Volume 15 Number 2	March 1999

Scientific Simulations Gas Phase Nucleation Multi-Component/ Multi-Phase Materials Estimating Hospital Quality Future Symposium Colloquium Series Special Seminars Visitors Supercomputing '98 Research Reports ontaminant particles are currently the major source of yield loss in the manufacture of computer chips. Particles that land on a wafer during manufacture can interrupt an electrical current path or destroy the precise planarity required for multilevel circuits. The source of these particles is not ambient dust--this is controlled effectively by clean rooms--but the actual process through which the chips are manufactured. Most particles smaller than about 100 nanometers are generated by gas phase nucleation, initiated by the same chemistries and conditions used to deposit thin films during circuit fabrication. As electronic devices become smaller, the critical size above which a particle causes 'killer' defects is also shrinking, with the critical size of these particles anticipated to reach fifty nanometers within the next few years. Numerical models of particle nucleation and growth in semiconductor process environments are being developed by Professors Steven Girshick, Uwe Kortshagen, and Peter McMurry in the Department of Mechanical Engineering at the University of Minnesota. Professors Girshick and McMurry are Associate Fellows of the Supercomputing Institute. These investigators are working with graduate students Sandeep Nijhawan, Song-Moon Suh, Milind Mahajan, and Upendra Bhandarkar in addition to a postdoctoral research associate, Mark Swihart. This work is part of a larger effort including experimental studies that involve Professor Stephen Campbell of the Electrical and Computer Engineering Department at the University of Minnesota. The research is supported by grants from the Semiconductor Research Corporation, Advanced Silicon Materials, Inc., the National Science Foundation, and the Supercomputing Institute. Developing these models is a challenging problem involving chemically reacting flow and/or plasmas, chemical clustering, and aerosol dynamics. Recent work has focussed on nucleation in silane, one of the most widely used process gases. For example, the figure below shows simulation results for the dominant reaction paths to particle nucleation during thermal decomposition of silane under conditions typical of the production of high-purity polysilicon rods. The silicon hydrides and reactions represent only a small fraction of the total reaction mechanism considered. The ability to conduct these simulations required substantial new development of the database of thermochemical and kinetic properties of silicon hydrides. Predicted paths to particle nucleation in silane. Numbers next to the arrows indicate net reaction rates in units of 10-15 mol cm-3 s-1. These clustering models are coupled to aerosol dynamics models, which allow predictions of particle growth by surface reactions, coagulation, and transport. Simulations have been conducted to study the effects of temperature, pressure, silane concentration, and carrier gas. Comparisons of these simulations with experimental results are quite encouraging. The clustering mechanism for silane is being extended to the more complicated case of plasma-activated chemistry. In addition, more complex chemistries are being considered, such as the silane-oxygen system used for deposition of silicon oxide dielectric films.

HOME | BULLETINS | CONTACT US | PREVIOUS ARTICLE | NEXT ARTICLE | THIS ISSUE