Bridging Scales in Heterogeneous Catalysis: A Multiscale Computational Framework for Operando Conditions

The accurate modelling of heterogeneous catalytic processes under operando con- ditions remains one of the most formidable challenges in computational chemistry and materials science. Traditional approaches employing static surface models fail to capture the dynamic nature of active sites, adsorbate-induced reconstruc- tions, and temperature-dependent phenomena that characterise real catalytic sys- tems. This work presents a comprehensive multiscale computational framework that bridges density functional theory (DFT) calculations with kinetic Monte Carlo (kMC) simulations, enhanced by a high-fidelity machine learning potential (MLP), to achieve realistic timescales commensurate with experimental turnover frequen- cies. We demonstrate the methodology through the exemplar case of CO oxidation on Pt(111). The framework reveals previously unidentified dynamic phenomena, including a non-monotonic dependence of activation energy on coverage driven by dynamic site creation and temperature-induced surface restructuring. The frame- work successfully predicts turnover frequencies within an order of magnitude of experimental values across a wide temperature range, a significant improvement over static models. Our results indicate that incorporating surface dynamics and realistic reaction conditions fundamentally alters predicted catalytic behaviour, sug- gesting that static models may systematically misrepresent catalytic mechanisms. This methodology opens new avenues for rational catalyst design based on operando behaviour rather than idealised surface structures.