Over 90% of traffic accidents are caused by human error. Therefore, the realization of autonomous driving could save countless lives and drastically reduce the associated financial expenses. Moreover, the collective behavior of self-driving cars would avoid traffic jams and thus reduce fuel consumption and greenhouse gas emissions. The majority of concepts is based on Light Detection And Ranging (LiDAR), which is the most precise method to measure distances. Matched to the 95% of commercial LiDAR systems based on laser wavelengths of mostly 905nm, siliconbased photo sensors are used. Avalanche photo diodes (APD) are the only sensor solution in mass production [6]. Due to an internal multiplication mechanism based on impact ionization, high signal-noise-ratios (SNR) are achieved and provide the required resolution of low signals from more than 100m distant targets. Currently none of the LiDAR technologies meet the reliability requirements of the automotive industry concerning the aging of installed components. Consequently, autonomous driving cannot yet be realized for public use. Very little is known about the aging of APDs in general and nothing at all in the field of automotive LiDAR. In order to provide novel insights into APD aging that help designers to achieve more robust sensors and thus to enable a step closer to the realization of autonomous driving, it was the aim of this thesis prepared in the industrial environment to reveal the underlying physical aging mechanisms and their effects on the function of APDs in automotive LiDAR application. At first, a novel APD degradation model was developed encompassing a wide range of processes, treating numerous fundamental aspects of negative oxide charge generation and Si:SiO2 interface trap generation. So far, no model is known covering the kinetics of APD degradation comprehensively in such deep detail. Due to the feedback between degradation phenomena and sensor internal fields and currents, a coupled problem arose. It was tackled by a sophisticated numerical iteration approach which was tailor-made and solved this problem self-consistently in a tandem procedure combining the simulation of sensor degradation and the Silvaco Atlas device simulator. This led to novel insights into the APD degradation behavior. The generation of negative oxide charges was identified to cause a drift of the impact ionization rate in the sensor edge. The generation of interface traps promotes the accumulation of negative oxide charges by their supply of thermally generated dark current. In this way, degradation is about 14% faster. In order to reflect not only the causal relations of APD degradation, the model was calibrated on experimental degradation data. With the calibrated degradation model and its self-consistent simulation approach an elaborated powerful tool was available. Stress experiments have been performed on test sensors under a variation of operation conditions and on APDs. APDs of the studied design are currently tested and installed in automotive LiDAR modules. The entire set of experimental results found its complete physical interpretation in conjunction with the degradation model which achieved an excellent agreement. Thereby, numerous novel insights were revealed: The extent of degradation is induced by the properties of the sensors oxide layer. The degradation pace increases with temperature, voltage and intensity of illumination whereas the impact of temperature is particularly strong due to the significant participation of the dark current during degradation. The oxygen vacancy was proven to be the dominant trap in the oxide layer of the studied sensors. An empirical distribution of individual sensor properties was achieved. In some cases, the impact ionization rate in the sensor edge increased which indicates a major problem, as noise increases when the generation- recombination processes in the sensor become more pronounced during degradation. In order to estimate the impact of the degradation induced increase of noise on the LiDAR application, the empirical distribution of individual sensor properties was extrapolated to the tail where sensors are very prone to degradation. Furthermore, the available noise models were extended to cover the effect of degradation. Application of the calibrated APD degradation model revealed, that the APD noise is highly effected and even triples during aging. The origin was exclusively assigned to the edge contribution. There, the avalanche breakdown of the edge dark current caused by degradation is the main initiator. Consequently, for the first time ever, the signal-noise-ratio (SNR) degradation mode of APDs in LiDAR application was identified. During degradation, the SNR of small signals from 100m distant objects degrades to a value below 1, where even theoretically a resolution is impossible. Finally, the picture of APD degradation was completed by the estimation of lifetime. In the case of the most severe conditions in LiDAR operation, it amounts to only 1000 h, which falls much below the requirements of the automotive industry of several decades.

No description available.