Atomic Layer Deposition Optimization Using Convolutional Neural Networks

Abstract

Atomic layer deposition (ALD) is a chemical engineering process used to coat surfaces with a thin film. It is a versatile process able to deposit a wide range of films using different chemical reagents. When developing novel ALD processes, a technician must determine the dosing time of each reagent. To accelerate this development process, we trained convolutional neural networks to predict the reagent saturation times of novel ALD reactions given the reagent dosing times and film growth rates of example reactions. We generated two kinds of models. Single reaction models made predictions based on a single example ALD reaction. Multiple reaction models made predictions based on ten example reactions using the same reagents with different dosing times.