Algorithm prediction of lifetime under high-power microwave based on T-type field plate HEMTs

Area of Science:

• Semiconductor Device Physics
• Machine Learning Applications
• Reliability Engineering

Background:

• High electron mobility transistors (HEMT) are crucial for high-power microwave (HPM) applications.
• Predicting HEMT device lifetime under HPM stress is vital for system reliability.
• Traditional methods often lack accuracy and efficiency in lifetime prediction.

Purpose of the Study:

• To develop a deep learning model for predicting HEMT device lifetime under HPM stress.
• To evaluate the performance of the deep learning model against traditional machine learning techniques.
• To establish a data-driven approach for assessing HPM reliability in HEMT devices.

Main Methods:

• Construction of a deep learning algorithm network model.
• Simulation of HPM stress effects on HEMT using TCAD technology.
• Comparative analysis with Support Vector Machine, Decision Tree, K-NN, Ridge, and Linear Regression.

Main Results:

• The deep learning model achieved a relative prediction error percentage below 15%, with most values under 5%.
• The deep learning algorithm demonstrated superior performance compared to five traditional machine learning methods.
• The model provides a data-driven tool for understanding HPM reliability and device lifetime.

Conclusions:

• Deep learning offers a highly accurate and efficient method for predicting HEMT device lifetime under HPM stress.
• The developed model can assist researchers in quickly obtaining device lifetime data and aid in device design.
• This approach signifies a broad future prospect for deep learning in semiconductor device reliability prediction.