Microwave Avalanche Diodes: Simon M. Sze, Robert Ryder
Microwave Avalanche Diodes: Simon M. Sze, Robert Ryder
8, AUGUST 1971
Microwave AvalancheDiodes
SIMON M. SZE, MEMBER, IEEE, AND ROBERT M RYDER, FELLOW, IEEE
Abstract-Microwave avalanche diodes of various types ( IMPATT, Section VI presents the operational principle of the TRAPATT mode
T R A P A T . etc.) can generate power sufficient for microwavereceivers
and briefly discusses other modes of operation. Designs and state-
and some transmitters. This brief review summarizes mechanisms of
operation, power output, efficiency, noise, and some important fea- of-the-art performances of avalanche diodes as of late 1970 are
tures of design and fabrication. summarized in Section VII.
11. DRIFTV m m AND AVALANCHE MULTIPLICATION
I. INTRODUCIION
There are two fundamental physical processes most pertinent to
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
ELECTRIC FIELD ( V l C m )
-
Fig. 1 . Drift velocity of charge carriers. At high fields lo5 V/cm, the
scattering limited velocity c. is nearly independent of field [5]-[7].
TABLE I
MATERIAL PARAMETERS
Semiconductor Ge Si GaAs
Thebasicequationwhichdescribesavalanchemultiplication
when it is initiated by electrons is [I21
jowa dx = 1.
sided abrupt junctions (and Schottky barrier contacts) are shown BACKGROUND DOPING, NE ( c m - 3 )
[I31 in Fig. 3. It is interesting to note that at a given background Fig. 3. Breakdown voltage, breakdown field, and depletion-layer
doping the breakdown voltage increases with the semiconductor width for abrupt p-n junctions [13].
energy bandgap(Vs-E2/2), and for a given semiconductor the value
of V' decreases with increasing doping (V' NB 3/4). Also shown in
N
300 GHz (or wavelength range from1 mm to 1 m). The diodes have
Fig. 3 are the maximum electricfields at breakdown. For Si the maxi- been fabricated from Ge, Si, and GaAs and can probably be con-
mum field is 3 x 10s V/cm at IO" m ~ - ~
and increases slowly to structed from other semiconductors should the need arise. WATT
6 x IO5 V/cm at lo1' diodes offer potentially inexpensive, reliable, compact, and moder-
ately efficientmicrowave sources. We shall nowconsiderthe
In. m h s
t OF IMPATT OPERATION mechanisms of WATT operation.
MATT diodes are at present the most powerful CW solid-state
sources of high-frequency microwave power; up to1 W at 50 GHz A . Basic Considerations
has been realized (see later Fig. 28). The MAIT diode has covered Many electrondischarge devicesmay be considered to be
thecompletemicrowavefrequencyspectrumfrom 300 MHz to analyzable into the following functionalparts:
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
a) a source of carriers ;
b) a region where the current of the discharge is modulatedby
the RF signal ;
c) a working region where the carriers produce useful power
output by interacting with the RF field;
d) one or more collectors which dispose of the spent carriers.
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
SZE AND RYDER: MICROWAVE AVALAKCHE DIODES 1143
L
DISTANCE
-G
DC CURRENT DENSITY, J,
(bl
0 I 2r -G .
TRANSIT ANGLE = B = w W / V ,
(bl
Fig. 5. Read diode. (a) Avalanche and drift regions. (b) Idealized
negative resistanceversus transit angle.
For example,foraIO-GHzoperationinsilicon,thedepletion
width should be about 5 pm.
FREQUENCY
B. General Considerations [151 (C)
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
1144 PROCEEDINGS OF THE IEEE, AUGUST 1971
J= OHMIC
CONTACT
T
89.
i 120
ISOTHERMAL I-V
40
\ r0 Ac
20 0
(bl
v
The capacitanceassociatedwithadepletionregion of W is
sented for various ac voltages [18]. Also indicated on the plot are C=E,.A/W. Using the transit-time frequency defined in (6), (7) can
the calculated efficiencies. Note that generally the susceptance in- be rewritten as [19]
creases with increasing ac voltage amplitude whereas the negative
conductance decreases. The calculated result shows efficiencies as P, f = Eivf/8nXc (8)
high as 18 percent. Such high efficiency, however, has not been
obtained experimentally fromdiodes operated in the WATT mode. where X, is the reactance (2nfC)-'. Assuming we are limited to
The limitations on efficiency and power output will now be con- some minimum circuit impedance, (8) predicts that the maximum
sidered. power which can be given to the mobile carriers decreases as l/fz.
