366 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 45, NO. 2, FEBRUARY 1998 Hot Electron Degradation of the DC and RF Characteristics of AlGaAs/InGaAs/GaAs PHEMT’s Mattia Borgarino, Student Member, IEEE, Roberto Menozzi, Yves Baeyens, Member, IEEE, Paolo Cova, Student Member, IEEE, and Fausto Fantini, Member, IEEE Abstract— This paper reports on hot electron (HE) degradation of 0.25-m Al0:25 Ga0:75 As/In0:2 Ga0:8 As/GaAs PHEMT’s by showing the effects of the hot electron stress on both the dc and rf characteristics. The changes of dc and rf behavior after stress turn out to be strongly correlated. Both can be attributed to a decrease of the threshold voltage yielding different effects on the device gain depending on the bias point chosen for device operation and on the bias circuit adopted: a fixed current bias scheme will minimize the changes induced by the stress. The work also presents a study of the dependence of device degradation on the stress bias condition. Index Terms— FET’s, high-speed circuits/devices, microwave devices, microwave FET’s, millimeter-wave devices, millimeterwave FET’s, MODFET’s, reliability. I. INTRODUCTION G ALLIUM ARSENIDE (GaAs)-based pseudomorphic HEMT’s (PHEMT’s) are the fastest commercially available 3-terminal devices to date, and a large number of interesting applications—both low-noise [1], [2] and power [3]–[6]—has recently been reported in the open literature. On the reliability front, for a few years, researchers concentrated their efforts on high-temperature storage or life test stressing (see, for instance [7]), and only recently did the attention shift toward room-temperature, high drain bias stress experiments designed to study possible hot electron (HE)related instabilities. The short gate necessary for microwave or millimeter-wave operation and the low bandgap energy of the InGaAs channel indeed make electron heating and impact ionization an issue the designer must take into consideration to achieve proper operation and long life of PHEMT devices and MMIC’s. In particular, it was reported that, as a consequence of hot electron stressing, the PHEMT dc characteristics may show transconductance compression [8], threshold voltage instability [9]–[11], breakdown walkout [11]–[13], and “power slump” [14]. These effects were generally attributed to trap creation and/or charge accumulation under the gate or at the interface between semiconductor and passivation in the gatedrain region. Off-state, high drain-gate bias stress experiments were carried out as well [15], [16] with similar results. Manuscript received May 22, 1997; revised August 6, 1997. The review of this paper was arranged by Editor J. Xu. This work was supported in part by the British Council and NATO. M. Borgarino, R. Menozzi, P. Cova, and F. Fantini are with the Dipartimento di Ingegneria dell’Informazione, Università di Parma, 43100 Parma, Italy. Y. Baeyens was with KULeuven/IMEC, Mercierlaan, B-3001 Heverlee, Belgium. He is now with Fraunhofer Institut IAF, Freiburg, Germany. Publisher Item Identifier S 0018-9383(98)00937-X. However, as far as the rf effects of such stress conditions are concerned, very limited data are available in the literature. The published works generally focused on high power devices, therefore no indication was given on the possible effects of the stress on the PHEMT small-signal parameters. For instance, in [8] 2 GHz contours in the – plane are compared before and after the stress, while other studies were concerned about the degradation of the device power output. We have recently reported on HE-induced instabilities of PHEMT -parameters measured up to 50 GHz and their correlation with the dc effects of the stress [17]. This point has a big practical relevance, since it gives indications on the possibility of limiting the study of device hot electron degradation to the relatively cheap and quick dc characterization. In addition to this, this paper provides a study of long-term recovery of stressed devices that leads us to a new interpretation of the experimental results and completes the preliminary picture drawn in [17] with a detailed discussion of possible degradation mechanisms. Moreover, we investigate here for the first time the quantitative dependence of the HE-induced degradation on the bias point where the PHEMT’s are stressed, another topic which was never studied systematically so far. In the next section, we give a description of the PHEMT’s under test and illustrate the stress conditions and the characterization techniques adopted. Section III presents the experimental results, that are discussed in Section IV, and