To estimate the probability of failure for a deployed population, when the laboratory aging of a sample certified population produces zero failures, an appropriate model is required. The widely used exponential (constant hazard rate) model, and a declining-hazard-rate Weibull model are contrasted. On balance the use of the Weibull model seems preferred because: (i) It is consistent with laboratory aging studies on many types of electronic components; (ii) It is less sensitive to the choices for parameters (e.g., thermal activation energy) whose values are not known accurately; and, (iii) It may yield more conservative (i.e., larger) values for the relevant hazard rate than the exponential model. The unavoidable uncertainties connected with the use of either model are discussed. An example is given of a plausible, but in fact mistaken, use of the exponential model that leads to disastrously incorrect reliability predictions.

From electron-irriadiation studies, it is known that point defects can strongly affect the electrical properties of GaAs-based MESFET's, MODFET's, solar cells, resonant-tunneling diodes, MIMIC circuits and other devices. As an example, 1 × 1016 cm−2, 1 MeV electrons reduce the transconductance in a 1 μm by 200 μm MESFET by nearly an order of magnitude. Fortunately, annealing at 350 °C for 10 min. can largely restore the device performance, although not without some adverse effects. Point defects, or simple point defect complexes, can also exist in as-grown GaAs and affect devices in several different ways. For example, an As-rich stoichiometry can lead to an abundance of Ga vacancies, and thus to a higher Si donor activation in implanted MESFET devices; however, it can also promote an increase in unwanted impurities which sit on the Ga site. As another example, extremely high (> 3 × 1019 cm−3 ) concentrations of As arntisites, which are found in MBE GaAs grown at 200°C, lead to very unusual electrical and optical properties, and make possible a highly useful buffer material anid a very fast photoconductive switch. However, there are also adverse effects here, such as slow-transients in some MODFET devices, which may result from defect diffusion. Thus, the effects of point defects in GaAs devices must be understood.

The emission of visible light from the high-field region near the drain of GaAs/AlGaAs MESFETs has been used to study the quality of fabricated transistors. Inhomogeneities or bright spots in the emission have proven to indicate the presence of defects in the gate/drain spacing or surface contaminants, such as re-deposited gate material. Processing abnormalities of 0.5 μm diameter are easily seen. The time-variation of emission from the brightest spots is characteristic of the formation of microplasmas the location of which will often predict the site of destructive device breakdown.

In this paper the effects of DX levels on the properties of AlxGa1−xAs and certain heterojunction device structures are summarized. Studies of alloy effects, variations of the local atomic environment near the donor, are reviewed and microscopic models of the DX center are discussed in terms of these and other recent experimental results.

The properties of deep donor states (DX's centers) in III-V alloys are reviewed in relation to their influence on the device characteristics and limitations. Because of the systematic research being performed on AlGaAs, most of the information presented refers to such material. The electron thermal emission and capture properties of the DX's are then related to the DC and noise characteristics in heterojunction transistors. The optical properties of DX centers indicate a clear difference between unipolar and bipolar device performance at low temperatures. The technical efforts to avoid DX centers will also be described.

Deep level admittance spectroscopy (DLAS) of the Sn-DX centers in Alx Ga1−x As:Sn (0.2 < x < 0.6) shows three peaks SNi, SN2 and SN3. The SN3 peak is identified to be related to the dominant peak of the Sn-DX center observed in the conventional DLTS technique. The SN1 and SN2 peaks are not easily seen in DLTS. A careful analysis of the DLAS data shows that the three peaks are not due to independent (chemically distinct) defects related to Sn, but they are caused by the multiple sates of the same DX center.

Conspicuous enhancement of dislocation glides under minority carrier injection is reviewed. Systematic investigation of dislocation glide velocity under electron-beam irradiation in various semiconducting crystals showed that the radiation enhanced dislocation glide (REDG) effect exhibits features expected from the recombination enhanced defect motion (REDM) mechanism. Theoretical analysis, based on the kink diffusion model for the elementary process of dislocation motion, shows that the REDG effect requires enhancement of double-kink formation. An experimental attempt provided an evidence for the smallest double kink formation is the only process that is enhanced by carrier injection. In light of the knowledge available at present, a guide line in device design to minimize degradation in the dislocation glide mode is suggested.

