日志 Ingrid Schwarz

An Empirical Composition Dependent Model of Dopant Diffusion Coefficients in Si, Si1-x Gex and Ge Material Systems

Previously published fast empirical models for diffusion coefficients in silicon-germanium (Si1-x Gex) [1][2] were not applicable to high germanium content x≥0.5 and hence did not properly extend towards germanium. For some dopants, diffusion coefficients become very small and hence this model cannot be applied to devices containing silicon-germanium with high germanium content or devices containing silicon, silicon-germanium and germanium

TCAD Simulation of Leakage Through Threading Dislocations in GaN-based pn-diodes

Gallium nitride (GaN)-based devices for power electronics show superior performance in comparison to silicon carbide and silicon-based devices [1]–[3]. The development of vertical devices, like pn-diodes and power HEMTs results in higher power density and voltage handling. One of the key parameters of this technology is the dislocation density. This is lower in free-standing GaN-on-GaN epitaxy than in heteroepitaxial GaN growth on different substrates like SiC or Si, but still has a density of 104-106 cm-2 [4]. The diode reverse leakage seems to be related to the dislocation density, and it can be modelled with a Poole-Frenkel or a hopping conduction mechanism [5]. The Poole-Frenkel model is already implemented in the trap-assisted tunnelling model in Silvaco Atlas [6]. For the leakage in threading dislocations a variable-range hopping (VRH) model has been implemented in the simulator, based on Ref. [7].

Hints, Tips and Solutions – DeckBuild Remote VM Setup

This document describes how to setup a virtual environment suitable for the deckbuild remote mode.
The host system to consider shall be Windows (7 or 10). The goal is to use the windows version of the TCAD GUI tools TonyPlot and DeckBuild and to run any simulation (Victory, Athena, Clever) on a Linux Virtual Machine (VM).