "Ga(Al)As Molekularstrahlepitaxie fur Submikronbauele-
mente," in Proc. of MIOP 87, (Wiesbaden, Germany), S. 6B/5, (1987).
[13] P. Verlangieri and M. Schneider, "Microfabrication of GaAs Schottky diodes for mul-
tipliers, mixers, and modulators," Int. J. of Infrared and Millimeter Waves, Vol. 6,
(1985), Nr. 12, S. 1191-1201.
[14] W. Peatman and T. Crowe, "Design and fabrication of Q.5(im GaAs Schottky barrier
diodes for low-noise Terahertz receiver applications," Int. J. Infrared and Millimeter
Waves, Vol. 11, (1990), Nr. 3, S. 355-365.
[15] M. Schneider and R. Linke, "Low-noise millimeter-wave mixer diodes prepared by
molecular beam epitaxy (MBE)," Applied Physics Letters, Vol. 31, (1977), Nr. 3,
S. 219-221.
[16] N. Keen, A. Grub, H. Hartnagel, J. Freyer, H. Grothe and R. Zimmermann, "New
submillimeter-wave Schottky-barrier mixer diodes: First results," in Proc. of lQ th
Int. Conference on Infrared & Millimeter-Waves, (Lausanne, Schweiz), (1991).
Third International Symposium on Space Terahertz Technology Page 643
Electrical and Infrared Properties of Thin Niobium Microbolometers near T c
E.N. Grossman, J.E. Sauvageau, and D.G. McDonald ^^3 ""^3 2
Electromagnetic Technology Division ' i
National Institute of Standards and Technology IN \3 O ]m */j* i «7
Boulder, CO
Abstract
Niobium microbolometers approximately 1 /xm wide x 2 /im long x 10 nm thick have been
i
J integrated at the feeds of equiangular spiral antennas made of 200 nm thick Nb. The device's
current-voltage characteristic and infrared responsivity as a function of DC bias voltage were mea-
sured over a range of temperature spanning approximately ± 2 % around T c . The greatest voltage
responsivity occurs well below T c , in a regime where the I-V curve is significantly hysteretic due
to self-heating and resembles the I-V curve of a superconducting microbridge. i , ) 6 "\
The idea of constructing sensitive terahertz-frequency bolometers by combining the principle of
antenna-coupling with that of superconducting transition-edge thermometry has recently received
much attention 1,2 . The idea is to use a lithographic antenna to concentrate the infrared power into
an area much smaller than a square wavelength. At the feed of the antenna are integrated both
an absorbing load (simply a film that is conjugately matched to the antenna impedance at the IR
frequency), and a superconducting transition-edge thermometer, used to monitor the changes in the
absorber's temperature as varying IR powers are applied to it. The absorber and the thermometer
can physically be the same thin-film device, since at IB. frequencies, well above the superconductor's
gap frequency, its impedance is just that of the normal metal. The device is temperature-biased
at the midpoint of the thermometer's resistive transition. A small bias current is applied to it and
its resistance changes monitored by measuring the voltage developed across it. Previous analyses
of these devices have electrically modelled the thermometer in a very simple way, namely as a
lumped (i.e. spatially homogeneous) temperature-dependent ohmic resistance. That is, it has
been assumed that near T c the temperature and current dependent voltage may be separated as
V(I, T) = IR(T). This method of analysis naturally grew out of earlier, quite successful, work 3 on
large area, surface- absorbing bolometers. However, there is some reason to expect that very small
bolometers may behave differently from large ones. In particular, as the superconducting transition
Page 644 Third International Symposium on Space Terahertz Technology
temperature is approached from below, (l-i->0, where t = T/T c ), both the London penetration
depth and the Ginzburg-Landau coherence length diverge, as (1 — t)~ 1 / 2 . At quite realistic values
of (1 — t), one or both length scales may become comparable to or larger than the lateral dimensions
of a small bolometer at the feed of a lithographic microantenna. This can create qualitatively new
physical effects. For example, if the bolometer is comparable in width to the penetration depth,
flux flow conduction is strongly affected because the current distribution due to a single magnetic
vortex is highly perturbed by the bolometer edges. If the the bolometer is comparable in size to the
coherence length, a weak link, or localized region of depressed superconductivity, is created. Such
finite-size effects have been recognized for a long time 4 as important to the the electrical properties
of small microbridges near T c .
Another method of creating a superconducting microbolometer is to monitor the temperature-
induced change in the thermometer's inductance 5,6 . The inductance change is due to the variation
in the pentration depth with temperature discussed above, and is not confined to a very narrow
range of temperature near T c . Such an inductive thermometer can be formed from exactly the same
physical device as the resistive thermometer, namely a thin superconducting film patterned so as
to contact and span the two antenna terminals. Aside from not requiring such precise temperature
control as a transition-edge bolometer, operating well below T c with an inductive thermometer
allows the existence of a large population of unbroken Cooper pairs, which provides a separate,
non-thermal mechanism for far-infrared detection, namely direct pair-breaking. Previous work 7
has indicated that this mechanism can be much faster than bolometric detection, at least in hard
superconductors, such as niobium. Our research on such photoinductive detectors relies on the use
of an integrated DC SQUID (superconducting quantum interference device) to convert changes in
the device's inductance to measurable voltage changes. Unfortunately, we have not yet succeeded in
reliably fabricating a DC SQUID integrated with a photoinductive film and a lithographic antenna.
We have therefore investigated some of the properties of the thin detectors near T c , where the
resistive mode of operation obviates the need for a SQUID.
The devices consist of pure niobium films 10 nm thick, patterned by conventional optical
lithography into bolometers nominally 1 /on wide by 2 /xm long. The bolometer lies at the feed
of a self- complementary spiral antenna with a 65° opening angle, formed from 200 nm thick nio-
bium. (See figure 1.) Reliable, low resistance electrical contact between the antenna feed and the
thin bolometer is accomplished by DC magnetron sputtering the two films successively, with an Al
etch-stop layer between them, in a single deposition run, without breaking vacuum. The antenna,
etch-stop layer, and bolometer are then patterned by a sequence of plasma and wet etches. At low
Third International Symposium on Space Terahertz Technology
Page 645
temperatures, the thin (10 nm) Al etch stop layer is rendered superconducting by the proximity
effect from the much thicker Nb antenna lying on top of it, and can therefore be neglected elec-
trically. The DC normal-state resistance of the bolometers is approximately 20 Cl. At frequencies
well above the Nb energy gap frequency 2A/h sa 700 GHz, but well below the normal electron
relaxation frequency vp/l e = 48 THz 8 , the impedance of the bolometer, both above and below T c ,
should approximate the DC normal-state resistance, (vp is the Fermi velocity and l e the electronic
mean free path.) This implies a loss of about 1.8 dB due to impedance mismatch with the antenna,
if the antenna impedance can be approximated by the quasi-static value of about 75 fi appropriate
to a self-complementary design on a silicon substrate.
10nm
Figure 1. Photograph of the antenna-coupled microbolometer
The diced chips, 1 cm square, were placed circuit-side up against a Si hemisphere, and the
infrared radiation coupled in through the hemisphere and substrate, in the conventional "reverse
microscope" configuration 9 . Mechanical pressure was applied from the circuit side to hold the chip
in place against the hemisphere and to maintain an optical quality contact between the polished
Si surfaces of the substrate and hemisphere. The very wide antenna beam 1 was refocussed by an
off-axis elliptical mirror operating at an angular magnification of 6.3 to a point inside the cryostat
where a cold field stop was placed to reduce to a manageable level the total background power
incident on the chip (i.e. the power in spatial modes other than the one coupled to the antenna).
Page 646
Third International Symposium on Space Terahertz Technology
This is necessary because the hemisphere/chip assembly is mounted on a thermal platform whose
thermal resistance to the heat bath is fairly high in order to allow convenient temperature regulation
and control. Any substantial optical background power would therefore raise the platform's (and
chip's) temperature excessively. The refocussed beam from the elliptical mirror is then collimated
by a cold lens before emerging through the cryostat window. (See figure 2.) The material of the
collimating lens may be chosen to define a broad spectral bandpass and optical power level that
is desired. In the optical measurements described here, a ZnSe lens was used to define a broad,
short wavelength (A < 20 fun) bandpass. The optical input was mechanically chopped between
simple blackbody sources at 300 K and 77 K. Using handbook values 11 for the transmission of
Si and ZnSe, the power coupled into the single spatial mode of the antenna is calculated from
P = ' h ' J£° * 1 j 1 ac , where x = hv/kT is the dimensionless frequency, h is Planck's constant,
k Boltzmann's constant, T(x) the transmittance of the combination of Si and ZnSe lenses, and T
the absolute temperature. This yields a power difference P(3Q0K) - P(77K) = 1.86 nW. When
the reflection losses at the surfaces of the ZnSe and Si lenses and the KRS-5 cryostat window are
included, a total incident optical signal power of 0.65 nW is obtained. The power-weighted mean
frequency of the optical input is 28 THz (wavelength 11 /zm).
s v .v ^-V V-V^^-M-V
ZnSe collimating lensc
Cold field stop c^^ \
Elliptical mirror
Si hemisphere
Bolometer chip
77 K Blackbody
300 K Optical chopper
KRS-5 Cryostat window
Resistive heater
- — Teflon piston
Ge resistance thermometer
Thermal standoff
///, 4 K work surface
Figure 2. Experimental setup for the infrared responsiviiy measurements
The device was electrically biased through the antenna arms by a 100 CI output impedance
Third International Symposium on Space Terahertz Technology Page 647
source. The voltage was monitored (in a 4-terminal configuration) at the outer periphery of the
antenna arms. Since the antenna is formed of much thicker Nb than the bolometer, the critical
temperature of the antenna arms is some 0.8 K higher than that of the bolometer. Therefore, over
the entire temperature range investigated, the antenna arms formed (at DC and audio frequencies) a
superconducting short circuit, and the measured voltage was just that across the bolometer element
itself. As shown in figure 2, a cold transformer, separately calibrated at 4 K, could be switched
into the circuit to raise the signal voltage level and improve the match to the high impedance room
temperature amplifier which followed. To obtain high sensitivity from a small superconducting
detector (i.e. one which is not a long meander line), it is essential to provide such an impedance
transformation. In many cases, however, we did not use the transformer in order to avoid certain
ambiguities in interpretation of the data. For example, it turned out that the bias-dependent
dynamic impedance of the device varied over a huge range; at its highest values, the RC rolloff
frequency created between the post-transformer device impedance and the filter capacitance on the
signal lines became comparable with the optical chopping frequencies (10 - 400 Hz). Obviously,
this would cause spurious systematic structure in the curve of optical response versus bias, which
was avoided by measuring the voltage before the transformer, where the signal impedance is much
lower.
The basic data consist of curves of demodulated optical signal voltage and DC current versus
DC bias voltage, obtained at various temperatures above and below the superconducting transition.
(Since the dynamic impedance of the device is frequently comparable to the 100 CI bias source
impedance, the choice between DC voltage and current as the independent variable is more or less
arbitrary.) The set of data with the most complete temperature coverage is shown in figure 3. It
was taken with the cold transformer switched out of the circuit. As the DC voltage was slowly
swept, the DC current and demodulated optical signal were recorded simultaneously. During all the
measurements, the temperature of the platform was monitored with a Ge resistance thermometer
and actively stabilized, with a precision of about 0.2 mK, by a commercial temperature controller
driving a simple resistive heater. The critical temperature of the bolometer as defined by the
onset of a supercurrent (i.e. vanishing zero-voltage resistance) was 7.978 ± .006 K. The critical
temperature as defined by a zero-voltage resistance that is half the normal state resistance was
7.992 ± .010 K. The data cover a range in reduced temperature from about 2.4 % below to +2.2 %
above the critical temperature, a much wider range than envisaged for "transition-edge" bolometers
in the previous analyses 1,2 .
The most obvious feature of the data is the huge increase in infrared responsivity at low voltage
Page 648
Third International Symposium on Space Terahertz Technology
(00
100
S00 1000 1500
Voltage ((iV)
2000
T 3
7.928 K
to
F
so
>
c
60
o
K)
c
o
o.
40
<0
o
OC
oe
20
S00 1000 1500
Voltage (iiV)
60 ^
- 40
20
20013
250
T : 7.821 K
01
§ 200
>
c
w 150
e
0)
c
O 100
a
a)
O
C 50
500
1000
1500
Voltage (jiV)
100
so
^60 '
40
20
2000
500 1000 1500
Voltage (jiV)
300
500 1000 1500
Voltage (nV)
- 100
120
Voltage (yV)
150
100
500 1000 1500
Voltage (jiV)
500 1000 1500
Voltage (jiV)
Third International Symposium on Space Terahertz Technology
Page 649
CD
>
500 1000 1300 aooo
Voltage (|iV)
500 1000
Voltage (nV
1500
2000
500
Voltage (jiV)
— 80
100
■ 60 3
C
o
20
20ott
E
>
oc
oc
500 1000 1500
Voltage (|iV)
3
o
500 1000 1500 2000
Voltage (nV)
CO
E
o
n
c
o
a
CO
o
Voltage (nV)
500 1000 1500
Voltage (jiV)
2000
Page 650 Third International Symposium on Space Terahertz Technology
and low temperature. (Note the changes in scale of the IB, response from graph to graph in figure
3.) This is partly due to the fact that, with a 100 fl load line (nearly a current source on the scale
of the figure) and voltage detection, the largest infrared signals will be obtained where the device's
dynamic impedance is the greatest. Thus, for the graphs at T < 7.856 K, the large increase in
infrared responsivity corresponds with a large increase in dV/dl at V < 1000 fiV. Indeed, at these
low temperatures, both the dynamic impedance and the IR responsivity apparently diverge as the
bias approaches the point of instantaneous "dropback" to the zero- voltage branch. As this point
is approached, it becomes increasingly difficult to bias the device stably. In these measurements,
bias stability appears to be the practical limitation on increasing the responsivity by this means.
At temperatures T < 7.892 K, the I-V curve is hysteretic, undoubtedly due to ohmic self-heating
(see below). Only the descending branch is plotted in fig. 3. For T = 7.892 K and 7.856 K, the
hysteresis loop encloses only the small horizontal "step" visible at V < 100/zV in the I-V curve,
while at lower temperatures, the difference between the critical current (on the ascending branch)
and the "dropback" current increases.
The data of figure 3 broadly describes the behavior of the IR responsivity over a large area in
parameter space, the "big picture" , but it does not give a very accurate idea of the responsivity
in the region of optimum device performance. Other data we have taken exhibit much higher
responsivities than those of figure 3; values as high as 4.8 x 10 4 V/W after the transformer and
7800 V/W before the transformer have been measured. Even these figures do not really reflect the
device's capabilities, however, since the test conditions were not optimum for the unexpectedly high
device impedances in this operating range. Firstly, the capacitance values for the cold blocking
capacitor and the cryostat feedthru capacitors turned out to be inappropriate for attaining the
full advantage of the transformer, and secondly, the operating temperature was not varied over a
fine enough grid to find the optimum operating point of the bare bolometer. Nonetheless, these
responsivity figures in themselves demonstrate technologically interesting levels of sensitivity for
the bolometer.
The device's dynamic impedance cannot provide the entire explanation for the observed de-
pendence of IR responsivity on bias visible in figure 3. This is clear from the data at T > 7.928 K,
where the dynamic impedance increases at higher bias voltages, yet the strong increase in IR re-
sponsivity at low voltage persists. An effect which certainly must be important in understanding
this data is local ohmic heating. This creates temperature gradients on the chip, in the immediate
vicinity of the bolometer, which of course are not probed or regulated by the Ge thermometer and
temperature control unit mounted on the thermal platform. At the biases typical in figure 3, this is
Third International Symposium on Space Terahertz Technology Page 651
the dominant effect, since the ohmic heating powers are in the range of tens of nW, which translates
into hundreds of mK temperature rise, for typical thermal conductances of 1-5 xlO -7 W/K 12 .
In the previous treatments of superconducting microbolometer performance 1 * 2 , ohmic heating
is incorporated by assuming the bolometer itself remains isothermal, (i.e. by ignoring any spatial
temperature gradients across the bolometer) and by imposing a condition on the DC bias, that
Jvr~ h G t ' < 1> w here G is the thermal conductance of the bolometer to the heat bath. As long as this
condition for avoiding thermal runaway is met, the device is treated as ohmic, with a temperature
dependent resistance, implying that dV/dT = V ^ j T '< and the infrared responsivity should be
linearly proportional to DC voltage. The regime in which this treatment is valid covers a small
region near the origin of the graphs for T > 7.98 K.
For the regime at lower temperature, where the measured infrared responsivity is higher, an
apparently more appropriate treatment is that developed by Skocpol, Beasley, and Tinkham 13,14
to describe self-heating effects and phase-slip centers in microbridges (see figure 7 in ref. 13 for
example). This model explicitly includes spatial temperature gradients. In fact, it assumes that in
general, part of the bridge can be normal and part superconducting. The voltage across the bridge
at any particular substrate temperature and bias current is determined by the length of the region
where T > T c , which is in turn determined by a self- consistent balance between the ohmic heat
generated in that region and the power conducted away from it, which depends on the geometry.