The mechanism of light-induced reactivation (LIR) of shallow substitutional acceptors in high-purity p-type hydrogenated GaAs has been investigated. Photoluminescence was used to determine the dependence of the rate and extent of this effect on photon energy, illumination intensity, as well as on sample temperature and chemical composition. At a sample temperature of 1.7 K a sharp threshold in the photon energy, Et, has been observed at about 7.5 meV below the bandgap energy of GaAs. This energy corresponds approximately to the onset of acceptorbound exciton absorption in the material. For photon energy E < Et, only a weak reactivation effect is observed. The efficiency of reactivation increases dramatically for E > Et, and for sufficiently large values of (light intensity).(illumination time) product the LIR process saturates. Both the extent of the subthreshold effect and the saturation level that is attainable with E > Et are independent of the photon energy, excitation power and exposure time in the investigated range of these quantities. For E > Et the initial LIR rate depends on the square of the light intensity, indicating a bimolecular reaction in terms of the photo-generated carrier densities. The observed strong dependence of the saturation level on the sample temperature during LIR is found to be consistent with the relative binding energies of different acceptor-hydrogen passivating complexes in GaAs. Based on these results, it is proposed that LIR of acceptors is electronically stimulated via recombination-enhanced vibrational excitation of acceptor-hydrogen complexes.

We detail experiments showing that acceptor passivation by atomic hydrogen in p-type GaAs is unstable to either illumination, forward bias annealing or reverse bias annealing. The long-term stability of operation of devices employing hydrogen in or near the active region of FETs or quantum-well lasers is therefore questionable. The systematics of acceptor reactivation during minority carrier injection or reverse bias annealing are presented.

The electron traps in Si-implanted active layers(n∼1017 cm−3) have been studied by capacitance and conductance DLTS techniques in conjunction with different anneal conditions, which include rapid thermal anneals at different temperatures and furnace anneals with a Si3N4 cap or capless in an AsH3 atmosphere. As compared to the electron traps in as-grown bulk n-GaAs (n∼4×1016 cm−3), nearly the same electron traps, i.e. EL2, EL3, ELM, EL5, EL6 and EL9 can be observed in the Si-implanted layers. Through a comparison with the annealing behavior of the main elect ron traps in bulk n-GaAs, the processing associated origins of some of the traps(EL2, EL3, EL4, EL5 and EL9) observed in Si-implanted GaAs layers have been determined. For some Si-implanted GaAs capped with Si3N4 and furnace annealed, traps EL3 and EMA dominate the trap EL2. In such layers it is found that emission due to EL3 is reduced while emission from EL12 is augmented by increasing the filling pulse width from 10 μs to 5×103 μs. This phenomenon can be explained in terms of a defect reaction enhanced by electron capture, showing a metastability or bistability.

We have studied the atomic scale relaxation of multilyered semiconductor systems through interdiffusion, and its implications for layer and hence device stability. Our technique combines chemical lattice imaging and vector pattern recognition to measure interdiffusion coefficients as small as 10-20 cm2/s at single AiGaAs/GaAs interfaces. Our results can be summarized as follows. (a) The interdiffusion coefficient and hence layer stability depend strongly on the distance from the top surface. (b) The strong influence of the surface stems from its role as the source of rative point defects (interstitials, vacancies) involved in the diffusion process. (c) The study of the initial transients in the diffusion process can directly yield the migration and formation energies of these defects as a function of their charge state. (d) The non-linear nature of the diffusion process can have serious consequences for the low temperature stability of multilayered materials and devices. We also discuss the implications of our results for the design and fabrication of stable multilayered systems.

