3x13 Yama no Susume | 1x836 Acacias 38 | El Puente HDTV 720p AC3 5.1

Reliability of compound semiconductor devices

Reliability of compound semiconductor devices

Microelectron. Reliab.,Vol. 32, No. 11, pp. 1559-1569, 1992. Printed in Great Britain. 0026--2714/9255.00+ .00 Pergamon Press Ltd RELIABILITY OF C O...

1MB Sizes 4 Downloads 159 Views

Microelectron. Reliab.,Vol. 32, No. 11, pp. 1559-1569, 1992. Printed in Great Britain.

0026--2714/9255.00+ .00 Pergamon Press Ltd

RELIABILITY OF C O M P O U N D SEMICONDUCTOR DEVICES FAUSTO FANTINI Universit~ di Parma, Dipartimento di Ingegneria dell'Informazione, Parma 43100, Italy and FABRIZIO MAGISTRALI ALCATEL Telettra, Quality and Reliability, Vimercate 20059, Italy Abstract--This paper reviews the reliability of III-V semiconductor devices with particular attention to the failure mechanisms typical of these structures. Instability effects at the surface of various FETs have been examined and the problems related to the metallurgies employed. For optical devices, the degradations caused by the growth of defects in the substrate and at the mirrors have been considered. The difficulties in the accelerated reliability evaluation have also been examined.


Compound semiconductor devices have had important applications during the last 20 years, due to their superior performance and peculiarities with respect to silicon; however, their economic importance is still much lower than was frequently forecast. The reasons for this situation are complex, but they probably stem mainly from the problems that are encountered in compound semiconductor technology and in the related high costs and low yields, in addition to the higher cost of the material. The difficulties of the technologies have cast doubts on their long term reliability too, so that several studies [1-13] have been performed, beginning in the mid-1970s, on the reliability and failure mechanisms of compound semiconductor devices. The large differences in the devices that can be manufactured using III-V semiconductor materials, including GaAs MESFETs, heterojunction transistors and all types of optoelectronic devices, makes a review of their reliability particularly intriguing and the number of differences is probably higher than the number of points found in common. However, some comparable features can be found in the importance of crystal defects, in the lack of a natural and stable surface passivation and finally in the impossibility of building a mono-metal system. It is worth mentioning that serious studies were performed during the 1970s on MESFETs [14-19] and in the field of optoelectronics devices, especially for GaAs-based laser diodes [20-33]; more recently, attention was focused on the InP-based LEDs and lasers [34--44] and some attention has been paid to the problems of heterojunction transistors [45-50]. In this review, we will consider the problems of the different classes of devices separately, and we will focus on the more common problems. First, the reliability of standard GaAs MESFETs will be reviewed, then we will consider the recent problems of

the heterojunction transistors (mainly HEMTs). Finally, we will examine the reliability of optoelectronic devices. GaAs MESFETs

The MESFETs are the most widespread of the III-V components, being the most mature of these products. Their reliability has been under study since the 1970s and their failure mechanisms are now quite well understood [1,3-7,9, 13-19]. Therefore, reliability figures that can be obtained from field operations are quite satisfactory and comparable to those of silicon devices of similar complexity [13]. Let us review the failure mechanisms that are limiting the MESFET reliability. ESD, burn out and surface effects The MESFET, a sketch of which is shown in Fig. 1, that fails in the field is very difficult to analyse because these are usually completely destroyed [19, 51]. This effect is due to the very low robustness intrinsic to the structure of the devices, which feature very short ( < 1/~m) and large channels (frequently > 100/~m, depending on the power) and very short distances between the gate and source or drain (about 1/~m). The first time, this destruction of the device was considered as a specific failure mechanism and was termed burn out. Later on, it was recognized that burn out is only a failure mode and must have a particular cause. One of the major causes of destruction of MESFETs is the application of an electrostatic discharge (ESD). In effect, if you stress the MESFETs according to the human body model (HBM), as is required in many standards, the threshold values for the failures are very low, in the order of a few hundred volts, putting these devices into a higher level of sensitivity to ESD and rendering the handling



1560 Source


S.L Substrate

Fig. 1. Sketch of a GaAs MESFET.

precautions mandatory. The high energy involved in the ESD phenomenon frequently induces major damage in the devices, similar to that caused by surcharge effects, due to pulses that are quite common in the RF applications [51-54]. However, burn out can also occur in the absence of any overstress effect. The cause of the failure is usually a short-circuit between the electrodes under bias. The origin of the short is located either on the surface of the channel between the gate and drain, or at the interface between the channel and the buffer layer [54].

Fig. 2. SEM picture of the interface between the gate and GaAs for a virgin (a) and failed (b) device, obtained by back-etch.