finally we draw a few conclusions in Section V. II. EXPERIMENTS The devices under test are Al Ga As/In Ga As/GaAs PHEMT’s designed and fabricated at IMEC for millimeter-wave low-noise applications. The layer structure features, from bottom to top, a 1- m thick undoped GaAs buffer, 13-nm undoped In Ga As channel, 5-nm undoped Al Ga As spacer, Si -doping cm lightly doped cm 30-nm Al Ga As Schottky layer, and 40-nm n cm GaAs cap. The PHEMT’s have 0.25- m long, 100- m wide, T-shaped Ti/Pt/Au gates patterned by -beam lithography. The gate is slightly offset toward the source so that the gate-source and gate-drain spacings are about 0.5 and 1 m, respectively. The gate recess is formed using a nonselective wet etch; consequently, the gate-source and gate-drain access regions are almost entirely covered by the cap layer. The ohmic contact metallization is Au/Ge/Ni/Au. A schematic cross section is shown in Fig. 1. Passivation is provided by a 200-nm thick 0018–9383/98$10.00 1998 IEEE BORGARINO et al.: HOT ELECTRON DEGRADATION OF DC AND RF CHARACTERISTICS Fig. 1. 367 Schematic cross section of the PHEMT’s under test. layer of SiN deposited at 250 C in 25 min by PECVD. The layer deposition conditions were optimized for minimal stress; on 2-in Si wafers the nitride induced stress was measured to be slightly compressive dyn/cm The index of refraction and dielectric constant of SiN were determined to be 1.86 and 7, respectively. The devices feature of about 600 mS/mm and unity a maximum extrinsic current gain cutoff frequency around 70 GHz. The gate-drain and gate-source breakdown voltages, measured and at 1 mA/mm leakage current, are V, respectively. The PHEMT’s were repeatedly characterized on-wafer using a Cascade coplanar probe station both for dc characteristics and -parameters. During the characterization the drain voltage never exceeded 3 V; the samples showed no degradation following the characterization steps and excellent long-term stability. We used an HP-4142 for the dc measurements, while the rf characterization up to 50 GHz was carried out using an HP-8510B network analyzer. In order to investigate possible HE-related instabilities, we performed room-temperature, dc, high drain bias, on-state accelerated stress experiments. Since high temperatures are known to inhibit electron heating due to enhanced phonon scattering, low temperatures represent a worst case condition for HE reliability; moreover, by limiting the device temperature at the value dictated by self-heating, it is possible to isolate hot electron effects from other degradation mechanisms that are accelerated by high temperatures (e.g., contact metal migration). Different bias conditions have been used for the stress, with ranging from 3.2–6 V, as will be illustrated in the following section. The presence of electron heating and impact ionization at a drain voltage just exceeding 3 V is testified by the typical bell shape of the – curve (see, for instance, [12]): when the dominant contribution to the reverse gate current is the flow of holes generated by impact ionization in the channel, the magnitude of increases with at low gate bias, due to the increase of channel electron concentration, but debecause of the reduction of the peak creases at high longitudinal electric field (roughly proportional to Since we found out that most of the change of the device characteristics takes place in the first seconds or minutes of stress, we generally set the stress duration at 5 min. Fig. 2. Output characteristics of one of the devices under test before (solid 5:5 V, VGS = 0 V. lines) and after (dashed lines) a 5-min stress at VDS VGS ranges from 0.4 V to 0.3 V with 0.1 V increments. = 0 Fig. 3. Transconductance measured at VDS = 2 V and VDS = 0:05 V, before (solid lines) and after (dashed lines) the stress of Fig. 2. The threshold voltage values reported in the following sections are calculated by a linear extrapolation of the versus curve measured at V. III. RESULTS A. Effects of the Stress on the dc Characteristics Fig. 2 shows PHEMT output characteristics before and after a 5-min stress at V, V. The stress produces a relevant drain current increase in the whole bias plane. This is due to a reduction of the device threshold voltage , as illustrated by Fig. 3, where we plotted, as a function of , the transconductance measured both in the linear and saturation region, before and after the stress. The curves are not distorted at all by the stress, and simply shift by about direction. 