Using Genetic Algorithms with Local Search for Thin Film Metrology

Metrology Mark Land Cognitive Computer Science Research Group Computer Science Department University of California at San Diego La Jolla, CA 92093-0114 John J. SIDorowich 7150 Aptos View Road Aptos, CA 95003 Richard K. Belew Cognitive Computer Science Research Group Computer Science Department University of California at San Diego La Jolla, CA 92093-0114 Abstract Nondestructively determining the essential parameters that describe the structure of a semiconductor wafer is a challenging inverse problem. We describe use of an optical inspection technology and show that it can be used e ectively in conjunction with genetic algorithms (GAs) and local optimization methods. We also use this concrete application to investigate GA/local search hybrids, compare them to simulated annealing, and investigate the value of the recombination operator relative to the \random crossover" variant suggested by T. Jones.