Photo-luminescence and transmission electron microscope studies of low- and high-reliability AlGaAs/GaAs HBTs (original) (raw)

Experimental procedure for the evaluation of GaAs-based HBT's reliability

Microelectronics Journal, 2001

This work describes the implementation of an experimental procedure to evaluate the reliability of Heterojunction Bipolar Transistors (HBT) on a GaAs substrate. It is based on the separation of aging test accelerating factors applied on two test vehicles: HBT and Transmission Line Model (TLM) structures associated with emitter, base and collector layers. To identify the physical origin of the degradation mechanism, analysis techniques are used: EDX, SEM and TEM observations for which a new sample preparation method has been worked out. Three different technological fabrication processes, are investigated: AlGaAs/GaAs double-mesa HBT, GaInP/GaAs self-aligned HBT and GaInP/GaAs fully planar HBT. These investigations have revealed two major failure mechanisms: the degradation of SiN±GaAs interface correlated with the increase of emitter-to-base leakage current of HBT submitted to combined bias and temperature stresses; the detachment of Ge/Mo/W emitter ohmic contact related to the base and collector current decrease for high level injection in forward regime.

High reliability InGaP/GaAs HBT

IEEE Electron Device Letters, 1998

Excellent long term reliability InGaP/GaAs heterojunction bipolar transistors (HBT) grown by metalorganic chemical vapor deposition (MOCVD) are demonstrated. There were no device failures (T = 10 000 h) in a sample lot of ten devices (L = 6:4 m 2 20 m) under moderate current densities and high-temperature testing (Jc = 25 kA/cm 2 ; Vce = 2.0 V, Junction Temp =264 C). The dc current gain for large area devices (L = 75 m 275 m) at 1 kA/cm 2 at a base sheet resistance of 240 ohms/sq (4 2 10 19 cm 03 @ 700Å) was over 100. The dc current gain before reliability testing (L = 6:4 m 2 10 m) at 0.8 kA/cm 2 was 62. The dc current gain (0.8 kA/cm 2) decreased to 57 after 10 000 h of reliability testing. The devices showed an f T = 61 GHz and f max = 103 GHz. The reliability results are the highest ever achieved for InGaP/GaAs HBT and these results indicate the great potential of InGaP/GaAs HBT for numerous low-and high-frequency microwave circuit applications. The reliability improvements are probably due to the initial low base current at low current densities which result from the low surface recombination of InGaP and the high valence band discontinuity between InGaP and GaAs.

Breakdown behavior of GaAs/AlGaAs HBT's

IEEE Transactions on Electron Devices, 1989

Avalanche breakdown behavior at the collector junction of the GaAs/AIGaAs heterojunction bipolar transistor (HBT) has been studied. Junction breakdown characteristics displaying hard breakdown, soft breakdown, and negative resistance breakdown behavior were observed and are interpreted by analysis of localized microplasma effects, uniform microplasma-free behavior, and associated current gain measurements. Light emission from the collector-base junction of the GaAs/AIGaAs HBT was observed and was used to investigate breakdown uniformity. Using a simple punchthrough breakdown model, the theoretical breakdown curves at different collector doping concentrations and thicknesses were computed and found to be in agreement with maximum breakdown voltages measured from devices displaying the most uniform junction breakdown.

Role of Neutral Base Recombination in High Gain AlGaAs/GaAs HBT’s

Neutral base recombination is a limiting factor controlling the maximum gain of AlGaAs/GaAs HBT's with base sheet resistances between 100 and 350 = : In this work, we investigate five series of AlGaAs/GaAs HBT growths in which the base thickness was varied between 500 and 1600Å and the base doping level between 2.9 2 and 4.7 2 10 19 cm 03 : The dc current gain of large area devices (L = 75 m 2 75 m) varies by as much as a factor of two at high injection levels for a fixed base sheet resistance, depending on the growth optimization. One of these series (Series TA) has the highest current gains ever reported in this base sheet resistance range, with dc current gains over 225 (@ 200 A/cm 2 ) at a base sheet resistance of 330 = : A high dc current gain of 220 (@ 10 kA/cm 2 ) was also confirmed in small area devices (L = 8 m 28 m). High-frequency tests on a separate set of wafers grown under the same conditions indicate these high current gains can be achieved without compromising the RF characteristics: Both high and normal gain devices exhibit an ft 68 GHz and fmax 100 GHz. By fitting the base current as a sum of two components, one due to recombination in the neutral base and the other in the space charge region, we conclude that an improvement in the minority carrier lifetime is responsible for the observed increase in dc current gain. Moreover, we observe a thickness-dependent variation in the effective minority carrier lifetime as the gains increase, along with a nonlinear dependence of current gain on base doping. Both phenomena are discussed in terms of an increase in Auger and radiative recombination relative to Hall-Shockley-Read recombination in optimized samples.