Well below T c , there is strong non-linearity in the I-V curve above the critical current due to the
rapidly changing size of the normal region of the bridge. Presumably, the large infrared response we
see in our devices well below T c is due to the EEL power, dissipated approximately uniformly across
the bolometer's surface area, changing the thermal balance in such way as to move the boundary
between the bolometer's normal and superconducting regions. Very small changes in IR power
apparently move this boundary very far. In some sense, the ohmic heating due to the bias current
adds some positive feedback to the process of driving the bridge normal with infrared heating.
Unfortunately, the theoretical results described in reference 13 cannot be directly applied to
our devices because of a difference in length scales. The microbridges discussed there, which were
fabricated mechanically rather than lithographically, were in general substantially larger than our
devices. On the other hand, the characteristic length scale which figures into their model is the
quasiparticle diffusion length. The devices in that study were formed of tin, a soft superconductor
with a much larger diffusion length than niobium. Therefore, it is very likely on theoretical grounds
that our devices can in fact be treated by that model, appropriately rescaled. More specific state-
ments, however, including any theoretical estimate of the optimal infrared responsivity available
Page 652 Third International Symposium on Space Terahertz Technology
from, this mechanism, must await the results of quantitative modeling.
In conclusion, we have measured the infrared responsivity of antenna-coupled, superconduct-
ing microbolometers as a function of DC bias and temperature, over a wider range than has been
envisaged in previous theoretical treatments. We find the highest responsivity occurs in a regime
well below T c , where the effects of ohmic heating due to the bias current are of dominant impor-
tance. A model developed to describe self-heating and phase slip centers in tin microbridges may
provide the most appropriate framework for analyzing the device's infrared response and electrical
properties in this regime.
We are grateful to Ron Ono for bringing the microbridge self-heating model to our attention.
This work was supported by the Innovative Science and Technology Agency of SDIO, and by NASA,
through the Office of Space Science and Applications.
Third International Symposium on Space Terahertz Technology Page 653
References
1. J. Mees, M. Nahum, and P.L. Richards, Appl. Phys. Lett, 59, p. 2329 (1991)
2. M. Nahum and P.L. Richards, IEEE Trans. Magn., MAG-27, p. 2484 (1991)
3. J. Clarke, G.I. Hoffer, P.L. Richards, and N.H. Yeh, J. Appl. Phys., 48, p.4865 (1978)
4. R.D. Parks, J.M. Mochel, and L.V. Surgent, Phys. Rev. Lett, 13, p. 331a (1964)
5. D.G. McDonald, Appl. Phys. Lett, 50, p. 775 (1987)
6. E.N. Grossman, D.G. McDonald, and J.E. Sauvageau, IEEE Trans. Magn., 27, p. 2677 (1991)
7. M. Johnson, Phys. Rev. Lett, 67, p.374 (1991)
8. The electron relaxation frequency is obtained from the free-electron expression u>l = 3 JS^ja" »
using a DC conductivity a (10 K) = 1.58 X 10 7 m -1 ft _1 , W.H. Henkels and C.J. Kircher, IEEE
Trans. Magn., MAG-13, p. 63 (1977), and an electron mean free path l e = 6.3 nm, S. Wolf, J.J
Kennedy, and M. Nisenoff, J. Vac. Sci. and Tech., 13, p. 145 (1975)
9. D.B. Rutledge, D.P. Neikirk, and D. Kasilingham, chap. 1 in Infrared and Millimeter Waves,
vol. 10, Academic Press, New York, (1983)
10. E.N. Grossman, J.E. Sauvageau, and D.G. McDonald, Appl. Phys. Lett, 59, p. 3225 (1991)
11. E.D. Palik (ed.) Handbook of Optical Constants of Solids, Academic Press, Orlando (1985),
and
E.D. Palik (ed.) Handbook of Optical Constants of Solids II, Academic Press, San Diego (1991)
12. E.N. Grossman, D.G. McDonald, and J.E. Sauvageau, Proceedings of the Second International
Symposium on Space Terahertz Technology, p. 407, JPL, Pasadena, CA (1991)
13. W.J. Skocpol, M.R. Beasley, and M. Tinkham, J. Appl. Phys., 45, p.4054 (1974)
14. W.J. Skocpol, M.R. Beasley, and M. Tinkham, J. Low Temp. Phys., 16, p. 145 (1974)
Page 654 Third International Symposium on Space Terahertz Technology
MEASUREMENTS OF THE SINGLE SIDEBAND SUPPRESSION FOR
A 650 GHZ HETERODYNE RECEIVER
/6?6S%3% S.Crewell&KNett N93~2< •
Institute of Remote Sensing
University of Bremen, FB 1
P.O. Box 330440
D-2800 Bremen 33, FRG
ABSTRACT
A large number of atmospheric trace gases, involved in the process of stratospheric ozone
depletion, show emission features in the submillimeter wavelength range (X=0. 1-1 mm).
High-resolution heterodyne techniques are a particularly useful tool in this spectral region as
vertical distribution of these species can be deduced. Here the receiver has to be operated in
the single sideband (ssb) mode preferably to avoid any interferences between the
contributions in both receiver sidebands. In the 625-655 GHz heterodyne receiver developed
at the University of Bremen a Martin- Puplett interferometer is used as a ssb-filter. A
laboratory set-up has been built up to measure the performance of this interferometer.
INTRODUCTION
Since the detection of stratospheric ozone depletion research activities have been focused on
the study of upper atmospheric chemistry. In the catalytic reactions destroying the ozone the
chlorine monoxide molecule is a key substance. Its influence in the chlorine and the
nitrogen cycle leading to the "Antarctic ozone hole" is described in [1]. CIO and also many
other chemically important molecules, for example HC1, N,0, OCIO, BrO.. , have
emission lines in the submillimeter region [2].
To detect the weak thermal emission of these molecules low-noise heterodyne receivers can
be used. They provide the possibility to observe the line shape of the molecules with a high
spectral resolution allowing to derive the vertical VMR (volume-mixing-ratio) profiles from
the pressure broadening of the line.
Due to the strong water vapor absorption in the troposphere submillimeter measurements of
atmospheric trace gases have to be performed from high flying aircraft.
A 625-655 GHz heterodyne receiver for airborne operation, SUMAS (Submillimeter
Atmospheric Sounder), has been developed at the University of Bremen in order to measure
C10,HC1,0 3 ,N,0.... which all have transitions in this narrow frequency range.
Third International Symposium on Space Terahertz Technology
Page 655
Atmospheric measurements have been performed successfully on the research aircraft
Falcon. The receiver has been further improved in a second project, SMS, with
international participation funded by the ESA/ESTEC (*). The goal of this project was to
develop spaceborne techniques for a submillimeter limb sounder and built up a
demonstration model for airborne operation.
For atmospheric measurements it is desirable to operate the receiver in a ssb mode because
emission lines, quite numerous in this frequency range, in the image band may interfere
with the lines in the signal band. Due to the high ratio of radio frequency (r.f.) to
intermediate frequency (i.f.) an external ssb-filter is preferable. In both radiometers a quasi-
optical Martin-Puplett interferometer is used and the performance of this has to be verified
over the bandwidth of interest.
In the following chapter the ssb transmission measurements for the HCl-frequency using the
SUMAS radiometer will be discussed as an example .
Receiver Design
The SUMAS radiometer as illustrated in Fig.l consists of a quasi-optical section (r.f.) and a
two-stage i.f. secton. In the r.f. part all elements are realized with a quasi-optical design.
signal
input
quasi-
optical
section/
calibration
unit
submm—
wave mixer
-0
1st i.f.
stage
11.08 GHz
i.f. output
3.7 ± 0.6 GHz
2nd i.f.
stage
microwave
mixer
local
oscillator
PL stabilized
636-641 GHz
reference
oscillator
2nd
local
oscillator
1 3.78GHz
Fig.l : Block diagramm of the SUMAS radiometer
(*) Estec Contract 8742/90/NL/PB
involves several industrial
Radiometer Physics, Germany)
Bern. Switzerland: , University of
"Limb Sounder Critical Technologies" This contract
companies (MBB-GmbH-Deutsche Aerospace, Germany;
and several scientific institutes (University of
Bremen. Germany; Chalmers University, Sweden;
University of Florence, Italy; Space Research organisation. The Netherlands).
Page 656 Third International Symposium on Space Terahertz Technology
Total power calibration is performed; a phase wobbler is used to avoid standing waves; two
cascaded Martin-Puplett interferometer serve for ssb-filtering and for diplexing of the local
oscillator (L.O.) and signal radiation, respectiwly. The L.O. is tunable between 636-641
GHz and is phase-locked to a highly stable reference. The quasi-optical Schottky diode
mixer operating at room temperature has a 3 X whisker forming an approximately gaussian
beam. In the 1st i.f. section at 11.08 GHz a low-noise HEMT cooled by liquid nitrogen
acts as a preamplifier. After additional amplifing the signal is down converted by a
microwave mixer to the second i.f. stage centered at 3.7 GHz. Here a post-amplifier and
filter are used to provide an i.f. output with a power level of 0-5 dbm and a bandwidth of
1.2 GHz. A detailed description of the radiometer can be found in [3] and [4].
Single Sideband Filter
For a quasi-optical ssb-filter two main principles, dual-beam and multi-beam
interferometer, exist. The Mach-Zehnder, an amplitude dividing interferometer, and the
Martin-Puplett interferometer, a polarisation rotating device, are examples for a two-beam
interferometer while the Fabry-Perot is the most common multi-beam interferometer (for
additional informations see [5]). The Martin-Puplett interferometer was found well suited to
meet our frequency requirements since it is easy to tune and to adjust. It consists basically
of two rooftop mirrors and a 45° grid. In our radiometer design an aditionally first grid
(90°) serves to define the vertical polarisation for our interferometer input and later to direct
the unwanted sideband to an absorbing load. An incoming vertical polarized signal is
splitted by the 45° grid in two beams with orthogonal polarisations, 45° and 1 35°
respectively. Then the beams are reflected at the rooftop mirrors in a way that the
polarisation rotates by 90°. Now the beam initially reflected by the 45° grid will be
transmitted and the other beam vice versa. The superposition of both beams is given by
their phase difference and therefore the power transmission is a function of the beams
pathlength difference A an the wavelength X :
P v ->v = 0.5 ( 1 + cos(2ttA/A)) (1)
In order to measure HC1 (f HCL =625.912 GHz) the L.O. is adjusted to 636.992 GHz and
the upper sideband frequency (^{,=648.072 GHz) should be suppressed. If the
interferometer should act as a ssb-filter P in (1) is required to be a maximum for the signal
wavelength (A=itiAhcl) ant ^ a minimum for the unwanted sideband (A=X us j, (2n+l)/2).
The pathlength difference can easy be tuned by a micrometer screw that translates one
rooftop mirror while the other is fixed. Although the relative accuracy of the micrometer is
good (Ax=2/xm) an absolute calibration has to be performed for each wavelength. For this
a signal corresponding to the frequency of interest is generated with the help of the
experimental set-up described in the next chapter and coupled to the radiometer input. The
Third International Symposium on Space Terahertz Technology
Page 657
11.0
E
e
c
.2 io.o
'55
o
a.
9.0
6
o
u
o
E 8.0
7.0
SSB-Filter Calibration
i l l i l l ! I
- lsb (625.912 GHz) maxima
•- usb (648.072 GHz) minima
x = 10.322 mm
m = 14
_J 1—1 L_J 1 1 1—1—
9 11
number
13
15 17
Fi g. 2: Calibration curves for the ssb-filter
r. f. signal is detected by the radiometers submillimeter mixer and can be observed at the i. f.
output with a spectrum analyzer or a powermeter. With this set-up the micrometer position
for maximum and minimum transmission can be determined for each sideband. The results
(Fig. 2) show two straight lines, one for the maxima of fya and the other for the minima of
fusb- Now the intersection of both curves give m=14 and A=6.706 mm, the micrometer
position for the ssb mode. The theoretical transmission curves given by (1) with this
micrometer settings yield a suppression of higher then 20 dB over the whole radiometer
bandwidth for the upper sideband and negligible losses for the atmospheric signal (see
Fig. 3).
SSB Transmission (theoretical)
-0.00
=a -o.oi V
t
-0.02 F-
t
-.
d = 6.706 mm
i/lo = 636.992 GHz
i/„ = 11.08 GHz
- -!0
-20
2 -0.03 ^
a t
■n r
-30 "A
5
-0.04
- -40
W
-0.05
-0.6
-0.4
-0.2 0.0
offset frequency
0.2
/ GHz
0.4
0.6
Fip. 3: Theoretical power transmission of the Martm-Puplett interferometer
solid curve : lower sideband (HC1) ; dashed curve : upper sideband
Page 658
Third International Symposium on Space Terahertz Technology
Experimental Setup
The experimental set-up {Fig. 4) used to measure the ssb-filter characteristic consists of
three parts : a source generating the coherent submillimeter signal, the whole radiometer
/
E
E
E
B
U
s
power
meter
or
spectrum
analyzer
-^g>
LO (0)
Multiplier
diplexer [_0(1 )
M1
M2
10(2)
LO(3) <S> M3
W
narrow — band
filter
Fig. 4 : Experimental set-up for measurements of the ssb-characteristics; dashed lines indicate the radiometer
with the integrated ssb-filter and a microwave output stage where the signal is analysed. A
microwave synthesizer(HP 82620) operating in the range 11.. 13 GHz is used as a local
oscillator LO(0). This signal is then multiplied by a quasi -optical schottky diode mixer
mounted on a linear translation stage .This allows to move the block orthogonal to the
axis of the radiometers antenna main lobe with a resolution of 0.5 /zm in x and z direction.
Additionally the multiplier signal can be matched to the receiver by a focusing (f=90mm)
Teflon lens and it is possible to optimize the lens distance by a further linear translator. A
coupling efficiency of more than 75 % can be attained theoretically by this set-up between
both beams. In the quasi-optical part of the receiver the beam propagates through the ssb-
filter and through the LO(l) diplexer and mixed down by the submillimeter mixer Ml of
the radiometer. The signal is converted down to the second i.f. inside the radiometer by a
microwave mixer M2 to the receiver output. In the analysing microwave stage a constant
i.f. signal of 150 MHz is generated by a further microwave mixer that is driven by another
tuneable microwave synthesizer LO(3). After narrowband-filtering the signal can be
detected by a spectrum analyzer or a powermeter.
Third International Symposium on Space Terahertz Technology
Page 659
For a defined setting of the ssb-filter the incoupled signal can be tuned to any frequency in
the lower or upper sideband including the whole bandwidth via the LO(0) . An experiment
computer with a IEEE-bus system controlling the LO(0) provides a simultaneous tuning of
the LO(3) that drives the Mixer 3 . The generated signal i.f. signal at 150 MHz is now
monitored on a spectrum analyzer and recorded also via IEEE.
Measurement Results
If the generated signal corresponds to the HC1 frequency largest power is obtained using the
49th harmonic of the LO(0) signal. Then a dynamic range of about 23 dB can be achieved
for the measurements of the single sideband characteristics. The relative transmission curves
for both sidebands with a given setting of the ssb-filter are obtained by normalyzing the
measurement curves to the data measured for a pathlength difference A=0. In addition, the
output signal without the generated submillimeter radiation was recorded at the beginning
and the end of the measurements in order to eliminate baseline and drift effects of the
radiometer. The results shown in Fig. 5, here with a resolution of 50 MHz, indicate that the
lower sideband is not affected by the setting of the ssb-filter corresponding to the theoretical
calculations (Fig. 3) while the upper sideband is suppressed more than 20 dB over the whole
bandwidth. This measurement gives a minimum value for the suppression of the unwanted
sideband because it is limited by the strength of the generated signal. An improvement of
this is expected by a further optimization of the multipliers antenna properties. In order to
reduce the noise in the curves (Fig. 5) a new measurement procedure with a sampling of
data points will be applied during future measurements.
To improve the dynamic range a measurement has been performed using the high-resolution
chirp Transform Spectrometer (CTS) on loan from the Max-Planck Institute for Aeronomy
2.0 i
-3.0 -
m
•o
-8.0
B-13.0
6
CO
C
£ -18.0
-2J.0 -
-28.0 '■■■' '
i t i i i i i i i i i i i i i i i i i i i
HCL, 19 March 92
A=6.706 mm
f = 625.912 GHz (HCI)
f = 648.072 Ghz (usb)
-0.60 -0.40 -0.20 0.00 0.20 0.40
offset frequency / GHz
Fig. 5 : Measured power transmission of the Martin- Puplett interferometer
solid curve : lower sideband (HCI) ; dashed curve : upper sideband
0.60
Page 660 Third International Symposium on Space Terahertz Technology
to analyze the i.f. output. The receiver, here the SMS, was operated in a radimetric mode
using the quasi-optical calibration unit (Fig.l) and adjusted in order to measure CIO. A
submillimeter signal corresponding at the center frequency of the image sideband was
generated and observed with a CTS resolution of 400 KHz resulting in a side suppression of
about 32 dB. This measurement required an integration of 4- 5 .
