IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 44, NO. 4, APRIL 1997 535 Effect of Doping on the Reliability of GaAs Multiple Quantum Well Avalanche Photodiodes Ilgu Yun, Student Member, IEEE, Hicham M. Menkara, Yang Wang, Ismail H. Oguzman, Student Member, IEEE, Jan Kolnik, Kevin F. Brennan, Gary S. May, Member, IEEE, Christopher J. Summers, and Brent K. Wagner Abstract— The effect of various doping methods on the reliability of gallium arsenide/aluminum gallium arsenide (GaAs/AlGaAs) multiple quantum well (MQW) avalanche photodiode (APD) structures fabricated by molecular beam epitaxy is investigated. Reliability is examined by accelerated life tests by monitoring dark current and breakdown voltage. Median device lifetime and the activation energy of the degradation mechanism are computed for undoped, doped-barrier, and doped-well APD structures. Lifetimes for each device structure are examined via a statistically designed experiment. Analysis of variance (ANOVA) shows that dark current is affected primarily by device diameter, temperature and stressing time, and breakdown voltage depends on the diameter, stressing time, and APD type. It is concluded that the undoped APD has the highest reliability, followed by the doped-well and doped-barrier devices, respectively. To determine the source of the degradation mechanism for each device structure, failure analysis using the electron-beam induced current method is performed. This analysis reveals some degree of device degradation caused by ionic impurities in the passivation layer, and energy-dispersive spectrometry subsequently verifies the presence of ionic sodium as the primary contaminant. However, since all device structures are similarly passivated, sodium contamination alone does not account for the observed variation between the differently doped APD’s. This effect is explained by dopant migration during stressing, which is verified by free carrier concentration measurements using the capacitance–voltage ( – ) technique. I. INTRODUCTION G ALLIUM arsenide/aluminum gallium arsenide (GaAs/AlGaAs) multiple quantum well (MQW) avalanche photodiodes (APD’s) are of interest as an ultra- low-noise image capture mechanism for high-definition systems. In this application, the image capture stage must have sufficient optical gain to enable very sensitive light detection, but at the same time, the gain derived during detection must not contribute additional noise. Various APD structures, including doped-barrier, doped-well, and undoped devices have been fabricated, and these structures are all being considered as candidates for the high-definition system imaging application. The effect of the different doping techniques on device performance is critical. An investigation Manuscript received October 7, 1996. The review of this paper was arranged by Editor P. K. Bhattacharya. This work was supported by NASA under Contract NAGW-2753. I. Yun, Y. Wang, I. H. Oguzman, J. Kolnik, K. F. Brennan, and G. S. May are with the School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0250 USA. H. M. Menkara, C. J. Summers, and B. K. Wagner are with the Georgia Tech Research Institute, Georgia Institute of Technology, Atlanta, GA 30332- 0250 USA. Publisher Item Identifier S 0018-9383(97)02358-7. of the relative advantages and disadvantages of each device structure as pertaining to long-term, low-noise performance is therefore warranted. APD performance is enhanced by minimizing the excess noise generated by carrier multiplication. This excess noise is reduced when the ratio of the ionization rate of electrons to that of holes (or vice-versa) is large. Chin et al. first proposed a means of artificially enhancing the ratio of electron-to-hole ionization coefficients through use of a MQW structure in the GaAs/AlGaAs material system [1]. Later, Brennan analyzed the use of the doped quantum well APD as a photomultiplier [2], and Aristin et al. evaluated various MQW APD structures, including the undoped, doped-barrier, and doped-well devices [3]. These new structures enable very low noise and high-speed APD’s. However, the noise performance of MQW APD’s is limited by dark currents due to both thermionic emission and field-assisted tunneling of carriers out of quantum wells. Therefore, increased dark current can severely limit the long- term reliability of these devices. Reliability assessment of avalanche photodiodes has been performed by several authors. Sudo et al. conducted acceler- ated life tests on germanium APD’s to measure their failure rates under practical use conditions [4]. This author also used bias temperature tests and the light-beam induced current method to evaluate lifetime and analyze the failure modes of InP/InGaAs APD’s [5], [6]. Kuhara likewise investigated the long-term reliability of InGaAs/InP photodiodes passivated with polyimide films [7], and Bauer and Trommer performed a similar investigation on devices passivated with silicon nitride [8]. Finally, Skrimshire et al. performed accelerated life tests on both mesa and planar InGaAs photodiodes for comparison purposes [9]. In this paper, accelerated life testing of undoped, doped- barrier, and doped-well APD device structures has been con- ducted with the objective of estimating long-term device reliability. Since an increase in dark current results in a reduc- tion of the APD signal-to-noise ratio (SNR) and breakdown voltage determines the operational voltage range of the device, these two parameters represent the most sensitive indicators of the characteristic degradation in these devices. Thus, dark current and breakdown voltage were the parameters monitored in this study. Degradation in these parameters was investigated via high temperature storage tests and accelerated life tests, and the results of these tests were used to estimate device lifetime by assuming an Arrhenius-type temperature dependence [10]. Using the median device lifetime and its standard deviation as 0018–9383/97$10.00  1997 IEEE