Uniformity optimization techniques for rapid thermal processing systems

This paper presents two efficient robust methods for uniformity optimization of rapid thermal processes. Both of these methods involve the reuse of empirical response surfaces linking zone powers to measured process data created on a baseline system. The first method uses fossilized gain matrices from the baseline system, while the second method involves customization of the baseline response surface for each system. The approaches use the response surfaces for iterative modification of zone powers to reduce the process nonuniformity on successively processed wafers. These methods are applied to the optimization of rapid thermal oxidation processes on several lamp-heated rapid thermal processing systems. Most of the uniformity improvement is obtained with the first two optimization runs; in some instances, the process is optimized to less than 1% 1-sigma nonuniformity with the use of just two wafers. Because the response surfaces from the baseline system can be reused for all similar systems, considerable savings in time and wafers are realized