This electronic limit is dominant at high frequencies ( 2100 GHz
N.POWERAND E n c w c r LIMITATIONS in Si).
A . Power-Frequency Limitation-Electronic B. Thermal Limitation on Power Output
Because of the inherent limitations of semiconductor materials At lower frequencies,the CW performance of an IMPAT diode
and theattainableimpedancelevels in microwave circuitry, the is mainly limited by thermal considerations, i.e., the
bypower which
maximum output power at a given frequency of a single diode is can be dissipated in a semiconductorchip. A typical device mounting
limited. The limitations on semiconductor materialsare 1) the criti- arrangement [20] is shown in Fig. 8(a). The cylindrical semiconduc-
cal electric field E,,, at which the avalanche breakdown occurs and tor wafer is metallized and attached to a larger metal stud which
2) the scattering-limitedvelocity us, which is the maximum attain- serves as a heatsink. The heat is generatedin the high field regionin
able velocity in the semiconductor. the semiconductor. The thermal resistance R, (in units of "CIW),
The maximum voltage that can be applied across a semiconduc- consists of three components: 1) the effective thermal resistance of
tor sample is limited by the breakdown voltage which, for a uniform the semiconductor, approximately W/2Aus,where A is the diode
avalanche, is given by V,=E,W where W is the depletion width. area and usthe semiconductor thermal conductivity,2) the contact-
The maximum current that can be carried by the semiconductor ing area component whichis proportional to 1/A, and 3) the spread-
sample is also limited by the avalanche breakdown process,because ing resistance component of the heat sink 1/2Duc where D is the
of the fact that the current in the space charge region causesan in- diameter.
crease of the electric field (from Poisson's equation). the With
maxi- The powerP , which can be dissipated in the diode,must be equal
mum field again givenby E,,,, we obtain I,= E,E,u,.A/W. Therefore, to the heat power which can be transmitted to the heat sink, there-
the upper limit on the power input is given by the product of V,,, fore, P = A T / R , where AT is the temperature difference between the
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
SZE ASTJ RYDER: MICROWAVE
AVALANCHE DIODES 1145
junction and the heat sink, and R , is the total thermal resistance. If
the reactance 27rj-C (where C=A&JW)is maintained constant, and
the major contribution to the thermal resistance is from the semi-
conductor, then we obtain for a given temperature increase A T :
pf = - c AT
E,
= constant
BREAKDOWN FIELD
conductor depends on the maximum allowed temperature rise and ,BREAKDOWN FIELD
minimum attainable circuit impedance. /'-
\
C. Limitations on Eficiency w \
a I \
Efliciency is the ratio of ac power output to the power delivered
by the dc voltage or current source; therefore
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
1146 PROCEEDINGS OF THE IEEE, AUGUST 1971
META
' EPI LAYER
-SUBSTRATE
8
-
v)
P=O.W il-cm
f = 1 0 0 GHz
12 ,
0 2
I
4
I
6
I
Fig. 13. Skin effect. (a) Current flow co&ed to a surface lamina of thick-
ness 6 causing nonunifonnity and resistive. loss in the diode.(b)Cal-
culated voltagedrop in substrate at 100 GHz for several diode diameters
D . (After De Loach, [27].)
causing a reduction of the efficiency. Physically, the difEculty results D. Bumout from Filament Formation
because a higher saturation current causes the avalanche to build Burnout may occur not only if the diode is overheated but also,
up too soon, reducing the avalanche phase delay again. The minor- more insidiously, if the camer current fails to be uniformly dis-
ity injection [25] from P poor ohmic contact will also increase the tributed over the diode area, but instead concentrates intofilaments
reverse saturation current andthus reduce the efiiciency. of locally high intensity. Such untoward behavior can often result
The effects of an unswept epitaxial layer [26]and the minority when the diode has a dc negative conductance, because then the
injection current [26] are shown in Fig. 12. The unswept layer of local region of greatest current density has also the lowest break-
thickness d (shown in the insert)gives rise to a series resistance downvoltage. For this reason, p-i-n diodes are prone to easy
whichreduces the terminal negativeresistance. In the example burnout. Themoving camer space charge of drift regions tends to
shown the dciency drops from 10 to 2 percent with 3 pm of un- preventlow-frequencynegative conductance (10) and therefore
swept layer. helps to prevent filamentary burnout. Some diodes whichhave
The skin effect [27] is shown in Fig. 13. As the operating fre- positive dc conductance at low currents may develop negative dc
quency of an WAIT diode is increased into the millimeter wave conductance and burnout at high currents.
