First-Principles Modeling in Heterogenous Catalysis: Where Are We?

First-principles microkinetic modeling has become a cornerstone of heterogeneous catalysis, linking atomistic energetics from density functional theory (DFT) to macroscopic observables such as turnover frequencies and selectivity. Despite notable successes in mechanistic elucidation and catalyst screening, predictive accuracy relative to experiments remains limited. In this Perspective, we critically examine the origins of inaccuracy in first-principles modeling by systematically analyzing uncertainty along the full modeling workflow, from active-site representation and reaction network design to electronic structure approximations, entropy treatments, microkinetic assumptions, and reactor coupling. We distinguish structural-model uncertainty from electronic-structure uncertainty and highlight how common simplifications, i.e., mean-field approximations, single-site occupancy assumptions, zero-conversion modeling, and slab-based active-site representations, can introduce mechanistic bias. We further evaluate recent advances, including beyond-mean-field kinetic frameworks, particle-based microkinetic models, higher-level electronic structure methods beyond the generalized gradient approximation (GGA), machine-learning acceleration, and Bayesian uncertainty quantification. While these developments significantly reduce specific error sources, they do not eliminate compounded uncertainties across scales. We argue that achieving truly predictive first-principles catalysis modeling requires integrated frameworks that combine realistic surface representations, rigorous uncertainty propagation, and reactor-level modeling. This Perspective provides a structured roadmap for advancing accuracy in computational catalysis.