Reliability of compound semiconductor devices In the first instance, the surface short-circuit occurs more easily in the absence of surface passivation, and when the passivation is carried out by SiO2 rather than by SiN. In effect, the failure mechanism is caused by the surface oxidation of the semiconductor, that preferentially forms GaO: the excess As is metallic and creates low resistance paths between the metals [55, 56]. An alternative explanation of the short circuit is the electromigration of the Au from the drain along the GaAs surface to the gate [57]. This is a diffusion mechanism and is sustained by the temperature and the electric field. In both cases a stable passivation system greatly increases the reliability of the devices [56]. The subsurface location of the short circuit has been shown by the presence of a deep-level trap enriched region at the channel/buffer interface, where a positive temperature coefficient exists, bridging the gate and drain regions. Hot electrons generated in the vicinity of the drain area can be pushed to flow through this region causing a positive feedback effect to begin [54]. Besides giving rise to these catastrophic effects, the GaAs surface is also responsible for other degradation phenomena. Firstly, the oxidation of the surface and the presence of surface traps can cause a reduction of the region between the gate and source/drain and a degradation of gm and /ass. Moreover, it also affects the reliability of the devices. During accelerated stress tests, an increase of transconductance was observed, together with a reduction of the gate junction breakdown voltage (or the increase of the gate leakage current). A tentative activation energy of 1 eV was calculated, and this value appears to be reasonable for reactions involving modifications of the GaAs surface [58]. Another effect of the surface states is the degradation of the dynamic characteristics of the devices, in particular for low VDs values, due to the slowness of the state occupancy to follow the gate voltage variations [59]: this effect has been called gate-lag.

Degradation of the Schottky contacts The degradation of the Schottky contacts has been considered the most dangerous failure mode of MESFETs, because it directly affects the active part of the device [60]. The stability of the interface is considered to be guaranteed for those materials that are employed in self-aligned technologies, as WSi and WN are, for example. However, the more conventional metal systems, such as Al- and Au-based multilayers, have demonstrated different degrees of sensitivity to the thermal stresses. In the case of the ?d-based multilayer, the interaction with the GaAs caused an increase of the barrier height during the first hours of life-testing, but this variation was saturated quite rapidly and did not induce any further degradation [61].


On the contrary, when Au-based gate-systems were employed, large degradations of the device characteristics were observed: decreases in Iass, increases of the Vp and channel resistance. This phenomenon has been justified by the reduction of the active channel (gate sinking), due to the penetration of the Au into the GaAs, through the barrier layer (Ti, TiPt, T i W . . . ) [62], as shown in Fig. 2, where a back-etched [63] picture of the gate is shown. No activation energy was calculated for the gate-sinking effect, due to the very highly localized and non-homogeneous phenomenon. A different explanation uses the compensation effects of Au or H, that reduce the channel doping [64]. Besides the metal-semiconductor interdiffusion, the ?d-gate devices suffer from the electromigration of the gate fingers that is induced by the high current density [61]. Obviously the risk in the real application is reduced by the fact that a high gate current is reached only in pulsed conditions.

Ohmic contact degradation The stability of the ohmic contacts obtained by alloying Au-Ge-Ni layers with GaAs has always been a concern: the first studies in the mid-1970s indicated a resistance increase as the main cause of failure [14-15]. In reality it does not seem to be true. Although the exact knowledge of the contact formation and operation still has some dark spots, their stability at room temperature has not been questioned recently [9, 60]. On the contrary, during the stress tests performed at very high temperatures the increase of the ohmic contacts resistance still is the predominant failure mechanism [65]. This can be easily explained by the high values of the activation energies ( > 1.SeV) that were found for this failure mechanism: it is well known that these mechanisms are accelerated by increasing temperatures and therefore they are dominant at the temperatures higher than 200°C employed in the accelerated tests. However, it is very dangerous to extrapolate from



[1000/K] 2.6 -




_ 125 I05 90




3.2 i 3.4

E. = I


Ea= 1.6 oV


I iK


0.5 eV%%%%%%~









"[' ~0


Time [hi

Fig. 3. Arrhenius plot from accelerated tests on power MESFETs, displaying two activation energies.



98-~Fallures go l 70 _

. 20


Corrosion and purple plague HAST ill

.//X/ HAST








I I0





Time [hi

Fig. 4. Log-normal plot of cumulative failure distributions obtained on the Al-gate and Au-gate power MESFETs submitted to temperature humidity bias tests (85% R.H.; 85°C) and highly accelerated stress tests (85% R.H.; 125°C). these high temperature data to the operating conditions: as shown in Fig. 3, a failure mechanism with a low activation energy can be masked at high temperatures by the ohmic contacts degradation, but can be decisive at lower temperatures. The exact physical mechanism responsible for the resistance increase is not yet clear, but can be attributed to the movement of the different species involved and in particular to the Ga out-diffusion [60]. Also the ohmic metals are subjected to electromigration, in particular in power devices where high drain currents are employed [17, 66]; however, the mechanism is limited by the use of large Au stripes that are not very sensitive to this phenomenon.

The MESFET devices are not particularly sensitive to the corrosion effects, mainly because the power devices are protected by their dissipation. However, the very high electric fields present between the electrodes can accelerate the corrosion phenomena in the presence of a high humidity; in this case the A1 gate devices are more sensitive to corrosion due to possible galvanic effects [3] (see the results of accelerated tests in Fig. 4). A failure mechanism that seems to be related to the particular ohmic metal system is the growth of Ni needles that can short circuit the metals [67], as shown in Fig. 5. The presence of AI and Au in the same chip suggests the possible formation of the purple plague compound and this does actually occur, but only at very high temperatures if a good barrier system is interposed between the two metals [3]. HEMTs

The devices exploiting the characteristics of heterojunctions between semiconductors with different gaps are becoming more and more popular. However, we only have preliminary data on the HEMTs (high electron mobility transistors, also known as MODFETs--modulation doped FETs).The structure of HEMTs (shown in Fig. 6, for the assessed choice of materials) is very similar to that of MESFETs, therefore their reliability is affected by the same failure mechanisms; moreover some failure

Fig. 5. SEM picture of a corroded ohmic contact.