Conclusions
For the performance of atmospheric measurements it is desirable to operate the receiver in a
ssb mode. The characteristics of a Martin-Puplett interferometer used as a ssb-filter in a
625-655 GHz heterodyne receiver has been measured by a laboratory setup. When the ssb-
filter is adjusted to measure at the HC1 frequency, for example, a suppression of more than
20 dB was obtained over a bandwidth of 1.2 GHz. Up to now the measurement is limited
by the strength of the generated submillimeter signal which is expected to be improved by
optimizing the antenna. A measurement using a narrow band filter shows that the
performance of the Martin-Puplett interferometer at the center frequency gives a
suppression of about 32 dB.
References
[1] J.C.Farman, B.G.Gardiner, J.D.Shanklin, "Large Losses of Total Ozone in Antarctica
reveal seasonal C10 x /NO x interaction", Nature, Vol. 315, 16 May 1985
[2] J. W. Waters, "Microwave Limb-Sounder of Earth's Upper Atmosphere", Attn. Res., Vol.
23, pp 391-410, 1989
[3] H.Nett, S.Crewell, K.Kiinzi, "Heterodyne detection of stratosperic trace gases at sub-
mm frequencies", IGARSS 9 7, Digest, Helsinki 91
|4] H.Nett, S.Crewell, K.Kiinzi, "A 625-650 GHz Heterodyne Receiver for Airborne
Operation", 16th Int. conf. on Infrared and Millimeter Waves Symposium, Digest, Vol.
1576, pp 460-461, 1991
[5] P. F. Goldsmith, Infrared and Millimeter Waves. Vol. 6, Ed.Button.K.S.. Academic Press.
New York, pp 277-342, 1982
Third International Symposium on Space Terahertz Technology Page 661
N93-27781
InGaAs/InP HETERO EPITAXIAL SCHOTTKY BARRIER ^SS^~3 3
DIODES FOR TERAHERTZ APPLICATIONS
Udayan V. Bhapkar, Yongjun Li, and Robert J. Mattauch
Semiconductor Device Laboratory
n Department of Electrical Engineering
The University of Virginia
Charlottesville, VA 22903-2442
ABSTRACT
This paper explores the feasibility of planar, sub-harmonically pumped, anti-parallel
|InGaAs/InP heteroepitaxial Schottky diodes for terahertz applications. We present calculations
'of the (I-V) characteristics of such diodes using a numerical model that considers tunneling. We
also present noise and conversion loss predictions of diode mixers operated at 500 GHz, and
obtained from a multi-port mixer analysis, using the I-V characteristics predicted by our model.
Our calculations indicate that InGaAs/InP heteroepitaxial Schottky barrier diodes are
expected to have an I-V characteristic with an ideality factor comparable to that of GaAs
Schottky diodes. However, the reverse saturation current of InGaAs/InP diodes is expected to be
much greater than that of GaAs diodes. These predictions are confirmed by experiment. The
mixer analyses predict that sub-harmonically pumped anti-parallel InGaAs/InP diode mixers are
expected to offer a 2 dB greater conversion loss and a somewhat higher single sideband noise
temperature than their GaAs counterparts. More importantly, the InGaAs/InP devices are
predicted to require only one-tenth of the local oscillator power required by similar GaAs diodes.
This work has been supported by the National Science Foundation under Grant ECS-9113123 and by NASA through the University of
Michigan under Grant Z-25251.
Page 662 Third International Symposium on Space Terahertz Technology
I. Introduction
GaAs Schottky diodes are frequently used as mixer elements in heterodyne receivers for the
few hundred gigahertz to few terahertz frequency range [1], At present a major limitation on
these devices for space-based applications is the difficulty in obtaining sufficient local oscillator
(LO) power from solid state sources; the maximum available LO power decreases sharply with
increasing LO frequency. One approach to overcoming this limitation is to use sub-
harmonically pumped, anti-parallel diode pairs, which halves the frequency at which the LO
power is needed to a range where obtaining sufficient LO power is less of an obstacle. Standard
GaAs Schottky diodes have a large turn-on voltage, and consequently require a substantial
applied bias to minimize LO power requirements and conversion loss. Unfortunately, it is not
feasible for each diode to be biased individually in an anti-parallel configuration. To reduce the
LO power requirement, InGaAs has been proposed as a material for use in sub-harmonically
pumped, anti-parallel diode structures [2]. Schottky barriers formed from InGaAs have a height
that decreases with increasing indium mole fraction. The resulting lower turn-on voltage of
these diodes suggests that they will require smaller LO voltages, and therefore power, for
optimum performance. An added benefit of using InGaAs instead of GaAs is its superior
mobility, which will lead to a lower series resistance, which in turn will reduce the conversion
loss.
Of crucial importance to both the conversion loss and noise of the mixer is the diode I-V
characteristic. We have previously reported on a Schottky diode current-voltage analysis that
considers electron tunneling and image force lowering [2]. In diodes with epitaxial layers doped
to greater than about 5.0 x 10 16 cm -3 , electron tunneling significantly affects the diode ideality
factor.
Third International Symposium on Space Terahertz Technology Page 663
We present the results of conversion loss and mixer noise calculations using the I-V
characteristics obtained from our diode model. A single diode equivalent circuit was used to
model diode mixers in the sub-harmonically pumped, anti-parallel configuration. We have used
a computer program by P. Seigel to perform the analysis [3], which we have modified to use the
current-voltage model discussed in this paper, rather than the standard thermionic-emission
model.
We also discuss fabrication technology currently being developed for a planar, anti-parallel
InGaAs/InP diode.
n. Diode Model
A. Anti-Parallel Planar Diodes
An electron micrograph of a GaAs anti-parallel planar diode is shown in Figure 1 [1]. The
Ino.5Gao.5As/InP anti-parallel planar diodes being designed in this study will have an identical
geometry. A cross-section of such a diode is shown in Figure 2. A circuit model of a planar
diode showing the major parasitics is shown in Figure 3. The junction capacitance, Cj, and the
series resistance, R s , are the dominant parasitics at high frequencies, and should be minimized
for optimum conversion performance. The effects of the pad-to-pad capacitance, Cpp, and the
finger inductance, L s , are smaller, and have not been considered in the mixer calculations
presented in this paper. Furthermore, the junction conductance itself may deviate from ideal
exponential behavior, particularly at high forward or reverse bias, and this can also affect the
conversion performance. This effect is considered in the Schottky I-V analysis presented in this
paper.
Page 664 Third International Symposium on Space Terahertz Technology
The series resistance of a Schottky diode can be reduced by increasing the conductivity of
the epitaxial layer and/or the substrate, or by reducing the thickness of the undepleted epitaxial
layer. The epilayer conductivity can be increased by increasing its doping concentration,
however, this involves a trade-off: the diode ideality is also decreased. The use of InGaAs,
which has a greater electron mobility than GaAs, allows for a high conductivity epilayer, and
avoids this trade-off, as we shall show later. In addition, the epilayer in InGaAs diodes can be
made much thinner than in GaAs diodes. This is because the epilayer is generally made slightly
thicker than the zero-biased depletion depth, which is proportional to the square root of the
Schottky barrier height. Therefore, in equally doped material, the zero-biased depletion depth in
Ino.53Gao.47 As is about half that in GaAs.
The zero-biased junction capacitance of InGaAs diodes is greater than that of comparable
GaAs diodes because the junction capacitance is, to first order, inversely proportional to the
depletion depth. However, at the current densities reached in normal operation, the typical
junction capacitances of the diodes are expected to be comparable. This is because the depletion
depth depends on the remaining barrier, that is, the Schottky barrier height minus the applied
voltage, and the current density is roughly proportional to the exponential of the remaining
barrier.
B. Schottky I -V Model
We shall briefly outline the quantum-mechanical transmission current-voltage model,
which is described in greater detail elsewhere [4], and is largely based on the work of Chang,
Crowell, and Sze [5,6].
The current density is calculated through a numerical evaluation of the equation
Third International Symposium on Space Terahertz Technology
Page 665
J(V) =
RT
J dEn
*(E n ,V)[F s (E n ,V)-F m (E n )]
(1)
where R is the Richardson constant in the semiconductor, T is the temperature, k is Boltzmann's
constant, x is the electron transmission coefficient, F s and F m are the semiconductor and metal
distribution functions respectively, V is the applied bias, E„ is the component of the incident
electron energy normal to the metal, and E^ is the minimum allowed electron energy. E m ;n
corresponds to the conduction band minimum in the metal or the semiconductor, depending on
the applied bias.
The electron transmission coefficient is obtained through a one dimensional solution of
Schrodinger's equation covering the region of the Schottky barrier, including the effects of
image force lowering. The transmission coefficient has been shown to vary sharply with the
electron energy, and ranges from near zero for electrons with energies much below the barrier
maximum, to near unity for electrons with energies a few kT greater than the barrier maximum
[4].
We have used a drifted- Max wellian to model the electron energy distribution, which is
given by [7]
F =
m
*v*
(27tkT)
Vi
exp
-m (v n - v d y
2kT
(2)
where v n is the normal component of electron velocity, and v^ is the drift velocity, defined as
Vd =
qNj'
(3)
where Nj is the concentration of the ionized donors in the epitaxial layer outside the depletion
region. The current density and drift velocity are calculated iteratively using (1-3) until they
Page 666 Third International Symposium on Space Terahertz Technology
converge to the desired level of accuracy.
C. I-V Predictions
Figures 4 and 5a-b show the I-V characteristics of the metal-semiconductor junctions
(denoted by Vj) of GaAs and Ino.5Gao.5As diodes respectively, as predicted by the algorithm
discussed in Section B. Vj. Table 1 shows the diode parameters used. The reverse current in a
Ino.5Gao.5 As diode is predicted to be several orders of magnitude greater than in a GaAs diode.
However, it is still nearly two orders of magnitude smaller than the maximum forward current.
Thus, while the reverse current is not negligible, it is not expected to degrade mixer performance
drastically, as we shall demonstrate in the next section of this paper. It is also noteworthy that
the voltage-dependent ideality factors of the two diodes at a forward current density of
10 4 A/cm 2 (which corresponds to a current of 0.08 mA in a 1 |im diameter diode) are nearly
identical.
Figures 5a-b also show the overall I-V characteristic of a whisker-contacted
Ino.5Gao.5As/InP diode fabricated in our laboratory, and includes the effects of the series
resistance. These whisker-contacted diodes are research devices only, and will be superseded by
planar devices. The data agree well with our model, in which the only adjustable parameter is
the Schottky barrier height.
in. Mixer Analysis
The conversion loss and noise calculations presented in this paper have used the multiple
reflection algorithm developed by Held and Kerr [8]. A FORTRAN language computer program
by P. Seigel, known as GISSMTX [3], (with a few modifications) was used to perform the
Third International Symposium on Space Terahertz Technology Page 667
calculations. The modifications to the program include the use of the I-V model described
above, instead of the thermionic emission model, as well as changes to permit the calculation of
subharmonically pumped mixer performance using the single diode equivalent circuit.
A . Held and Kerr Mixer Analysis
The Held and Kerr mixer analysis is based on the assumption of a large signal LO source,
upon which is superimposed a small signal RF. The time-dependent conductance of the diode is
determined through a non-linear analysis. The diode waveform is then resolved into the small
signal admittance, and is represented in the frequency domain through its Fourier coefficients.
These coefficients are then used to calculate the noise and conversion performance of the diode
mixer circuit. The analysis assumes a knowledge of the diode parameters as well as the
embedding impedances at the mixing frequencies. The analysis is amply described elsewhere,
and therefore we shall not outline the details of the analysis [8].
The mixer performance is in general affected by the embedding impedances presented to
the diode by the mixer block at several sideband and LO harmonic frequencies. These
parameters can be obtained through a characterization of the diode mount; however, our
investigation has not yet progressed to that point. Due to the absence of information on the
embedding impedances, we have chosen to simplify and standardize our analysis by representing
all higher order mixing frequencies as short circuits. In general, the use of such an
approximation will slightly underestimate the conversion loss and noise temperature. The
embedding impedances we have assumed are given in Table 2.
The performance of sub-harmonically pumped, anti-parallel diodes was estimated through
use of the single diode equivalent circuit [9]. In this circuit model, the odd-harmonic embedding
impedance is equal to twice that presented to a single diode, and the even-harmonic embedding
Page 668 Third International Symposium on Space Terahertz Technology
impedances are set to zero.
B. Calculations
Figure 6 shows the predicted upper- sideband (USB) conversion loss of GaAs and
Ino.5Gao.5As single diode mixers at 505 GHz, as a function of the available LO power. Zero
applied bias is assumed. The diode parameters used in the mixer analysis are given in Table 1.
The Ino.5Gao.5As diode will offer a minimum conversion loss of about 7 dB with Plq equal to
0.2 mW, in comparison to the GaAs diode, which will offer about 8 dB conversion loss with
Plo equal to nearly 2 mW. The superior predicted conversion loss of the Ino.5Gao.5As diode is
due primarily to its lower series resistance, and its lower LO power requirement with zero bias is
due to its lower barrier height.
Figures 7 and 8 show the predicted upper-sideband (USB) conversion loss and noise
temperature respectively, of sub-harmonically pumped, anti-parallel GaAs and Ino.5Gao.5As
diodes, as a function of the available LO power per diode. The total LO power required by the
diode pair is therefore twice the amount shown. An LO frequency of 250 GHz and a signal
frequency of 505 GHz were assumed. The minimum conversion loss of the Ino.5Gao.5As diode
pair is predicted to be about 1 1 dB with 0.2 mW of total available LO power, compared to about
9 dB with about 2 mW of available LO power for the GaAs diode pair. The minimum USB
noise temperatures of the Ino.5Gao.5As and GaAs diode pairs are predicted to be about 2000 K
and 1300 K respectively. Thus, Ino.5Gao.5As/InP diode anti-parallel mixers are expected to
reduce the LO power requirement by at least an order of magnitude compared to that of similar
GaAs diode mixers, while increasing the conversion loss and noise by no more than 2 dB and 50
percent respectively. The RF performance of Ino.5Gao.5As/InP diodes in an anti-parallel
configuration is expected to be somewhat degraded from that of single diodes of the same
Third International Symposium on Space Terahertz Technology Page 669
material and with similar parameters. We believe this is due to the relatively high reverse
saturation current of these diodes. However, this drawback is small in comparison to their
primary advantage: they will require LO sources at frequencies half of the signal frequency.
IV. Diode Fabrication
The objective of this facet of our work was to develop a device fabrication technology
which will enable us to produce predictable planar, anti-parallel Ino.53Gao.47 As mixer diodes
with reliable electrical characteristics. The fabrication procedure for planar, anti-parallel diodes
on Ino.53Gao.47 As is very similar to that on GaAs. A highly abbreviated outline follows.
1. Active Layer Thinning: The active layer of Ino.53Gao.47 As was intentionally grown
thicker than the theoretical zero-bias depletion thickness. This allowed us to adjust the
actual epitaxial layer thickness by using an electrochemical thinning technique. The
actual epilayer thickness was measured using a standard C(V) profiling technique but
with a 10 MHz frequency to allow the component of current through the space-charge
capacitance to be dominant. This provided us a way to optimize the I-V characteristic by
changing the actual active layer thickness.
2. Oxide Deposition. A thin (6000 A) layer of SiC>2 was pyrolytically deposited on the
active layer.
3. Ohmic Contact Formation. An ohmic contact was formed by electroplating Sn-
Ni/Ni/Au on n" 1-1 " Ino.53Gao.47 As and subsequently alloying at 400° C. The TLM pattern
test has shown that the Sn-Ni/Ni/Au ohmic contact on Ino.53Gao.47 As is about
2-3 x 10" 6 Qcm 2 , which is one order of magnitude better than that on GaAs.
4. Anode Definition. Standard photolithography and reactive ion etching were
employed to define anode windows in the Si02 layer. The anode metals (Pt and Au)
were DC electroplated through these windows onto the underlying active Ino.53Gao.47 As
layer to form the Schottky diodes.
5. Anode Contact Finger. Conductive thin films of chromium and gold were first
deposited over the entire wafer through use of a sputtering system. Next, the fingers
were electroplated over a photolithographically defined region. Both dry and wet etching
were used to remove the thin chromium/gold film covering the wafer, leaving the contact
fingers in place.
Page 670 Third International Symposium on Space Terahertz Technology
6. Surface Channel Etching. A final photolithography step defines the surface channel.
Buffered hydrofluoric acid was first used to remove the SiC»2 in the surface channel.
