Multiscale reaction science and engineering of hydrogen-based direct reduction ironmaking

Hydrogen-based direct reduction （HyDR）, in which green hydrogen replaces fossil reductants, is one of the core technological routes for deep decarbonization of the steel industry. The process spans multiple scales, from atomic interfacial reactions to multiphysics coupling in the shaft furnace reactor, while the correlation between its dynamically evolving multiscale behavior and the macroscopic reaction behavior remains insufficiently understood, constituting a key scientific problem that must be resolved to refine the theoretical framework of hydrogen metallurgy. Following an "electron/atom-micro/meso-macro reactor" framework, this review systematically summarizes the full-chain, multiscale advances in the HyDR of iron oxides. At the atomic/electronic scale, density functional theory （DFT） reveals the facet-dependent dissociative adsorption and interfacial reaction of H₂ and CO on iron-oxide surfaces and the corresponding rate-determining steps, while large-scale molecular dynamics driven by deep-learning potentials reproduce at DFT-level accuracy the defect evolution of wüstite and the nucleation-crystallization pathway from an "amorphous iron intermediate" to crystalline iron. At the micro/meso scale, X-ray computed tomography and environmental transmission electron microscopy combined with deep learning quantitatively decouple the dynamic evolution of the pore-crack network and confirm a vacancy-mediated,fluid-like aggregation mechanism of amorphous iron, on which basis the applicability limits of the unreacted shrinking-core, grain and random-pore models are clarified. At the macro scale, three numerical frameworks for the shaft furnace, namely the porous-media continuum model, the Eulerian two-fluid model, and computational fluid dynamics-discrete element method （CFD-DEM） coupled approach, are systematically analyzed. In view of the strongly endothermic nature of pure hydrogen reduction, thermal-management and process-optimization strategies based on coordinated thermal parameters are proposed. Finally, future directions are outlined for cross-scale coupling, general-purpose machine-learning potentials, in situ quantitative characterization, and industrial data validation to refine the theoretical framework of hydrogen metallurgy and provide theoretical support for advancing HyDR from mechanistic understanding to engineering practice.