Heterojunction IMPATT diodes: using new material technology in a classic device.
There are many current and potential applications of impact avalanche and transmit time (IMPATT) diodes in both the commercial and military sectors, since these devices are the preeminent solid-state choice for generating RF power in the 5 to 100 GHz frequency range. Current in-field applications include a commercial X-band weather-avoidance radar transmitter, telecommunications power amplifiers in K- and Ku-bands, missile seeker transmitters in X-band and threat simulators in C- through Ku-bands. These implementations of IMPATT diodes provide peak power levels in the 20 to 200 W range, and experimental combiners have been designed and built that exceed 1 kW of peak RF power. To realize many of these (and future) applications, improvements in IMPATT performance were needed.
The IMPATT diode utilizes two phenomena to generate RF power by effecting a 180 |degrees~ phase shift between the AC voltage and current waveforms, once DC power is provided to the diode and it is properly mounted in a resonant cavity or circuit. Figure 1 shows the single-drift high-low profile device, which is the configuration discussed in this paper. In this configuration, the P/N junction is located between the P-side and the avalanche layer, which is N+ doped. The device features highly-doped P++ and N++ contact layers, upon which ohmic metal contacts are deposited, a P+/N+ junction, a short N+ avalanche layer and a relatively long, lightly-doped N- drift layer. These diode structures were grown by molecular beam epitaxy (MBE), fabricated by standard photolithographic techniques and packaged for mounting in the resonant cavity, as shown in Figure 2. The package body is usually threaded for mounting. The diode is mounted upside down so that a positive voltage applied to the package lid is reverse bias for the P/N junction, as IMPATT diodes are always operated in reverse avalanche breakdown.
The basic theory of IMPATT operation has not changed much since the device was first introduced,|1~ and is best understood by referring to the waveform sketches in Figure 3. If a positive DC voltage slightly less than the avalanche breakdown voltage is applied to the N-side of the diode, and an AC voltage is superimposed, then as the AC voltage increases to its maximum value, a large pulse of electron-hole pairs is generated near the P/N junction as the junction electric field reaches a strength sufficient to cause impact ionization. This is shown as the injected current waveform. There is a time lag between the AC voltage peak and the injected current pulse, since the rate of ionized carrier production is proportional to the carrier density. Therefore, some time is required for the avalanche process to build up to its peak. In the idealized device this time lag is 90 |degrees~, and as the produced electrons move away from the junction, they begin to transit the drift layer, whose length is such that the electron pulse reached the N++ layer after a time equal to one-half of an RF period. Therefore, the drift layer length is the primary means of controlling the fundamental frequency of oscillation for an IMPATT diode. Since the avalanche and drift layers are depleted of mobile carriers by the applied DC voltage, the electron pulse induces an external current during its transit of the drift layer. The device has caused a 180 |degrees~ phase shift between the AC voltage and current waveforms, and therefore, has achieved a negative resistance effect, that is, a DC bias power is converted to RF power.
In order to compare the operation of the heterojunction IMPATT (HJ-IMPATT) diode with standard devices, some of the design guidelines applicable to standard, single-drift, high-low profiles are reviewed. The first-order theory of operation requires that for maximum DC-to-RF conversion efficiency the DC voltage drop across the avalanche layer must be minimized compared to the drop across the drift layer, since only in the drift layer is the AC current antiphase to the AC voltage. In addition to this consideration, the ideal device confines the region of impact ionization to the avalanche layer, and thus, the electric field must decrease below the ionization threshold at the avalanche/drift interface. These requirements place constraints on the doping levels and layer thicknesses, which have been optimized by numerous theoretical and experimental studies. For example, for optimum CW operation in the 5 to 18 GHz range, values|2~ have been provided that have resulted consistently in conversion efficiencies above 15 percent even in a large-scale production environment.
The major refinement to the original theory of operation|1~ is a dynamic mode of operation that has become known as the premature collection mode (PCM), which has been described previously.|3~ High-low-type IMPATTs are always operated in PCM, and the conversion efficiency can exceed 20 percent if the designer has allowed for PCM in the device structure. Higher efficiency is realized in PCM because of the large-signal modulation of the edge of the space-charge depleted layer by the AC voltage. As the depletion zone edge moves back towards the P/N junction, it meets the outgoing ionized carrier pulse, which is then collected earlier than in the small-signal case where there is no depletion zone edge modulation. The designer allows for PCM by utilizing a somewhat longer drift layer, which accommodates the depletion zone edge modulation. Practical IMPATT designs represent a compromise between these design guidelines and end-user requirements, which include pulsed operation and temperature stability, as well as other considerations that require deviations from the optimum profile design.
