3D Numerical simulations of vesicle and inextensible capsule dynamics

Alexander Farutin (1) , Thierry Biben (2) , Chaouqi Misbah (1)

1 LIPhy - Laboratoire Interdisciplinaire de Physique [Saint Martin d’Hères]
2 ILM - Institut Lumière Matière [Villeurbanne]

Vesicles are locally-inextensible fluid membranes while inextensible capsules are in addition endowed with in-plane shear elasticity mimicking the cytoskeleton of red blood cells (RBCs). Boundary integral (BI) methods based on the Green's function techniques are used to describe their dynamics, that falls into the category of highly nonlinear and nonlocal dynamics. Numerical solutions raise several obstacles and challenges that strongly impact the results. Of particular complexity is (i) the membrane inextensibility, (ii) the mesh stability and (iii) numerical precisions for evaluation of the boundary integral equations. Despite intense research these questions are still a matter of debate. We regularize the single layer integral by subtraction of exact identities for the terms involving the normal and the tangential components of the force. In addition, the regularized kernel remains explicitly self-adjoint. The stability and precision of BI calculation is enhanced by taking advantage of additional quadrature nodes located in vertices of an auxiliary mesh, constructed by a standard refinement procedure from the main mesh. We extend the partition of unity technique to boundary integral calculation on triangular meshes: We split the calculation of the boundary integral between the original and the auxiliary mesh using a smooth weight function, which takes the distance between the source and the target as the argument and falls to zero beyond a certain cut-off distance. We provide an efficient lookup algorithm that allows us to discard most of the vertices of the auxiliary mesh lying beyond the cut-off distance from a given point without actually calculating the distances to them. The proposed algorithm offers the same treatment of near-singular integration regardless if the source and the target points belong to the same surface or not. Additional innovations are used to increase the stability and precision of the method: The bending forces are calculated by differential geometry expressions using local coordinates defined in vicinity of each vertex. The approximation of the surface in vicinity of a vertex is obtained by fitting with a second-degree polynomial of local coordinates. We solve for the Lagrange multiplier associated with membrane incompressibility using two penalization parameters per suspended entity: one for deviation of the global area from prescribed value and another for the sum of squares of local strains defined on each vertex. The proposed advancement is to vary the penalization parameters at each time step in such a way, that the global area of each membrane be conserved and the sum of squares of local strains be at minimum. This optimization is achieved by solving a linear system of rank three times the number of entities involved in the simulation. If no auxiliary mesh is used, the method reduces to steepest descent method thanks to the explicit self-adjointness of the regularized single-layer kernel in the boundary integral equation. Inextensible capsules, a model of RBC, are studied by storing the position in the reference configuration for each vertex. The elastic force is then calculated by direct variation of the elastic energy. Various nonequilibrium physical examples on vesicles and capsules will be presented and the convergence and precision tests highlighted. Overall, a good convergence is observed with numerical error inversely proportional to the number of vertices used for surface discretization, the highest order of convergence allowed by piece-wise linear interpolation of the surface.

Mots clés: vesicles; capsules; Stokes flow; boundary integral method; singularity subtraction

Domaines: Biophysique [physics.bio-ph]

Alexander Farutin, Thierry Biben, Chaouqi Misbah. 3D Numerical simulations of vesicle and inextensible capsule dynamics. Journal of Computational Physics, 2014, 275, pp.539-568. ⟨10.1016/j.jcp.2014.07.008⟩. ⟨hal-00841996⟩