Abstract

Photodiodes fabricated with an InGaAs absorbing region are widely used in optical receivers for the 1.3-1.6μm-wavelength region. We report here the first comprehensive reliability study of planar InGaAs photodiodes. These devices are fabricated using selective area diffusion to define the junction area: a SiN_x film seals the junction perimeter from the ambient. Accelerated testing using thermal and nonthermal stresses has been used to determine relevant failure mechanisms, failure statistics, activation energy, and procedures for eliminating early failures (infant mortality). Photodiodes fabricated with an InGaAs absorbing region are widely used in optical receivers for the 1.3-1.6 μm-wavelength region. We report here the first comprehensive reliability study of planar InGaAs photodiodes. These devices are fabricated using selective area diffusion to define the junction area: a SiN_x film seals the junction perimeter from the ambient. Accelerated testing using thermal and nonthermal stresses has been used to determine relevant failure mechanisms, failure statistics, activation energy, and procedures for eliminating early failures (infant mortality). Over 120 photodiodes were aged at — 10-V bias at various temperatures ranging from 85 to 250°C. In contrast to previous work on mesa diodes, there is no change in dark current /« or any other parameter, but after an aging time which may extend to several thousand hours, /a increases very rapidly resulting in a well-defined end of life. Failures are shown to follow lognormal statistics. Median lives are strongly temperature dependent, being characterized by an activation energy of ~1.1 eV. Device failure Is shown to be correlated to localized leakage paths at the junction perimeter. The same mode of failure is induced by operating devices at high reverse bias, the reciprocal of the medial life varying exponentially with reverse bias. These data are used to estimate median life at use conditions. The sudden failure characteristic is insidious because early failures cannot be predicted In advance from initial aging rates or by correlation with initial parameters. High reverse bias was found to be particularly useful in inducing /a changes In some (presumed to be weak or flawed) devices. By combining overvoltage with high temperatures, we have developed a purge which identifies, for removal, those devices which are likely to fail early in service. The effects of purges of increased severity will be discussed. The effectiveness of a purge can be corroborated by a subsequent life test. Figure 1 shows life test results which illustrate the dramatic reduction in early failures for purged devices. Details will be presented. TUG3 Table I. Characteristics of the Hlah N.A. Fibers Developed TUH3 Fig. 1. (a) GaAs Schottky barrier photo detector structure; (b) equivalent eiectricai circuit of photodetector. TUH3 Fig. 2. Microwave fiber-optic iink. TUH3 Fig. 3. (a) Microwave opticai link frequency response; (b) photodetector frequency response. (Soiid line is caicuiated response based on measured circuit parameters.)

© 1984 Optical Society of America