ML-Driven Investigation of Non-Ideal effects in Multi-Gate MOSFET’s characteristics for Sub 10 nm Technologies

As transistor technologies continue to shrink to advanced nodes like 7nm, 5nm, and beyond, designers face increasing challenges due to various non ideal effects. Issues such as subthreshold leakage, gate leakage, parasitic ca pacitance, mobility degradation, and variations from the manufacturing pro cess make it difficult to accurately predict and optimize power in MOSFET based circuits. In this study, we focus on a specific type of multigate MOS FETarchitecture, such as FinFET or GAA-FET. The devices were calibrated using TCAD simulations along with physical validation to ensure accuracy. Standard power estimation methods often fail to account for non-ideal effects, limiting their accuracy at advanced nodes. This work proposes a compact hybrid machine learning model that improves power estimation by including these effects. With a hybrid DRC framework (Deep Neural Network, Ran dom Forest, CatBoost), the model was trained on simulation data generated by TCAD and executed within Visual Studio. To evaluate performance, the proposed hybrid model was created based on four other models, RF, DNN, A Neural Network (ANN), and a hybrid XGBoost model. The experimental data showed that the proposed hybrid had a maximum prediction accuracy of 96.31 %, which is a significant improvement over traditional methods. This level of performance demonstrates the capability of hybrid ML mod els to improve power estimation in the VLSI design flow, which can shorten development cycles and improve design dependability. The Random For est algorithm was selected for its robustness to overfitting and capacity to handle high-dimensional datasets. The combined architecture allows model generalization and interpretability.