Reliability of compound semiconductor devices ~OU/'Ce


n - GoAs

n ÷ -A/~Go1_~As

undoped - A~x Go l_xAs

undoped- Ga,4s

Fig. 6. Structure of a GaAIAs/GaAs HEMT.

mechanisms associated with the heterojunctions have been forecast [45-50]. The majority of experiments have been performed on AlGaAs/GaAs structures. For these devices the occurrence of degradation of ohmic contacts at very high temperatures was confirmed, together with the purple plague phenomenon [50, 68]. The Sehottky contacts have a less direct influence on device degradation than in the case of MESFETs, due to their distance from the active channel, however the barrier height still determines the device characteristics: different effects were found for two metal systems [50]. In the case of A1/Ni, out-diffusion of the Ni caused an increase of the barrier height and the decrease of loss. For AI/Ti gates the reaction between the two metals only increases the gate resistance. A clear understanding of the failure mechanisms associated with the presence of the heterojunction is still not available: the change in the electron density in the channel and the diffusion of the dopant (Si) from the A1GaAs to the channel was suggested, but has yet to be demonstrated. In effect, the majority of works support the stability of the structures under very high stress [50, 69]. Some results are also now available on pseudomorphic InGaAs/GaAs HEMTs [48, 70]: here too the picture is not yet clear, but a source of degradation seems to be the generation of traps in the strained InGaAs layer due to the growth of edge dislocations along the [110] direction [49].

extension of the channel/substrate depletion region due to a close contact, that changes the occupancy state of the interface traps [73-75]. Although a degradation can be observed over short times, it seems to be unlikely that this phenomenon will affect the long-term reliability of the devices. OPTOELECTRONICS DEVICES The optoelectronics devices make the conversion between electrical and optical signals and vice versa. According to their application, they may be divided into displays, lasers for optical-disks, emitters and receivers for optical communications. The reliability targets for these classes of devices are very different. Usually the displays need only to be visible and are more affected by problems related to specific packaging structures; the lasers employed in consumer applications have lower reliability requirements and their operating life is not expected to be very long. On the contrary, all the devices employed in fiber optic systems must be very reliable, due to the high cost of the repairs, in particular within submarine cable applications [76-78]. RECEIVERS For optoelectronic converters, the influence of the interaction between photons and material is reduced. In effect, the photon flux is generally very low. Therefore, the failure mechanisms are not different from those encountered in electronic devices, except that the semiconductor material is usually InGaAs/InP. According to the structure we may have PIN diodes and avalanche photo diodes (APD). The main problems reported are related to the stability of the interface between the passivation (usually SiN) and the semiconductor. It generally results in the increase of the dark current, i.e. the diode leakage current without illumination. The origin of the increase in the current may be traced back to contamination problems or changes in the occupancy of interface states. The activation energies reported for these mechanisms are higher than 1.5 eV, leaving some doubts as to the extrapolation at low temperatures [79, 80]. The increase in the reverse current can also be ascribed to localized breakdown phenomena due to plasma effects or to the migration of Au from the metallizations [8, 10].



When integrated circuits are manufactured by use of the same technologies as those employed for discrete devices, the failure mechanisms are the same as described before [71-73]. The only failure mechanism that is typical of ICs is the so-called side-gating. This is the reduction of the active channel due to the

Emitting devices for fiber optic applications may be classified according to their material, structure or operation. As far as the materials are concerned, two types of wafer substrates are employed: GaAs and InP. GaAs is used in the lower wave-length applications (


Fig. 7. EL image of DLDs crossing the active area of a GaA1As laser. 1.3-1.55 #m range. At present all the devices employ heterojunctions and there are several possible structures and combinations of materials. Let us just mention that GaA1As is the compound more frequently grown over the GaAs and that new applications require InGaAs. The quaternary InGaAsP is the most common compound grown on InP. Lattice matching is generally the reason for the choice, although in optical devices, the strained structures are also becoming more and more popular. Two types of devices are manufactured: LEDs (light emitting diodes) take advantage of the spontaneous photon emission, typical of the direct bandgap materials, whereas those that use the stimulated emission are obviously lasers. The lasers, due to their more complicated structures and highly concentrated optical power are more sensitive to both chip- and package-related failure mechanisms, saying nothing of the specific failure modes due to the laser structure. In the following section we will consider all the failure mechanisms with particular attention on the effects on the lasers.

