CutVEM: Conforming virtual element method on embedded domains with shape-agnostic element agglomeration

The virtual element method (VEM) is a stabilized Galerkin method that is robust and accurate on general polygonal meshes. This feature makes it an appealing candidate for simulations involving meshes with embedded interfaces and evolving geometries. However, the method can yield poorly conditioned stiffness matrices in such scenarios due to meshes having cut cells. We propose a novel element agglomeration algorithm for the virtual element method to address this issue. The agglomeration algorithm renders the VEM robust over planar polygonal meshes, particularly on finite element meshes cut by immersed geometries. The algorithm relies on the element stability ratio, which we define using the extreme eigenvalues of the element stiffness matrix. The resulting element agglomeration criterion is free from nebulous polygon quality metrics and is defined independently of polygon shapes. The algorithm proceeds iteratively and element-wise to maximize the minimum element stability ratio, even at the expense of degrading elements with better ratios. Crucially, element agglomeration alters the number of elements, not the degree of freedom count. The resulting method, which we label as CutVEM, retains node locations of cut elements unchanged, and yields discretizations that conform to embedded interfaces. This, in turn, facilitates straightforward imposition of boundary conditions and interfacial constraints. Through detailed numerical experiments that sample varied element-interface intersections, we demonstrate that CutVEM enjoys dramatically improved condition numbers of global stiffness matrices over the VEM. Furthermore, simulations of prototypical heat conduction problems with Dirichlet and Neumann boundary conditions on domains with immersed geometries show that element agglomeration does not noticeably degrade solution accuracy and that CutVEM retains the VEM's optimal convergence rate.