– spectives: Use of Statistical Approaches
Understanding ALD Through Modeling

In the case of atomic layer deposition (ALD), artificial intelligence does not appear to be a substitute for existing modeling approaches, but rather a bridge between already well-established approaches. ALD combines self-limiting surface chemistry, transport phenomena dependent on reactor geometry, and often multiple process objectives: growth rate, uniformity, conformality, selectivity, chemical purity, and layer stability. Conventional approaches therefore remain indispensable: DFT or AIMD for elementary chemistry; mechanistic models or kinetic Monte Carlo simulations for surface growth; and CFD for the reactor scale. However, directly linking these approaches remains costly, both in terms of computation time and parameterization effort. In this context, AI serves as a toolkit capable of accelerating certain computational building blocks, inferring hidden parameters from experimental...