Accessing Free-Energy Barriers from Unbiasing Dynamics: Delocalized Transition States in Heterogeneous Catalysis

Accurate determination of free-energy barriers remains a central challenge in heterogeneous catalysis. Conventional approaches rely on static potential-energy surfaces or biased molecular dynamics, which often underestimate entropic contributions and distort the true transition-state character. Here we introduce Screened Standard Molecular Dynamics (SSMD), an unbiased computational strategy that initiates standard MD trajectories from quasi-transition states to enrich sampling in the delocalized transition region without artificial constraints or biasing potentials. Reaction free-energy barriers are subsequently extracted from these trajectories through confidence-interval force screening followed by mean-force integration. Benchmarking on CO oxidation and ethane dehydrogenation over Pt(111), we find that SSMD yields smooth, converged free-energy profiles at an order of magnitude lower computational cost than constrained methods. SSMD systematically predicts slightly lower barriers than biased approaches, with deviations increasing at elevated temperatures. We show that the transition state is more delocalized than previously thought, a feature SSMD uniquely captures but conventional schemes severely underestimate, accounting for the observed barrier differences. It is worth noting that SSMD requires the projection of forces onto an appropriate reaction coordinate, which necessitates data analysis and processing and may therefore introduce some degree of bias. The convergence and sensitivity tests presented demonstrate the numerical robustness of these choices for the systems studied.