Advances in quantum computing algorithms and hardware are unlocking novel solutions to numerous classically intractable problems in the physical and computational sciences. In the context of chemical screening of precursors for etch processes in integrated circuit fabrication, it is shown in this work that two critical simulation tasks—(1) generating high-accuracy ab initio chemistry data, and (2) training molecular dynamics potentials—can both be reliably performed using present-day hybrid quantum–classical approaches, namely, variational quantum eigensolvers (VQEs) and variational quantum learning models (VQLMs). First, an adaptive-VQE approach is employed to demonstrate the accuracy of quantum algorithms for bond dissociation curves and to provide a heuristic demonstration of computational scaling behavior relative to brute-force classical methods. Next, a VQLM is trained to develop a quantum machine learning–based interatomic potential to probe the chemical influence of F atoms on the dissociation of the Si–Si bond, a well-studied and highly important reaction motif in etch chemistry. It is then shown that the VQLM—following appropriate hyperparameter optimization—achieves comparable performance with the underlying ab initio training data. Through this work, we illustrate how quantum computing approaches may provide valuable tools for research and development in the semiconductor fabrication industry as the technology matures.
Iyer et al. (Fri,) studied this question.