(320h) Experimental and Computational Fluid Dynamics Study of a Confined Jet with Low Reynolds Number and High Jetting Velocity

Chemical vapor deposition (CVD) is a chemical process used to produce high-purity, high performance solid materials. CVD processes are typically conducted in a confined chamber.  Gases are injected into the chamber using one single jet or multiple jets. The deposition (thickness) profile is affected by the flow pattern inside the deposition chamber, which depends primarily on the jetting behavior of the injection jets.  For the particular CVD process of interest, the corresponding gas jets have unusually high velocity (100-1000 m/s) and high Mach number but relatively low Reynolds number (100-1000).  Whether the jet is in the laminar, transitional, or turbulent flow regime can have significant impact on the jetting behavior thus the flow pattern.  Generally speaking, the jetting energy (jet velocity) dissipates much faster in the turbulent flow regime than in the laminar flow regime.   Study of such jetting condition is rare and hard to find in the public domain.  Understanding the jetting behavior is critical for process improvement such as increasing throughput, improving deposition uniformity and reducing defects and failure rates.  This paper aims to describe the flow behavior of confined gas jets under atmospheric and low pressure process conditions.   Helium at the room temperature is used to mimic the low-Re and high-Ma jetting conditions in CVD reactor.  In order to understand the jetting behavior in CVD reactors, a combined experimental and computational approach is adopted.   This is a crucial step for model validation as well as providing insights to the process.

     Laser light Mie scattering is utilized to study the overall flow transition behavior for the confined gas jet configuration.  Gas phase particle image velocimetry (PIV) is then utilized for further illustration of the local flow behavior of the confined jet.  The flow transition behavior observed in the experiment is then studied using computational fluid dynamics (CFD).  Several popular RANS turbulence models are compared including the k-ε, k-ω SST Transitional models and the large eddy simulation (LES) model.  The results suggest that with proper definition of the inlet flow boundary conditions, LES model is able to capture the flow transition behavior well while k-ε and k-ω SST model fail to do so.  Compared with k-ε model, k-ω SST model predicts slower velocity decay in the initial stage and faster velocity decay in the later stage for the transitional flow.