Short channel effects in graphene-based field effect transistors targeting radio-frequency applications

TL;DR

This paper presents a comprehensive model for short channel effects in graphene FETs, highlighting how electrostatics and velocity saturation influence high-frequency performance, with predictions of THz operation at nanoscale channels.

Contribution

The work introduces a self-consistent 2D electrostatic and transport model for GFETs, capturing short channel effects and predicting high-frequency limits with experimental validation.

Findings

High maximum oscillation frequencies above 1 THz for channels below 100 nm.

Short channel effects degrade output conductance and cutoff frequency scaling.

Minimizing gate series resistance is crucial for THz performance.

Abstract

Channel length scaling in graphene field effect transistors (GFETs) is key in the pursuit of higher performance in radio frequency electronics for both rigid and flexible substrates. Although two-dimensional (2D) materials provide a superior immunity to Short Channel Effects (SCEs) than bulk materials, they could dominate in scaled GFETs. In this work, we have developed a model that calculates electron and hole transport along the graphene channel in a drift-diffusion basis, while considering the 2D electrostatics. Our model obtains the self-consistent solution of the 2D Poisson's equation coupled to the current continuity equation, the latter embedding an appropriate model for drift velocity saturation. We have studied the role played by the electrostatics and the velocity saturation in GFETs with short channel lengths L. Severe scaling results in a high degradation of GFET output conductance. The extrinsic cutoff frequency follows a 1/L^n scaling trend, where the index n fulfills n < 2. The case n = 2 corresponds to long-channel GFETs with low source/drain series resistance, that is, devices where the channel resistance is controlling the drain current. For high series resistance, n decreases down to n= 1, and it degrades to values of n < 1 because of the SCEs, especially at high drain bias. The model predicts high maximum oscillation frequencies above 1 THz for channel lengths below 100 nm, but, in order to obtain these frequencies, it is very important to minimize the gate series resistance. The model shows very good agreement with experimental current voltage curves obtained from short channel GFETs and also reproduces negative differential resistance, which is due to a reduction of diffusion current.