Lattice defects As the emission of photons is caused by the radiative recombination, the presence of defects that induce traps in the forbidden band-gap is obviously causing a reduction of the emission. These defects may have been present in the lattice since the

manufacturing of the devices and they are active when they cross the active region. They usually follow a specific plane orientation, so that they are called dark line defects (DLD) [2, 8, 81]. These defects are observed when rapid degradation of both LEDs and lasers occur; they originate from the intrinsic defects of the lattice and propagate by successive climbs of dislocation through a mechanism of recombination-enhanced defect motion [81-84]. The GaA1As lasers are the most sensitive to this failure mechanism due to the larger energy released in the non-radiative recombination process; however the rapidity itself helps when screening the defective devices by high current stresses. In effect, the dependence on the current density is much more important than that on the temperature, which can be expressed by an activation energy of about 0.2 eV [20]. The aspect of DLDs crossing the active stripe of a GaAIAs/GaAs laser is shown in Fig. 7. In InGaAsP/InP devices, no rapid degradation due to DLDs was generally experienced. Also, the gradual degradation that is observed in both GaAs and InP based devices can be traced back to lattice defects. Again, the failure mechanism starts with the nonradiative recombination at point defects (mainly interstitials), that generate new point defects; the defects migrate and condensate to form defect clusters and microloops. The source of the defects is located at the buried hetero-interfaces where the

Reliability of compound semiconductor devices InGaAsP P-InP



active layer





n-lnP P-InP n-lnP

Fig. 8. Localizationof the defectiveareas of a InGaAsP/InP laser [12].

non-planar LPE (liquid phase epitaxy) growth is performed [38, 40-43], as shown in Fig. 8. These defects cause a reduction of the injected carrier lifetime in the active region and an increase of the leakage current in the blocking junctions. In some cases the degradation seems to be activated by the electromigration of the metal interstitials (In, Zn)

[851. Sometimes, the source of defects can be traced back to metals coming from the contacts, in particular Au. In this case, the degradation shows up with the formation of dark areas, called dark spot defects (DSD), and was demonstrated for InP diodes [8, 86].


Mirror degradation Fabry-Perot laser operation is based on the existence of mirrors at the end of a cavity. GaA1As/GaAs devices are sensitive to the surface oxidation that determines a kind of corrosion, called facet erosion. The damage can also be started by the optical power at the facet exceeding a threshold value. In this second case the degradation is generally much more sudden [2, 8, 87-89]. The stronger optical absorption occurring in the damaged area causes local heating of the crystal and shrinkage of the forbidden band-gap, which forces further optical absorption. The positive feedback may induce a thermal runaway with local melting or the growth of the defective area from the surface into the active area, as shown in Fig. 9. Recently, facet overheating during high power operation was measured for A1GaAs single quantum well lasers [90]. From these data it seems that catastrophic optical damage occurs at a facet temperature of about 140°C above room temperature, the facet overheating being strongly dependent on the operating current. InGaAsP/InP lasers are not sensitive to this failure mechanism, due to the better stability at their surfaces, on the contrary, this failure mechanism is

Fig. 9. EL image of the dark area originating at the mirror of a GaAIAs laser.



Fig. 10. EL image of the damage induced by ESD in two different InGaAsP lasers (dashed lines point to the position of the mirrors). expected to be dangerous for the new lasers used for optical amplification [91]; however, a degradation starting from the mirrors was found to be the failure mechanism induced by ESD in InGaAsP/InP buried crescent lasers [44]; in this case the effect is not the catastrophic burn-out, but the generation of nonradiative defects well confined inside the active layer, which propagate from the mirrors towards the inner part of the chip, as shown in Fig. 10. CONCLUSIONS The research undertaken in order to understand the specific failure mechanisms of the III-V semiconductor devices and to improve their reliability has made progress and we can assert that very reliable devices can be manufactured. However, the differences in the degree of maturity among the various technologies are still considerable and the applications for new devices require careful attention.

Sudden ~-~Cz~ ~ s ° l d e r

and heat/sink



degradationdue to


DSD.DLD ~ I generation degradation ~of I

C~ t~

p-side electrode degradation

\ I I

~ b u r y i n glayer il ( / ~ c h e d part degradation TM / ~ a c t i v e region and its vicinity



~ 0

i 10 4

~ 10 5

l l0 s

m 10 7


AGING TIME (h) Fig. 1I. Estimated degradation models for InGaAsP/InP lasers [42].

In the field of MESFET devices and integrated circuits, a high reliability has already been demonstrated in field operations, whereas prudence must be applied in the extrapolation of the data from the accelerated tests, due to the presence of high activation energy phenomena, in particular for the degradation of the ohmic contacts. The same consideration is valid for the HEMTs, where new phenomena related to the carrier confinement could appear, but are at this moment considered to be unlikely. We still require data concerning the reliability of HBTs. In the field of optoelectronic devices, the confidence in the reliability of InP-based lasers seems to be very high and nobody has yet reported results as poor as those from the case of GaAs-based devices. In the case of lasers, the reliability evaluation performed by means of accelerated testing is complicated by the impossibility of keeping the device working correctly at high temperatures, due to the increase of the threshold currents. For GaAs-based lasers, an activation energy of 0.7 eV was calculated from the very early experiments at the Bell Laboratories [22], and employed later on without any real correlation with the physical degradation mechanisms. However in general, the results of these extrapolations have not been considered to be satisfactory. A complete characterization of the reliability of InGaAsP/InP lasers performed by the research group of NTT Atsugi Laboratory gave the results summarized in Fig. 11, where the importance of the various failure mechanisms during their lifetime, and their rapidity, are reported.

Reliability of compound semiconductor devices T h e new interest in the reliability of I n G a A s / G a A s lasers a n d the i m p o r t a n c e o f short-wavelength lasers for c o n s u m e r application is creating a p u s h for more studies o n the stability o f the mirrors, which is the m o r e c o n t e n t i o u s issue for these devices.