65 mV in the negative Even though the stress is performed at room-temperature, self-heating may result in much higher channel temperatures, due to the high thermal resistivity of GaAs. It is thus important to ascertain whether or not self-heating plays a role in the observed device instability. With this aim in mind, we have 368 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 45, NO. 2, FEBRUARY 1998 Fig. 4. Threshold voltage values measured on ten devices before the stress, after a 5-min stress at various VDG values, and after a few weeks of room-temperature storage without bias. stressed two devices placed close to each other on the same wafer under different bias conditions corresponding to the same value of the dissipated power. The low drain bias stress condition A ( V, V) yielded no shift, while a stress at high drain bias point B ( V, V) produced a mV, even though both conditions correspond to the same dissipated power of 80.3 mW. It was thus demonstrated that the change of must not be ascribed to device self-heating. Moreover, temperature- and bias-accelerated interdiffusion processes such as “gate sinking” [18], would be very hard to imagine under the present short-time, room-temperature, low-gate current density stress conditions. A partial recovery generally takes place after the stress is over, as shown by Fig. 4. Here we have reported the values of ten devices before the stress, after a 5-min stress under various bias conditions, and after a few weeks of device storage (with no bias applied) at room temperature. There is a tendency of to increase slightly during this storage; however, the initial value of the threshold voltage is never recovered. B. Effects of the Stress on the rf Characteristics At microwave and millimeter-wave frequencies, the shift curves shown in Fig. 3 reflects into a variation of of the , as shown in Fig. 5, where we plotted the device gain the dependence of the magnitude of measured at 20 GHz, both for V and V. Since at V the hot electron treatment results in an increase of increases after the stress the transconductance (Fig. 3), (Fig. 5). On the contrary, at V is lower after the stress (see Fig. 3), and decreases accordingly (Fig. 5). The reduction of obviously implies a decrease of the device current and power gains; for instance, at V, V the unity current gain cutoff frequency decreases from 71 GHz to 66 GHz, the maximum frequency of oscillation from 105 GHz to 90 GHz. Fig. 6 illustrates the , measured at frequency dependence of the magnitude of V, V, both before (solid line) and after (dashed line) the stress of Fig. 2. Fig. 5. Drain bias dependence of the magnitude of S21 measured at 20 GHz for VGS 0 V (open circles) and VGS = 0:3 V (full circles), before (solid line) and after (dashed line) the stress of Fig. 2. = Fig. 6. Frequency dependence of the magnitude of S21 measured at VGS = 0:3 V, VDS = 2 V before (solid line) and after (dashed line) the stress of Fig. 2. After the stress, jS21 j has been measured also at VGS = 0:065 V to compensate for the 65 mV VT decrease and get the same IDS as before the stress (crosses). 0 Also at rf, as was the case for the dc results, the device decrease. In degradation can be simply described as a order to demonstrate this statement, we measured the poststress -parameters adjusting in such a way as to get the same drain current we had before the stress at V shift); under these (thus compensating the stress-induced conditions (Fig. 6, crosses) practically does not change after the stress. This means that from a practical standpoint the device behavior will or will not change after the stress depending on the kind of circuitry that sets the PHEMT bias point: in particular, a biasing arrangement whereby the device drain current is kept constant (which has also the advantage of overcoming the issue of the dispersion of among nominally and identical devices) will automatically compensate the therefore minimize the stress effect, both at dc and rf. C. Stress Bias Dependence of the Degradation We performed 5-min stress experiments on a set of devices using different values of drain and gate bias so as to cover BORGARINO et al.: HOT ELECTRON DEGRADATION OF DC AND RF CHARACTERISTICS Fig. 7. Threshold voltage shift as a function of the drain-gate bias used during the stress. Each point corresponds to a different device, and the duration of the stress was 5 min in all cases. The VGS stress values are 0.3 V (full circles), 0 V (open squares), and 0.3 V (full triangles). 