Optimization of cleaning process in semiconductor gas delivery system by computational fluid dynamics simulation

Abstract

Purging specialty gas delivery systems in semiconductor industries are essential for ensuring worker safety and complying with stringent environmental regulations. The cleaning process impacts production costs due to the required resources and time. This study introduced and evaluated various purging methods, combining vacuum, vacuum with purge, pressurization, and pressure-gradient purge steps. Considering the density and viscosity of various specialty gases used in semiconductor manufacturing, such as gaseous NH3, HCl, WF6, and SiCl4, were selected as the representative specialty gases for the system. Considering the utilization of the semiconductor industries, ultra-high purity nitrogen was selected to clean all components of the gas delivery system, including gas channels, filters, pressure regulator and diaphragm valves. Efficiency was measured by monitoring residual gas concentration changes during different purge-step configurations, the number of purge configuration, and step durations. The dynamic behavior of specialty gases was analyzed based on geometric factors and chemical properties. Findings indicated that cleaning dead-end zones was particularly challenging and impacting overall system efficiency. Frequent purge with a short step time proved more effective in removing residual chemicals than the present conventional vacuum purge process with a continuous N2 supply. Employing a pressure gradient between the inlet and outlet enhanced cleaning efficiency. These insights provide valuable guidelines for optimizing purging processes to reduce cleaning time and resource consumption in semiconductor manufacturing.