We introduce an approach for solving the incompressible Navier-Stokes equations on a forest of Octree grids in a parallel environment. The methodology uses the p4est library of Burstedde et al. (2011) [15] for the construction and the handling of forests of Octree meshes on massively parallel distributed machines and the framework of Mirzadeh et al. (2016) [54] for the discretizations on Octree data structures. We introduce relevant additional parallel algorithms and provide performance analyses for individual building bricks and for the full solver. We demonstrate strong scaling for the solver up to 32,768 cores for a problem involving O (6.1 x 10(8)) computational cells. We illustrate the dynamic adaptive capabilities of our approach by simulating flows past a stationary sphere, flows due to an oscillatory sphere in a closed box and transport of a passive scalar. Without sacrificing accuracy nor spatial resolution in regions of interest, our approach successfully reduces the number of computational cells to (at most) a few percents of uniform grids with equivalent resolution. We also perform a numerical simulation of the turbulent flow in a superhydrophobic channel with unparalleled wall grid resolution in the streamwise and spanwise directions. (c) 2020 Elsevier Inc. All rights reserved.

We introduce a numerical framework that enables unprecedented direct numerical studies of the electropermeabilization effects of a cell aggregate at the meso-scale. Our simulations qualitatively replicate the shadowing effect observed in experiments and reproduce the time evolution of the impedance of the cell sample in agreement with the trends observed in experiments. This approach sets the scene for performing homogenization studies for understanding the effect of tissue environment on the efficiency of electropermeabilization. We employ a forest of Octree grids along with a Voronoi mesh in a parallel environment that exhibits excellent scalability. We exploit the electric interactions between the cells through a nonlinear phenomenological model that is generalized to account for the permeability of the cell membranes. We use the Voronoi Interface Method (VIM) to accurately capture the sharp jump in the electric potential on the cell boundaries. The case study simulation covers a volume of (1mm)3 with more than 27,000 well-resolved cells with a heterogeneous mix of morphologies that are randomly distributed throughout a spheroid region.

We analyze the accuracy of two numerical methods for the variable coefficient Poisson equation with discontinuities at an irregular interface. Solving the Poisson equation with discontinuities at an irregular interface is an essential part of solving many physical phenomena such as multiphase flows with and without phase change, in heat transfer, in electrokinetics, and in the modeling of biomolecules’ electrostatics. The first method, considered for the problem, is the widely known Ghost-Fluid Method (GFM) and the second method is the recently introduced Voronoi Interface Method (VIM). The VIM method uses Voronoi partitions near the interface to construct local configurations that enable the use of the Ghost-Fluid philosophy in one dimension. Both methods lead to symmetric positive definite linear systems. The Ghost-Fluid Method is generally first-order accurate, except in the case of both a constant discontinuity in the solution and a constant diffusion coefficient, while the Voronoi Interface Method is second-order accurate in the L∞-norm. Therefore, the Voronoi Interface Method generally outweighs the Ghost-Fluid Method except in special case of both a constant discontinuity in the solution and a constant diffusion coefficient, where the Ghost-Fluid Method performs better than the Voronoi Interface Method. The paper includes numerical examples displaying this fact clearly and its findings can be used to determine which approach to choose based on the properties of the real life problem in hand.

We introduce an approach for simulating epitaxial growth by use of an island dynamics model on a forest of quadtree grids, and in a parallel environment. To this end, we use a parallel framework introduced in the context of the level-set method. This framework utilizes: discretizations that achieve a second-order accurate level-set method on non-graded adaptive Cartesian grids for solving the associated free boundary value problem for surface diffusion; and an established library for the partitioning of the grid. We consider the cases with: irreversible aggregation, which amounts to applying Dirichlet boundary conditions at the island boundary; and an asymmetric (Ehrlich–Schwoebel) energy barrier for attachment/detachment of atoms at the island boundary, which entails the use of a Robin boundary condition. We provide the scaling analyses performed on the Stampede supercomputer and numerical examples that illustrate the capability of our methodology to efficiently simulate different aspects of epitaxial growth. The combination of adaptivity and parallelism in our approach enables simulations that are several orders of magnitude faster than those reported in the recent literature and, thus, provides a viable framework for the systematic study of mound formation on crystal surfaces.

We present an approach to simulate the diffusion, advection and adsorption–desorption of a material quantity defined on an interface in two and three spatial dimensions. We use a level-set approach to capture the interface motion and a Quad/Octree data structure to efficiently solve the equations describing the underlying physics. Coupling with a Navier–Stokes solver enables the study of the effect of soluble surfactants that locally modify the parameters of surface tension on different types of flows. The method is tested on several benchmarks and applied to three typical examples of flows in the presence of surfactant: a bubble in a shear flow, the well-known phenomenon of tears of wine, and the Landau–Levich coating problem.