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
SZE AND RYDER: MICROWAVE AVALANCHE DIODES 1147
TRANSIT-TIMEFREPUENCY=6GHz
c
I I I I I I
0 2 4 6 8 10 12
FREOUENCY ( G H z )
Fig. 15. Noise measure for GaAs IMPATT diodes. (After Blue,
TABLE 11
NOISE MEASURE
OF IMPATT DIODES
Semiconductor Ge Si GaAs
t Small-signal’
noise measure (dB)
Large-signalb oscillator noisemeasure
(dB)
30
40
40
55
25
35
a Forsmall-signal amplifier,the noise measureis M = T/Towhere
To =290”K.
Oscillator FM noise is given by
kToMB lI2
* L a = (&)(T)
wheref, =oscillator frequency,Qex, =external circuit Q, M = t h e largesignal
FREOUENCY (GHz) noisemeasure,B=measurementbandwidth, and Po =oscillatorpower
output.
Fig. 14. Calculatedequivalentnoisevoltage versus frequency.The peak
at the avalanche frequency and the dip at twice the transit time frequency
arereally reflectionsofcorrespondingchanges in diodeimpedance. a particular silicon diode [31] is shown in Fig. 14. The two theories
The output noise power varies more smoothly with frequency. (After
agree qualitatively in showing a peak of noise voltage at the av-
Gummel and Blue, [3 13.)
alancheresonantfrequency w,, falling off rapidly at higherfre-
quencies. Noise can therefore be reduced somewhat by operating
V. NOISEBEHAVIOR well above the avalanche frequency and keeping the current low.
The noise level ofWATT oscillators is very high, corresponding These conditions are in conflict with the conditions favoring high
to amplifier noise figures of 2 0 4 0 dB. The reason for such high power and efficiency so that tradeoffs are necessary to optimize for
noise is that an avalanche discharge consists of a large number of particular applications.
carriers which cross the interaction space as a group, but whose Fig. 15 shows theoretical and experimental results of the noise
time position “jitters” in phase from one cycle to the next because measure in aGaAs WAIT diode [32]. At thetransit-time frequency
of the randomway in which the avalanche grows from few a initial (6 GHz) the noise measure is about 32 dB. The minimum noise
ionizing events. Therefore this “jitter” noise is much greater than measure of 22 dB, however, is obtained at about twice the transit-
shot noise, in which the individualcarrier transits are random. The time frequency. Oneimportant feature of the GaAs noise measure
avalanche noise can be greatly reduced by driving the oscillator is that it is substantially lowerthan thatfor Si WATT diodes.
with a smaller quieter “locking” oscillator which triggers the av- Table I1 shows some of the state-of-the-art noise measures [33] of
alanche more uniformly in successive cycles. Ge, Si, and GaAs WATT diodes. The amplifier and oscillator
The noisehasbeenquantitativelyanalyzedinsmall-signal noises in the table are for a lossless circuit at the frequency corre-
theories, analytically by Hines [u)] and numerically by Gummel sponding to maximum oscillator efficiency,withoutharmonic
and Blue [31]. For a high-gain amplser under simplified assump tuning.
tions, the noise figure is [30] The main reasonfor the low-noise behavior in GaAs is that for
a given field the electronand hole ionizationrates are essentially the
NF=1+ 4 V*Ik To same in GaAs while in Si theyare quite different. From (2) it can be
4m7:(02 - wf) shown that to obtain a large multiplication factor M the average
distance of ionization l/(a) is about equal to W, (the avalanche
where ’V isvoltage and T’ the transit-time across the avalanche width) if a,= ap, but is about equal to WJIn M if a, >> ap. So that
region, andm is an exponent which describes approximately the field for a given W, there haveto be considerably more ionization events
dependence of impact ionization, i.e., a=a,(E/E,)”’. Theplot of in Si, which results in higher noise.
mean-square noise voltage per unit bandwidth versus frequency for The relation between power output and FM noise measure is
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
1148 PROCEEDINGS OF THE IEE, AUGUST 1971
32 t
!2gi
26
t
23 t
26 X) 34 46
38 42 50
NOISE
MEASURE (dB)
Fig. 16. Noise versus output power for a phase-locked oscillator. Locking
powerwas held constant at +4 dBm. Contours of constant load im- DISTANCE -
pedance Z and constant diode current I are shown. (After Irvin et ol., (b)
WI.)
t
shown in Fig.16 for some Si and GaAs 6-GHz WAIT diodes [MI.
