Reducing Ultra-High-Purity (UHP) Gas Consumption by Characterization of Trace Contaminant Kinetic and Transport Behavior in UHP Fabrication Environments

摘要

Trends show that the fraction of the world's population with electronic devices using modern integrated circuits is increasing at a rapid rate. To meet consumer demands: less expensive, faster, and smaller electronics; while still making a profit, manufacturers must shrink transistor dimensions while increasing the number of transistors per integrated circuit; a trend predicted by Gorden E. Moore more than 44 years prior. As CMOS transistors scale down in size, new techniques such as atomic-layer deposition (ALD) are used to grow features one atomic layer at a time. ALD and other manufacturing processes are requiring increasingly stringent purities of process gases and liquids in order to minimize circuit killing defects which reduces yield and drives up manufacturing cost. Circuit killing defects caused by impurity incursions into UHP gas distribution system can come from a variety of sources and one of the impurity transport mechanisms investigated was back diffusion; the transport of impurities against convective flow. Once impurity incursions transpire, entire production lines are shut down and purging with UHP gas is initiated; a process that can take months thus resulting in tens of millions of dollars in lost revenue and substantial environment, safety, and health (ESH) impacts associated with high purge gas consumption. A combination of experimental investigation and process simulation was used to analyze the effect of various operational parameters on impurity back diffusion into UHP gas distribution systems. Advanced and highly sensitive analytical equipment, such as the Tiger Optics MTO 1000 H2O cavity ring-down spectrometer (CRDS), was used in experiments to measure real time back diffusing moisture concentrations exiting an electro-polished stainless-steel (EPSS) UHP distribution pipe. Design and operating parameters; main and lateral flow rates, system pressure, restrictive flow orifice (RFO) aperture size, and lateral length were changed to impact the extent of back diffusing impurities from a venting lateral. The process model developed in this work was validated by comparing its predictions with data from the experiment test bed. The process model includes convection, molecular diffusion in the bulk, surface diffusion, boundary layer transport, and all modes of dispersion; applicable in both laminar and turbulent flow regimes. Fluid dynamic properties were directly measured or were obtained by solving Navier-Stokes and continuity equations. Surface diffusion as well as convection and dispersion in the bulk fluid played a strong role in the transport of moisture from vents and lateral branches into the main line. In this analysis, a dimensionless number (Peclet Number) was derived and applied as the key indicator of the relative significance of various transport mechanisms in moisture back-diffusion. Guidelines and critical values of Peclet number were identified for assuring the operating conditions meet the purity requirements at the point of use while minimizing UHP gas usage. These guidelines allowed the determination of lateral lengths, lateral diameters, flow rates, and restrictive flow device configurations to minimize contamination and UHP gas consumption. Once a distribution system is contaminated, a significant amount of purge time is required to recover the system background due to the strong interactions between moisture molecules and the inner surfaces of the components in a gas distribution system. Because of the very high cost of UHP gases and factory downtime, it is critical for high-volume semiconductor manufacturers to reduce purge gas usage as well as purge time during the dry-down process. The removal of moisture contamination in UHP gas distribution systems was approached by using a novel technique dubbed pressure cyclic purge (PCP). EPSS piping was contaminated with moisture, from a controlled source, and then purged using a conventional purge technique or a PCP technique. Moisture removal rates and overall moisture removal was determined by measuring gas phase moisture concentration in real time via a CRDS moisture analyzer. When compared to conventional purge, PCP reduced the time required and purge gas needed to clean the UHP gas distribution systems. However, results indicate that indiscriminately initiating PCP can have less than ideal or even detrimental results. An investigation of purge techniques on the removal of gas phase, chemisorbed, and physisorbed moisture, coupled with the model predictions, led to the testing of hybrid PCP. The hybrid PCP approach proved to be the most adaptable purge technique and was used in next phase of testing and modeling. Experiments and modeling progressed to include testing the effectiveness of hybrid PCP in systems with laterals; more specifically, laterals that are "dead volumes" and results show that hybrid PCP becomes more purge time and purge gas efficient in systems with increasing number and size of dead volumes. The process model was used as a dry-down optimization tool requiring inputs of; geometry and size, temperature, starting contamination level, pressure swing limits of inline equipment, target cleanliness, and optimization goals; such as, minimizing pure time, minimizing purge gas usage, or minimizing total dry-down cost.