0 a range of spanning from 3 V to 6 V. One may expect to be a good accelerating factor for the hot electron stress because, as we pointed out above, a simplified device model predicts a linear relationship between the drain-gate bias and the peak longitudinal electric field in the channel, which is directly related to electron heating and impact ionization. The results summarized by Fig. 7 support this statement. Although a little noisy (but one should consider that each point in Fig. 7 corresponds to a different device, hence some scatter must be expected simply as a consequence of the process tolerances) the data show a roughly linear dependence between and the drain-gate stress voltage (i.e., the channel longitudinal electric field). The relationship with the gate leakage current (that has a strongly nonlinear dependence on instead has more of an exponential nature (i.e., is approximately proportional to the logarithm of IV. DISCUSSION A fully recoverable threshold voltage decrease following hot electron stress tests performed on commercial devices was was attributed to a build-up of reported in [9], [10]. positive charge in the semiconductor region underneath the gate, namely a compensation of trapped electrons by some of the holes generated by impact ionization in the channel and flowing toward the gate. This interpretation was consistent with the results of high temperature storage tests, that yielded an activation energy for which supported the hypothesis of charge exchange by DX-centers in the AlGaAs [9]. Although the results presented in this work share some features with the findings of [9], they require a more articulate interpretation. In particular, positive charge trapping in the AlGaAs layer cannot be the only degradation mechanism taking place here, since it could produce only a temporary change of once the stress is over, the charge distribution in the deep levels must get back to the equilibrium conditions in a matter of hours or days, as found in [9]. Nevertheless, this kind of phenomenon can account for the (small) temporary portion of that ranges between 5 369 mV and 10 mV (see Fig. 4). We observed a slight 10 mV) decrease of also after a 36 h high temperature 180 C storage performed on some of the devices. This finding supports the hypothesis that deep levels under the gate may contribute to the observed device degradation; in this case thermally activated electron detrapping is responsible for the threshold voltage change. It is worth pointing out that the modest amount of shift that can be attributed to deep levels is consistent with the -doped nature of the AlGaAs layer, which reduces the number of DX defects [19], [20]: higher concentrations of deep levels, hence larger recoverable shifts, should be expected in the case of uniform, high doping donor layer. As far as the permanent part of is concerned, we believe that hole trapping at the interface between the gate metal and the AlGaAs or in the very thin oxide interface layer (always present in Schottky contacts unless they are deposited in situ on cleaved semiconductor surfaces) is the degradation mechanism to be blamed, in a way similar to that of Silicon MOSFET’s undergoing hot electron stressing. As briefly anticipated above, the positive charge storage may (whether temporary or permanent) that gives rise to originate either from capture of some of the holes that are generated by impact ionization in the channel and flow toward the gate, or from field-aided tunneling of electrons out of the traps, or from a combination of the two. The study of the on the stress bias described in Section II, dependence of where we saw that the shift scales linearly with the draingate bias used during the stress, suggests that the degradation mechanism is connected with the electric field more than with the gate current: if the dominant mechanism leading to positive charge storage were the capture of some of the holes flowing to the gate, we would expect a linear dependence of on , which is not the case. However, one should keep in mind that the hole flow from the channel to the gate due to impact ionization is not the only component of the gate current (the other being the ordinary reverse bias gate-drain leakage); it is therefore possible that both the field and the gate current contribute to the change. In order to get a better understanding of the physics of device degradation we measured the device threshold voltage and gate leakage current after a series of subsequent stress V, V (resulting in a total stress steps at time of 1 h), and kept monitoring these quantities from time to time after the stress was over. The time dependence of is shown in Fig. 8, where we notice that: 1) shifts almost entirely in the first seconds of the stress; 2) as the stress saturates and eventually starts being recovered goes on, s); and 3) the recovery proceeds after the end of the s) and appears to be over for times in the stress range of 10 –10 s. The second point deserves a few words of explanation, since the degradation mechanism proposed above cannot account for a recovery for the temporary part of that takes place during the stress itself. The answer lies in the presence of some “breakdown walkout” during our stress experiment, as demonstrated by the decrease (not shown here), with increasing stress time, of both the drain and gate current measured at the stress bias point V, V). 