Machine-Learning-Assisted Buried-Window FET Sensors for High-Reliability and High-Sensitivity Applications

TL;DR

A new type of sensor using advanced transistor design and machine learning improves biosensing performance for high-sensitivity applications.

Contribution

The DBW-FET design with dual buried windows and machine-learning optimization enhances biosensor performance and CMOS compatibility.

Findings

The DBW-FET shows higher drain current and lower subthreshold swing than conventional FETs.

Machine learning optimization improved current sensitivity by 20–25% with low leakage.

The design is CMOS-compatible and suitable for nanoscale biosensing applications.

Abstract

This paper presents a novel Double Buried-Window Junctionless Field-Effect Transistor (DBW-FET) designed for high-sensitivity, label-free biosensing applications. The proposed device integrates two buried windows, one N-type and one P-type, beneath the active channel within the buried oxide layer, along with two nanocavities serving as biomolecular recognition sites. The dual buried windows form two depletion regions that enhance electrostatic coupling, suppress short-channel effects, and improve biomolecular sensitivity. Numerical simulations using Silvaco TCAD Atlas were performed to investigate device performance under various biomolecular binding conditions. Results show that the DBW-FET exhibits higher drain current, lower subthreshold swing, and improved sensitivity compared with a conventional junctionless FET (C-FET). Furthermore, a machine-learning-assisted optimization framework employing Gaussian Process Regression (GPR) and Bayesian Optimization (BO) was implemented to identify optimal buried window parameters. The optimized design achieved a 20–25% improvement in current sensitivity while maintaining low leakage. These findings demonstrate that the proposed DBW-FET offers a promising and Complementary Metal-Oxide-Semiconductor (CMOS)-compatible architecture for next-generation nanoscale biosensors.

Figures