Sensitivity Prediction in Triple-Material Dielectric-Stack Gate-All-Around (TM-DS-GAA) MOSFET Biosensors Using Random Forest Machine Learning

This work presents the prediction of sensitivity in a Triple-Material Dielectric-Stack Gate-All-Around (TM-DS-GAA) MOSFET using dielectric modulation approach. The variation in dielectric values is considered as the exposure of biomolecules. The dielectric stack comprises of low- and high-K materials as insulators. An analytical model of surface potential has been created. The variations in surface potential, threshold voltage and drain current are investigated for high-k materials such as TiO 2 , ZrO 2 and Al 2 O 3 . The proposed device’s sensitivity can be estimated by using threshold voltage and the On-current to Off-current ratio (I on /I off ) ratio variation.The Random Forest Machine Learning (RF-ML) method has been applied to predict the sensitivity. The performance of RF-ML technique has been explored with Mean Squared Error (MSE) and coefficient of determination (R 2 -score). The MSE of RF-ML is achieved as 0.087 and R 2 -score as 99.8%. It shows an excellent match between RF-ML and device simulation. Furthermore, the sensitivity of TM-DS-GAA MOSFET biosensor has been evaluated against the Dielectric-Modulated Junctionless (DM-JL) MOSFET which exhibits a notable improvement.The proposed device can be used in advanced biosensing applications to detect biomolecules.