Then, the conducting Ino.53Gao.47 As between the pads was removed by using H3PO4 :
H2O2 : H2O, which provides the desired etch profile for different InGaAs crystal
directions, thus allowing the finger to be undercut.
Early results of ohmic contact formation have shown a problem associated with alloying:
the oxide near the anode region was damaged after alloying. This problem was solved by
changing the plating parameters and the alloying temperature without changing the quality of the
contact. The most crucial step in device fabrication is anode formation. Early attempts at anode
formation on planar, anti-parallel Ino.53Gao.47 As resulted in a low breakdown voltage, as well as
instability and nonuniformity of the I-V. These problems were overcome by optimization of the
plating parameters. Profiles resulting from several chemical etchants have been investigated
with respect to the desired profile in forming the surface channel, which undercuts the anode
fingers and protects the anode region. The H3PO4 : H2O2 : H2O family has been found to
provide the desired results.
V. Conclusion
In this paper we have presented I-V calculations of Ino.5Gao.5As/InP Shottky barrier diodes.
The model we have used considers electron tunneling, image force barrier lowering, and the
effect of a drifted Maxwellian electron distribution. The model has been shown to agree well
with experimental data on whisker-contacted Ino.5Gao.5As/InP and GaAs diodes fabricated in
our laboratory. The I-V characteristics of Ino.5Gao.5As/InP diodes have been shown to be
similar to those of GaAs diodes, but are displaced in voltage and have a higher reverse saturation
current.
Third International Symposium on Space Terahertz Technology Page 671
The I-V characteristics obtained from our model were used to predict the mixer
performance of both single and sub-harmonically pumped, anti-parallel diodes. A modified form
of Seigel and Kerr's analysis was used. The calculations show that in unbiased operation,
Ino.5Gao.5As/InP single diode mixers will offer conversion performance equal to that of
comparable GaAs diode mixers, and require only one-tenth the LO power. Furthermore, sub-
harmonically pumped, anti-parallel Ino.5Gao.5As/InP diode mixers are expected to offer
performance nearly as good as that of the best GaAs diode mixers, but will require one-tenth as
much LO power to achieve their optimum performance.
The fabrication techniques for anti-parallel, Ino.5Gao.5As/InP planar diodes with surface
channels have been extensively investigated. Difficulties in anode plating and ohmic contact
formation have been resolved, and suitable chemical etchants necessary for the fabrication
sequence have been found.
REFERENCES
[1] T.W. Crowe and W.C.B. Peatman, "GaAs Schottky Diodes for Mixing Applications
Beyond 1 THz," Proc. Second Int. Symp. on Space Terahertz Tech., Feb. 1991, pp. 323-
339.
[2] U.V. Bhapkar, T.A. Brennan, and RJ. Mattauch, "InGaAs Schottky Barrier Mixer Diodes
for Minimum Conversion Loss and Low LO Power Requirements at Terahertz
Frequencies," Proc. Second Int. Symp. on Space Terahertz Tech., Feb. 1991, pp. 371-388.
[3] P.H. Seigel and A.R. Kerr, "The Measured and Computed Performance of a 140-220 GHz
Schottky Diode Mixer," IEEE Transactions, Vol. MTT-32, Feb. 1984, pp. 1579-1590.
[4] U.V. Bhapkar and R.J. Mattauch, "The Numerical Simulation of the Current- Voltage
Characteristics of Heteroepitaxial Schottky Barrier Diodes," not yet published.
[5] C.R. Crowell and S.M. Sze, "Quantum-Mechanical Reflection of Electrons at Metal-
Semiconductors Barriers: Electron Transport in Semiconductor-Metal-Semiconductor
Structures," Journal of Applied Physics, Vol. 37, No. 7, June 1966, pp. 2683-2689.
[6] C.Y. Chang and S.M. Sze, "Carrier Transport Across Metal-Semiconductor Barriers,"
Solid-State Electronics, Vol. 13, pp. 727-740.
Page 672
Third International Symposium on Space Terahertz Technology
[7] T. Vogelsang and W. Hansen, "The Electron High-Energy Distribution Function: A
Comparison of Analytical Models with Monte Carlo Calculations," /. Appl. Phys., Vol. 69,
No. 6, 1991, pp. 3592-3595.
D.N. Held and A.R. Kerr, "Conversion Loss and Noise of Microwave and Millimeter- Wave
Mixers: Part I ~ Theory," IEEE Transactions, Vol. MTT-26, Feb. 1978, pp. 49-55.
[8]
[9] S.A. Maas, Nonlinear Microwave Circuits, Artech House, 1988, pp. 232-237
Table 1. Diode parameters.
diode materials
GaAs
Ino.5Gao.5As/InP
anode diameter
0.5 urn
0.5 urn
ohmic contact width
50 urn
50 urn
ohmic contact length
50 urn
50 ^im
chip thickness
125 urn
125 urn
active layer thickness
0.110 um
0.065 urn
<t>B
0.950 eV
0.277 eV
active layer doping
1.5 x 10 17 cm -3
1.5 x 10 17 cm -3
buffer layer doping
2.0 x 10 18 cm -3
2.0 x 10 18 cm -3
^jo
1.22 fF
2.10 fF
*~jmax
10.0,fF
10.0 fF
Isat
7.85 x 10 -17 A
8.25 x 10" 7 A
imax
5.0 x 10" 3 A
5.0 x 10 -3 A
n
1.12
1.09
R s
15.0 Q
9.0 Q
Table 2. Mixer embedding impedances.
corf
75 a
'"image
75 Q
COLO
50 Q
COif
matched
Third International Symposium on Space Terahertz Technology
Page 673
Figure 1. Electron micrograph of GaAs anti-parallel planar diode.
Si
2
\
Anode
./
Anode Finger v
<A
Anode Pad
T5
V
Ohmic Contact Pad
\ ourrace r~
n +
n ++ Ino.53Gao.47As
I Channel /
\ (Air) /
n""" Ino.53Gao.47 As
\ 1
semi-insulating InP
Figure 2. Cross-section of anti-parallel, planar InGaAs/InP diode (not to scale).
Page 674
Third International Symposium on Space Terahertz Technology
app
■OTV-
Cpp
c i
A/V
Re
Figure 3. Circuit model of planar diode.
2\ in4
J (A/cm z ) 10 4 r
Figure 4. Forward current-voltage characteristic of GaAs Schottky diode.
Third International Symposium on Space Terahertz Technology
Page 675
• ^ff**
10 5
I
<***
10 4
/o
yo
J (A/cm 2 ) 10 3
v app
10 2
-J
...Vj
'■
o
o data
10 1
3
>
10°
1
i
1
0.0
0.1
0.2
0.3
V(V)
Figure 5a. Forward current- voltage characteristic of Ino.5Gao.5As/InP Schottky diode.
1000
500-
J (A/cm 2 )
-500-
-1000
Y
v app
000 data
I:
■ ■ ■ •
-0.3
-0.2 -0.1
V(V)
0.0 0.1
Figure 5b. Reverse current- voltage characteristic of Ino.5Gao.5As/InP Schottky diode.
Page 676
Third International Symposium on Space Terahertz Technology
(10 5 A/cm 2 ) 2
Figure 5c. Current-voltage characteristic of the heterojunction
in an Ino.5Gao.5As/InP Schottky diode.
20
15
L USB (dB) 10 -
5-
10
1-2
V
Ino.5Gao.5AS
GaAs
—1 1 I ■ ■ ■ 1 « I t t 1 |_
10
1-1
10 u
10 1
Plo (mW)
Figure 6. Conversion loss (USB) versus LO power of single
Schottky diodes at 505 GHz.
Third International Symposium on Space Terahertz Technology
Page 677
••■'-'s
20
15-
L USB (dB) 10
10"
\J
Ino.5Gao.5As
GaAs
1 1
10 _1 10°
Plo (mW)
10 1
Figure 7. Conversion loss (USB) versus LO power per diode of sub-harmonically
pumped, anti-parallel Schottky diodes at 505 GHz.
10'
Tusb(K) 10 3
10 2
Ino.5Gao.5As
GaAs
Iff
,-2
-i — ' ■ ■ ■ ■ I
10" 1 10°
Plo (mW)
1 1 1 1 1 1
10 1
Figure 8. Noise temperature (USB) versus LO power per diode of sub-harmonically
pumped, anti-parallel Schottky diodes at 505 GHz.
Page 678 Third International Symposium on Space Terahertz Technology
N93-27 t&Zi
* A BROADBAND THz RECEIVER FOR LOW
V BACKGROUND SPACE APPLICATIONS
C. Hagmann, D. J. Benford, A. C. Clapp, P. L. Richards, and P. Timbiet
Department of Physics, University of California, Berkeley CA 94720
ABSTRACT
have developed a sensitive bolometric receiver for low background
applications. In a 10 % bandwidth at 1 THz, this receiver is
approximately 100 times more sensitive than a quantum limited heterodyne
receiver with a 1 GHz IF bandwidth. This receiver is designed to be used for
the long wavelength band (200-700 (im) in the MIPS instrument on NASA's
SIRTF satellite. The bolometers are cooled to 100 mK by an adiabatic
demagnetization refrigerator. Roughly 60 g of cesium chrome alum salt is
partially demagnetized to 100 mK, followed by a slow regulated downramp to
compensate for the heat leak. The hold time- of the ADR system is about 18
hours with a temperature stability of AT rms = 10 |iK. The composite bolometers
have electrical reponsivities of 10 9 V/W and electrical NEP's of about 3xl0" 17
W/VHz. The bolometer signals are read out by JFET preamplifiers located on
the helium plate and operated at 120 K. We have addressed a number of space
qualification issues, such as the development of an analog magnet controller,
construction of a cryogenic shake-table for bolometers and selection of the
paramagnetic salt CCA which can survive a bakeout at 50 °C. The receiver is
scheduled to be flown in the spring of 1992 on a balloon telescope. This flight
has a dual purpose. One is to provide a realistic test of the capabilities of the
new receiver. The other is to search for anisotropics in the cosmic microwave
background on scales of a few degrees.
f Present address: Dept. of Physics, Brown University, Providence RI 02912
Third International Symposium on Space Terahertz Technology Page 679
INTRODUCTION
Since the highly sucessful flight of the Infrared Astronomical Satellite
(IRAS) in 1982, much work has been devoted towards the development of a
new generation of low background receivers. NASA's Space Infrared
Telescope Facility (SIRTF) is a LHe cooled long-life instrument, which will
allow background or confusion limited observations for wavelengths ranging
from 2-700 urn. The long wavelength band covers the spectral range of 200-
700 u.m, which is 0.4-1.5 THz. Scientific targets for the long wavelength
capability of SIRTF include quasars, brown dwarfs, protogalaxies and the
cosmic microwave background.
In principle either direct detectors or heterodyne receivers could be used at
this wavelength. For broadband photometry, direct detectors are more
sensitive than heterodyne receivers due to their wider bandwidth and lack of
quantum noise. The baseline instrument for long wavelength photometry on
SIRTF is a bolometric receiver operating at 0.1 K and cooled by an adiabatic
demagnetization refrigerator (ADR). A prototype of this receiver has been
constructed and will be flown on a balloon borne telescope as a realistic test of
its flight worthiness.
REFRIGERATOR
The 100 mK refrigerator [1], shown in Figure 1, operates by adiabatically
demagnetizing a paramagnetic salt. The ADR cycle starts by thermally
shorting the 100 mK stage to the helium bath cold plate at T=1.5 K. This is
accomplished with a mechanical heat switch actuated by a superconducting
solenoid located on the helium plate. In the 'ON' position, a current of 100
mA actuates the solenoid, which forces two jaws to clamp a cold finger
Page 680 Third International Symposium on Space Terahertz Technology
attached to the 100 mK stage. The resulting pressure contact has a thermal
conductance of about 10 mW/K at T=1.5 K. In the 'OFF position, a spring
pushes the jaws apart.
The field for magnetizing the paramagnetic salt is produced by a
conduction cooled superconducting solenoid with a rated central field of 2.5 T
for 6 A of current. The magnet is surrounded by a ferromagnetic shield made
of vanadium permendur to reduce the stray field to insignificant levels and to
increase the field homogeneity within the coil. This shielding material was
chosen because of its high saturation flux density and small remanence.
Our ADR uses the hydrated paramagnetic salt cesium chrome alum CsCr
(S0 4 ) 2 -12 H 2 (CCA) as its working substance. The magnetic ions Cr 3+ (/=3/2)
in this salt have a density of 2.1 xlO 21 cnv 3 and order at around T= 20 mK due
to their weak magnetic interaction. To make good thermal contact between
the salt and the rest of the 100 mK stage, the salt is grown directly on a
skeleton of 200 gold wires. The 60 g of salt in the pill is approximately 0.1
mole. The salt is corrosive and has a tendency to dehydrate in air. To prevent
any degradation, the salt pill is sealed in a stainless steel can; the gold wires are
brought outside through a hardsoldered seal and are hardsoldered to a copper
cold stage.
The salt pill is suspended from the surrounding 1.5 K cold plate with two
sets of six Kevlar ropes on each end. The suspension can be made very stiff
because of the high tensile strength of Kevlar; the fundamental resonant
frequency is about 200 Hz. The heat leak through the Kevlar suspension is
measured to be 0.25 uW.
During isothermal magnetization, the magnetic field is ramped up to 2.5
Tesla and then held constant while the cold stage is allowed to equilibrate for
about 30 minutes. The chromium spins align in the external field and the
entropy is reduced by about 70 %. The heat switch is then opened and the
magnetic field is ramped down, cooling the stage adiabatically. When the stage
temperature reaches T= 100 mK, the fast downramp is stopped and a PC based
temperature controller [2] maintains T=100 mK by slowly reducing the field,
thereby compensating for the heat leak through suspension and wiring. The
hold time has been measured to be 18 hours with a temperature stability of
Third International Symposium on Space Terahertz Technology Page 681
AT rms = 3 |iK. For SIRTF and balloon use, a compact analog PID controller has
been developed which has achieved temperature stabilities of about 10 (iK.
After the magnetic field has been reduced to zero, the heat switch is closed
and the refrigerator is cycled again. The duty cycle of the refrigerator is more
than 95 %.
BOLOMETERS
We are currently testing and optimizing composite bolometers
constructed in similar ways to the devices used at and above temperatures of
300 mK [3,4]. A schematic picture of a bolometer [4] is shown in Figure 2. The
main elements of the bolometer are a radiation absorber, a thermistor to sense
the deposited power and a weak thermal link to the 100 mK plate.
The absorber consists of a = 600 A thick film of Bi evaporated on a 2mmx
2mmx 35|im diamond substrate. The absorptivity of this structure has been
measured to be about 50 % for frequencies from 15-250 cm" 1 and normal
incidence [5]. The bolometer is enclosed in an optical cavity to maximize the
optical efficiency. The substrate is supported by two strands of = 10 (im thick
Nylon fiber under tension.
For the temperature sensing element we use a neutron transmutation
doped chip of germanium (200umx 200fimx 200)im) [6]. Ohmic contacts are
provided on two faces by boron implantation and metallization with layers of
Pd and Au. At helium temperatures, the doped germanium is in the hopping
conduction regime with a resistance which varies as
R = Roexp((A/T)o:5). (1)
The constants R and A depend on geometry and doping. Electrical
connections are made by attaching two 6 |j.m thick graphite fibers to the
contact pads with silver epoxy. The thermal conductance of the bolometer is
Page 682 Third International Symposium on Space Terahertz Technology
dominated by the Nylon with G = 4xl0- n W/K at T= 100 mK for 12 strands.
The time constant of the bolometer is given by x = C/G (C is the bolometer
heat capacity) and measured to be about 30 ms. Our measured x 's turned out
to be an order of magnitude larger than is extrapolated from values of C and
G measured at 300 mK. Sources of this discrepancy are being investigated.
The bolometer sensitivity is typically expressed as a noise equivalent
power which has the contributions
NEP2 = NEP2 photon + 4kT2 G + 4kT2 R/| S p + eV I S P (2)
where e n is the voltage noise of the JFET preamp and S is the voltage
responsivity given by
S(o))=I-(dR/dT) • IG + icoCH (3)
and I is the bias current and g)/2tc is the chopping frequency.
We have measured electrical NEP's of 3x1 0' 17 W/VHz for a chopping
frequency of 6 Hz. We are planning to measure the absorbed power NEP by
replacing the Bi film absorber with a meander line bismuth heater. Other
groups have seen indications of an electric field dependent thermistor
resistance and an increasing thermal resistance between phonons and
electrons in the NTD germanium at our operating temperatures. Either effect
would give an absorbed power NEP which is larger than the electrically
measured NEP.
