A genetic algorithm approach to manage ion implantation processes in wafer fabrication

The management of ion implantation processes is one of several challenging problems in scheduling wafer fabrication facilities. A complicating factor is the fact that there are sequence dependent set-ups (e.g. species changes). Because of the set-ups, it is sometimes desirable to leave an implanter idle (if another lot requiring this species will arrive soon) rather than to change the set-up. We study the use of a genetic algorithm (GA) to assign the jobs to machines where the First In First Out (FIFO) dispatching rule is used to schedule the individual machines. This approach is compared to the use of a commonly used dispatching policy-set-up avoidance. The parameters of the genetic algorithm (population size, crossover probability, and mutation probability) are analysed using response surface techniques to find combinations that allow the algorithm to determine a relatively good solution in a short CPU time.