Degradation in optical and electrical properties has been observed for high-purity and high-mobility p-type GaAs layers which contain significant concentrations of an unidentified shallow acceptor-like defect, labeled “A”, that is frequently incorporated in crystals grown by molecular beam epitaxy. Low-temperature photoluminescence and variable temperature Hall-effect measurements were employed to monitor the aging process in samples stored for about one year at room temperature. Profound changes in the exciton recombination spectra, indicative of increasing concentration of the “A” defect, have been accompanied by a decrease in hole mobility and an increase in carrier concentration. These results are discussed in the context of the acceptor-pair defect model, originally proposed by Eaves and Halliday [J. Phys. C: Solid State Phys. 17, L705 (1984)].

The results in the paper show that high defect density appears during rapid thermal annealing. The spreading of the defect distribution increases with increasing of the annealing time. So the leakage current increases up. It is found that the PN junction leakage current for short annealing time(5s) and for hot implantation are obviously lower than that of RT implantation. The mechanism of the leakage current reduction is discussed.

Material issues in III-V alloy semiconductors and our current understanding of degradation in III-V semiconductor lasers and LED's are systematically reviewed.

Generation of defects and thermal instability are among these issues for these systems. Defects introduced during crystal growth are classified into two types: interface defects and bulk defects. Defects belonging to the former type are stacking faults, V-shaped dislocations, dislocation clusters, microtwins, inclusions, and misfit dislocations, and the latter group includes precipitates and dislocation loops. Defects in the substrate can also be propagated into the epi-layer. Structural imperfections due to thermal instability are also found. They ame quasi-periodic modulated structures due to spinodal decomposition of the crystal either at the liquid/solid interface or growth surface, and atomic ordering which also occurs on the growth surface through migration and reconstruction of the deposited atoms.

Three major degradation modes, rapid degradation, gradual degradation, and catastrophic failure, are discussed. For rapid degradation, recombination-enhanced dislocation climb and glide are responsible for degradation. Differences in the ease with which these phenomena occur in different hetero-structures are presented. Based on the results, dominant parameters involved in the phenomena are discussed. Gradual degradation takes place presumably due to recombination enhanced point defect reaction in GaAlAs/GaAs-based optical devices. This mode is also enhanced by the internal stress due to lattice mismatch. However, we do not observe this mode in InGaAsP/InP-based optical devices. Catastrophic failure is found to be due to catastrophic optical damage at a mirror or at a defect in GaAlAs/GaAs DH lasers, but not in InGaAsP/InP DH lasers. In each degradation mode, the role of defects in the degradation and methods of elimination of degradation are discussed.

Optical degradation of long wavelength (1.30–1.55 micron) laser diodes during normal operation or accelerated aging test caused by lattice structural deterioration of the active region materials and mirror facet damage were investigated in detail. Extrinsic dislocation loops of 1/2<100>{010) types were observed in gradually degraded channeled-substrate-buried-heterostructure (CSBH) lasers. These dislocation loops, originated at the sidewall interfaces outside the active region, grew into the active region in the direction of minority carrier injection. A great enhancement of the loops' growth rate was observed after they entered the active region, indicating a nonradiative recombination enhanced defect reaction under the strong optical field. Furthermore, the <100> oriented extrinsic dislocation loops were confirmed to be dark-line-defects (DLDs). On the other hand, strong nonradiative recombination centers, created by mirror facet damage, or pre-existed internally inside the cavity, resulted localized melting on the {111} planes. The propagation of the localized melt-patch along the laser beam direction created a wormlike defect along the laser cavity, which degraded the laser device catastrophically. Various types of grown-in defects for CSBH and etched-mesa-buried-heterostructure (EMBH) laser devices are described and their effects on the device performance are demonstrated.

A series of experiments have been performed to verify the nature of hole trap A and B in GaAs proposed by Zou and revised by Zhou (one of the present authors). The relative concentration of these two Craps is responsible for the degradation behavior of the diodes. Accordingly, the degradation mechanism in GaAs LED can be reasonably deduced based on thermodynamic consideration.

