Guard Ring Breakdown Voltage Prediction Using Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for simulating and optimizing the performance and reliability of silicon carbide (SiC) power devices with multiple floating guard rings are time-consuming and lack thorough evaluation of reliability, despite the potential of machine learning (ML) integration with technology computer-aided design (TCAD).
Innovation Solution
A reinforcement learning (RL) model is trained to predict the breakdown voltage (BV) of semiconductor devices with guard rings by maximizing a reward function based on manufacturing parameters, using a training dataset to learn the dependence of implantation dose and energy on BV.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the number and width of guard rings are increased to improve reliability, then the breakdown voltage prediction accuracy improves, but the wafer area occupied increases
Solution Approach 1:
The patent applies parameter changes by optimizing the implantation dose and energy of guard rings as key parameters. By adjusting these parameters within specific ranges (dose: 1×10^12 to 1×10^14 atoms/cm², energy: 100 to 1000 keV), the model achieves accurate breakdown voltage prediction without requiring increased guard ring quantity or width, thus maintaining compact wafer area while improving reliability assessment accuracy
2Reliability
If TCAD simulation is used to evaluate device performance, then the reliability assessment is thorough, but the design process time increases significantly
Solution Approach 1:
The patent implements preliminary action by pre-training the reinforcement learning model using TCAD simulation data to establish the relationship between guard ring parameters and breakdown voltage. Once trained, the model can rapidly predict breakdown voltage without requiring repeated TCAD simulations during the design process. This preliminary training phase captures the complex physical relationships, enabling fast subsequent predictions that reduce design cycle time while maintaining thorough reliability assessment
Solution Approach 2:
The patent applies copying by creating a simplified computational model (reinforcement learning model) that replicates the behavior of the complex TCAD simulation system. The RL model learns to copy the TCAD simulation's ability to predict breakdown voltage by training on its output data, then uses this copied knowledge for rapid predictions without invoking the full TCAD simulation process, thereby reducing computation time while preserving assessment accuracy
Data Source
AI summary
A method is used for training a reinforcement learning (RL) model to predict a breakdown voltage (BV) of a semiconductor device with a guard ring. The method comprises determining a set of structural parameters of the semiconductor device; preparing a training dataset formed by a plurality of manufacturing parameters of the semiconductor device, wherein the plurality of manufacturing parameters comprise a dose concentration and at least one dose energy of implanting a guard ring (GR) on the semiconductor device; and training the RL model using the training dataset by maximizing a reward function calculated based on a between a predicted BV value generated by the RL model and a target BV value corresponding to the plurality of manufacturing parameters.