CFD Modelling and Sensitivity-Guided Design of Silicon Filament CVD Reactors

TL;DR

This study uses CFD modeling and sensitivity analysis to understand and optimize silicon filament CVD reactors, revealing how geometry, flow, and temperature regimes influence film growth and uniformity.

Contribution

It introduces a validated 3D CFD model that elucidates the interplay of transport phenomena and reaction kinetics in filament CVD, guiding reactor design and operation.

Findings

Filament geometry and thermal gradients define growth regimes.

Reducing filament diameter can triple growth rates in high-temperature regimes.

Flow and thermal asymmetries significantly impact film uniformity.

Abstract

Filament-based chemical vapor deposition (CVD) for silicon (Si) coatings is often treated as a straightforward adaptation of planar deposition, just with a cylindrical substrate. But this overlooks a fundamental shift in how transport phenomena and reaction kinetics interact. In filament CVD, the filament is not just a substrate; it is the dominant heat source and flow disruptor all in one. In this work, we ask: What really governs Si film growth on filaments? Using a validated three-dimensional computational fluid dynamics (CFD) model, we show that filament geometry, thermal gradients, and flow-induced buoyancy do not merely affect uniformity -- they define the very regimes (reaction-limited, transition, diffusion-limited) in which deposition occurs. Our model, validated against three independent experimental studies (R-squared = 0.969), reveals how temperature, flow rate, and reactor layout fundamentally shift the growth dynamics. We show that reducing filament diameter from 2 mm to 500 microns does not just increase local surface area -- it can triple growth rates (up to approximately 39 microns per minute), but only in carefully tuned high-temperature regimes. Similarly, multi-filament setups do not simply scale deposition, they reshape flow fields and compound thermal asymmetries, sometimes improving uniformity, other times worsening it. These effects are not evident without modeling the interplay between transport and kinetics, something experiments alone cannot resolve. To bridge physical understanding with design actionability, we apply global sensitivity analysis via Polynomial Chaos Expansion (PCE) and Sobol indices. These reveal how in reaction-limited regimes, temperature dominates; in diffusion-limited regimes, reactant transport takes over.