370 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 45, NO. 2, FEBRUARY 1998 Fig. 8. Time dependence of the threshold voltage shift of a device stressed 5:5 V. The stress ends at 3600 s, followed by room-temperature at VDG storage. = This phenomenon was observed on different PHEMT samples [12], [15], and is thought to be due to electron trapping at the surface of the gate-drain access region giving rise to a decrease of the peak electric field in the channel; this reduces the and decrease. So, due impact ionization rate, therefore to “breakdown walkout” both the mechanisms that we have indicated above as possible reasons for the storage of positive charge in deep traps under the gate, namely the electric field and the gate current, decrease with time during our stress, until a condition is reached where hole release becomes dominant, and recovery begins. The behavior of the gate leakage current is pretty much the as illustrated by Fig. 9, except that its same as that of recovery is practically complete. Thus, it is reasonable to link with the increase of gate leakage, the temporary part of whereas the permanent portion of the shift is not accompanied change relative to the pre-stress conditions. by any relevant This observation supports the interpretation given above of the being due to positive charge storage in deep recoverable centers of the AlGaAs layer, since an increase of the positive charge concentration in the AlGaAs results in an enhanced conduction band bending underneath the gate, which in turn produces an increase of the tunneling component of the gatedrain leakage. The data of Fig. 10 provide this theory with further support: after 5 min of the stress of Figs. 8 and 9 not only the gate leakage has increased (dashed line) with respect to the pre-stress situation (solid line), but its dependence has turned into a nice exponential, which indicates that the tunneling component is at this stage dominant. Once the trapped positive charge has been released and the recovery points sit again on top of the pre-stress is over (crosses), the curve. Finally, we draw further indications that the observed degradation mode is due to charge accumulation effects below the gate, and not to surface-related phenomena, from similar HE experiments performed on InP HEMT’s. We have observed shifts in nonpassivated devices [21], whereas a change of the device surface condition in SiN-passivated HEMT’s produces a totally different degradation mode, namely, a Fig. 9. Time dependence of the gate leakage current shift of a device stressed at VDG = 5:5 V. The stress ends at 3600 s, followed by room-temperature storage. Fig. 10. VGS dependence of the gate leakage measured before the stress (solid line), after 5 min of the stress of Figs. 8 and 9 (dashed line) and after seven days of the following room-temperature storage (crosses). transconductance and rf gain compression at high [22], [21]. We may thus conclude that the SiN passivation of the PHEMT’s studied here does not play a role in the HE degradation process. V. CONCLUSIONS In this paper, we have shown results of hot electron stress experiments performed on GaAs-based pseudomorphic HEMT’s designed for low noise applications at millimeterwave frequencies. The dc effect of the stress is a reduction of the device threshold voltage; a small fraction of the shift turned out to be temporary, most of it was permanent. No additional degradation mode was observed at frequencies up to 50 GHz. The degradation of device performance, therefore, will largely depend on the operating bias point, and on the bias circuitry adopted; in particular, it was demonstrated that a bias circuit that provides a constant current drive will variation. The devices were minimize the impact of the stressed under a wide range of bias conditions, and we found shift) a roughly linear dependence of the degradation on the gate-drain voltage used during the stress, i.e., on the BORGARINO et al.: HOT ELECTRON DEGRADATION OF DC AND RF CHARACTERISTICS channel peak electric field. The experimental results can be explained assuming that