Adaptive Methods for Multiscale and Multiphysics Simulation of Hydrogen Applications

The transition to carbon-neutral energy systems requires combustion technologies that significantly reduce emissions. Hydrogen is a particularly promising energy carrier because it enables carbonfree operation. However, its low molecular weight and high diffusivity can cause thermodiffusive instabilities. Its low density requires high injection velocities, which lead to strong compressibility effects and underexpanded jets. Additionally, practical combustors exhibit a wide range of spatial and temporal scales, from acoustics and coherent structures to flame fronts and turbulent–nonturbulent interfaces, all of which can interact with the new physics introduced. This thesis addresses these challenges by developing and validating a multi-physics Large-Eddy Simulation (LES) model for hydrogen applications. A two-equation quasi-dynamic (QD) k–l_sfs closure, where both turbulent kinetic energy and an effective filter width are transported, is proposed. The model provides the memory and scale awareness needed to combine the Artificially Thickened Flame (ATF) model with Adaptive Mesh Refinement (AMR). Systematic tests in canonical flows and a premixed hydrogen jet flame show that the proposed closure reduces commutation errors at refinement interfaces and accurately reproduces fine-grid flame statistics on adaptive meshes. Building on this LES model, the thesis examines mixture formation in compressible hydrogen jets used in direct injection. A compressible multi-species solver is used to investigate the onset of turbulence and mixing in underexpanded hydrogen jets, reconciling previously conflicting numerical and experimental results. Complementary direct numerical simulations of a temporally evolving hydrogen jet are then employed to analyze turbulent–non-turbulent interfaces (TNTIs) under strong density and viscosity stratification, offering new insights into the relationship between interface geometry, vorticity, and scalar transport, and evaluating suitable interface markers for filtered fields. Finally, the thesis revisits multi-regime (MR) hydrogen combustion, where premixed, non-premixed, and stratified burning modes coexist. A regime-aware ATF strategy is proposed, in which ATF is selectively deactivated in non-premixed regions identified by a Takeno flame index, avoiding artificial thickening of non-premixed flames while retaining ATF and turbulence–chemistry closure where appropriate. Applied to an MR burner configuration, the combination of QD k–l_sfs, MR-ATF, and AMR reproduces overall flame topology, regime distributions, and thermo-chemical statistics with good accuracy. This validates the proposed LES model across a wide range of test cases and advances understanding of hydrogen-mixture formation.