Acknowledgements--The authors wish to acknowledge M. Vanzi for supplying the pictures of the failed devices, P. Gandolfi for the drawings, M. Fukuda for the permission to reproduce Figs 8 and I 1 and J.-M. Dumas for supplying unpublished data.

REFERENCES 1. J. E. Davey and A. Christou, Reliability and degradation of active III-V semiconductor devices, In Reliability and Degradation (edited by M. J. Howes and D. V. Morgan), pp. 237-300. Wiley, NY (1981). 2. D. H. Newman and S. Ritchie, Reliability and degradataion of lasers and LEDs, In Reliability and Degradation (edited by M. J. Howes and D. V. Morgan), pp. 301-362. Wiley, NY (1981). 3. J. C. Irvin, The reliability of GaAs FETs, In GaAs FET Principles and Technology (edited by J. V. DiLorenzo and D. D. Khandelwal), pp. 349-400. Artech House, Norwood, MA (1982). 4. A. Christou, Report on the 1982 GaAs device reliability workshop, 20th A. Proc. Reliab. Phys. 1982, pp. 276-277, IEEE (1982). 5. A. Christou, Reliability problems in state-of-the-art GaAs devices and circuits, Qual. Reliab. Engng Int. 5, 37-46 (1989). 6. S. Adams, D. Dening, R. Finke, D. LaCombe and L. Rider, GaAs reliability and test, In Microelectronic Reliability, Volume I: Reliability, Test and Diagnostics (edited by E. B. Hakim), pp. 241-311. Artech House, Norwood, MA (1989). 7. D. E. Riva and G. Clerico Titinet, Reliability of compound semiconductor microwave field effect devices, In Microelectronic Reliability, Volume II: Integrity Assessment and Assurance (edited by E. Pollino), pp. 195-258. Artech House, Norwood, MA (1989). 8. P. Montangero and A. Piccirillo, Reliability of optoelectronic components, In Microelectronic Reliability, Volume II: Integrity Assessment and Assurance (edited by E. Pollino), pp. 259-310. Artech House, Norwood, MA (1989). 9. F. Magistrali, C. Tedesco and E. Zanoni, Failure mechanisms of GaAs MESFETs and low-noise HEMTs, In Semiconductor Device Reliability (edited by A. Christou and B. A. Unger), pp. 211-267. Kluwer, Dordrecht (1990). 10. S. P. Sim, A review of the reliability of III-V optoelectronic components, in Semiconductor Device Reliability (edited by A. Christou and B. A. Unger), pp. 301-319. Kluwer, Dordrecht (1990). 11. W. T. Anderson and A. Christou, Status of compound semiconductor device reliability, in Semiconductor Device Reliability (edited by A. Christou and B. A. Unger), pp. 423-437. Kluwer, Dordrecht (1990). 12. M. Fukuda and T. Ikegami, Optoelectronics reliability, Proc. ESREF'90, pp. 393~,04, Tecnopolis CSATA N.O. (1990). 13. P. Brambilla, F. Fantini, F. Magistrali and M. Sangalli, Life tests and field results of GaAs FETs, Ann. Telecommun. 45, 617-624 (1990). 14. D. A. Abbott and J. A. Turner, Some aspects of GaAs MESFET reliability, IEEE Trans. Microwave Theory Techs. MTT-24, 317 321 (1976).


