Numerical Simulation on Electronic Performance of MOS Diodes based on Artificial Intelligence

In the realm of electronic engineering, MOS diodes are pivotal for the functionality of various electronic devices, typically analyzed through costly and time-intensive physical experiments. This study employs a Convolutional Neural Network (CNN) to efficiently simulate and predict the I- V characteristics of MOS diodes, thus enhancing research and development productivity while reducing costs. We collected a comprehensive dataset from semiconductor manufacturing labs, covering a wide range of operating conditions to train and test the CNN model. The model demonstrated exceptional accuracy, with a correlation coefficient reaching 0.98 and robust performance across different noise levels — maintaining a correlation coefficient above 0.85 even with 20% noise. These findings underline the CNN model's capacity not only to match but also to surpass traditional physical and empirical approaches, providing a more efficient, accurate tool for semiconductor device analysis. The implications of this research extend beyond MOS diodes, suggesting potential applications across various semiconductor devices, promising to significantly advance device design and optimization strategies.