We present a Voronoi Interface approach to the study of cell electropermeabilization. We consider the nonlinear electropermeabilization model of Poignard et al. [20], which takes into account the jump in the voltage potential across cells' membrane. The jump condition is imposed in a sharp manner, using the Voronoi Interface Method of Guittet et al. [14], while adaptive Quad/Oc-tree grids are employed to automatically refine near the cells boundary for increased accuracy. Numerical results are provided to illustrate the accuracy of the methods. We also carry out simulations in three spatial dimensions to investigate the influence of shadowing and of the cells shape on the degree of permeabilization.

We present scalable algorithms for the level-set method on dynamic, adaptive Quadtree and Octree Cartesian grids. The algorithms are fully parallelized and implemented using the MPI standard and the open-source p4est library. We solve the level set equation with a semi-Lagrangian method which, similar to its serial implementation, is free of any time-step restrictions. This is achieved by introducing a scalable global interpolation scheme on adaptive tree-based grids. Moreover, we present a simple parallel reinitialization scheme using the pseudo-time transient formulation. Both parallel algorithms scale on the Stampede supercomputer, where we are currently using up to 4096 CPU cores, the limit of our current account. Finally, a relevant application of the algorithms is presented in modeling a crystallization phenomenon by solving a Stefan problem, illustrating a level of detail that would be impossible to achieve without a parallel adaptive strategy. We believe that the algorithms presented in this article will be of interest and useful to researchers working with the level-set framework and modeling multi-scale physics in general.

We introduce a simple method, dubbed the Voronoi Interface Method, to solve Elliptic problems with discontinuities across the interface of irregular domains. This method produces a linear system that is symmetric positive definite with only its right-hand-side affected by the jump conditions. The solution and the solution's gradients are second-order accurate and first-order accurate, respectively, in the L∞ norm, even in the case of large ratios in the diffusion coefficient. This approach is also applicable to arbitrary meshes. Additional degrees of freedom are placed close to the interface and a Voronoi partition centered at each of these points is used to discretize the equations in a finite volume approach. Both the locations of the additional degrees of freedom and their Voronoi discretizations are straightforward in two and three spatial dimensions.

We present a numerical method for solving the incompressible Navier–Stokes equations on non-graded quadtree and octree meshes and arbitrary geometries. The viscosity is treated implicitly through a finite volume approach based on Voronoi partitions, while the convective term is discretized with a semi-Lagrangian scheme, thus relaxing the restrictions on the time step. A novel stable implementation of the projection step is introduced, making use of the Marker And Cell layout for the data. The solver is validated numerically in two and three spatial dimensions.

We present a local level-set method based on the hash table data structure, which allows the storage of only a band of grid points adjacent to the interface while providing an O(1) access to the data. We discuss the details of the construction of the hash table data structure as well as the advection and reinitialization schemes used for our implementation of the level-set method. We propose two dimensional numerical examples and compare the results to those obtained with a quadtree data structure. Our study indicates that the method is straightforward to implement but suffers from limitations that make it less efficient than the quadtree data structure.

Author(s): Guittet, Arthur | Advisor(s): Gibou, Frederic | Abstract: This dissertation focuses on numerical simulation methods for continuous problems with irregular interfaces. A common feature of these types of systems is the locality of the physical phenomena, suggesting the use of adaptive meshes to better focus the computational effort, and the complexity inherent to representing a moving irregular interface. We address these challenges by using the implicit framework provided by the Level-Set method and implemented on adaptive Quadtree (in two spatial dimensions) and Octree (in three spatial dimensions) grids. This work is composed of two sections.In the first half, we present the numerical tools for the study of incompressible monophasic viscous flows. After a study of an alternative grid storage structure to the Quad/Oc-tree data structure based on hash tables, we introduce the extension of the level-set method to massively parallel forests of Octrees. We then detail the numerical scheme developed to attain second order accuracy on non-graded Quad/Oc-tree grids and demonstrate the validity and robustness of the resulting solver. Finally, we combine the fluid solver and the parallel framework together and illustrate the potential of the approach.The second half of this dissertation presents the Voronoi Interface Method (VIM), a new method for solving elliptic systems with discontinuities on irregular interfaces such as the ones encountered when simulating viscous multiphase flows. The VIM relies on a Voronoi mesh built on an underlying Cartesian grid and is compact and second order accurate while preserving the symmetry and positiveness of the resulting linear system. We then compare the VIM with the popular Ghost Fluid Method before adapting it to the simulation of the problem of the electropermeabilization of cells.