Machine Learning Approach for Prediction of Point Defect Effect in FinFET

As Fin Field Effect Transistor (FinFET) scales aggressively, even a single point defect becomes a source of performance variability. The point defect is inevitably introduced not only by process damage such as epitaxial growth and ion implantation but also by cosmic rays. Technology computer-aided design (TCAD) is able to simulate the characteristics of the device with the defect. In this work, a machine learning algorithm is tested if it can reproduce the TCAD results. The impact of point defect in bulk FinFET is used as test vehicle to validate the machine-learning algorithm. TCAD is used first to generate a massive number of current-voltage characteristics dataset. The TCAD dataset is then exclusively divided into groups for machine learning training, validation and test. The trained model provides high accuracy test results within 1 % error, showing the possibility to expedite failure analysis cycle via machine learning.

Automated summary

As FinFET scales aggressively, even a single point defect can cause performance variability. A machine learning algorithm is tested to replicate TCAD results, with the impact of a point defect in bulk FinFET used for validation. The trained model shows high accuracy test results, indicating potential for expediting failure analysis cycle through machine learning.