The diodes were evaluated in a single tuned coaxial resonator circuit
in which the load resistance presented to the resonator was incre-
mentally varied by using interchangeable impedance transformers
Z . At maximum power output the noiie measure is relatively poor.
Lower noise measure can be realized at the expense of a slightly
reduced power output. We note again that at a given power level DISTANCE
(say 1 W) the GaAs IMPATT diode is about 10 dB quieter than a Si (C)
WAIT diode.
Fig. 17. TRAPATT modeoperation(schematic). (a) Current density versus
time. (b) Idealized electric field rising with time. (c) Avalanche shock
VI. MECHANISMS OF TRAPAITOPERATION AND front which sweeps through the diode with velocity v , greater than the
OTHERMODE^ OF OPERATION scattering-limited velocity0,. (After Clarfeine et ol., [&I, De Loach and
Scharfetter, [41].)
A . TRAPATT Mode
The trapped plasma or TRAPATT mode is a high-power high-
or upon differentiation
efficiency mode. To date thehighest pulse power of 1.2 kW has been
obtained at 1.1 GHz (5 diodes in series) [35] and the highest effi-
ciency [36] of 75 percent has been obtained at 0.6 GHz. However,
the TRAPAIT operation is a rather complicated manner of oscillation where v, is the avalanche zone velocity and represents the velocity
and requires good control of both device and circuit properties. In at which the leading edge of the avalanche region progressedthrough
addition, the TRAPAIT mode generally shows considerably higher the diode, Fig.17(c). For example, for NA= 10'' and J = 104
noisemeasure than WAIT mode; and the upper operating fre- A/cmz, uI= 6 x lo7 cm/s which is much larger than the scattering-
quency seems to be practically l i i t e d to below the millimeter wave limited velocity. Thus the avalanche zone [39] (or avalanche shock
region. front) will quickly sweepacross most of the diode, leaving the diode
To understand the initiating of the WAIT mode, we consider filled by a highly conducting plasma of holes and electrons whose
a current step which is applied at t = O to a n+-p-p+diode, Fig. 17(a). space charge depressesthe voltage to low values [37], [38].
If at t =0 the maximum electric field of the diode is smaller than the The field and carrier density calculated numerically for a particu-
critical field E, at breakdown, the field profile will at first move up lar moment is shown [42] in Fig. 18(a). The corresponding voltage
with time as shown in Fig. 17(b). This is because [37l-[41] and currentwave forms are shown in Fig. 18(b) and (c), respectively.
dE (The dot indicates the situationshown on Fig. 18(a)). We note that
J = E~--' the voltage at the beginning of the cycle can become considerably
dt
larger thanthe steady-state breakdown voltage.After the av-
Hence the field can be expressedas alanche zone passes through the device the voltage drops to low
values, but soon rises again as the field polarizes the space charge.
qNA
E(x, t ) = E , - -x -
Jt
+ Because of the dependence of the driftvelocity on field at low fields
ES E, (Fig. l), the electrons and holes will drift at velocities determined by
where N A is the doping concentration of the p region. Thus the the low-field mobilities, and the time of transit of the carriers can
value oft at which the electric field reachesE , at a given distance x become much longer than W/us.
into the depletion region isobtained by setting E(x, t )=E,, yielding Thus the GAIT mode can operate at comparatively low fre-
quencies since the discharge time of the plasma, i.e., the ratio Q/Z
of its charge to its current, can be substantially greater than the
nominal transit time W/vsof the diode at high fields.Therefore, the
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
SZE AKDRYDER MICROWAVEAVALANCHEDIODES 1149
ELECTRONS 1000 I
I
A A
y\ TRAPATT
I
DISTANCE
(a)
I I I 1 1
05 I 2 5 10 20 50 100
t FREQUENCY ( G H Z )
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
1150 PROCEEDINGS OF THE IEEE, AUGUST 1971
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
SZE AND RYDER: MICROWAVE AVALANCHE DIODES 1151
-THEORETICAL
GaAS
0 EXPERIMENTAL FOR N , = 6 X 1 0 1 5 c ~ - 3
TABLE IV
DEVICE PARAMETERS
AND PERFORMANCE
OF EXPERIMENTAL
SILICON IMPATT DIODES
Frequency
Range 8-10 GHz 5- GHz’ 10&110 GHz
mi/,
t I I
lo20 DOUBLE
DRIFT REGION
0.53W
@P N N+
/ I I 0 0 . 7 ~I
1 0 0 200 300 400
JUNCTION TEMPERATURE PC)
(a)
- DOUBLE DRIFT
CALCULATED ( f =&,)
P El
1 1 . I I I I I I
4 6 8 IO 12 11
84 16 i
CURRENT DENSITY (KILO-AMPS /crn2)
(b)
Fig. 25. Poweroutputand efficiency ofsingle-driftversusdouble-drift
IMPATT oscillators at5 0 GHz. (a) Power output for two individual diodes.