the hot electron stress leads to hole accumulation in deep traps of the AlGaAs underlying the gate (which accounts for the temporary part of the shift), and in the interface/oxide layer between the gate metal and the semiconductor (permanent portion of the shift). The positive charge storage may come either from field-enhanced electron detrapping, or from capture of holes generated by impact ionization in the channel. ACKNOWLEDGMENT The authors wish to thank M. van Rossum of IMEC, Belgium, for his encouragement and support, G. Vannini and A. Santarelli of the University of Bologna, Italy, for the use of the RF equipment, and W. De Raedt and P. Richardson of IMEC for device fabrication. REFERENCES [1] Y. Itoh, Y. Horiie, K. Nakahara, N. Yoshida, T. Katoh, and T. Takagi, “A V -band, high gain, low noise, monolithic PHEMT amplifier mounted on a small hermetically sealed metal package,” IEEE Microwave Guided Wave Lett., vol. 5, pp. 48–49, Feb. 1995. [2] Y. Itoh, K. Nakahara, T. Sakura, N. Yoshida, T. Katoh, T. Takagi, and Y. Ito, “W -band monolithic low noise amplifiers for advanced microwave scanning radiometer,” IEEE Microwave Guided Wave Lett., vol. 5, pp. 59–61, Feb. 1995. [3] M. Aust, H. Wang, M. Biedenbender, R. Lai, D. C. Streit, P. H. Liu, G. S. Dow, and B. R. Allen, “A 94-GHz monolithic balanced power amplifier using 0.1-m gate GaAs-based HEMT MMIC production process technology,” IEEE Microwave Guided Wave Lett., vol. 5, pp. 12–14, Jan. 1995. [4] J. C. L. Chi, J. A. Lester, Y. Hwang, P. D. Chow, and M. Y. Huang, “A 1-W high-efficiency Q-band MMIC power amplifier,” IEEE Microwave Guided Wave Lett., vol. 5, no. 1, pp. 21–23, Jan. 1995. [5] S.-W. Chen, P. M. Smith, S.-M. J. Liu, W. F. Kopp, and T. J. Rogers, “A 60-GHz high efficiency monolithic power amplifier using 0.1-m PHEMT’s,” IEEE Microwave Guided Wave Lett., vol. 5, no. 6, pp. 201–203, June 1995. [6] Y.-L. Lai, E. Y. Chang, C.-Y. Chang, T. K. Chen, T. H. Liu, S. P. Wang, T. H. Chen, and C. T. Lee, “5 mm high-power-density dual-delta-doped power HEMT’s for 3 V L-band applications,” IEEE Electron Device Lett., vol. 17, pp. 229–231, May 1996. [7] A. Christou, J. M. Hu, and W. T. Anderson, “Reliability of InGaAs HEMT’s on GaAs substrates,” in Proc. Int. Reliab. Phys. Symp., 1991, pp. 200–205. [8] Y. A. Tkachenko, C. J. Wei, J. C. M. Hwang, T. D. Harris, R. D. Grober, D. M. Hwang, L. Aucoin, and S. Shanfield, “Hot-electron induced degradation of pseudomorphic high-electron mobility transistors,” in Proc. Microwave and Millimeter-Wave Monolithic Circuits Symp., 1995, pp. 115–118. [9] C. Canali, P. Cova, E. De Bortoli, F. Fantini, G. Meneghesso, R. Menozzi, and E. Zanoni, “Enhancement and degradation of drain current in pseudomorphic AlGaAs/InGaAs HEMT’s induced by hot-electrons,” in Proc. Int. Reliab. Phys. Symp., 1995, pp. 205–211. [10] G. Meneghesso, C. Canali, P. Cova, E. De Bortoli, and E. Zanoni, “Trapped charge modulation: a new cause of instability in AlGaAs/InGaAs pseudomorphic HEMT’s,” IEEE Electron Device Lett., vol. 17, pp. 232–234, May 1996. [11] Y. C. Chou, G. P. Li, K. K. Yu, P. Chu, L. D. Hou, C. S. Wu, and T. A. Midford, “Hot carrier reliability in high-power PHEMT’s,” in Proc. GaAs IC Symp., 1996, pp. 46–49. [12] R. Menozzi, P. Cova, C. Canali, and F. Fantini, “Breakdown walkout in pseudomorphic HEMT’s,” IEEE Trans. Electron Devices, vol. 43, pp. 543–546, Apr. 1996. [13] Y. C. Chou, G. P. Li, Y. C. Chen, R. Lai, and D. C. Streit, “Reliability and alleviation of premature on-state avalanche breakdown in deep submicron power PHEMT’s,” in Proc. Int. Reliab. Phys. Symp., 1997, pp. 261–267. [14] J. L. Muraro, F. Coppel, and G. Gregoris, “Hot electron effect on HFET devices: How to assess reliability of high power amplifiers?,” in Proc. GaAs Reliab. Workshop, 1996, pp. 16–17. 371 [15] Y. C. Chou, G. P. Li, K. K. Yu, C. S. Wu, P. Chu, L. D. Hou, and T. A. Midford, “Off-state breakdown walkout in high-power PHEMT’s,” in Proc. GaAs IC Symp., 1996, pp. 42–45. [16] R. E. Leoni and J. C. M. Hwang, “Effects of reverse gate-drain breakdown on gradual degradation of power PHEMT’s,” in Proc. GaAs IC Symp., 1996, pp. 31–33. [17] R. Menozzi, M. Borgarino, P. Cova, Y. Baeyens, and F. Fantini, “The effect of hot electron stress on the dc and microwave characteristics of AlGaAs/InGaAs/GaAs PHEMT’s,” Microelectron. Reliab., vol. 36, no. 11/12, pp. 1899–1902, 1996. [18] C. Canali, F. Castaldo, F. Fantini, D. Ogliari, L. Umena, and E. Zanoni, “Gate metallization ‘sinking’ into the active channel in Ti/W/Au metallized power MESFET’s,” IEEE Electron Device Lett., vol. EDL-7, pp. 185–187, Mar. 