This paper presents results of a study of the degradation of commercially available GaAs and AlGaAs light emitting diodes subjected to neutron bombardment at a TRIGA reactor. Devices were characterized using current-voltage and light output measurements prior to and following a sequence of neutron irradiations and after high temperature annealing. A model is derived which can be used to determine the lifetime damage constant product, τoK, if the light output measurements as a function of IMeV equivalent neutron fluence are made at a fixed operating current. For current levels smaller than approximately 1 ma, τoK and operating current is logarithmic with τoK decreasing as current increases. Annealing at temperature up to 275°C recovers some of the neutroninduced damage but does not affect the validity of the model.

High resolution maps of the InGaAs photoluminescence intensity over whole 50mm diameter PIN detector wafers grown by low pressure MOCVD were obtained using a scanning photoluminescence (PL) system. The leakage current of PIN detectors was found to decrease exponentially with increased PL intensity. This correlation was quantitatively valid for both variation on a given wafer and for wafer-to-wafer variations. The density of morphological features on the surface of InGaAs(P)/InP heterostructures also was found to decrease exponentially with increased PL intensity, where a simple linear relationship between leakage current and feature density was then determined. The features are likely a manifestation of substrate defects which propagate from the surface of poor quality substrates.

The inherent complexity of defect processes in III-V's and the consequent difficulties with ab initio and semi-empirical methods are recalled. A potential solution using massive Monte Carlo simulation on microcomputers is suggested. Evidence for the validity of the Ballistic Model, BM, of atomic diffusion in III-V's is noted. According to the BM the effect of strain (in the absence of any electrostatic, population, or recombination effect) is to increase the rate at which a given mobile atom hops where the sample is compressed. For the case of misfit strain at a (100) junction, we note that the anisotropy of the elastic constants implies that some planes running into the bulk will be compressed whichever the sign of the misfit. This implies that misfit strain of either sign should increase the observed rate of interdiffusion, in the absence of other effects. We also recall the importance (demonstrated at low T) of recombination enhancement of atomic diffusion, RED. Devices are processed at temperatures where the thermal rate of recombination is very high and often operated at high levels of injection. The interaction of strain with RED is clearly important and complicated. III-V crystals have the further complication of being piezoelectric. The active piezoelectric axes are <111>, so a pure <100> strain does not produce a field. However, the accommodation a device makes to misfit at (100) junctions can generate a strong field, which may fluctuate with bias voltage.

Because of different band-edge lineups, strain conditions, and growth orientations, various strained-layer superlattice (SLS) materials can exhibit qualitatively new physical behavior in their optical properties. We describe two examples of new physical behavior in SLS: strain-generated electric fields in polar growth axis superlattices and strained type II superlattices. In SLS, large electric fields can be generated by the piezoelectric effect. The fields are largest for SLS with a [111] growth axis; they vanish for SLS with a [100] growth axis. The strain-generated electric fields strongly modify the optical properties of the superlattice. Photogenerated electron-hole pairs screen the fields leading to a large nonlinear optical response. Application of an external electric field leads to a large linear electrooptical response. The absorption edge can be either red or blue shifted. Optical studies of [100], [111], and [211] oriented GaInAs/GaAs superlattices confirm the existence of the strain-generated electric fields. Small band-gap semiconductors are useful for making intrinsic long wavelength infrared detectors. Arbitrarily small band gaps can be reached in the type II superlattice InAs/GaSb. However, for band gaps less than 0.1 eV, the layer thicknesses are large and the overlap of electron and hole wavefunctions are small. Thus, the absorption coefficient is too small for useful infrared (IR) detection. Including In in the GaSb introduces strain in the InAs/GaInSb superlattice, which shifts the band edges so that small band gaps can be reached in thin-layer superlattices. Good absorption at long IR wavelengths is thus achieved.