Machine learning based prediction of I-V and transconductance curves for 3D multichannel junctionless FinFET

In this study, we have proposed the utilization of machine learning techniques to predict the I-V characteristics and transconductance curve of a multichannel junctionless FinFET design. Various machine learning models were employed to predict the current-voltage (I-V) and transconductance (g(m)) curve by training the models with data generated from Technology Computer Aided Design (TCAD) simulations for the 3D multichannel junctionless FinFET. Specifically, three different machine learning models were constructed using the random forest (RF), linear regression (LR), and decision tree regression (DTR) algorithms. These models aimed to uncover hidden relationships and establish correlations between different physical parameters. This approach offers valuable insights into extracting several short channel effects parameters, such as threshold voltage (V-th), on-state current (I-on), off-state current (I-off), and subthreshold swing, from the trained dataset. The results indicate that both the random forest and decision tree regression models achieved a similar level of accuracy in predicting the I-DS-V-GS and g(m) curves when compared to the TCAD simulations for the proposed device structure. The RF, LR, and DTR models were evaluated using metrics such as root mean squared error (RMSE) and error rate. The RF-ML and DTR-ML models exhibited 99% accuracy with an error rate below 1%, even when trained with only 20% of the dataset (i.e., 255 training samples). This work primarily aims to demonstrate the integration of machine learning techniques with device simulation tools to advance technology for new device development.