15. T. Irie, I. Nagasako, H. Kohzu and K. Sekido, Reliability study of GaAs MESFET's, IEEE Trans. Microwave Theory Techns. MTT-24, 321-328 (1976). 16. R. E. Lundgren and G. O. Ladd, Reliability study of microwave GaAs field-effect transistors, 16th A. Proc. Reliab. Phys. 1978, pp. 255-260. IEEE (1978). 17. I. Drukier and J. F. Silcox Jr, On the reliability of power GaAs FETs, 17th A. Proc. Reliab. Phys. 1979, pp. 150-155. IEEE (1979). 18. M. Omori, J. N. Wholey and J. F. Gibbons, Accelerated active life test of GaAs FET and a new failure mode, 18th A. Proc. Reliab. Phys. 1980, pp. 134-139. IEEE (1980). 19. H. Fukui, S. H. Wemple, J. C. Irvin, W. C. Niehaus, J. C. M. Hwang, H. M. Cox, W. O. Schlosser and J. V. Di Lorenzo, Reliability of improved power GaAs fieldeffect transistors, 18th A. Proc. Reliab. Phys. 1980, pp. 151-158. IEEE (1980). 20. M. Abe, I. Umebu, O. Hasegawa, S. Yamakoshi, T. Yamaoka, T. Kotani, H. Okada and H. Takanashi, High-efficiency long-lived GaA1As LEDs for fiberoptical communications, IEEE Trans. Electron. Dev. ED-24, 990-994 (1977). 21. L. R. Dawson, V. G. Keramidas and C. L. Zipfel, Reliable, high-speed LEDs for short-haul optical data links, Bell Syst. Techn. J. 59, 161-168 (1980). 22. R. L. Hartman and R. W. Dixon, Reliability of DH GaAs lasers at elevated temperatures, Appl. Phys. Lett. 26, 239-242 (1975). 23. W. B. Joyce, R. W. Dixon and R. L. Hartman, Statistical characterization of the lifetimes of continuously operated (A1, Ga) As double-heterostructure lasers, Appl. Phys. Lett. 28, 684-686 (1976). 24. S. Ritchie, R. F. Godfrey, B. Wakefield and D. H. Newman, The temperature dependence of degradation mechanism in long-lived (GaA1) As DH lasers, J. Appl. Phys. 49, 3127-3132 (1978). 25. K. Mettler, H. D. Wolf and K. H. Zschauer, Calculation of the homogeneous degradation of injection laser parameters from initial degradation rates, IEEE J. Quantum Electron. QE-14, 819-826 (1978). 26. H. Ishikawa, T. Fujiwara, K. Fujiwara, M. Morimoto and M. Takusagawa, Accelerated aging test of Ga~_~AlxAs DH lasers, J. Appl. Phys. 50, 2518-2522 (1979). 27. K. Mizuishi, N. Chinone, H. Sato and R. Ito, Acceleration of the gradual degradation in (GaAI)As doubleheterostructure lasers as an exponent of the value of the driving current, J. Appl. Phys. 50, 6668-6674 (1979). 28. M. Ettenberg and H. Kressel, The reliability of (AlGa)As CW laser diodes, IEEE J. Quantum Electron. QE-16, 186-196 (1980). 29. H. Imai, K. Hori, M. Takusagawa and K. Wakita, Activation energy of degradation in GaA1As double heterostructure laser diodes, J. Appl. Phys. 52, 3167-3171 (1981). 30. M. J. Robertson, Effects of 140 Mbit/s operation on degradation of GaA1As DH lasers, Electron. Lett. 19, 88-89 (1983). 31. B. A. Dean and M. Dixon, The functional reliability of semiconductor lasers as optical transmitters, Semicond. Semimetals 22C, 153-191, Bell Telephone Inc. (1985). 32. P. Durand and F. Gobin, Sp6cificit6 de la fiabilit6 des composants opto61ectroniques pour fibre optique utilis6s dans le secteur des T616communications, rOnde Electrique 68, (3), 76--88 (1988). In French. 33. L. Battistella, F. Magistrali, M. Vanzi and F. Fantini, Life test and failure analysis results on laser diodes for fiber optic communications systems, Proc. 6~me Colloque Int. de Fiabilitd et de Maintenabilitd, pp. 233-239 (1988). 34. S. Yamakoshi, M. Abe, O. Wada, S. Komiya and T. Sakurai, Reliability of high radiance InGaAsP/InP



36. 37. 38.


40. 41.

42. 43. 44.



47. 48.

49. 50.




LEDs operating in the 1.2-1.3/~m wavelength, IEEE J. Quantum Electron. QE-17, 167-172 (1981). K. Mizuishi, M. Hirao, S. Tsuji, H. Sato and M. Nakamura, Reliability of InGaAsP/InP buried heterostructure lasers, Jpn J. Appl. Phys. 21-1, 359-364 (1982). T. Ikegami, K. Takahei, M. Fukuda and K. Kuroiwa, Stress tests on 1.3 /~m buried-heterostructure laser diode, Electron. Lett. 19, 282-283 (1983). Y. Nakano, G. Iwane and T. Ikegami, Screening method for laser diodes with high reliability, Electron. Lett. 20, 397 398 (1984). E. Oomura, H. Higuchi, Y. Sakakibara, R. Hirano, H. Namizaki, W. Susaki and K. Ikeda, InGaAsP/InP buried crescent laser diode emitting at 1.3 /~m wavelength, IEEE J. Quantum Electron. QE-20, 866-873 (1984). F. R. Nash, W. J. Sundburg, R. L. Hartman, J. R. Pawlik, D. A. Ackerman, N. K. Dutta and R. W. Dixon, Implementation of the proposed reliability assurance strategy for an InGaAsP/InP, planar mesa, buried heterostructure laser operating at 1.3 a m for use in a submarine cable, AT&T tech. J. 64, 809-859 (1985). M. Fukuda and G. Iwane, Degradation of active region in InGaAsP/InP buried heterostructure lasers, J. Appl. Phys. 58, 2932-2936 (1985). M. Fukuda and G. Iwane, Correlation between degradation and device characteristic changes in InGaAsP/InP buried heterostructure lasers, J. Appl. Phys. 59, 1031-1037 (1986). M. Fukuda and T. Ikegami, Reliability and degradation mechanism of InGaAsP/InP semiconductor lasers, Annls Telecommun. 45, 625~29 (1990). O. Ueda, Degradation of III-V opto-electronic devices, J. Electrochem. Soc. 135, 11C-22C (1988). F. Magistrali, D. Sala, G. Salmini, F. Fantini, M. Giansante and M. Vanzi, ESD-related latent failures of InGaAsP/InP laser diodes for telecommunication equipment, 29th A. Proc. Reliab. Phys. 1991, pp. 224-233, IEEE (1991). A. Christou, W. W. Tseng, M. Peckerar, W. T. Anderson, D. M. McCarthy, F. A. Buot, A. B. Campbell and A. R. Knudson, Failure mechanism study of GaAs MODFET devices and integrated circuits, 23th A. Proc. Reliab. Phys. 1985, pp. 54-59, IEEE (1985). A. Belhadj, P. Audren, C. Vuchener, J. Paugam and J. M. Dumas, Caractrrisation et rrsultats de fiabilit6 de transistors ~ haute mobilit6 +lectronique (HEMT), Annls TblOcommun. 45, pp. 584-590 (1990). In French. W. T. Anderson and A. Christou, Reliability of advanced LAMBE FETs and HEMTs, Proc. ESREF'90, pp. 471-483, Tecnopolis CSATA N.O. (1990). N. L. Saunders, A. A. Rezazadeh, D. V. Morgan, H. Thomas and L. G. Hipwood, The long term stability of n-AIGaAs/InGaAs/GaAs pseudomorphic HEMTs, Proc. ESREF'90, pp. 485-492, Tecnopolis CSATA N.O. (1990). A. Christou, J. M. Hu and W. T. Anderson, Reliability of InGaAs HEMTs on GaAs Substrates, 29th A. Proc. Reliab. Phys. 1991, pp. 200-205, IEEE (1991). C. Canali, F. Magistrali, M. Sangalli, C. Tedesco, E. Zanoni, G. Castellaneta and F. Marchetti, Reliability aspects of AIGaAs/GaAs HEMTs, 29th A. Proc. Reliab. Phys. 1991, pp. 206-213, IEEE (1991). C. Canali, F. Castaldo, F. Fantini, D. Ogliari, M. Vanzi, M. Zicolillo and E. Zanoni, Power GaAs MESFET: reliability aspects and failure mechanisms, Microelectron. Reliab. 24, 947-955 (1984). J. J. Whalen, M. C. Calcatera and M. L. Thorn, Microwave nanosecond pulse burnout properties of GaAs MESFETs, IEEE Trans. Microwave Theory Techqs MTT-27, 1026-1031 (1979).