Heterojunction IMPATT Diodes
The HJ-IMPATT diode design takes advantage of the properties of the abrupt p/N heterojunction, as well as the large bandgap of |Al.sub.x~|Ga.sub.1-x~As and the effects following from this. Figure 4 shows the conduction band of a GaAs/|Al.sub.0.3~|Ga.sub.0.7~As p/N abrupt heterojunction with doping levels of |N.sub.A~ = |10.sup.17~, |N.sub.D~ = |10.sup.16~ |cm.sup.-3~; a portion of the bandgap difference appears as a 0.25 to 0.3 eV conduction band discontinuity. This discontinuity produces a number of effects, one of which is a substantially larger peak junction electric field that can be as much as 80 percent larger than in a standard junction with equivalent doping levels.|4~
A detailed review of IMPATT diode operation and theory has been published previously.|5~ While many contributions have been made to the device theory, the original paper still contains the basis for understanding the IMPATT diode. Referring to Figure 3, the average power developed by the IMPATT diode is given by
|Mathematical Expression Omitted~
|I.sub.max~ = magnitude of the current pulse
|V.sub.RF~ = peak RF voltage
|omega~ = fundamental frequency (rad/s)|6~
Parameters x and y are the off and on times (in radian units), respectively, of the current pulse. The assumptions here are a sinusoidal RF voltage, implying that the diode is mounted in a resonant high Q cavity and an ideal rectangular current pulse. This idealized device requires an ionized carrier pulse of very short duration compared to the RF period, and neglects diffusion effects as the pulse transits the drift layer. By integrating the current pulse over one period to yield the DC current,
|Mathematical Expression Omitted~
Using |P.sub.DC~ = |V.sub.DC~|I.sub.DC~ and the conversion efficiency definition of |eta~ = |P.sub.RF~/|P.sub.DC~ yields
|Mathematical Expression Omitted~
|eta~ = efficiency
Equation 3 can be maximized when the current pulse starts at y = 1 (so |omega~t = |pi~ or 90 |degrees~ after the AC voltage peak), and stops at a somewhat earlier value of x = 1.74 instead of at x = 2. This is the basis for the early collection of the carrier pulse, that is, the premature collection mode. For these ideal conditions and a 50 percent RF voltage modulation (|V.sub.RF~/|V.sub.DC~ = 0.5), the maximum diode DC-to-RF efficiency is 1/|pi~ or 32 percent.
The DC voltage drop across the device in reverse bias is shared by the drift and avalanche layers, but the voltage across the avalanche layer itself reduces efficiency. An estimate of this effect|7~ is given by
|Mathematical Expression Omitted~
|V.sub.av~ = voltage drop across the avalanche layer
|V.sub.ext~ = voltage across the entire diode
For |V.sub.av~ |is less than~ |is less than~ |V.sub.ext~, the previous result was obtained, or 32 percent for the maximum efficiency. However, |V.sub.av~ cannot be made arbitrarily small, since the junction electric field must be sufficiently high to sustain impact ionization. However, the efficiency drops rapidly as |V.sub.av~ increases relative to |V.sub.ext~. A number of other factors contribute to lower device efficiency, such as contact and bulk material resistances, and thermal effects.
There are several factors restricting the design of IMPATT diode profiles. Heterojunction technology allows further tailoring of the device beyond varying the layer dopings and thicknesses to enhance performance. Figure 5 shows the energy band structure of a prototype heterojunction IMPATT diode as predicted by the semiconductor simulation program SEDAN under zero external bias. In this case, the avalanche layer (between 1.2 and 1.5 |micrometer~) is |Al.sub.0.3~|Ga.sub.0.7~As, with N-type doping of 2 x |10.sup.17~ |cm.sup.-3~. The conduction band spike is at 1.2 |micrometer~. In this device there are two heterojunctions, a p/N type at 1.2 |micrometer~ and a N/n type at 1.5 |micrometer~. The GaAs p-side of the junction is doped at 2 x |10.sup.17~ |cm.sup.-3~, and extends from 0.6 to 1.2 |micrometer~. The drift layer is GaAs n-doped at 4 x |10.sup.15~ |cm.sup.-3~, extending from 1.5 to 4.5 |micrometer~. SEDAN modeling also predicts that this device at zero applied bias would have a peak junction electric field of over 290 kV/cm, almost twice as large as in a standard P/N junction with the same doping levels.