The bolometer voltage is read out by a JFET pair (Interfet NJ132L)
packaged in a light-tight can and heated to the optimum temperature of about
120 K. Typical noise voltages are 5-6 nV/VHz at 6 Hz. The JFET current noise is
.not significant for bolometer resistances of about 10MQ. We use a Kapton
stripline with Constantan conductors for the high impedance leads between
the bolometer and the JFETs to avoid microphonic noise, crosstalk, and
excessive thermal conductance.
Third International Symposium on Space Terahertz Technology Page 683
SPACE FLIGHT ISSUES
In order to make our receiver space qualifiable, a number of issues need to
be addressed. Since the SIRTF cryostat is likely to be baked during pump out,
all ADR components have to survive a simulated bake out test. Commonly
used paramagnetic salts such as FAA (ferric ammonium alum) and CPA
(chrome potassium alum) dehydrate in sealed containers at temperatures as
low as 35 °C. Bake out tests with CCA indicate allowed temperatures of at least
50 °C. The large size of the cesium ion leads to a tighter binding of the waters
of hydration in the crystal. An undesirable side effect is the very low solubility
of CCA in water, which complicates crystal growth.
Another important topic is survivability of the instrument during
launch. We have already carried out a shake test of the suspension system at
room temperature and at T= 77K. The tests show a fundamental resonance at
200 Hz with a Q of about 100. Since the flat vibration spectrum of the rocket is
amplified at that frequency, it should not coincide with any bolometer
resonances. To investigate the bolometer vibration spectrum of the
bolometer, we developed a cryogenic shake table based on an electromagnetic
vibrator immersed in liquid helium. The force is transmitted into a vacuum
can through a bellows. We can subject small mass objects such as bolometers
to accelerations of up to 60 g (rms).
We have run a reliability test on the mechanical heat switch by cycling it
for 4000 times which corresponds to 8 years in orbit. No degradation in switch
performance was detected.
An overall test of the system will be carried out during an upcoming
flight of a balloon borne telescope [7,8] for measuring the anisotropy of the
cosmic microwave background: In previous flights we used 3 He cooled
bolometers with NEP's of about lxlO" 16 W/VHz. The 100 mK bolometers are
expected to provide significantly lower noise. The balloon photometer has a
multimode broadband feedhorn, and dichroic filters which split the beam into
four bands at 3, 6, 9, and 12 cm' 1 (90, 180, 270, and 360 GHz). In addition, we use
quartz and glass bead filters to attenuate high frequency power. The overall
Page 684 Third International Symposium on Space Terahertz Technology
optical efficiency of the balloon photometer from cryostat window to detector
is approximatey 20 %.
CONCLUSIONS
We have developed a bolometric THz receiver for space applications
which operates at 100 mK and achieves an NEP of about 5xl0~ 17 W/VHz. To
compare this with the potential sensitivity of heterodyne receivers, we can
calculate the noise equivalent temperature (NET) for both types of receiver.
Assuming a 10 % bandwidth at 1 THz, the diffraction limited Rayleigh-Jeans
power is P s = 2kTB and NET = NEP/2kB = 1.8xl0" 5 K/VHz. For a quantum
noise limited heterodyne receiver with an IF frequency of 1 GHz we calculate
an NET = hv/kVB (F = 1.5xl0" 3 K/VHz. This particular example gives a 100
times greater sensitivity for the bolometric receiver.
At present the status of the bolometer channels on SIRTF is uncertain
due to budgetary limitations. A great opportunity would be missed and much
valuable science lost for the SIRTF mission if the bolometer channels were to
be removed.
ACKNOWLEDGEMENTS
This work was supported by NASA grants NSG-7205, FD-NAGW-2121
and JPL contract 958764, and by the Center for Particle Astrophysics through
NSF cooperative agreement AST 8809616.
Third International Symposium on Space Terahertz Technology Page 685
REFERENCES
[1] P.T. Timbie, G.M. Bernstein, and P.L. Richards, Cryogenics 30, 271, (1990).
[2] G. Bernstein, S. Labov, D. Landis, N. Madden, I. Millet, E. Silver, and P.
Richards, Cryogenics 31, 99, (1991).
[3] A.E. Lange, E. Kreysa, S.E. McBride, P.L. Richards, and E.E. Haller, Intl. J.
Infraared and Millimeter Waves, 4(6), 689, (1983).
[4] D.C. Alsop, C. Inman, A.E. Lange, and T. Wilbanks (in preparation).
[5] J. Clarke, G.I. Hoffer, P.L. Richards, and N.H. Yeh, /. Appl. Phys., 48(12),
4865, (1977).
[6] E.E. Haller, Infrared Phys., 25, 257, (1985).
[7] M.L. Fischer, D.C. Alsop, E.S. Cheng, A.C. Clapp, D.A. Cottingham, J.O.
Gunderson, T.C. Koch, E. Kreysa, P.R. Meinhold, A.E. Lange, P.M. Lubin, P.L.
Richards, and G.F. Smoot, Ap.j. (in press).
[8] D.C. Alsop, E.S. Cheng, A.C. Clapp, D.A. Cottingham, M.L. Fischer, J.O.
Gunderson, E. Kreysa, A.E. Lange, P.M. Lubin, P.R. Meinhold, P.L. Richards,
and G.F. Smoot, Ap.j. (submitted).
Page 686
Third International Symposium on Space Terahertz Technology
400-800 M m
Lens + Winston cone
.1 K stage
with
Bolometers
Thermal
Bus
200-400 M m
Lenses + Winston cones
(2X2 array)
■ Vacuum
Jacket
Primary
Shield
Secondary
Shield
Vacuum
Jacket
Figure 1: Adiabatic demagnetization refrigerator
Third International Symposium on Space Terahertz Technology
Page 687
Nylon Support
Fibers
1 cm
XTD Ce
Thermistor
Graphite
Contact
Fibers
Absorber
35 um Diamond Substrate
i mm
Figure 2: Sketch of a composite bolometer
Page 688 Third International Symposium on Space Terahertz Technology
&>06 7 ' AlGaAs/GaAs Quasi-Bulk Effect Mixers: Analysis and Experiments
N93-27783
f\ K. S,Yngvesson, J.-X. Yang, F. Agahi, D. Dai, C. Musante, W. Grammer.
and KM. Lau
ABSTRACT
The lowest noise temperature for any receiver in the 0.5 to 1
THz range has been achieved with the bulk InSb hot electron mixer,
which unfortunately suffers from the problem of having a very narrow
bandwidth ( 1-2 MHz) We have demonstrated a three order of
magnitude improvement in the bandwidth of hot electron mixers, by
using the two-dimensional electron gas (2DEG) medium at the
hetero-interface between AlGaAs and GaAs. We have tested both in-
house MOCVD-grown material, and MBE material, with similar
results. The conversion loss (Lc) at 94 GHz is presently 18 dB for a
mixer operating at 20K, and calculations indicate that Lc can be
decreased to about 10 dB in future devices. Calculated and measured
curves of Lc versus PLO, and IDC, respectively, agree well. We argue
that there are several different configurations of hot electron mixers,
which will also show wide bandwidth, and that these devices are
likely to become important as low-noise THz receivers in the future.
I. INTRODUCTION
In our contribution to the Second International Symposium on
Space Terahertz Technology [ 1 ], we discussed the potential
advantages of bulk and "quasi-bulk" ( surface-oriented, or quasi-two-
dimensional ) mixer devices for the THz frequency range. A review of
the receiver noise temperatures achieved in this range is given in
Figure 1. This graph has been updated compared with the one in [1],
using data given in presentations at the 1992 symposium. In the
frequency range above 500 GHz, the lowest noise temperatures are
still the ones reported for the InSb hot electron mixer [2]. Also, the
frequency dependence of the receiver noise temperature of this
mixer is much less s*eep than that for either the SIS or the Schottky
diode mixers. We attr bute these advantageous features to the bulk
nature of the InSb m xer, which leads to a very small parasitic
reactance. , In contrast, severe requirements have to be imposed on
the size of SIS or Schottky devices for use at these frequencies.
This work has been supported by the National Aeronautics and Space
Administration, under grant NAGW-1659
The authors are with the Department of Electrical and Computer
Engineering, University of Massachusetts, Amherst, MA 01003
Third International Symposium on Space Terahertz Technology Page 689
The InSb mixer has one major draw-back, i.e. its very narrow
bandwidth (about 1 MHz ), which has limited its use in practical
systems. A similar narrow bandwidth was obtained for a bulk GaAs
mixer [3]. As originally pointed out by Smith et al.[4], it does not
appear that this limitation should apply to all hot electron mixers,
however, and our earlier paper [1] gave preliminary results which
supported this idea. We demonstrated conversion with at least 1 GHz
bandwidth in a mixer which employs a nonlinear element based on
the two-dimensional electron gas (2DEG) medium at a
heterojunction interface formed between AlGaAs and GaAs. Our
continued investigation of this mixer at 94 GHz, reported in the
present paper, has yielded a 1.7 GHz bandwidth, while the
conversion loss has been improved from 30 dB to 18 dB. We have
also obtained good agreement between measured and calculated
conversion loss.
In this paper, we review the properties of hot electron mixers,
especially those employing the 2DEG medium, and present our
measured and calculated data. Two different modes of operation have
been attempted: (1) Type I: T his is the type of mixer which we have
demonstrated. It operates at temperatures up to about 100K, and
requires LO power of the order of a milliwatt. (2) Ty pe II: The Type
II mixer was proposed by Smith et al. [41, operates at about 4K, and
requires the use of a moderately large magnetic field There are two
distinct magnetically biased hot electron mixers: Type Ila is tuned to
cyclotron resonance with the help of a fairly low magnetic field
( < IT ). So far, this type of device has been demonstrated as a
detector, but not as a mixer. The LO power is predicted to be in the
microwatt range, or lower. We have also proposed a mixer (Type lib) ,
which uses a somewhat larger magnetic field, 1-4 Tesla, i.e. in the
region in which Shubnikov-DeHaas oscillations are seen for the DC
resistance. Again, only detection has so far been demonstrated.
We briefly compare the 2DEG mixers with other potentially
interesting hot electron mixers using thin film superconductors,
which also employ a quasi-two dimensional geometry. We argue that
there is a large class of different hot electron media, which can be
used for mixing with bandwidths of about 1 GHz. The potential
extension of two-dimensional hot electron mixers to THz
frequencies is finally assessed.
II. FREQUENCY CONVERSION IN HOT ELECTRON MIXERS
Basic Mechanism
Hot electron mixers employ a nonlinear (electron) bolometer
device of the type shown in Figure 2. In electron bolometers, the
electron gas is heated by the applied power (DC and /or RF) to a
temperature, T e , above the lattice temperature, To. The advantage of
an electron bolometer, compared with a standard bolometer, which
relies on heating of the lattice, is that the specific heat of the
p 6go Third International Symposium on Space Terahertz Technology
electrons is much smaller than that of the lattice As discussed in
greater detail in [1], the basic nonlinearity which gives rise to mixing
in hot electron mixers is associated with the fact that the resistance
(Rb) of the device is a function of the electron temperature, and
therefore also a function of the total power applied (P). A typical
curve of Rb= Rb(P) for a 2DEG device is shown in Figure 3.
Response Time and Bandwidth
The maximum rate at which the resistance of a hot electron
bolometer can vary is determined by the response time of the
electron gas. This response time is in turn given by the effective
time for energy transfer from the electron gas to the lattice, which is
basically the same as the energy relaxation time, x e . The bandwidth
of the mixer extends up to a frequency given by [1]:
B - 1/ 27tx e (1)
Values of t e are wellknown for a number of different media. In
bulk semiconductors, t e is typically in the range 10" 12 to 10* 10
seconds, and both measurements and theory have been reviewed in
standard reference works [5,6]. The much longer x e ( 10* 7 to 10 -6 s
), which is found in bulk InSb and GaAs hot electron mixer devices,
is due to a bottle neck for the energy relaxation process in these
devices from the conduction band to the donor levels [2]. It Is
worthwhile to explore in some further depth why bulk
semiconductors are limited to such a narrow bandwidth when used
as hot electron mixers, and why no wider bandwidth devices have
been developed during the last thirty years. The energy relaxation
time which is typically calculated refers to energy transfer to the
lattice from a hot electron distribution within the conduction band,
as shown in Figure 4. At temperatures higher than what corresponds
to the splitting between the donor levels and the conduction band
(roughly 50-100K), almost all donors become ionized, and the only
energy relaxation processes which are relevant are those within the
condcution band, which thus are quite fast. At the much lower
temperatures, which are used in bulk InSb or GaAs mixers, the
donor states are occupied, and the much slower relaxation to these
states dominates, as also shown in Figure 4. A wide band mixer
employing bulk GaAs may be designed at the higher temperatures,
but does not appear to have been attempted. One reason may be that
too much heating of the lattice occurs, see the discussion below
which compares "three-dimensional" and "two-dimensional" devices.
For the 2DEG medium at an AlGaAs/GaAs interface, Sakaki et al
measured x e -values of 10 -10 to 10" 9 s for temperatures below 10 K
[7]. The shorter energy relaxation time compared with bulk GaAs at
these temperatures apparently occurs because the donors and the
electrons are well separated due to the modulation-doped structure
of the material, hence energy relaxation within the band (actually
sub-band(s)) dominates. There is thus a very real advantage to using
the heterostructure configuration in this case, in that a much wider
Third International Symposium on Space Terahertz Technology Page 691
bandwidth should be feasible, as compared with bulk GaAs. At
somewhat higher temperatures, 50 to 150 K, Shah found energy
relaxation times of about 10* 11 to 10- 10 s [8]. At these temperatures,
Shah found evidence for some lengthening of x e due to accumulation
of optical phonons at the specific energy of about 36 meV ( a so-
called "hot phonon" effect ). Due to this phenomenon, x e is shorter in
material with a lower surface density of electrons, n s t while at the
lower temperatures the opposite dependence on n s is obtained [7],
In both temperature ranges, the 2DEG medium should achieve
bandwidths of the order of 1 GHz or greater, providing the solution
for the longstanding problem of the narrow bandwidth of bulk
semiconductor mixers. The 2DEG material systems have the further
advantage that many different combinations of materials are feasible.
They are the subject of an intensive research effort at present, so
that considerable information is available on most aspects of physics
and fabrication.
Recently, hot electrons have also been studied in thin films of
superconductors [9,10]. These studies have shown that the energy
relaxation time for Nb is such that bandwidths of the order of 100
MHz are feasible, while wider bandwidths should be possible in NbN,
as well as in YBCO. Intrinsic mixer conversion loss of close to dB
and bandwidth of 40 MHz were deduced from measurements at
about 140 GHz.
A distinct mode of operation of a superconducting film mixers
is being pursued by Grossman et al., see another paper at this
conference. A range of materials and modes of operation for hot
electron mixers with much improved bandwidth capability thus is
now available. Only detailed future studies will show which of these
materials that is optimum for a THz hot electron mixer.
Two-Dimensional versus Three-Dimensional Geometry
The original bulk hot electron mixer device consisted of a bar
of InSb across a waveguide. The device impedance needs to be
matched to the microwave circuit, and this can be accomplished by
adjusting the length to cross-sectional area ratio, once the carrier
concentration and and mobility are given. The minimum size is
determined by fabrication considerations, as well as the need to fit
the device into the waveguide. Devices used in practise [2] contain
close to 10 10 electrons. In the 2DEG case, the device resistance (Rb)
is
R B = (L/W) * (1/ensji) (2)
Here. L and W are the length and width of the active region, n s is
the density in carriers per cm 2 , and \i the mobility in cm 2 /Vs. It is
easy to adjust L/W by using photolithography, and a number of
devices with the same resistance which matches the microwave
circuit are possible. Specifically, the total number of electrons in a
Page 692 Third International Symposium on Space Terahertz Technology
typical device is smaller ( ~ 10 6 ). Less power is then required to
drive the device nonlinear by heating the electrons, since the power
loss per electron at a given temperature is basically a materials
constant. This fact probably explains why a bulk GaAs hot electron
device at 50-100 K may be difficult to realize.
Conversion Loss
The conversion loss for bulk semiconductor mixers was
derived by Arams et al. [11J. We have extended their treatment to
include some further effects, such as the finite reflection loss of a
practical mixer ( To = 1 -I To I 2 ) [12]. The equivalent circuit of the
device is shown in Figure 5. When the conversion loss, Lc, is
optimized with respect to the ratio Pdc/Plo. as well as the IF load
impedance, Rl, one finds 112] :
( 3)
We have introduced Rbo. which is the device resistance at an
equivalent operating point for which the DC power is equal to the
total dissipated power (P ) at the mixer operating point, i.e. P = Plo
+ Pdc- Also, the factor C is defined as (compare Figure 2):
C = dR/dP ( 4 )
The microwave circuit impedance Z , typically 100 ohms, is used to
calculate T . The optimum value for Lc is 6 dB, which should be
compared with 3 dB for a double-sideband Schottky-barrier diode
mixer. SIS-mixers can have some conversion gain.
