Comparison of Single-Particle Monte Carlo Simulation with Measured Output Characteristics of an 0.1µm n-MOSFET

F M Bufler,A Schenk,Y Asahi,N Inada,W Fichtner,C Zechner

doi:10.1080/106551402100012327

Abstract

A comparison between non-selfconsistent single-particle Monte Carlo (MC) simulations and measurements of the output characteristics of an 0.1 µm n-MOSFET is presented. First the bulk MC model, which features a new simplified treatment of inelastic acoustic intravalley scattering, is validated by comparison with experimental literature data for mobilities and velocities. The dopant distribution of the MOSFET is obtained from a 2D process simulation, which is calibrated with SIMS and electrical measurements and fine-tuned by a comparison of the measured transfer characteristics in the subthreshold regime with a coupled Schro¨dinger drift-diffusion (DD) simulation. Then the quantum effect is replaced by a shift of the work function and the DD, hydrodynamic (HD) and MC models are adjusted to reproduce the measured drain current in the linear regime. The results of the three models in the non-linear regime are compared without further adjustment to the measured output characteristics. While good agreement is found for the MC model, the on-current is significantly overestimated by the HD model and underestimated by the DD model.

Highlights

As metal-oxide semiconductor field-effect transistors (MOSFETs) are scaled into the sub 0.1 mm regime, the question arises to what extent the saturation drain current increases and which simulation models are able to take the involved quasiballistic transport accurately into account [1]
The investigated device structure is a LDD n-MOSFET with a gate length of LG 1⁄4 0:12 mm and a metallurgical gate length of approximately Leff 1⁄4 0:085 mm: The non-uniform p-type channel doping is in the order of 1018 cm23 and the maximum doping in source and drain is about 3 £ 1020 cm23: The oxide thickness is 2.6 nm
The transport models have to agree at bulk level, i.e. especially the fielddependence of the drift velocity and the dopingdependence of the low-field mobility have to be the same. The former criterion is fulfilled via the CaugheyThomas parameterization of the velocity-field characteristics in the DD and HD models, while the latter is ensured by a doping-dependent adjustment of the Monte Carlo (MC) impurity scattering rate [15] to the measured mobility data [20] that are used in the classical device simulation [21]

Summary

Introduction

As metal-oxide semiconductor field-effect transistors (MOSFETs) are scaled into the sub 0.1 mm regime, the question arises to what extent the saturation drain current increases and which simulation models are able to take the involved quasiballistic transport accurately into account [1]. Realistic 0.1 mm MOSFETs involve non-uniform substrate doping and maximum doping levels above 3 £ 1020 cm: Apart from the lack of experimental verification, there remains some uncertainty to what extent the results from those structures can be extrapolated for realistic devices and, in particular, how realistic devices can be simulated. It is the aim of this paper to investigate the issue of quasiballistic transport and saturation drain current for a real state-of-the-art MOSFET. After adjusting the transport models in the linear regime to the measured drain current, the non-linear regime of the output characteristics is simulated without

Objectives

Methods

Results

Conclusion