RF Power and Efficiency Results
All devices were characterized in a Kurokawa single diode waveguide cavity as free-running oscillators. Proper impedance matching in a cavity of this type is facilitated by a wide selection of transformer rings, which act as quarter-wavelength transformers. Ring impedance values were available from 5 to 25 |omega~. The cavity also includes an adjustable backplane short and a waveguide sliding stub tuner for impedance match adjustment.|11~
TABULAR DATA OMITTED
TABULAR DATA OMITTED
Table 1 lists the results of 10-piece samples taken from both standard and HJ-IMPATT diode wafers, the standard performance values taken from wafers representing the best optimized profile for the frequency band of interest. Pertinent DC parameters are listed in Table 2. Of particular interest are the first two listed Ku-band results. The HJ-IMPATT profile in this case was virtually identical to the standard profile with respect to doping levels, layer thickness, and breakdown voltage range. The sole difference is the addition of the heterojunction. This first Ku-band HJ-IMPATT diode (and to some extent the first K-band HJ-IMPATT diode profile) proved vital in the development of the first-order theory of operation.
The K-band HJ-IMPATT diodes exhibited a 0.4 to 1.3 dB power improvement over standard devices. Although this was encouraging, it was not enough to justify the added difficulty that these profiles present in growth and fabrication. The first Ku-band profile produced disappointing results as the devices consistently ran at a lower frequency, but this result provided a valuable clue as to the effect of the heterojunction. As an understanding of these new devices was gained, the next attempt, the optimized drift Ku-band profile, produced diodes that were superior, with 0.5 to 2 dB higher power and substantially greater efficiency. Samples from this optimized drift profile have been compared with standard devices in production oscillator modules over the temperature range from -25 |degrees~ to 60 |degrees~ C, and show a 0.7 to 1 dB power improvement, with up to 2 dB improvement at room temperature and maximum current drive.|12~ The optimized drift HJ-IMPATT diode profile was unintentionally 5 to 7 V lower in breakdown voltage and still exhibited superior power and frequency performance. A properly designed device with a breakdown voltage range from 20 to 22 V is expected to achieve a power increase of over 2.5 dB.
Initially a number of HJ-IMPATT diode profiles were grown by MBE based on well-established standard CW IMPATT profiles in K- and Ku-bands, and the resulting 2" diameter wafers were fabricated into diodes with the standard process. The first surprising result came when testing these prototype HJ-IMPATT diodes on a curve tracer. The diodes had such low reverse leakage current that during the curve tracer's 60-Hz sweep there was not sufficient enough charge to cause the device to go into avalanche breakdown. The applied voltage was increased 1.2 to 1.5 times the nominal breakdown voltage before the avalanche effect actually began, as shown in Figure 6. This phenomenon is seen on every HJ-IMPATT device, and has never been noted in standard devices. In this regard, the HJ-IMPATT diodes resemble gas discharge rectifier diodes, requiring a very high applied voltage to initiate breakdown.
The HJ-IMPATT diodes indeed exhibit 2 to 3 orders of magnitude less leakage current prior to avalanche breakdown, as shown in Figure 7, which also shows the typical reverse I-V behavior for a standard device. The figure also shows another HJ-IMPATT diode anomaly, the exceptionally sharp onset of avalanche breakdown. Standard IMPATT diodes require 0.7 to 0.8 V to transition to full avalanche breakdown, with more voltage required in poor-quality devices. In the HJ-IMPATT diodes, the transition into breakdown requires less than 0.05 V. Figure 7b shows additional examples.
Several mechanisms produce the leakage current. Two of these mechanisms, depletion-zone hole-electron generation and tunneling, depend on a material's intrinsic doping |n.sub.I~, and hence the bandgap energy. Since the depletion zone extends through the larger bandgap |Al.sub.0.3~|Ga.sub.0.7~As material, these sources of leakage current are substantially reduced. A third mechanism, the thermal generation of minority carriers in the quasi-neutral zones, takes place at or near the edges of the space-charge depleted layer on both sides of the P/N junction. It can be shown that the generation rate will be much larger on the p-side of a p/N heterojunction because this is the small bandgap side.|8~ Thus, electrons generated just at the edge of the p-side depletion zone (which is only 0.2 to 0.4 |micrometer~ from the junction) are acted upon by the electric field. With a sufficient reverse bias, these electrons accumulate enough energy between collisions to achieve impact ionization. Since these thermally generated electrons cause the onset of avalanche breakdown, an explanation of the exceptionally sharp avalanche transition in HJ-IMPATT diodes is now possible.
