Optimization study of a probe chuck for semiconductor wafers using genetic algorithm and deep reinforcement learnings

Abstract

The probe chuck is an inspection device assessing the thermal durability of semiconductor wafers in various temperature environments before shipping. It is most important to ensure that the temperature of the chuck upper surface, on which the wafers are placed, is uniform. This study presents an axisymmetric chuck model to improve surface temperature uniformity in both radial and circumferential directions. The local distribution of the flow path height in the axisymmetric chuck was adjusted to make the chuck upper surface with a constant wall heat flux to simultaneously become as uniform temperature as possible. Three optimization algorithms, namely the genetic algorithm (GA), deep q-network (DQN), and actor-critic (AC) were applied. The optimized shape of the flow pathway, improved temperature uniformity, pressure drop, and local heat transfer coefficient profile by three different optimization algorithms are presented in detail. As a result, the surface temperature difference was significantly reduced from 7.137 K in the existing spiral model to 0.682 K. The optimal axisymmetric chuck could reduce surface temperature differences up to 90 % compared with the conventional spiral chuck.