Reliability-Aware Machine Learning Prediction for Multi-Cycle Long-Term PMOS NBTI Degradation in Robust Nanometer ULSI Digital Circuit Design

This paper presents a comprehensive framework to enhance the reliability of nanometer Ultra Large Scale Integrated (ULSI) digital circuits by employing advanced machine learning models to predict multi-cycle, long-term PMOS Negative Bias Temperature Instability (NBTI) degradation. The study utilizes the 32nm technology node, analyzing degradation trends at 300K and 400K using the HSPICE MOSRA tool. Machine learning techniques, including Support Vector Regression (SVR), Random Forest (RF), K-Nearest Neighbor (KNN), Artificial Neural Network (ANN), and Adaptive Neuro-Fuzzy Inference Systems (ANFIS), are employed on datasets comprising critical parameters such as supply voltage (VDD), threshold voltage (Vth), and oxide thickness (Toxe). Experimental results reveal that smaller nodes like 32nm experience up to 11.04% threshold voltage degradation at 400K after 10,000 cycles, compared to only 2.16% for the 250nm node at 300K. ANN models achieved a Mean Squared Error (MSE) as low as <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$1.32\times 10^{-9}$</tex> at 300K for the 32nm node, with a predictive accuracy exceeding 99%. Regression analyses demonstrate polynomial and power models achieving <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex> values of 1.00 in several scenarios, effectively capturing non-linear degradation dynamics.