53. W. T. Anderson, F. A. Buot, A. Christou and Y, Anand, High power pulse reliability of GaAs power FETs, A. Proc. Reliab. Phys. 1986, pp. 144-149, IEEE (1986). 54. F. A. Buot, W. T. Anderson, A. Christou, K. J. Sleger and E. W. Chase, Theoretical and experimental study of subsurface burnout and ESD in GaAs FETs and HEMTs, A. Proc. Reliab. Phys. 1987, pp. 181-190, IEEE (1987). 55. F. Capasso and G. F. Williams, A proposed hydrogenation/nitridization passivation mechanism for GaAs and other III-V semiconductor devices, including lnGaAs long wavelength photodetectors, J. Electrochem. Soc. 129, 821-824 (1982). 56. J. M. Dumas, D. Lecrosnier and J. F. Bresse, Investigation into GaAs power MESFET surface degradation, A. Proc. Reliab. Phys. 1985, pp. 39-44. IEEE (1985). 57. K. H. Kretschmer and H. L. Hartnagel, Interelectrode metal migration on GaAs, A. Proc. Reliab. Phys. 1987, pp. 102-106. IEEE (1987). 58. C. Canali, E. Corti, B. Gabrielli, F. Magistrali, A. Paccagnella, M. Sangalli and C. Tedesco, Change of g=Or) and breakdown voltage induced by thermal annealing of surface states in power MESFETs, IEEE Trans. Electron. Dev. ED-37, 1163-I 165 (1990). 59. J. M. Dumas, F. Garat and D. Lecrosnier, Development of gate-lag effect on GaAs power MESFETs during aging, Electron. Left. 23, 139-141 (1987). 60. E, Zanoni, A. Callegari, C. Canali, F. Fantini, H. L. Hartnagel, F. Magistrali, A. Paccagnella and M. Vanzi, Metal-GaAs interaction and contact degradation in microwave MESFETs, Quality Reliab. Engng Int. 6, 29-40 (1990). 61. C. Canali, F. Fantini, A. Scorzoni, L. Umena and E. Zanoni, Degradation mechanisms induced by high current density in Al-gate GaAs MESFETs, IEEE Trans. Electron. Dev. ED-34, 205-211, (1987). 62. 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 MESFETs, IEEE Electron. Dev. Lett. EDL-7, 185-187 (1986). 63. F. Fantini, F. Magistrali, D. Ogliari, M. Sangalli and M. Vanzi, Back-etch: an effective tool for characterization and failure analysis of MESFET devices, Proc. ISTFA 1988, pp. 235-241, ASM (1988). 64. W. O. Camp Jr, R. Lasater, V. Genova and R. Hume, Hydrogen Effects on Reliability of C_vaAsMMICs, In Semiconductor Device Reliability (edited by A. Christou and B. A. Unger), pp. 471-477. Kluwer, Dordrecht (1990). 65. N. Kornilios, K. Tsagaraki, J. Stoemenos and A. Christou, Reliability of GaAs MESFETs on GaAs and Si Substrates, Proc. ESREF'90, pp. 455-463, Tecnopolis CSATA N.O. (1990). 66. C. Canali, F. Chiussi, F. Fantini, L. Umena and M. Vanzi, Electromigration effects in power MESFET rectifying and ohmic contacts, Electron. Lett. 23, 364-365 (1987). 67. F. Magistrali, D. Ogliari, M. Sangalli and M. Vanzi, Degradation mechanism of GaAs MESFET devices in high humidity conditions, Proc. ISTFA 1989, pp. 141-151, ASM (1989). 68. P. Conti, F. Fantini, G. Gallo, R. Menozzi and E. Pollino, High temperature storage tests on commercial HEMTs, Proc. ESREF'91, pp. 491-498, IXL Bordeaux (1991). 69. W. T. Anderson, A. Christou, F. A. Buot, J. Archer, G. Bechtel, H. Cooke, Y. C. Pao, M. Simons and E. W. Chase, Reliability of discrete MODFETs: life testing, radiation effects, and ESD, 26th A. Proc. Reliab. Phys. 1988, pp. 96-101, IEEE (1988). 70. P. Audran, J. M. Dumas, S. Mottet, C. Vuchener, J. Paugam and M. P. Favennec, Stability of interfaces

Reliability of compound semiconductor devices

71. 72.


