Limitations of the intrinsic cut-off frequency to correctly quantify the speed of nanoscale transistors

TL;DR

This paper critically examines the limitations of using the intrinsic cut-off frequency ($f_T$) to measure the speed of nanoscale transistors, highlighting the inaccuracies of quasi-static estimations due to displacement currents.

Contribution

It establishes conditions under which the quasi-static approximation for $f_T$ fails and provides numerical analysis showing significant discrepancies in ballistic transistors.

Findings

Quasi-static $f_T$ can be overestimated by an order of magnitude.

In some cases, $f_T$ cannot be defined due to the gate current being smaller than the drain.

Displacement currents significantly affect the accuracy of $f_T$ measurements.

Abstract

The definition of the intrinsic cut-off frequency ($f_T$) based on the current gain equals to one (0 dB) is critically analyzed. A condition for the validity of the quasi-static estimation of $f_T$ is established in terms of the temporal variations of the electric charge and electric flux on the drain, source and gate terminals. Due to the displacement current, an electron traversing the channel length generates a current pulse of finite temporal width. For electron devices where the intrinsic delay time of the current after a transient perturbation is comparable to such width, the displacement currents cannot be neglected and the quasi-static approximation becomes inaccurate. We provide numerical results for some ballistic transistors where the estimation of $f_T$ under the quasi-static approximation can be one order of magnitude larger than predictions obtained from a time-dependent numerical simulations of the intrinsic delay time (including particle and displacement currents). In other ballistic transistors, we show that the gate current phasor can be smaller than the drain one at all frequencies, giving no finite value for $f_T$.