Presentation Information
[8p-B11-14]Efficient Optimization of Oxide Fill Process Conditions Using Topography Simulation and Robust Bayesian Optimization
〇Takuyo Nakayama1, Satoru Yokota1 (1.Kioxia Frontier Tech. R&D)
Keywords:
stable process condition search,oxide trench fill,Bayesian Optimization
In plasma processing for memory fabrication, a key challenge is that the processing profile is difficult to stabilize due to variations in the spatial distribution of reactive species in the plasma. To efficiently search for stable process conditions, we developed a robust Bayesian optimization method that accounts for such variations. We verified its effectiveness through shape simulations for oxide film fill in trench structures by plasma CVD. The results showed that while the optimal conditions obtained by conventional Bayesian optimization led to voids inside the trench, no such voids were observed under the conditions obtained by robust Bayesian optimization, demonstrating that stable conditions can be efficiently explored.