1986. [19] M. Mizuta, M. Tachikawa, H. Kukimoto, and S. Minomura, “Direct evidence for the DX center being a substitutional donor in AlGaAs alloy systems,” Jpn. J. Appl. Phys., vol. 24, no. 2, pp. L143–L146, Feb. 1985. [20] P. M. Mooney, “Deep donor levels (DX centers) in III–V semiconductors,” J. Appl. Phys., vol. 67, no. 3, pp. R1–R24, Feb. 1, 1990. [21] R. Menozzi, M. Borgarino, Y. Baeyens, K. van der Zanden, M. Van Hove, and F. Fantini, “The effect of passivation on the hot electron degradation of lattice-matched InAlAs/InGaAs/InP HEMT’s,” in Proc. 9th Int. Conf. Indium Phosphide and Related Mater., 1997, pp. 153–156. [22] R. Menozzi, M. Borgarino, Y. Baeyens, M. Van Hove, and F. Fantini, “On the effects of hot electrons on the DC and RF characteristics of lattice-matched InAlAs/InGaAs/InP HEMT’s,” IEEE Microwave Guided Wave Lett., vol. 7, no. 1, pp. 3–5, Jan. 1997. Mattia Borgarino (S’97) was born in Parma, Italy, in 1968. He received the Laurea degree in electronic engineering (cum laude) from the University of Parma, Italy, in 1993. After serving in the army, he started pursuing the Ph.D. degree in the Department of Information Technology, University of Parma. His current research interests include the reliability of AlGaAs/GaAs and InGaP/GaAs HBT’s and of GaAs and InP HEMT’s, and the numerical simulation of electromigration. Roberto Menozzi was born in Genova, Italy, in 1963. He received the Laurea degree (cum laude) in electronic engineering from the University of Bologna, Bologna, Italy, in 1987, and the Ph.D. degree in information technology from the University of Parma, Parma, Italy, in 1994. After serving in the army, he joined a research group at the Department of Electronics, University of Bologna. Since 1990, he has been with the Department of Information Technology, University of Parma, where he became a Research Associate in 1993. His research activities have covered the study of latch-up in CMOS circuits, IC testing, and the dc, rf, and noise characterization, modeling, and reliability evaluation of compound semiconductor and heterostructure electron devices such as GaAs MESFET’s and GaAs and InP HEMT’s. Yves Baeyens (S’89–M’96) was born in Asse, Belgium, on May 30, 1969. He received the M.S. and Ph.D. degrees in electrical engineering from the Catholic University of Leuven, Belgium, in 1991 and 1997, respectively. His Ph.D. research was performed in cooperation with IMEC, Leuven, Belgium, and treated the design and optimization of coplanar InP-based dual-gate HEMT amplifiers, operating up to W-band. Since December 1996, he has been a Visiting Scientist at the Fraunhofer Institute for Applied Physics, Freiburg, Germany. His research interests include the technology, design, and reliability of HEMT MMIC’s for microwave and opto-electronic applications. 372 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 45, NO. 2, FEBRUARY 1998 Paolo Cova (S’97) was born in Milano, Italy, in 1966. He received the Laurea degree in electrical engineering in 1992, and the Ph.D. degree in information technology in 1996, both from the University of Parma, Parma, Italy. His research interests are in the field of the reliability evaluation of III–V compound semiconductor devices, with particular reference to pseudomorphic HEMT’s and 980-nm pump lasers. He is currently working on the study of hot electron stress phenomena in GaAs-based FET’s and on the reliability issues of power transistors for railway traction. Fausto Fantini (S’71–M’74) was born in Bologna, Italy, in 1946. He graduated in electronic engineering in 1971 from the University of Bologna. In 1973, he joined the Quality and Reliability Department, Telettra S.p.A., Vimercate, Milano, Italy, where he worked on the reliability of semiconductor devices and established the laboratory of failure analysis. From 1987 until 1990, he was Associate Professor of Electronics, Scuola Superiore di Studi Universitari e di Perfezionamento S. Anna, Pisa, Italy. In 1990, he joined the University of Parma, Parma, Italy, as Full Professor of Microelectronics. Since 1992, he has been Director of the Research Center on Materials and Information Technologies (MTI) at the same university. His research interests cover various aspects of semiconductor device physics and reliability, including corrosion, electromigration, and metal/semiconductor interaction. He has authored or coauthored three books and over 100 research articles and international conference papers; he organized three summer schools on failure physics (1980, 1987, and 1991), and the first ESREF in Bari (1990). Mr. Fantini is a member of AEI and AICQ (Italian Association for Quality).