Direct numerical simulation of three-dimensional thermo-diffusive unstable premixed flame

To address the global warming issue, the transition toward carbon-free energy resources is one of the main strategies, and hydrogen has emerged as a promising alternative fuel, particularly when harvested from renewable resources. However, for hydrogen to be widely adopted as a fuel, various industries must facilitate its integration. A major challenge in hydrogen combustion is its intrinsic instabilities, which arise in lean mixtures and significantly impact flame behavior. These instabilities not only contribute to flame destabilization but also pose critical safety concerns, necessitating a deeper understanding of these instabilities. This thesis explores the fundamental dynamics of thermo-diffusive (TD) and Darrieus–Landau (DL) instabilities in low Lewis number (below the critical Lewis number) mixtures, focusing on lean hydrogen-air premixed flames through high-fidelity Direct Numerical Simulations (DNS). With the latest trends in High-Performance Computing (HPC) centers increasingly adopting accelerators like GPUs, it has become essential to leverage their computational power for large-scale simulations. To take advantage of these advancements, this thesis employs NekRS, a high-order spectral element solver optimized for GPU acceleration that enables efficient and high-fidelity simulations. A simplified one-step deficient reactant kinetic model and a low-mach-number formulation are employed to reduce computational demand for large-scale flames. To better understand the effect of intrinsic instabilities in realistic flame configurations, this study performs three-dimensional DNS of a thermo-diffusive unstable flame. The flame characteristics and domain size used in these simulations are adapted directly from an established two-dimensional DNS dataset, enabling a clear assessment of how introducing the third dimension influences overall flame behavior. The results indicate that three-dimensional flames exhibit higher consumption speeds than their 2D counterparts, and enhanced local curvature effects, leading to increased local reactivity and higher superadiabatic temperature. Furthermore, the dispersion relation of premixed flames characterizes the growth rate of each wavelength perturbation and provides insights into flame front stability. Comparing dispersion relation with velocity spectra reveals the wavenumber with the maximum growth rate influences flame dynamics, and aligned with peaks observed in velocity spectra and inverse curvature. This consistency underscores the role of specific wavenumbers in shaping flame behavior and turbulence generation. Additionally, to examine the effect of different species on the dispersion relation between 2D and 3D flames, a detailed kinetic DNS of a lean hydrogen flame was conducted using the multi-wavelength perturbation method. The comparison of 2D and 3D dispersion relations confirms agreement in the linear regime. In conclusion, this thesis highlights the critical role of three-dimensional effects in the intrinsic instabilities of premixed flames. Compared to 2D cases, 3D unstable flames exhibit enhanced consumption speeds, broader reaction-rate distributions, and intensified flame–flow interactions due to increased surface wrinkling. These findings provide a foundation for advancing the understanding of multi-dimensional combustion instabilities to improve predictive models for lean hydrogen flames.