(b) Range of efficiency for four diodes of each type. (After Seidel et al.,
[591.)
IO
-E
z a
1200 -
0
L
1.0 z
0 1000 -
L
-
/
Y
n
I
I-
t
- BOO-
POWER
t-
3
F
3
I
0.I I .o IO 100 0
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
SZE AND RYDER : MICROWAVE AVALANCHE DIODES 1153
‘\
I. I/f xX 1
I
X PULSE diode, by Irvin et al. [MI,
conditions, as mentioned
exhibits to date the highest C W output
power from a single chip. The highest power result under pulse
previously,was
obtained
by
Liu
+i et aZ. [35].
t ‘ 1 ~ DIODES
3 ON DIAMOND
I
ACKNOWLEDGMENT
The authors would like to acknowledge the contributions of
numerous colleagues at Bell Telephone Laboratories. In particular
they wish to thank B. C . Dehach, Jr., R.Edwards, W. J. Evans,
J. C. Irvin, T. Misawa, D. L. Scharfetter, and T. E. Seidel for their
published and unpublished work, helpful discussions, and critical
reading of the manuscript.
REFERENCES
[l] R. L. Johnston, B. C. De Loach, Jr., and G. B. Cohen, “A silicon diode
microwaveoscillator,” Bell Syst.Tech. J., vol. 44, Feb. 1965, pp.
369-372.
[21 W. T. Read, “A proposed high-frequency negative resistance diode,”
Bell Syst. Tech. J., vol. 37, Mar. 1958, pp. 401-446.
[3] H. J. Prager, K. K. N. Chang, and S. Weisbrod, “High-power, high-
efficiency silicon avalanchediodesatultra high frequencies,” Proc.
IEEE (Lett.), vol. 5 5 , Apr. 1967, pp. 586-587.
[41 B. C. De Loach, Jr., “Modes of avalanche diodes and their associated
circuits,” IEEE J. Solid-stateCircuits, vol. SC-4, Dec. 1969, pp.
376-384.
[5] S. M. Sze, Physics of SemiconductorDevices. New York: Wiley,‘
1969.
[6] J. G. Ruch and W. Fawcett, “Temperature dependence of the transport
properties of gatsdetermined by the Monte Carlo method,” to be
published.
171 V. L. Dalal,“Hole velocity in p-gats,” Appl.Phys.Lett., vol. 12,
June 15, 1970.
I I I l l I I I l l I x 1 1 I~ [8] E. M. Conwell, High Field Transport in Semiconductors. New York:
I 10 100 1000 Academic Press, 1967.
FREQUENCY ( G k z ) [9] C. Y. Duh and J. L. Moll, “Temperature dependence of hot electron
drift velocity in silicon at high electric fields,” Solid-state Electron.,
Fig. 28. State-of-the-art performance, power versus frequency, for I M P A ~ vol. 1 1 , 1968, p. 917.
diodes in 1965 and 1970. Ilf slope is expected for thermal limitation; [lo] G. A. Baraff, “Distributionfunctionsandionizationratesforhot
lif’ for electronic limitation ([19], [34], [48], [53], [aO], [62], [63]). electrons in semiconductors,” Phys. Reo., vol. 128, 1962, p. 2507.
[ l l ] C. R. Crowell and S. M. Sze, “Temperature dependence of avalanche
multiplication in semiconductors,” Appl.Phys.Lett., vol. 9, 1966,
p. 242.
[121 J. L. Moll, Physics of Semiconductors. New York: McGraw-Hill,
1964.
[13] S. M. Sze and G. Gibbons, “Avalanche breakdown voltages of abrupt
and linearly graded p-n junctions on Ge, Si, GaAs and Gap,” Appl.
Phys. Lett., vol. 8, 1966, p. 1 1 1.
