Combining Generative Adversarial Network and First-Principles Based Microkinetics for Heterogeneous Catalyst Design

Microkinetic analysis based on density functional theory (DFT) was combined with a generative adversarial network (GAN) to enable artificial proposal of heterogeneous catalysts based on the DFT-calculated dataset. The approach was applied to the NH3 formation reaction on Ru-Rh alloy surfaces as an example. The NH3 formation turnover frequency (TOF) was calculated by DFT-based microkinetics. Specifically, six elementary reactions (N2 dissociation, H2 dissociation, NHx (x = 1–3) formation, and NH3 desorption) were explicitly considered, and their reaction energies were evaluated by DFT. On the basis of TOF values and atomic compositions, new alloy surfaces were generated using the GAN. This approach successfully generated the surfaces not included in the initial dataset but have higher TOF values. The N2 dissociation reaction was more exothermic for the generated surfaces, leading to higher TOF. The present study shows that automatic improvement of catalyst materials is possible by using the iterative steps of DFT calculation and sample generation by GAN.