In order to compare calculated and measured conversion loss,
one needs to perform a slightly different calculation than the one in
(3). which assumes that the mixer has been optimized at all points of
a curve of, say, Lc versus Plo. An actual measured curve would
represent Lc as a function of either Plo or Idc. with the other
variable held constant. To perform this calculation, one finds it
necessary to iteratively search for the correct operating point [12].
Data calculated using this method will be presented in the
experimental section.
m. MATERIALS GROWTH AND DEVICE FABRICATION
2DEG Device Structure
A simple two-terminal device structure has been developed,
which is shown in Figure 6. The sequence of epitaxial layers is the
Third International Symposium on Space Terahertz Technology Page 693
same as is used for AlGaAs/GaAs HFET devices. Mesas are etched
with a height of about 1.5 micrometers. The 2DEG sheet is then
located within the mesa, which is surrounded by semi-insulating
GaAs for isolation of the devices. The metalization of the devices
consists of a standard sequence of evaporated layers for forming
AuGe ohmic contacts, and the device pattern is then defined by a lift-
off process. Typically, gold plating was used for building up the
metalization to sufficient thickness. There is a thin top layer of highly
doped GaAs which facilitates the formation of good ohmic contacts.
This layer has to be etched off after the devices have been defined.
The wafer was finally thinned to about 125 micrometers, and a small
chip cut out by first scribing the wafer. The IV-characteristic was
measured by probing the individual devices, which have sufficiently
large contacts pads.
Materials Growth
The epitaxial layers are grown with low pressure MOCVD on
undoped semi-insulating GaAs (100) substrates oriented 2 degrees
off towards the (110) planes to achieve better morpology of the epi
layers. The sources used were trimethylaluminum, trimethylgallium,
100% arsine, and silane as the n-type dopant. Further details of the
process can be found in our paper at the previous conference. We
have also utilized material grown by MBE, courtesy of Dr. D. Masse',
Raytheon Company, Lexington, MA. The maximum mobility for both
types of materials is quite similar, as shown in Figure 7.
TV-Characteristics
The Type I mixers were biased to the region indicated in Figure 7,
where it may be noted that the mobility is a strong function of lattice
temperature in this region. It may be assumed that the mobility as a
function of electron temperature follows essentially the same curve
[ 8]. We may then conclude that the IV-characteristic will tend to
"saturate", as the electrons heat up when the voltage is increased.
We show measured curves att low temperatures for three different
devices in Figure 8. Dimensions and other data for the devices are
given in the caption of this figure. The curves can be modeled with
an analytical expression, which is useful in calculating data such as
conversion loss:
= (^)tan>V)
(5)
IV. EXPERIMENTS
Experimental setup
Two different experimental setups were used: (1) A liquid
helium dewar with built-in superconducting magnet. (2) A
Page 694 Third International Symposium on Space Terahertz Technology
mechanical refrigerator which could provide temperatures to about
20 K. In both cases, the devices were mounted as flip-chips on a
circuit etched on a Duroid 5880 substrate, which was inserted into a
split-block mixer mount. The Duroid substrate is not ideal for low-
temperature experiments, but was used for ease of fabrication and
assembly. Improved mixer circuits on silicon substrates are under
development [13], and will be used in the future. RF and LO power
were fed to the mixer through a stain-less steel waveguide, which
was connected to the mixer block. The circuit inside the mixer
block was essentially a finline transition, with the device soldered
across the narrow part of the finline. One side of the finline had to be
insulated, to allow the device to be biased. The IF was extracted from
the device through a simple transition to a coaxial cable.
Measurements on the Type I mixer
The Type I mixer was typically measured at about 20 K. The
conversion loss is essentially independent of the lattice temperature
over a wide range, up to about 100K, so the exact temperature of
operation is not important. The RF power (from a GUNN source) at
the mixer block input was carefully calibrated with a power meter.
The IF power was measured with a spectrum analyzer, which was
also calibrated with a power meter. The measured numbers
represent the conversion loss from the input of the block to the IF
output connector near the block. As the LO, we either made use of
another GUNN source, or (for higher power) a BWO.
The first measurement of the Type I mixer was performed in
the 4.2 K setup. It should be realized that the electron temperature
is about the same, independent of the lattice temperature: we
estimate a value of T e = 85 - 90K at the operating point. The
conversion loss in the first measurement was 30 dB. The normalized
response versus IF frequency is displayed in Figure 9, which also
shows the normalized response of InSb and GaAs bulk mixers as
measured by others [2,3]. It is clear that the 2DEG mixer provides a
roughly three orders of magnitude increase in bandwidth compared
with previous hot electron mixers. The 3 dB bandwidth is 1.7 GHz.
Subsequent measurements at about 20 K have consistently shown the
same bandwidth, but the minimum conversion loss has been
improved to 18 dB.
The conversion loss has been calculated for two devices, with
IV-curves as shown in Figures 8(a) ( "4.2K mixer") and 8(b) (19K
mixer"), respectively. Figure 10 shows the calculated conversion
loss as a function of LO power (curves), compared with the measured
data (points). The agreement is good, and indicates the validity of
using our theoretical approach. In Figure 11, calculated and
experimental data for Lc versus Idc are compared. Again, the
agreement is good, with a small shift of the current scale at low
currents. The minimum conversion loss Is well predicted. The
increased conversion loss for the 19K mixer at high bias currents
may be due to lattice heating: the dissipated DC power at the highest
current (10 mA) is about 200 mWI One general trend in comparing
Third International Symposium on Space Terahertz Technology Page 695
different devices, is that a device with low maximum (saturated)
current requires less LO power to drive to optimum conversion loss.
It is clearly also advantageous to have a high initial slope for the IV-
curve. The latter requirement can be satisfied by making the device
wide, but this will lead to a larger saturation current. It si preferable
to use a narrower device, and to maximize the initial slope by
choosing material with high mobility, as well as by keeping the
contact resistance as low as possible In the devices tested so far, the
contact resistance appears to limit the value of the initial slope, and
future devices are expected to yield considerably lower conversion
loss. Another effect, which we have noticed, is that the initial
resistance increases after the device has been soldered into the
circuit. It is thus reasonable use the calculated conversion loss for
the best device which we believe is feasible to fabricate, as an
indication of the performance which should be feasible. The
optimum conversion loss for such a "next generation** device is given
in Figure 12. Note that both the conversion loss and the optimum LO
power have been improved considerably.
Measured Results for the Type U Device
As mentioned earlier, the Type II device is expected to operate
at about 4.2K, and be biased by a magnetic field. We have
demonstrated a Type Ila detector at 94 GHz, similar to the one
described by Smith, Cronin et al. (4J, as shown in Figure 13. The
peak response occurs at about the expected magnetic field for
cyclotron resonance .The measured responsivity is about 1V/W, which
is much lower than obtained by Cronin's group ( about 250 V/W )[14].
The reason for the lower responsivity is very much an open question
at this stage; it may, for example, be due to differences in the
materials used. We have measured a larger responsivity (5-10 V/W),
at both 94 and 238 GHz, in a newly discovered mode (Type lib) at
higher magnetic field, see Figure 14. The device in this case was
severely mismatched to the microwave circuit due to its very high
resistance at high magnetic fields ( =» 2kflt ). We can therefore
estimate that the responsivity in this mode may be increased to at
least 50-100 V/W. The frequency independence of the responsivity
up to 238 GHz is also noteworthy. The detection mechanism is again
not yet known [12], but it is likely to involve transitions between
(spatially) extended and confined states, which have been studied
extensively in other work on the Shubnikov-deHaas and Quantum
Hall effects.
We attempted to demonstrate mixing in both of the above Type
II devices, but were not successful in doing this. One indication that
this negative result is not unexpected can be obtained by using the
following expression for the optimum conversion loss, derived in
[111:
— MiAW
(6)
Page 696 Third International Symposium on Space Terahertz Technology
where the quantity in square brackets can also be written:
lcP °' IcSo (7)
Here, 9*o is the responsivity of the device as a detector at an
equivalent bias current Io = J^ 3 - • For a responsivity of 10 V/W at a
V Rbo
bias current of 10 |iA, we predict I< = 86 dB, which would not have
been measurable. Further work on Type II mixers then needs to
emphasize attaining an increased responsivity in the detector mode.
V. POTENTIAL OF HOT ELECTRON MIXERS FOR THZ
FREQUENCIES
General Discussion
We have seen in the previous section that it is reasonable to
assume that the conversion loss of the Type I mixer may be
decreased toward the 10 dB level. In order to estimate how these
results at 94 GHz will scale with frequency, we must investigate how
the absorption of the 2DEG varies with frequency. As discussed at the
previous conference, high mobility (i.e. long momentum relaxation
time, x m ) media, such as the 2DEG, will show evidence of charge
carrier inertia at lower frequencies than one expects for typical
semiconductors. At the electron temperature of 90 K, representative
of the operating points for 2DEG Type I mixers, one may expect the
cox m = 1 point to occur at about 100 GHz. The equivalent circuit of
the device ( see Figure 15 ) then incorporates an inductance with a
reactance of close to 100 ohms. At higher frequencies, it will become
necessary to resonate out this reactance by inserting a monolithic
capacitor in series, as also shown in this figure. Of course, other
circuits may be used to match to the 2DEG device, such as a back-
short in a waveguide, etc, but monolithic circuit should yield the
widest bandwidth. A different approach may be to find other material
combinations which show the required nonlinearity, but with a lower
maximum mobility, so that the reactance is minimized. We are
pursuing microprobe measurements of 2DEG devices at frequencies
up to 40 GHz in order to build up a base of information for designing
matching circuits for 2DEG mixers at the higher frequencies [13].
Estimated Noise Temperature of Hot Electron Mixers
We may roughly estimate the expected noise temperature of a
hot electron mixer by assuming that a conversion loss of 10 dB is
Third International Symposium on Space Terahertz Technology Page 697
achievable, and that the effective temperature of the device is equal
to its electron temperature. A cooled IF amplifier with a noise
temperature of 10K is also assumed. We then obtain:
Tr = Lc * T e + (Lc-1) * Ti F = 1000K (8)
This receiver noise temperature would be competitive at
frequencies In the 500GHz to ITHz range. Hot electron mixers
which operate at lower electron temperatures would have a potential
for even lower noise temperatures. This is one of the reasons for
continuing to pursue the development of the Type II mixers. It may
be noted, though, that even a mixer with Tr = 1000K at close to
ITHz, developed from the present 94 GHz prototype, may be a very
good choice, since it may only require cooling to 50-80K.
VI. CONCLUSION
For many years, hot electron mixers have been known to be
capable of achieving very low noise temperatures, even up to
frequencies close to ITHz, but with only about 1 MHz bandwidth. In
our work with a 94 GHz 2DEG hot electron mixer, operating at 20K,
we have for the first time demonstrated a hot electron mixer with a
bandwidth, which is sufficient for all typical applications in the THz
range. This type of hot electron mixer has the same advantages of
low parasitics as the InSb version, but with added features such as
potential for monolithic integration with the IF amplifier and
antennas, etc. Further development should demonstrate the
eprformance of this type of mixers at higher frequencies, and with
lower conversion loss. Noise temperature measurements will be ,
performed in the near future, which will enable us to make a firmer
estimate of the noise temperatures which can be achieved at higher
frequencies. Finally, we note that wide bandwidth hot electron
mixers may be the general rule, not the exception, as one may have
been tempted to think based on the earlier results achieved by InSb
and GaAs bulk mixers. Many different 2DEG media exist, which could
potentially be more optimum than the one we have started with.
Further striking evidence is provided by the results with
superconducting film hot electron mixers [9,10]. A final tabulation
compares different hot electron mixers, see Table 1. The evidence
from these efforts supports the notion that the hot electron mixers
will find a niche of applications in the THz range.
Vn. REFERENCES
[1] Yang, J.-X., Grammer.W., Agahi.F., Lau,K.-M., and Yngvesson,
KS. 'Two-Dimensional Electron Gas ("2DEG") Hot-Electron
Mixers for Millimeter Waves and Submillimeter Waves,"
Second International Symposium on Space THz
Technology, JPL, 1991, p. 353.
Page 698 Third International Symposium on Space Terahertz Technology
[2] Brown, E.R., Keene, J., and Phillips, T.G., "A Heterodyne
Receiver for the Submillimeter Wavelength Region Based on
Cyclotron Resonance in InSb at Low Temperature,"
Intem.J.Infrared and Millimeter Waves, 6,1121, (1985).
[3] Fetterman, H., Tannenwald. P.E., and Parker, CD., "Millimeter
and Far Infrared Mixing in GaAs," Proc.Symp.SMM Waves, PIB,
New York (1970).
[4] Smith, S.M., Cronin, N.J., Nicholas.RJ., Brummel, M.A., Harris,
J.J., and Foxon, C.T., "Millimeter and Submillimeter Detection
Using Gai- x Al x As/GaAs Heterostructures," Intem.J.Infrared and
Millimeter Waves, 8,793(1987).
[5] Seeger, K., "Semiconductor Physics", Fourth Ed., Springer
Verlag, Berlin (1989), see for example p. 102 and p. 198.
[6] Conwell, E.M., "High -Field Transport in Semiconductors," in
Solid-State Physics, Suppl., Vol.9, Academic Press .New York
(1967).
[7] Sakaki, H., Hirakawa, K., Yoshino, J., Svensson, S.P., Sekiguchi,
Y., Hotta, T., and Nishii, S. "Effects of Electron Heating on the
Two -Dimensional Magnetotransport in AlGaAs/GaAs
Heterostructures," Surface Science, 142,306 (1984).
[8] Shah, J., "Hot Carriers in Quasi-2D Polar Semiconductors,"
IEEE J.Qu. Electronics, QE-22, 1728 (1986).
[9] Gershenzon, E.M., Gol'tsman, G.N., Gousev, Y.P., Elant'ev, A.I.,
and Semenov, A.D., "Electromagnetic Radiation Mixer Based on
Electron Heating in Resistive State of Superconductive Nb and
YbaCuO Films," IEEE Trans, Magnetics, MAG-27, 1317 (1991).
[10] Kozyrev, A.B., Samoilova, T.B., Soldatenkov, O.I., and Vendik,
O.G., "Destruction of Superconducting State in Thin Film by
Microwave Pulse," Solid State Communications, 77,441 (1991).
[11] Arams, F., Allen. C, Peyton, B., and Sard, E.," Millimeter Mixing
and Detection in Bulk InSb," Proc. IEEE.54,612 (1966).
[ 1 2] Yang. J.-X., "AlGaAs/GaAs Two Dimensional Electron Gas
Devices: "Applications in Millimeter and Submillimeter Waves, "
Ph.D. Thesis, University of Massachusetts (to be publ.).
[13] Grammer, W., M.Sc. Thesis, University of Massachusetts ( to be
publ.).
[14] Cronin, N.J., University of Bath, Priv. Comm.
Third International Symposium on Space Terahertz Technology
Page 699
100 ooo
Figure 1 . Receiver noise temperatures in the MM and Sub-MM range,
versus frequency. The data for Schottky barrier diodes were
originally compiled by EX. Kollberg, and the SIS data by P.L.
Richards. The InSb noise temperatures are from Brown et al. [2]
.Some more recent data from this conference have been added ( see
arrows).
THZ
RADIATION
Figure 2. Schematic drawing of a generic electron bolometer device.
Page 700
Third International Symposium on Space Terahertz Technology
i
Conduction £ >
Band
Figure 3. Typical curve of R B = R B (P)
for a 2DEG device.
rie+Pco+P*— >
A
R<P)
2DEG
LPF
Figure 4. Energy relaxation proct
In a semiconductor.
Rl
Figure 5. Equivalent circuit
a hot electron mixer device.
Ohmic Contacts
M«UI EI«ctrod««
Figure 6. Structure of the 2DEG device.
Third International Symposium on Space Terahertz Technology
Page 701
10
in
> ,0'
U
.S io°
g 10*
2
10"
* » *UBE: r\, -4.5*10f' citT 2
a a QA166 (UOCVO): n, - 1 x 1 d 2 cm*
a a a aaooaaD a a oaoc
Bio* region-
10° 10' 10 2
LATTICE TEMPERATURE in K
10-
Figure 7. Mobility versus lattice temperature for two 2DEG devices.
16
14
= 10
I—
Z 8
bJ
Ql
cn 6
U 4
2 -
0.0
DEVICE l-V CHARACTERISTICS
-
\ t r - — r
T-4.2. 18. 19 K
-
-
(o)^^^-^"
-
-
s (b > ________
■ -
- />r
(c)
i i i i
~
0.2 0.4 0.6 0.8
VOLTAGE in Volt
1.0
(o) U8E. t-6 urn W-100 urn, O 4.2 K
(b) MBE. L-5 um W»50 urn, O 19 K
(c) A166 (M0CV0). L-43 um. W-20 um. O 18 K
Figure 8. Measured IV-curves for three 2DEG devices.
