Title:	CFD for CVD
DNr:	LiU-compute-2026-4
Project Type:	LiU
Principal Investigator:	Örjan Danielsson <orjan.danielsson@liu.se>
Affiliation:	Linköpings universitet
Duration:	2026-02-27 – 2026-07-01
Classification:	20306
Keywords:

Abstract

Chemical vapor deposition (CVD) is used in the production of semiconductor components, protective coatings for high temperature applications or hard coatings on machining tools. The CVD process involves several interacting physical phenomena: fluid flow, heat transfer and chemical reactions. Our research focuses on obtaining a detailed understanding of how these phenomena control the material quality and composition over the entire substrate. The knowledge is then used to improve and optimize the process conditions and reactor configurations for large-area substrates, faster deposition times, and supreme material quality. We are using Computational Fluid Dynamics (CFD) to study the CVD process. CFD is well suited to simulate mass and heat transfer in complex engineering applications. We can for example study how the turbulence intensity influence temperature fields and gradients in the reactor. It is also possible to add almost arbitrarily complex chemical models to the fluid flow. The main challenge here is the large differences in time scales between the chemical reactions and the fluid flow, which leads to numerically stiff equation systems. Solving these problems requires fine-detailed computational meshes and short time steps, and thus large computational times. In previous projects (LiU-2014-00036-10, LiU2015-00017-36, LiU2017-00089-14, and LiU-2018-28) it was shown that CFD simulations will give valuable insights into the CVD process, but that they also require a high degree of details to give useful results. So far, due to the heavy computational load, we have been forced to make several simplifications to our models. The aim of the present project is to make very detailed simulations of a CVD reactor to a level never seen before in the literature. We will do that by carrying out a tightly coupled high fidelity simulation campaign of the reactor, combining an industrial second order finite volume workflow in a commercial solver (ANSYS Fluent) with a higher order spectral element workflow in Neko to reach well resolved LES and DNS quality reference conditions. This will provide new understanding of the influence of upstream flow behavior on downstream gas composition and temperature profiles. It can also show the sensitivity of small geometrical reactor features to the process performance. The long-term goal is to construct a fully functional digital twin of a CVD reactor.