S. M. Sze, M. P. Lepselter, and R. W. MacDonald, “Metal-semicon-
ductor ImDatt diode,” Solid-state Electron., vol. 12, 1969, p. 107.
[I41 M. Gilded and M. E. Hines, “Electronic tuning effects in the Read
microwaveavalanchediode,” IEEETrans.ElectronDevices, vol.
ED-13, Jan. 1966, pp. 169-175.
H. K. Gummel and D. L. Scharfetter, “Avalanche region of Impatt
diodes,“ Bell Syst. Tech. J., vol. 45, Dec. 1966, pp. 1797-1828.
[I51 For a detailed study of the IMPATT physics, see, for example, [5] and
the following references :
J. E. Carroll, Hot Electron Microwave Generators. London: Edward
Arnold (Publishers) Ltd., 1970, chs. 8-10.
H.A. Watson, Ed.,Microwave Semiconductor Devices and their Circuit
Applications. New York: McGraw-Hill, 1969, ch. 15.
T. Misawa,“Impattdiodes,”in SemiconductorApplications, vol. 6,
SemiconductorsandSemimetals. NewYork:Academic Press, 1970.
W. J. Evans, “Avalanche diode oscillators,”Solid in State and Quantum
Electronics, W. D. Hershberger, Ed. New York: Wiley, 1971.
[16] T. Misawa, “Negativeresistancein p-n junctionsunderavalanche
Fig. 29. State-of-the-art power versus frequency, for TRAPATT breakdownconditions,” IEEETrans.ElectronDecices, vol. ED-13,
diodes (1351, [40], [661-[751). Jan. 1966, pp. 137-151.
171 S. M. Sze and W. Shockley, “Unit cube expression for space-charge
resistance,” Bell Syst. Tech. J., vol. 46, 1967, p. 837.
C . State of the Art 181 D. L. Scharfetter and H.K. Gummel, “Large-signal analysiso f a silicon
Readdiodeoscillator,” IEEETrans.ElectronDevices, vol. ED-16,
Jan. 1969, pp. 6&77.
A s u m m a r y of the present state of the art is given in Fig. 28 for W. J. Evans and G. I. Haddad, “A large-signal analysis of IMPATT
W A T T diodes and in Fig. 29 for TRAPATT diodes. Also shown in diodes,” IEEETrans.ElectronDeaices, vol. ED-15, Oct. 1968, pp.
Fig. 28 are the results obtained [19] in 1%5. We note there are two 708-7 11.
J. L. Blue, “Approximate large-signal analysis of Impatt oscillators,”
to three orders of magnitude improvement in power output and Bell Syst. Tech. J., vol. 48, Feb. 1969, pp. 383-396.
over oneorder of magnitude in extending the upper frequency limit. 191 B. C. De Loach, Jr., “Recent advances in solid-state microwave gen-
Authorized licensed use limited to: IEEE Xplore. Downloaded on March 27, 2009 at 07:25 from IEEE Xplore. Restrictions apply.
1154 PROCEEDINGS OF THE IEEE, AUGUST 1971
erators,” in Adcances in Microwaves, vol. 11. New York: Academic -, “CW millimeter-wave IMPATT diodes with nearly abrupt junc-
Press, 1967. tions,” Proc. IEEE (Lett.), vol. 56, Feb. 1968, pp. 234-235.
[20] C. B. Swan, T. Misawa,andL.P.Marinaccio,“CompositeavalancheR.A.ZettlerandA. M. Cowley, “Batchfabricationofintegrated-heat-
diode structures for increased power capability,” IEEE Trans. Electron sink Impatt diodes,” Electron. Lett., vol. 5 , Dec. 1969.
Decices, vol. ED-14, Sept. 1967, pp. 584-589. [48] R.Edwards, D. F.Ciccolella, T. Misawa, D. E. Iglesias, and V. Decker,
[21] G. Gibbons and T.Misawa, “Temperature and current distribution in “Millimeter-wave silicon Impatt diodes,” Int. Electron Devices Meet-
anavalanchingp-njunction,” Solid-StateElectron., vol. 11, 1968, ing, Oct. 29-31, 1969.
p. 1007. [49] D. De Nobel and H. G. Kock, “A silicon Schottky barrier avalanche
[22] R. H. Haitz, H. L. Stover, and N. J. Tolar. “A method for heat flow .
transit time diode.” Proc. IEEE (Lett.).,, vol. 57.
~
Nov. 1969. DD. 2088-
,~ ~ ~ ~~~~