Under reverse bias, electrons that have been thermally generated as described must surmount the conduction band discontinuity. In other words, the discontinuity acts as an energy filter, an effect previously described.|9~ Immediately prior to breakdown, only those electrons that have not lost kinetic energy due to elastic scattering can surmount the barrier. The carrier flux across the barrier is reduced according to the relation
|Mathematical Expression Omitted~
J = carrier flux
||psi~.sub.b~ = effective barrier height
The combination of reduced carrier flux over the discontinuity and the reduction in generation current from the N-side accounts for much lower reverse leakage current. This implies that the regions in the HJ-IMPATT diode that generate the precursor carriers (those that initiate the avalanche cascade) are confined to the immediate vicinity of the p/N heterojunction. In a standard device, by contrast, both the P- and N-side contribute leakage current. Therefore, the transition to avalanche breakdown is less distinct as carriers are generated from various locations in the avalanche layer and the p-side. The sharp transition always observed in the HJ-IMPATT diode results from a narrowly confined region of leakage current generation and reduced diffusion (thermally-generated) current. Study of the forward and reverse I-V characteristics showed good agreement between theory and experiment.|10~
Phase Noise and Large-Signal Impedance Measurements
Samples of standard and HJ-IMPATT diodes were subjected to double-sideband phase noise characterization, which showed that the HJ-IMPATT devices exhibited 3 to 6 dB lower phase noise content from 0 to 2 kHz frequency offset from the carrier frequency. Above 2 kHz, the HJ-IMPATT devices were comparable to standard devices. This experimental result, predicted in advance,|13~ is interpreted as evidence that the heterojunction is fundamentally altering the voltage and current waveforms as well as the energy distribution of the avalanching carriers.
A separate series of measurements were undertaken using a specially-designed low SWR impedance characterization fixture based on the APC-7 connector.|14~ The fixture uses a quarter-wave transformer adjacent to the device under test, thus minimizing impedance mismatch losses through the connector, since the fixture must be removed in order to measure the input |S.sub.11~ of a double-sleeve coaxial tuner that is used for final impedance matching. A number of standard and HJ-IMPATT devices were measured, which demonstrated that the optimized drift Ku-band HJ-IMPATT diode profile had a comparable conductance/area figure of merit, but at a significantly higher frequency.
A First-Order Theory of Operation
The standard Ku-band IMPATT epitaxial profile provided a baseline design for the first Ku-band HJ-IMPATT profile, with the doping levels and layer thicknesses used in the standard profile adopted for the heterojunction profile. In the first attempt at a Ku-band HJ-IMPATT profile, the MBE system produced a wafer that was within 3 percent in layer thicknesses, and within 5 percent of the target doping levels. However, these first Ku-band HJ-IMPATT diodes oscillated typically 7 percent lower in frequency when tuned for maximum RF power. Since the only significant difference in these devices was the heterojunction in the avalanche layer, it was concluded that the heterojunction in the avalanche layer, it was concluded that the heterojunction had altered the basic time constant of the avalanche layer. The next HJ-IMPATT profile featured a drift layer length that had been shortened by 10 percent compared to the first profile. This second attempt, the optimized drift profile, resulted in markedly superior devices despite the unintentionally low breakdown voltage. A similar pattern was observed retrospectively with the K-band HJ-IMPATT diode, which was actually the first heterojunction IMPATT diode growth.
The first-order theory now proposed requires that the heterojunction account for both the improved power results as well as the longer effective time constant in the avalanche layer that caused the slightly lower optimum frequency in the first Ku-band devices. Equation 3 and Figure 3 stipulated that for optimum efficiency, the avalanche cascade ideally must start exactly 90 |degrees~ after the AC voltage peak, or halfway into a cycle. This 90 |degrees~ phase delay is an approximation, as the avalanche process is not localized very well in time or position within the avalanche layer. The carriers that begin each cycle's avalanche cascade in a standard IMPATT diode are distributed in both energy and position, depending upon which generation mechanism creates them. Computer simulations of IMPATT diode voltage and current waveforms|15,16~ indeed show that the carrier pulse begins much earlier, at only 60 |degrees~ to 75 |degrees~ after the RF voltage peak instead of the ideal 90 |degrees~. Since the RF voltage is still positive, the early onset of the avalanche cascade (and hence the externally induced current) implies that the standard IMPATT device dissipates power for that portion of the AC cycle, resulting in lower average RF power. Using a simple computer model of the voltage and current waveforms, it is straightforward to determine that early onset of conduction current is a dominant factor in degrading IMPATT diode power and efficiency.|17~
Because of the heterojunction, the diode's I-V characteristics are clearly ideal, and since the region of the device that generates the precursor carriers (which initiate the avalanche cascade) is much narrower, the HJ-IMPATT diode's induced current waveform lag is closer to 90 |degrees~. Standard devices show in their I-V behavior significant pre-avalanche leakage current that initiates early conduction current in RF operation. Since the HJ-IMPATT diodes approach an ideal current-voltage waveform, their RF performance will be superior. This theory also accounts for the lower optimum frequency of the first Ku-band HJ-IMPATT diode profile, which was virtually copied from the standard profile. If one models the avalanche region as a resistor and inductor connected in series, the effect of incorporating the heterojunction is to increase the R-L time constant, since the avalanche cascade requires additional time to build to its peak (fewer precursor carriers), yielding a lower free-running frequency. This was compensated for in the optimized drift Ku-band profile where the drift layer was shortened by 10 percent compared to the standard length. Even with this, these devices should have oscillated in the 16 to 16.3 GHz range, if they had been standard devices.