Page 702
Third International Symposium on Space Terahertz Technology
CQ
T=4.2 K W-Band
Q
■ 2D6G
■ InSb
GaAs
*
-10 H — ' ■ ■ ' ""I —
.0001 .001
i i ' i H i i i 1 1 1 1 1 | i i i 1 1 1 1 1| t i i 1 1 1 1 1
.01 .1
IF frequency (GHz)
10
Figure 9. Normalized IF response for the 2DEG mixer, compared
with InSb and GaAs bulk mixers.
50
CD
-o
c
40
to
to
O
JO
-j
z
o
tO
20
oc
UJ
>
z
10
o
o
T 1 1 1 1 1 1 1 r
T-4.2. 19 K
The 4.2 K mixer
Cfto.u...q_ o a
The 19 K mixer
' ' I ' ■ ' '
0123456789 10
LO POWER in mW
Figure 10. Calculated and measured conversion loss for two 2DEG
devices, versus LO power.
Third International Symposium on Space Terahertz Technology
Page 703
T-4.2, 19 k
\Th« a.2 K mixtr
O 10
O
Tho 19 K mi«#r
2 4 6 8
BIAS CURRENT in mA
10
Figure 1 1 . Calculated and measured conversion loss for two 2DEG
devices, versus DC bias current.
O
0.05 0.1 0.15 0.2
INCIDENT LO POWER, mW
Figure 12. Calculated conversion loss for an "optimum" next
generation 2DEG device.
Page 704
Third International Symposium on Space Terahertz Technology
CYCLOTRON RESONANCE DETECTOR
>
6
u
<
>
T*4 2K I.buwOOImA
Figure 13. Detected voltage versus magnetic field for a Type Ila
2DEG device operating as a straight detector at 94 GHz.
MAGNETIC FIELO (Ta»Jo>
Figure 14. Measured detected voltage output versus magnetic field
from a Typellb 2DEG device, operating as a straight detector at 94
and 238 GHz, respectively.
Figure 15. Equivalent circuit of a 2DEG device, including a matching
capacitor in series.
Third International Symposium on Space Terahertz Technology
Page 705
COMPAWSON OF DIFFERENT HOT ELECTRON MIXERS
MIXER
TYPE
BAND-
WIDTH
CONVERSIO
N
LOSS
7R .
KELVIN
SEMI
COND.
InSb
1 MHz
= 12 dB
300-500
SEMI
COND.
2DEG
20/50 K
2 GHZ
18 dB
(meas.)
10 dB (proj.)
1000 ?
SEMI
COND.
2DEG
4.2K,
MAGN.
FIELD
1-2 GHZ
(proj. )
10- 13 dB
(proj.)
300-500?
SUPER
COND.
Nb (meas.)
100 MHZ
7.5 dB*
?
SUPER
COND.
NbN (proj.)
Several
GHz
? MIXING
MEAS. AT
1.5 GHz"
?
SUPER
COND.
YBCO
(proj.)
Several
GHz
12 dB
(meas. at 2
GHz) •••
?
SUPER
COND.
Nb, Nonl.
Ind.'*"
(proj)
200MHz
-1GHz
70.000
@3THz
Alloys
lower
•) Gershenzon et al.. IEEE Trans. Magnetics. MAG-27.1317
(1991) c°a
*•) O. Vendlk. Priv. Comm.
•••) Kolesov, Chaloupka. et al.. to be published. 1992
**••) Grossman et al. this conf.
Page 706 Third International Symposium on Space Terahertz Technology
ALL-SOLID-STATE RADIOMETERS FOR ENVIRONMENTAL STUDIES TO 700 GHZ
%£-^3
Ralph, Rudiger and Peter Zimmermann ^ 9 3
RPG Radiometer-physics GmbH
5309 Meckenheim, Germany
-27 784
ABSTRACT
We report results with an all-solid-state radiometer for measurements of
the CIO molecule at 649 GHz. The project is part of a program to provide
low-noise, low-weight, low-power radiometers for space operation, and
special effort has been expended on the development of high-efficiency
solid-state frequency multipliers and Schottky-barrier mixers with low
local oscillator power requirements.
The best measured system noise temperature was 1750 K with the mixer and
preamplifier cooled to 77 K. The mixer diode was easily pumped into sa-
turation, indicating that the design has excellent prospects of opera-
ting at higher frequencies - our present design goal being 1 THz. We com-
ment on the principal design features of such systems and will report on
stratospheric measurements performed with this system.
INTRODUCTION
All solid-state radiometers in the frequency range 60-560 GHz have been
reported [{] » [2] , [3j , {,4 J , and the measured parameters for the
highest frequency receivers are tabulated in Figure 1. The ever-increasing
Third International Symposium on Space Terahertz Technology p a g e jqj
importance of atmospheric investigations in the higher sub-millimeter
wavelength-range has led to a consolidated program to reach 1 THz with
such instruments- To this end a radiometer (s- Fig. 2) has been construc-
ted for the detection of the CIO molecule at 649 GHz; this is the sub-
ject of this paper.
The three principal sections of the system are the optics, the all solid-
state local oscillator and the low-noise mixer-preamplifier.
QUASI-OPTICS
At frequencies above 200 GHz the losses of waveguide components are too
high for sensitive radiometers, and quasi-optical systems become increa-
singly compact. The prime component requirements for coupling the anten-
na and local oscillator into the mixer are high efficiency feed-horns,
precisely machined ofset-mirrors and low-loss wire grids for polarising
couplers and filters. All mirrors are ellipsoids, whilst grids comprise
20 um diameter tungsten wire with 50 um spacings (between centres).
For IF bandwidths af/f < 0.5 the Martin-Puplett coupler is preferred.
Figure 2 shows the total front-end schematic, including a single-side-
band filter and path-length modulator (from University of Bremen, to be
reported at this conference). The quasi-optical' beams, beam-waist loca-
tions and sizes are also shown. The feedhorns in this receiver are dual-
mode horns as shown schematically in Figure 3 [_£"] . These horns have been
measured at lower frequencies and have in all cases symmetrical near gaus-
sian beam-patterns, to at least -15 dB. The losses of the optics plus feed-
horns were measured by connecting the output of the frequency multiplier
Page 708 Third International Symposium on Space Terahertz Technology
directly to the mixer, and then feeding the l.o. to the mixer via the
two horns and the quasi -optics: the loss was around 1 dB.
LOCAL OSCILLATOR
The local oscillator power requirement for a mixer depends strongly on the
diode quality and on the mixer mount- and coupling-losses between source
and mixer. Particularly in the submillimeter-range these parameters are not
easy to determine. By extrapolating from lower submillimeter-wavelength, we
estimate the l.o. power requirement to be in the range 100 uW to 1 mW for a
cooled mixer. The requirement was to build an all solid-state source for la-
ter space application.
In this paper the source comprises a Gunn oscillator, a waveguide cross-
coupler plus harmonic mixer, a waveguide frequency doubler and tripler. All
components are directly connected together, to reduce losses (Figure 4).
(a) Gunn-Oscillator
A Varian InP Gunn diode is mounted into a WR-8 waveguide (Figure 5). Bias
is supplied to the diode via a coaxial low-pass filter held by teflon spacers,
The frequency is pretuned by a cap-resonator, fine adjustment over several
GHz being provided by a dielectric tuning rod sliding under the cap. A wave-
guide back-short allows adjusting for maximum power. The power output is
shown in Figure 6.
Third International Symposium on Space Terahertz Technology Page 709
(b) Frequency Multiplier
There are several possibilities of attaining power at 650 GHz from a solid
source followed by a multiplier. Gunn oscillators deliver output power up
to 50 mW at frequencies up to 115 GHz. For this reason a sixtupler seems a
natural choice. Though we build one with —100 uW output power, it was very dif-
ficult to tune, as several idlers had to be optimised simultaneously. The
better choice is a doubler-tripler or tripler-doubler combination, where
each stage can be developed separately. Calculations using the program by
Siegel et al. [6]] showed that the doubler-tripler combination should re-
sult in higher efficiency; hence this combination was chosen.
To facilitate easy system construction and testing, especially for the case
of cascaded multiplication stages, Radiometer-physics developed in 1987
"in-line" waveguide frequency multipliers (Figure 7). This means, that input-
and output-waveguide are in-line, compared to the standard crossguide-type
structure.
A schematic for both multipliers is shown in Figure 8. Input power is coup-
led from the input waveguide via a probe and coaxial filter to the diode
which is part of the last section of this filter. The filter is of the low-
pass Tchebycheff-type and has been modelled a lower frequencies. Probe and
diode are located in the bend of input- and output-waveguides respectively.
The DOUBLER input waveguide dimensions are 2.0x1.0 mm, which are tapered to
2.0x0.5 mm at the probe. The dimensions of the output waveguide are 0.6x0.3 mm
Backshorts are provided for tuning. The diode is a 6P4 from University of
Virginia, with the following DC-parameters R s =.12 ohm , Cj = 18 fF,
•V br = 20 V. It is contacted by a whisker of length 0.3 mm and diameter 25 ym.
d-f
Page 710 Third International Symposium on Space Terahertz Technology
With 50 mW input signal an efficiency of 20% was achieved, compared with a
theoretical efficiency of 37%. Power was measured with a Hughes Thermistor
mount at the input. Output power was measured with an Anritsu power head
140-220 GHz with a tapered transition to match the input waveguide of the
multiplier
The TRIPLER input and output waveguide dimensions are 0.6x0.3 mm and
0.4x0.2 mm respectively. It operates with a diode 2T2 again from University
of Virginia. Its DC-characteristics are Cj = 6.0 fF, R s = 12 ohm, V D r = -11V.
The whisker is 0.2 mm long, with 12 ym diameter. An important feature of
the tripler is the idler tuning circuit in form of a stub waveguide with a
moving short. The output waveguide contains a short section of reduced-width
waveguide to prevent idler propagation to the output. Low frequency model-
ling was important in attaining optimum waveguide dimensions. The output po-
wer achieved was 300 uW, which means an efficiency of 3% compared to 17% the-
oretical.
The measurement of output power in the range 600-700 GHz is not trivial, and
caution should be taken in specifying power at such high frequencies - in
particular the power standards should be quoted. In our case relative mea-
surements were performed using the same Anritsu thermistor power-head as
for the doubler, with appropiate waveguide transitions. The absolute power
output was then measured with the Queen Mary College acoustic calorimeter
at 624 GHz, where an absolute calibration exists: this showed that the ab-
solute power from the l.o. input is about three times greater than the. An-
ritsu reading. Hence to a first approximation we can multiply all Anritsu
readings by factor 3. Absolute measurements were also performed at 690 GHz
with a similar frequency multiplier chain, whereby 30 uW were measured with
the Anritsu head and 90 uW with the QMC instrument - once again a factor 3.
Third International Symposium on Space Terahertz Technology Page 711
It is relevant to draw attention to the problems arising due to using
different lower frequency thermistor heads, this is illustrated in
Figure 10 for three measurements from a 345 GHz frequency multiplier. -
To show the possibilities of achieving l.o. power from solid-state sour-
ces in the submm-range we have included Figure 10.
MIXER
The mixer is shown schematically in Figure 11. The dual-mode horn is identi-
cal to that used for the l.o. output. The waveguide dimensions are 0.6x0.2 mm
and the contacting back-short is gold-plated beryllium-copper. The diode
chip (type 1T6 from the University of Virginia) is mounted on the coaxial
choke structure, as shown in Figure 12. The chip is turned down to a cylin-
drical form and gold plated on the cylindrical surface, so that the chip is
also the first choke-section. It is seen that the choke is supported by a.
Macor disc: the dimensions of this disc are important for obtaining a broad
i.f. bandwidth. The calculated impedance of the choke is shown in Figure 14
from d.c. to 1100 GHz. Figure 15 shows the measured output impedance of the
mixer across the i.f. band 9-11 GHz, without and with an i.f. matching trans-
former
The 1T6 diode has the following d.c. -characteristics: R s =-30 ohm
Cjo = 0-35 fF, V Dr = -6 V. The whisker is of gold-plated phosphorbronze,
diameter 7 pm; the total whisker length is 0.2 mm, and includes a loop to
take up slight changes in waveguide and choke dimensions on cooling the
mixer to 77 K. To test that there was adequate l.o. power the diode was
biased at room-temperature to 0.86 V, whereby the diode drew a current of
Page 712 Third International Symposium on Space Terahertz Technology
10 uA without l.o. and >600 uA with l.o.: hence there was adequate l.o.-
power for optimum mixer operation.
SYSTEM RESULTS
The system was tested using standard hot-cold foam absorbers from Emerson and
Cuming at the input to the diplexer for DSB- and at the input to the SSB-
filter for SSB-measurements. In the final configuration the cold image ter-
mination was replaced by an absorber for constructional reasons, without any
appreciable sacrifice of performance.
The cryogenic i.f. preamplifier was developed at Chalmers Technical University
and had a noise temperature of 35 K at 77 K ambient. All other i.f. compo-
nents preceding the spectrometer were commercially available.
Double-sideband system tests with a 2 GHz i.f. bandwidth yield the following
results:
Tamb Tsys Tm Lo
300 K 3800 K 1900 K 8.0 dB
77 K 1750 K (1500 K) (8.0 dB)
Tsys-values are measured directly, whilst T M and L^ are corrected for resi-
dual power mismatch of 0.15, and we assume the same conversion loss at
77 K ambient.
For single-sideband system tests the SSB-filter was adjusted by using a tu-
nable Gunn-oscillator and multiplier chain, identical to the l.o. chain and
a spectrum analyzer. The sideband rejection was at least 25 dB.
The result was:
Tusb = 3800 K at 77 K
Third International Symposium on Space Terahertz Technology Pa 8 e
Mere than adequate l.o. power was available to reach minimum Tsys. It
was noted that the noise temperature was fairly insensitive to changes
in diode current in the range 400-800 uA.
The system has been flown on an aircraft contracted by the University
of Bremen. A first uncorrected result shows the CIO line in Fig. 15.
PROSPECTS
It appears reasonable to expect spaceborne all -sol id-state radiometer system
to be feasible at least to 1 THz, using only waveguide techniques. Calcula-
tions using the computer model of Siegel and Kerr yield l.o. output powers
of 360 uW at 1 THz; taking into account losses this still implies that suf-
ficient l.o. power will be available at 1 THz to pump a Schottky-barrier
mixer diode; such a development program has commenced at Radiometer-physics.
Since completing the 650 GHz system new Schottky-barrier mixer diodes have
been reported [7l , [,Q~\ . Hence, with improvements in multiplier- and
mixer-technology it should be possible to operate such systems at even higher
frequencies.
ACKNOWLEDGEMENT
The authors express their thanks to:
Dr. Nett from University of Bremen and Dr. Albinson from Chalmers Institute
of Technology, Sweden for their helpful collaboration
Dr. Th. Crowe from University of Virginia for providing the excellent Scfiottky
barrier diodes
Dr. E. Armandillo for support under ESA contract: "Limb Sounder critical Re- '
ceiver Technologies for Remote Sensing of the Atmosphere".
Mr. F. Leipold of Radiometer-physics for his indefatigable support
Page 714 ■ Third International Symposium on Space Terahertz Technology
REFERENCES
1 P. Hartogh, G.K. Hartmann, P. Zimmermann
"Simultaneous Water Vapour and Ozone Measurements with Mlllimeterwaves
in the Stratosphere and Mesophere". IGRASS 3.-6. Juni Helsinki, 1991
2 F. Lewen, University of Cologne, private Communication
3 J. Hernichel, R. Schieder, J. Stutzki, B. Vowinkel, G. Winnewisser, P. Zimmermanr
"A 492 GHz Cooled Schottky Receiver for Radio-Astronomy"
Proceeding of Third Intern. Symposium on Space Terahertz Technology,
March 24-26, 1992, Michigan
4 R. Zimmermann, Ra. Zimmermann, P. Zimmermann
"All Solid-State Radiometer at 557 GHz", 21st European Microwave Conference
Stuttgart 1991
5 M.H. Picket, J.C. Hardy, J. Farhoomand
"Characterisation of a Dual Mode Horn for Submillimeter Wavelengths",
IEEE, MTT-32, 1984 (936-937)
6 P.H. Siegel, A. R. Kerr, W. Hwang
"Topics in the optimisation of millimeter-wave mixers."