74. 75. 76.




80. 81.

in pseudomorphic InGaAs HEMTs, Proc. ESREF'91, pp. 393-400, IXL Bordeaux (1991). D. Ogbonnah and A. Fraser, Reliability investigation of 1 micron depletion mode IC MESFETs, 24th A. Proc. Reliab. Phys. 1986, pp. 132-137, IEEE (1986). G. Kervarrec, J. M. Dumas, J. Y Boulaire and J. F. Bresse, Failure mechanisms study of a standard GaAs technology, Proc. ESSDERC'87, pp. 617~20, North Holland (1988). J. M. Dumas, Parasitic effects and degradation mechanisms of GaAs integrated circuits, Proc. Int. Conf. Quality Electron. Compon. 1989, pp. 120-127, IXL Bordeaux (1989). T. Itoh and H. Yanai, Stability of performance and interfacial problems in GaAs MESFETs, IEEE Trans. Electron. Dev. ED-27, 1037-1045 (1980). C. Kocot and C. A. Stolte, Backgating in GaAs MESFETs, IEEE Trans. Electron. Dev. ED-29, 1059-1064 (1982). Y. Iwamoto and H. Fukinuki, Recent advances in submarine optical fiber cable transmission systems in Japan, 1EEE J. Lightwave Technol. LT-3, 1005-1016 (1985). O. Mitomi, K. Kawano, T. Nozawa and M. Yanagibashi, Reliability of optical components for use in submarine optical fiber transmission system, IEEE J. Lightwave Technol. LT-5, 838-847 (1987). R. H. Murphy, Models used in undersea fibre optic systems reliability prediction, In Semiconductor Device Reliability (edited by A. Christou and B. A. Unger), pp. 147-160. Kluwer, Dordrecht (1990). S. R. Forrest, V. S. Ban, G. Gasparian, D. Gay and G. H. Olsen, Reliability of vapor-grown planar In0.53Ga0.47As/InP p-i-n photodiodes with very high failure activation energy, IEEE Electron. Dev. Lett. EDL-9, 217-219 (1988). H. Sudo and M. Suzuki, Surface degradation mechanism of InP/InGaAs APDs, IEEE J. Lightwave Technol. LT-6, 1496-1501 (1988). P. M. Petroff, Defects in III-V compound semi-

82. 83. 84. 85.


87. 88. 89. 90.

9 I.


conductors, Semicond. Semimetals 22A, 379-403, Bell Telephone Inc. (1985). L. C. Kimerling, P. Petroff and H. J. Leamy, Injectionstimulated dislocation motion in semiconductors, Appl. Phys. Lett. 28, 297-300 (1976). S. O'Hara, P. W. Hutchinson and P. S. Dobson, The origin of dislocation climb during laser operation, Appl. Phys. Lett. 30, 368-371 (1977). L. C. Kimerling, Recombination enhanced defect reactions, Solid-St. Electron. 21, 1391-1401 (1978). S. N. G. Chu, S. Nakahara, M. E. Twigg, L. A. Koszi, E. J. Flynn, A. K. Chin, P. B. Segner and W. D. Johnston, Defect mechanisms in degradation of 1.3-/~m wavelength channeled-substrate buried heterostructure lasers, J. Appl. Phys. 63, 611~23 (1988). A. K. Chin, C. L. Zipfel, F . Ermanus, L. Marchut, I. Camlibel, M. A . DiGiuseppe and B. H. Chin, The migration of gold from the p-contact as a source of dark spot defects in InP/InGaAsP LEDs, IEEE Trans. Electron. Dev. ED-30, 304-310 (1983). N. Chinone, H. Nakashima and R. Ito, Long-term degradation of GaAs-Ga I _~AlrA,s DH lasers due to facet erosion, J. Appl. Phys. 48, 1160-1162 (1977). F. R. Nash, R. L. Hartman, N. M. Denkin and R. W. Dixon, GaAs laser reliability and protective facet coatings, J. Appl. Phys. 50, 3122-3132 (1979). M. Fukuda, Facet oxidation of InGaAsP/InP and InGaAs/InP lasers, IEEE J. Quantum Electron. QE-19, 1692-1698 (1983). W. C. Tang, H. J. Rosen, P. Vettiger and D. J. Webb, Time development of A1GaAs single quantum well laser facet temperature on route to catastrophical breakdown, In Laser Diode Technology and Applications III, SPIE Proc. 1418, (edited by D. Renner), pp. 338-342, SPIE (1991). M. Okayasu, M. Fukuda, T. Takeshita and S. Uehara, Stable operation (over 5000 h) of high-power 0.98 #m InGaAs-GaAs strained quantum well ridge waveguide lasers for pumping Er3 +-doped fiber amplifiers, IEEE Photonics Technol. Lett. 2, 689~591 (1990).