Incorporating an abrupt GaAs/|Al.sub.0.3~|Ga.sub.0.7~As heterojunction at the P/N junction of the IMPATT diode produced unusual I-V characteristics and superior RF performance with up to 2 dB improvement in maximum RF power. A qualitative first-order theory of operation demonstrates how the physics of the heterojunction produces the observed DC and RF effects. Future work will focus on HJ-IMPATT profiles in X- through Ka-bands.
The author expresses his appreciation for the efforts of J. Franklin, who has performed all MBE growths, and to J. Harris of Stanford University, who was the author's principal thesis advisor. This article is dedicated as a memorial to F. Rosenbaum of Washington University, who provided insightful analysis of the HJ-IMPATT diode phenomenon, including the recommendation for optimizing the HJ-IMPATT diode's drift length. He predicted a number of the experimental effects and the author will always be grateful for his enthusiasm and support.
1. W.T. Read, "A Proposed High-Frequency, Negative Resistance Diode," Bell System Technology Journal, March 1958, pp. 401-447.
2. J. Pribitich, et al., "Design and Performances of Maximum-Efficiency Single- and Double-Drift-Region GaAs IMPATT Diodes in the 3 to 18 GHz Frequency Range," Journal of Appl. Phys., Vol. 49, No. 11, Nov. 1978, pp. 5584-5594.
3. R.L. Kuvas and W.E. Schroeder, "Premature Collection Mode in IMPATT Diodes," IEEE Transactions Electron Devices, Vol. ED-22, No. 8, August 1975, pp. 549-558.
4. M.J. Bailey, "Heterojunction IMPATT Diodes: Theory and Practice," Thesis for the Degree of Engineer, Stanford University, 1992, pp. 29-31.
5. S. Sze, Physics of Semiconductor Devices, Second Edition, Wiley, New York, 1981, pp. 566-612.
6. G. Haddad et al., "Basic Principles and Properties of Avalanche Transit-Time Devices," IEEE Trans. Microwave Theory Tech., Vol. MTT-18, No. 11, November 1970, pp. 752-771.
7. D.L. Scharfetter and H.K. Gummel, IEEE Trans. Electron Devices, Vol. ED-16, 1969, p. 64.
8. J. Harris, course notes (EE 428A, 428B) Department of Electrical Engineering, Stanford University, 1986-1989.
9. F. Rosenbaum, communication, 1990.
10. M.J. Bailey, "Heterojunction IMPATT Diodes," IEEE Tran., Elec. Devices, Vol. 39, No. 8, Aug. 1992, pp. 1829-1934.
11. K. Kurokawa, "The Single Cavity Multiple-Device Oscillator," IEEE Trans. Microwave Theory Tech., Vol. MTT-19, No. 10, October 1971, pp. 793-801.
12. S. Waltersdorff, general services engineering technical report, 1991.
13. F. Rosenbaum, op cit.
14. M.J. Bailey, thesis, pp. 68-72.
15. S. Sze, op cit., pp. 588.
16. R. Neece, "Material and Optical Effects on IMPATT Diode Operation," PhD thesis, NC State University, 1988.
17. M.J. Bailey, thesis, pp. 57-65.
Michael Jon Bailey received his BA degree in physics from the University of California, San Diego in 1979 and his MA degree in physics from the University of California, Santa Barbara, in 1981. In 1992, Bailey received the Degree of Engineer from Stanford University. From 1982 to 1987, he conducted applications engineering for a GaAs MESFET product line at Teledyne Microwave. Currently, Bailey is senior engineer at Litton Solid State in Santa Clara, CA. His present work is concentrated on extending the power and frequency performance of IMPATT diodes and components over the frequency range from 5 to 60 GHz.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Technical Feature; impact avalanche and transit time|
|Author:||Bailey, Michael Jon|
|Date:||Jun 1, 1993|
|Previous Article:||Progress in optical amplification.|
|Next Article:||Generalized microstrip-slotline transitions: theory and simulation vs. experiment.|