NASA Techn. Paper 2287, March 1984
7 N. Keen, A. Grub, H. Hartnagel, J. Freyer, H. Grote, Rii. Zimmermann
"New Submillimeter-Wave Schottky-Barrier Mixer, Diodes: First Results"
Revised late Paper, Stuttgart
8 T.W. Crowe, W.C.B. Peatmann
"GaAs Schottky Diodes for Mixing Applications byond 1 THz"
Proc. 2nd International Symposium on Space Terahertz Technology, Jet
Propulsion Lab., Febr. 1991 (323-339)
Third International Symposium on Space Terahertz Technology
Page 715
345 GHz L 2]
cooled 20 :<
490 GHz [3]
cooled 20 :<
557 GHz [*]
uncooled
only uncooled results; n./n. not measured
Fig. 1 : RESULTS OF SYSTEMS WITH MIXERS AND MULTIPLIERS
FROM RADIOMETER-PHYSICS
650 GHz
cooled 77 :<
IF:
1.5 * 0.25 GHZ
1.5 + 0.3 GHZ
1.4 + 0.3 GHZ
10 + 1 GHZ
T IF:
io - 20 :<
10 :<
50 :<
35 <
T MIX,DS3
n.m. -
n..-n. -
1200 :<
2000 '< *
L C,QS3
H..T1. -
n.oi. -
3 dB
8.3 dB*
T sys,0S3
350 K
550 :<
1600 K
1750 :<
(3800 < S33)
01
CONICAL
SECTION
02
03
I
7
CIRCULAR
GUIDE
CBC-ftECTANG.
TRANSITION
axo
=0.4x0.2 nnm
RECTANGULAR
GUIDE
U
01 = 2.96 inn
02 = 0.60 inn
03 = 0.47 urn
Li = 4.9 run
L2 = 3.0 nrn
L3 = 2.3 irm
L4 = 3.0 mn
Fig. 3: DUAL - MODE HORN (SCHEMATIC)
/ /( "J V^rror' SINGLE" SI DEB AND
FILTER
MOVABLE
MIRROR
GUNN-
OSC.
HARM.
MIXER
COUPLER
MULTIPLIER
A A
I — "^V emptied ' ,
4 . V\ nlrror i / I
elliptical
nlrror >
650 GHz CDDLED
SCHDTTKY RECEIVER
BLOCK DIAGRAM
DIMENSIONS IN MM
BEAM WAISTS'
A' 0.82 nn
B' 6.0 nm
O 4.65 nm
£• 4.0 rm
FIG. 2:
RECEIVER BLOCK
DIAGRAM
I
TO
5^
o
3
o
</)
5'
3
o->
en
I
3
Q
83
Third International Symposium on Space Terahertz Technology
Page 717
HARMONIC
GUNNHJSCH.L-»rOR "«£» DOUBLES
FREQUENCY TUNING
PQVER TUNING _
OUTPUT
FEEDHCIRN
MICROMETER
tO&a-KKJ GHz
Fig. 4: LOCAL OSCILLATOR
BIAS-CONNECTOR —
t-IYLDN
SCREW +.SP ACER — e
{
POWER TUNING
BACKSHORT
BERYLLIUM-COPPER
SPRING
CHOKE
STRUCTURE
TEFLON
SPACER
WR-8
WAVEGUIDE
DIELECTRIC ROD
_GUNN
DIODE
Fig.i": GUNN-OSCILLATOR
Page 718
paves
nW
io •*■
4S -
40 A
Third International Symposium on Space Terahertz Technology
H !-
■i ! —
I 98 100 102 104 106 108
Fig. 6: GUNN OSCILLATOR
OUTPUT POWER
FREO.
GHz
STAGEi
OUTPUT
TUNING
STAGE2
OUTPUT
TUNING
INPUT-
WAVEGUIDE
£1
1
OUTPUT- •
VAVEGUIDE
INPUT
TUNING
INPUT
TUNING
Fig. 7: ARRANGEMENT OF WAVEGUIDES
FOR IN-LINE MULTIPLIERS
Third International Sy7nposium on Space Terahertz Technology
Page 719
PC BIAS
WHBXOPqST
output/
ACXSH08T
TRIPLER
ONLY
Fig. 8: MULTIPLIER (SCHEMATIC)
specified 0.5 mW
over 330-345"GHz
Pout W*f*
Pcut WRS **
PcutWRTO*
Measurements at:
CHALMERS INSTITUTE
OF TECHNOLOGY, SWEDEN
RAOIOMETER-PHYSICS
Fig. 9: MEASUREMENTS OF OUTPUT POWER
WITH VARIOUS POWER METERS
Page 720
Third International Symposium on Space Terahertz Technology
•
10mU
i
1
- •
- •
■
• x2
5mW
—
•
■
■ x3
♦ x^
-
■
■
c x5
* x6
-
♦
* x3x2
® x2x3
x x2x^
1mW
3
500^iW
—
c
$ ®
-
t ♦
$
3
A A
i
I
mfinW
® i
iUU|l W i-
▲
c
50}dW
RADIOMETER
PHYSICS
MULTIPLIER
X
1 *
"* .
DATA
i
—
! *
*!
RESISTIVE
MJLTIPUCAT
EGN
KltiW
I
I.I.I,
CW.Y
1 , 1
200 300 400 500 600 700 GHz
Fig. 10 POWER OUTPUTS FROM VARIOUS MULTIPLIERS
Third International Symposium on Space Terahertz Technology
Page 721
BACK SHORT
WAVEGUIDE
BELLDV
FEEDHQPM
VHtSKEP PIN
Fig. 11: MIXER (SCHEMATIC)
■•2.0 Kacsr
D = Diode
I] = .125
1 = .100
L 2 = .250
2 = .100
L 3 = .235
3 = .355
L 4 = .240
4 = .100
L 5 =2.95
5 = .40
Fig. 12: MIXER CHOICE STUCTURE
Page 722
Third International Symposium on Space Terahertz Technology
Fig. 13: MIXER CHOKE RF- IMPEDANCE
Fig. 14: MIXER IF-IMPEDANCE 9-11 GHZ
A: before B: after matching
ni.o
"i — ■ — i — i — i — »~ »-~i — • — 1 — •"
"• ""■ — 1"--» — f — »—- f — i -- ! - t •— r" - 1- — i — ■ — i — • i — i— i • i i r~i r
a-
10x8 MHz
00.0
510 spectra integrated
(4sec. each)
w
79.0
78.0
4x40 MHz
I*-
_*J 5x80 MHz
77.0
Flight Kir una southwards
February C J^
76.0
"j-l_lJ._l_l_lJjJ_l.l-LlJ-l-l..lJ-Ll-i_l_LLIJJ^.l..i_Ll_r^
■0.60
■0.40 -0.20 0.00 0.20 0.40
offset frequency / GHz
0.60
3
S"
3
Crt
■a
o
<n
5"
o
3
</>
&
r>
(M
•S
re
-t
»
3-
o>
3-
3
o
I
Fig. 15: C10-LINE SPECTRUM
Page 724
Third International Symposium on Space Terahertz Technology
A 492 GHz Cooled Schottky Receiver for Radio- Astronomy
93-27785
J. Hernichel, R. Schieder, J. Stutzki, B. Vowinkel,
G. Winnewisser, P. Zimmermann *)
I. Physikalisches Institut
Universitat zu Koln
Ziilpicher StraBe 77
D-5000 Koln 41
Germany
^_^- Abstract
We developed a 492 GHz cooled GaAs Schottky receiver
driven by a solid state local oscillator with a DSB noise
temperature of 550 K measured at the telescope. The
receiver- bandwidth is asl.O.GHz. Quasi-optical mirrors
focus the sky and local oscillator radiation into the
mixer. Stability analysis via the Allan variance me-
thod shows that the total system including a 1 GHz
bandwidth acousto optical spectrometer built in Co-
logne allows integration times up to 100 sec per half
switching cycle. We succesfully used the receiver at the
KOSMA 3 m telescope on Gomergrat (3150m) located
in the central Swiss Alps near Zermatt during January-
February 1992 for observations of the 492 GHz, [CI]
3 Pi — * 3 Pq fine structure linie in several galactic sour-
ces. These observations confirm that Gomergrat is an
excellent winter submillimeter site in accordance with
previous predictions based on the atmospheric opacity
from KOSMA 345 GHz measurements.
Introduction
After the first 345 GHz observing run in winter 1988/89
[7] followed by three succesfull runs in the subsequent
winter periods, the 492 GHz receiver described here
and the observations carried out in January-February
1992 are the second step in our development toward
solid state LO Schottky receivers at 650 GHz and
800 GHz. The experience we made during the total
observing time of 120 days for 345 GHz and 35 days for
the 492 GHz receiver is that Gomergrat is an excellent
ground based site for submillimeter radioastronomy du-
ring the winter period. Together with Mauna Kea, Ha-
waii, Gomergrat is the only site where 492 GHz obser-
vations have been carried out.
Receiver Description
Based on the experience of the 345 GHz receiver [2] we
designed and built the receiver described here. The
front end section is shown schematically in Fig. 1. The
diplexer is a folded Fabry-Perot resonator using a mi
vable elliptical mirror; the semi transparent plates ai
wire grids with 125 lpi and 35 pm diameter wire; the
yield 83% reflection. A quasi-optical mirror arrai
gement focuses the sky and local oscillator radiatio
through a Potter horn into the mixer. The mirroi
are corrected for phase errors [5] and were produce
on a NC-milling machine at our Institute. Mixer an
IF-amplifier are cooled to 20 K. The bandwidth of th
HEMT amplifier is wl.OGHz at 1.4GHz center fre
quency; its noise temperature is «8K. Local oscilla
tor power is delivered by a 98.2 GHz Gunn oscillato
(60 mW) followed by a varactor quintupler, using a di
ode type 2T2 (J?, = 12fi,C ; =5fF) from the SDL, Uni
versity of Virginia. For an overview of typical multiplie
performance see [8]. The local oscillator is PLL stabi
lized. The Schottky- mixer is fundamentally pumped
has a coaxial IF-section and uses a 1T6 diode fron
University of Virginia (#,=20^,(7, =0.35 fF). The mi
xer is matched to the 50 fi preamplifier impedance bj
a simple coaxial transformer, without using a circula
tor. The receiver optics include two absorbing loadi
at ambient (295 K) and cold head (35 K) temperatures
The receiver design is very compact; it consists of twc
packages: a) the optic front plate and dewar, b) the
electronics, which fits in a 19" rack. A personal com:
puter is used to controle a) and b) fully remote. Fig. S
shows the optics and dewar part; the optic-plate diame-
ter is 32 cm and the weight is less than 45 kg, including
the cold head.
Receiver Characterization
The DSB receiver noise temperature was measured tc
be 500 K for a 600 MHz HEMT amplifier and 550 K
for the 1 GHz broadband HEMT amplifier, installed at
the telescope. Fig. 3 shows the receiver noise tempera-
ture versus mixer current for ambient and 20 K; coo-
ling thus reduces the noise temperature by a factor 01
2.8. The receiver noise temperature measured over the
QKSQJN&L PAGE IS
OF POOR QUALITY
Third International Symposium on Space Terahertz Technology
Page 725
1 GHz bandwidth at the telescope, using the Y-factor
method, is shown in Fig. 4: Receiver beam pattern mea-
surements were made in the laboratory; the 1/e opening
angle was measured to be 8.9° compared to the calcu-
lated value of 8.8°. Fig. 5 shows the two receiver beam-
pattern (horizontal and vertical scan) the Gauss-fit and
the 1/e-level which is the 13% power level. Similar to
earlier publications [3] we used receiver stability tests,
to verify the overall system performance. The Allan
variance method [6] adopted in Cologne, using a 1 GHz
bandwidth acousto optical spectrometer (AOS) yields
an integration time in the total power mode of 100 sec
per duty cycle (Fig. 6), without significant increase in
baseline structure.
Observing Run on Gornergart
We observed several galactic sources during the obser-
ving run in January- February 1992 at the KOSMA te-
lescope. Fig. 7 shows two highlights of the [C I] 3 Pi — *
3 Po 492 GHz lines from the galactic sources S 140 and
NGC 1977. The observations confirm that the KOSMA
location is an excellent winter submillimeter site in ac-
cordance with previous predictions based on the atmos-
pheric opacity from KOSMA 345 GHz measurements
[4]. The precipitable water vapor is below 1mm for
about 20% of the time in the winter period. Surpri-
singly the site does not show a significant variation of
transmission between day and night time. A Jupiter
map (Fig. 8) shows a clean beam. The half power dia-
meter of 54" results after deconvolution of Jupiter size
at the time of observation in a beam FWHM of 48",
consistent with the 47" predicted from the 12 dB edge
taper measured at the telescope. The telescope surface
was measured and adjusted by a holographic method
[1] to an rms of ss 38/im
Conclusion
We have designed and built a compact, cooled 492 GHz
Schottky mixer receiver with solid state LO. The re-
ceiver was installed on the KOSMA 3 m telescope on
Gornergrat during January-February 1992 and used to
observe several galactic sources. It is shown that the te-
lescope surface and the KOSMA site are well suited for
492 GHz observations. The receiver is build modular,
compact and lightweight and thus is usable for observa-
tions on other telescopes. The receiver noise tempera-
ture is 550 K DSB at the telescope over a simultaneous
bandwidth of ssl.O GHz.
Acknowledgements
We gratefully acknowledge the unconventional and ef-
ficient technical support during the observing run by
Ralf and Riidiger Zimmermann. The receiver develop-
ment was funded by the Bundesminister fur Forschung
und Technology (BMFT). The KOSMA 3 m- telescope is
operated by the University of Cologne and supported by
the Deutsche Forschungsgemeinschaft (DFG) through
grant SFB 301, as well as special funding from the Land
Nordrhein-Westfalen.
References
[1] Fuhr, W. ET al: Holographic alignment of the
KOSMA 3 m reflector, Univ. Cologne, in prepa-
ration.
[2] HERNICHEL, J.: Aufbau und Inbetriebnahme
des 345 GHz Empfangers fur das Kolner 3 m-
Radioteleskop, Diplom Thesis, Univ. Cologne,
(1989).
[3] Hernichel, J., Lewen, F., Matthes, K.,
Klumb, M., Rose, T., Winnewisser, G.,
Zimmermann, P.: SubmiJiimeter Receiver De-
velopment at the University of Cologne, Proc. of
Second Intern. Symposium on Space Terahertz
Technology, JPL, Pasadena, 641-647 (1991).
[4] Kramer, C, Stutzki, J.: Atmospheric Trans-
parency at Gornergrat, Technical Memorandum,
Univ. Cologne, (1990).
[5] Rose, T.: Aufbau eines SIS-Empfangssystem
mit quasioptischem Miscber, Diplom Thesis,
Univ. Cologne, (1989).
[6] Schieder, R., Tolls, V., Winnewisser, G.:
The Cologne Acousto-Optical Spectrometers, Ex-
perimental Astronomy 1, 101 (1989).
[7] Winnewisser, G., Zimmermann, P., Herni-
chel, J., Miller, M., Schieder, R., Un-
gerechts, H.: CO submillimeter observations
from Gornergrat, Astron. Astrophys. 230, 248-
251 (1990).
[8] Zimmermann, R., Zimmermann, R., Zimmer-
mann, P.: All Solid State Radiometers for Envi-
ronmental Studies to 700 GHz, this Conference.
^present address: Radiometer Physics, Bergerwiesen-
str.15, D-5309 Meckenheim, Germany
Primary
Mirror
Secon
Mirr
3
3
O
3
CO
"<
3
ts
o
5"
3
Q
3
in
-a
&
n
£
-«
&
3-
3-
3
O
£
492 GHz Receiver
lop view
side view
a*
a.
optic- plale
3
O
O
St
&
multiplier
harmonic
mixer
C iiiiii
oscillator
P 3
optic- plate
dewar
0">
o
S'
3
o
3
El
ci
a-
2
a-
a
o
optic 492 G J Iz
g
ho
Page 728
Third International Symposium on Space Terahertz Technology
3500
2000
2500
2000
1500
■'000
500
200
400 500
mixer current (uA)
800
1000
warm
cooled
receiver noise temperature versus mixer current
figure 3
Vetocily !>"»>/«)
200 -
492SO0
*920OO
Rest frequency (MHr)
receiver noise temperature versus frequency
figure 4
Third International Symposium on Space Terahertz Technology
Page 729
uu
go
1 1 1 1 J*
V ' I ■ ■ 1 1
V horizontal -9 —
A vert ica 1 -< —
so
♦ Gauss
70
\
60
\
50
-
\
40
ft
30
V\.
20
-
\^ l/e-level
10
-
\
>- k^Vf>Wf/ ' 1
L 1 ^^fc*^ — t » ,
-25 -20 -15 -10
-5 5
[Degreel
10 15 20 25
receiver beam pattern 492 GHz
figure 5
<r 2 [counts]
18E 3
-i 1 — i — i — i t i i 1 1 r-
» i
l>
I, I I I I I I I till
is
*'4±
_i i 1 1 1
j ■ ■ -i
r [sec]
Allan plot
figure 6
Page 730
Third International Symposium on Space Terahertz Technology
Vetnri'v (trm/O
figure 7
VHnrJtv (y™/*)
Jupiter beam map 492 GHz
10 15 20 (10)- 90% of peak
figure 8
Live Music Archive
Librivox Free Audio
Metropolitan Museum
Cleveland Museum of Art
Internet Arcade
Console Living Room
Books to Borrow
Open Library
TV News
Understanding 9/11