Multiscale Modeling
Multiscale modeling describes a system with separate models of phenomena at different length or time scales (for example atomistic catalyst models, reactor or device models, and system-level process, supply chain or economic models) that are connected through a small set of shared variables, so that information passes between scales. Each scale can then be handled with the methods and tools best suited to it, material choices can be judged by their system-level consequences, and system requirements can guide material design. Coupling the scales, especially for simultaneous optimization, requires every model to be expressed in a compatible mathematical form.
Atoms-to-systems optimization of formic acid dehydrogenation (2025)
Sunshine and coworkers linked a DFT, Wulff construction and microkinetic catalyst model, a CSTR flowsheet model and a techno-economic cost model through cross-scale variables such as turnover frequency, hydrogen yield, catalyst mass and price (Sunshine et al. 2025). Replacing each model by a linear model decision tree surrogate made the coupled problem a mixed-integer linear program. Co-optimizing all three scales lowered the hydrogen cost by about 40% relative to optimizing each scale sequentially with intuitive targets, and switched the preferred catalyst from the most active metal (Pt) to the cheapest (Cu). Summary: sunshine2025a.
Systems-to-Atoms framework for hydrogen infrastructure (2025)
The Systems-to-Atoms (S2A) framework couples CatMAP microkinetic models of formic acid dehydrogenation (facet rates combined into a nanoparticle turnover frequency through a Wulff construction) with a reactor sizing model and a bottom-up technoeconomic model of a hydrogen delivery chain (Yuan et al. 2025). Passing parameters across scales showed a result that no single-scale model would show. Raising the reaction temperature increased Pd activity by about three orders of magnitude but barely changed station cost, because the reactor and catalyst were under 3% of the total. Raising the pressure to 100 bar left Pd activity unchanged but saved about $2/kg H2 in downstream compression. Across scales, a pressure-sensitive and cheap Cu catalyst gave nearly the same station cost as Pd at 100 bar. Summary: yuan2025.