Semi-Classical Monte Carlo Simulation of Contact Geometry, Orientation, and Ideality on Nano-scale Si and III-V n-channel FinFETs in the Quasi-Ballistic Limit

Abstract

The effects of contact geometry and ideality on InGaAs and Si nano-scale n-channel FinFET performance are studied using a quantum-corrected semi-classical Monte Carlo method. Illustrative end, saddle/slot, and raised source/drain contacts were modeled, and with ideal transmissivity and reduced transmissivity more consistent with experimental contact resistivities. Far-from-equilibrium degenerate statistics, quantum-confinement effects on carrier distributions in real-space and among energy valleys, quasi-ballistic transport inaccessible through drift-diffusion and hydrodynamic simulations, and scattering mechanisms and contact geometries not readily accessible through non-equilibrium Green's function simulation are addressed. Silicon $\langle \hbox{110} \rangle$ channel devices, Si $\langle \hbox{100} \rangle$ channel devices, multi-valley (MV) InGaAs devices with conventionally-reported energy valley offsets, and idealized $\Gamma$-valley only $\left( \Gamma \right)$ InGaAs devices are modeled. Simulated silicon devices exhibited relatively limited degradation in performance due to non-ideal contact transmissivities, more limited sensitivity to contact geometry with non-ideal contact transmissivities, and some contact-related advantage for Si $\langle \hbox{110} \rangle$ channel devices. In contrast, simulated InGaAs devices were highly sensitive to contact geometry and ideality and the peripheral valley's energy offset. It is illustrative of this latter sensitivity that simulated $\Gamma$-InGaAs device outperformed all others by a factor of two or more in terms of peak transconductance with perfectly transmitting reference end contacts, while silicon devices outperformed $\Gamma$-InGaAs for all contact geometries with non-ideal transmissivities, and MV-InGaAs devices performed the poorest under all simulation scenarios.