A new provably stable weighted state redistribution algorithm (Q6598502)

The paper addresses the problem of stabilising finite volume methods on cut cells in Cartesian grids through state redistribution techniques, which have been a subject of growing interest in computational mathematics.\N\NThe scientific problem tackled by this paper is the numerical instability that arises when small cut cells intersect with solid geometries within a structured Cartesian grid, a common occurrence in simulations involving embedded boundaries. These small cut cells cause severe time step restrictions due to the Courant-Friedrichs-Lewy (CFL) condition and can lead to instability or loss of accuracy in the solution. Traditional methods, such as cell merging, can address this issue, but they introduce other challenges, including increased diffusion and computational complexity. The paper's contribution lies in presenting a modified state redistribution algorithm that maintains the accuracy and stability of the finite volume method while preserving the computational efficiency by avoiding global preprocessing steps.\N\NTo solve this problem, the authors build on previous work on state redistribution (SRD) algorithms. The traditional SRD method temporarily merges cut cells with neighbouring cells, redistributing the solution state across the combined volume in a conservative manner. This approach stabilises the numerical scheme by reducing the effect of small cell sizes. The new method introduced in this paper generalises the SRD approach by introducing weighted state redistribution. It allows for a continuous, volume-fraction-based stabilisation rather than the sharp cut-off used in previous methods, thus improving the monotonicity and stability of the solution without imposing strict time step constraints. The authors provide both theoretical analysis and numerical experiments to demonstrate the monotonicity and total variation diminishing (TVD) properties of their new method.\N\NIn terms of methodology, the paper first introduces a simple one-dimensional model to illustrate the core ideas behind the weighted redistribution scheme. The model employs a finite volume method with explicit time stepping, where small cut cells merge temporarily with their neighbours to redistribute the state variable in a way that ensures conservation and linearity. The weights assigned to each cell are adjusted dynamically based on the size of the cut cells, ensuring that the merging becomes progressively less significant as the cut cell size increases. The paper extends this framework to two and three dimensions, proving that the algorithm is conservative and monotone in most practical scenarios. Further, the authors provide proofs of the stability of the scheme under various configurations, including cases where the cut cells are adjacent to boundaries or irregular geometries. Monotonicity is rigorously established through a detailed examination of the coefficients in the redistribution formulas, ensuring that no new extrema are introduced in the solution.\N\NThe main findings of the manuscript are multi-fold. The new weighted state redistribution algorithm is shown to be provably monotone and TVD under a wide range of conditions, with the added benefit of shutting off continuously as the cut cell size approaches a threshold. This contrasts with earlier methods that applied a fixed stabilisation regardless of cell size. The authors demonstrate through a series of computational experiments, including tests in two and three dimensions, that their method outperforms previous approaches, especially in terms of reduced diffusion and improved accuracy in complex geometries. Specifically, they show that their algorithm works well in simulations involving supersonic vortex flow, shock diffraction, and acoustic pulse scattering in trefoil-shaped cavities, where small cut cells are prevalent. The new method achieves second-order convergence for both volume and boundary errors, a significant improvement over the original SRD.\N\NThe significance of this research lies in its potential impact on the field of computational fluid dynamics and other areas that rely on high-accuracy simulations on irregular grids. The weighted SRD algorithm allows for stable and accurate time stepping in simulations involving embedded boundaries, without the need for global preprocessing, which is often computationally expensive. This makes the method highly applicable in real-world simulations, such as those involving fluid-structure interaction, shock wave propagation, and complex geometries where traditional methods struggle with instability or excessive diffusion. The paper's contribution is thus a major step forward in improving the efficiency and accuracy of finite volume methods in challenging numerical environments.