From Ideal Models to Structural Ensembles: Predictive Realism for Molecular Simulation of Porous Materials

Molecular simulation of porous materials has traditionally relied on ideal, perfectly periodic structures as its basic representation. This choice enabled the development of transferable methods and provided a common reference for studies of adsorption, diffusion, and materials screening. Increasing experimental and computational evidence, however, shows that many porous materials behave as heterogeneous systems shaped by defects, local chemistry, interfaces, and working conditions, so that single structure models no longer correspond to the material that is actually probed. In this work, we argue that the main obstacle to predictive simulation is not the lack of advanced methods, but the way the structural object itself is defined. We introduce predictive realism as the requirement that molecular simulations represent the physically relevant variability of real materials whenever that variability affects measurable observables. In this sense, it refers not simply to agreement with experiment, but to the adequacy of the structural representation relative to the material probed under the relevant conditions. Using recent results from experiment and simulation, we show how multiple physically plausible structural realizations can lead to different measurable responses, so that ensemble-based descriptions can align molecular modeling with measured response in adsorption, transport, and mechanical stability. Rather than promoting maximal complexity or specific techniques, this framework emphasizes the selective inclusion of structurally relevant variability, quantitative validation, and reproducible workflows. By reframing prediction in terms of ensembles, distributions of observables, and uncertainty, it outlines a route toward molecular simulations that are both more realistic and more reliable for real materials under working conditions, and that can serve as a consistent standard for future studies. Molecular simulation of porous frameworks grew around a pragmatic choice: represent the solid as a perfectly periodic crystal. That decision matched the experimental resolution of the time, the scale that simulations could handle, and the community's need for shared workflows and intuition. With that representation, adsorption and diffusion could be treated quantitatively and reproducibly. The approach gave us common benchmarks, transferable input files, and a clean language to compare families of materials across laboratories.