Abstract
The physics-of-failure (PoF) technique is a practical approach to evaluate the reliability of semiconductor devices. However, the PoF approaches are usually insufficient in dealing with multi-mechanism failure and fitting the Monte Carlo (MC) sampling data. In our study, we propose an improved reliability evaluation method based on PoF technique and maximum entropy (MaxEnt) principle. The PoF models are used to generate time-to-failure samples of the failure mechanisms. Cumulative damage rules and competing failure rules are adopted to deal with multi-point and multi-mechanism failure and generate lifetime samples of the device. And the lifetime samples are fitted by MaxEnt distributions through the proposed fitting algorithm. The numerical examples given in the paper indicate that the MaxEnt distributions can describe the samples well and have a competitive advantage in dealing with multi-peak samples. A case study about a semiconductor device with multi-mechanism failure is presented to explain the workflow of the proposed reliability evaluation approach. The results show that the proposed MaxEnt distributions can yield reliable reliability evaluation results compared with Weibull and Lognormal distributions in the multi-mechanism failure process.
Highlights
The recent years have seen the rapid development of semiconductor devices in many fields such as 5G, Internet of things, artificial intelligence, etc
In this study, we proposed an improved reliability evaluation approach for semiconductor devices with multi-mechanism failure based on PoF technique and maximum entropy (MaxEnt) principle
The PoF technique utilizes PoF model to calculate TTF corresponding to the failure mechanism, and Monte Carlo (MC) sampling procedure, which considers the fluctuation of process and service conditions generate a batch of TTF sampling data
Summary
The recent years have seen the rapid development of semiconductor devices in many fields such as 5G, Internet of things, artificial intelligence, etc. PoF methods are recognized as physics-based approaches that consider failure mechanisms and processes to predict the lifetime of semiconductor devices. B. Wan et al.: Reliability Evaluation of Multi-Mechanism Failure for Semiconductor Devices mechanisms, such as negative-bias temperature instability (NBTI), hot carrier injection (HCI), time-dependent breakdown (TDDB) and electromigration (EM) [5]. Wan et al.: Reliability Evaluation of Multi-Mechanism Failure for Semiconductor Devices mechanisms, such as negative-bias temperature instability (NBTI), hot carrier injection (HCI), time-dependent breakdown (TDDB) and electromigration (EM) [5] For packaging failure such as the fatigue of solder joint, the damage caused by thermal cycling and vibration can be regarded as a cumulative form. PoF methods are practical in evaluating the reliability of semiconductor devices, but they do not go into details about the multi-mechanism failure process. E is Young’s modulus of solder; σu is the tensile strength of solder, n is the fatigue component depends on the material
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have