Design study of AlGaAs/GaAs HBTs

IEEE Transactions on Electron Devices, 1990

The frequency performance of AlGaAs/GaAs heterojunction bipolar transistors (HBTs) having different layouts, doping profiles, and layer thicknesses was assessed using the BIPOLE computer program. The optimized design of HBTs was studied, and the high current performances of HBTs and polysilicon emitter transistors were compared. It is shown that no current crowding effect occurs at current densities less than 1×105 A/cm2 for the HBT with emitter stripe width SE<3 μm, and the HBT current-handling capability determined by the peak current-gain cutoff frequency is more than twice as large as that of the polysilicon emitter transistor. An optimized maximum oscillation frequency formula has been obtained for a typical process n-p-n AlGaAs/GaAs HBT having base doping of 1×10 19 cm-3

Analysis of the surface base current drift in GaAs HBT's

Microelectronics Reliability, 1996

A field-induced degradation mechanism responsible for the surface base current drift in GaAs HBT's is studied on the basis of accelerated life-tests : two stress conditions are applied to HBT's and associated TLM structures. An EDX analysis performed on the base ohmic contact confirms the migration of metallic species which are assessed to spread into the extrinsic base region of AIGaAs/GaAs HBT's. The self-passivated GalnP/GaAs technology is expected to improve the HBT reliability.

Heavily Zn-doped graded-base AlGaAs/GaAs HBTs grown by MOCVD

IEEE Transactions on Electron Devices, 1991

Heavily Zn-doped graded-base AlGaAslGaAs HBT's, which had base doping concentrations of 3.5 x 5.5 X lo", and 8.1 x lOI9 cm-', were fabricated using MOCVD-grown epitaxial wafers. The maximum measured current gain was 42 with the base doping concentration of as high as 8.1 x The current gain decreased drastically from 96 to 42 when doping concentration increased from 3.5 x 1019 to 8.1 x 10" cm-'. Assuming that the emitter efficiency is unity, the minority electron lifetime in the base is estimated by the current gain by consideration of drift in the graded base. The dependence of the electron lifetime in the graded base on Zn doping concentration was similar to that in the uniform base reported in a previous paper.

THE ROLE OF SUBSTRATE DISLOCATIONS IN CAUSING INFANT FAILURES IN HIGH COMPLEXITY InGaP/GaAs HBT ICs

2007

A novel MSI circuit and reliability test method has been developed to measure the probability of infant failure in InGaP/GaAs HBTs. This circuit enables simultaneous reliability testing of 200 HBTs per circuit, and has been used routinely to test 12,800 HBTs/wafer (and sometimes 100, 000HBTs/wafer) to measure the probability of infant failure Pqf on each wafer. We have applied this MSI reliability test method to more than 100 production HBT IC wafers and many experimental wafers. With these tests, we have identified a new HBT failure mode, for which a very small fraction of HBTs (Pqf = 10 -3 to 10 -4 ) fail by beta degradation much sooner than the main HBT population. Moreover, the probability of infant failure Pqf is proportional to the substrate EPD, with a constant of proportionality approximately equal to the emitter area. This observation is consistent with a model where any HBT whose e-b junction is pierced by a substrate dislocation will suffer beta degradation much sooner than its companions. This data has motivated Agilent to procure substrates with the lowest possible EPD for the manufacture of high complexity HBT ICs with sufficient reliability for use in Agilent instrument products.