Existence, uniqueness, and approximation of a fictitious domain formulation for fluid-structure interactions (Q2154797)

The paper provides a summary over a computational model for the simulation of fluid-structure interaction problems based on a fictitious domain approach that evolved from immersed boundary methods to distributed Lagrange multiplier methods. The introduction clarifies that the object of study is a solid elastic body immersed in a Newtonian incompressible fluid and its dynamics. The authors describe the challenges involved with modeling these kinds of systems comparing Eulerian and Lagrangian approaches as well as monolithic and partitioned schemes, and justify their choice of a fictitious domain method with being unconditionally stable in time and allowing for larger displacements. The remainder of the section summarizes the structure of the paper. Section 2 is considered with the formulation of the model problem and the fictitious domain method. More precisely, the authors assume that the union of the domain for the fluid and the solid is fixed in time. The authors continue with deriving the continuous (fictitious domain) problem which is finally presented as Problem 1. A stability result is then give in Proposition 1. The third section deals with the existence and uniqueness of a linearization of Problem 1. First, the authors give a good overview over existing literature for existence and uniqueness for particular kinds of fluid-structure interaction problems. Then, the linearized formulation is given as Problem 2. In particular, a linear model for the elasticity is assumed, as well as a prescribed motion of the solid, and convective terms are neglected. For this problem, an existence and uniqueness result is given in Theorem 2. Section 4 discusses suitable time discretizations for Problem 1. First, in Problem 3, a simple (fully implicit nonlinear) backward Euler scheme is presented for which the authors proved a favorable stability result in another publication. Since dealing with the nonlinearity is expensive, another semi-implicit scheme using the structure in an earlier timestep is discussed and Proposition 3 gives a similar stability result. The authors also considered higher-order schemes, like a BDF2-type approach, and were able to prove a stability result at least for a linear model for the elasticity. The fifth section is concerned with the spatial discretization of the proposed continuous or semi-discretized problems using finite elements methods for a linear Piola-Kirchhoff tensor. The authors point out that the semi-implicit time discretization schemes require solving a stationary problem in each time step that has a common form, only varying by some coefficients. In particular, the weak formulation in Problem 4 exhibits a (double-)saddle point structure. Thus, existence and uniqueness depend on two inf-sup conditions. Section 5.1 goes on with actually describing the finite element discretization in Problem 5 (corresponding to Problem 4) clarifying that three fixed meshes are required: one mesh for the whole domain corresponding to the Eulerian variable, one mesh for the solid body corresponding to the Lagrangian variable, and one mesh for the solid body corresponding to the Lagrangian variable. Of course, the (double-) saddle point structure also applies here and one of the inf-sup condition necessary for showing stability follows from choosing usual Stokes pairs for the approximation of velocity and pressure. The remainder of the section is considered with the inf-sup condition for the operator \(\mathcal{A}\) depending on the inf-sup stability of the Lagrange operator in Assumption 1. It turns out that this assumption can be fulfilled for either choice of the Lagrange operator if the associated ansatz space for the Lagrange variables is contained in the one for the solid body as discussed in Proposition 7 and 8. However, the first case requires additional stability of the \(L^2\) projection onto the ansatz space for the solid body.\\ As a final analytical result, Theorem 10 states (quasi-)optimal error estimates in case Assumption 1 and usual compatibility conditions for the Stokes problem are fulfilled. Finally, Section 6 provides some numerical results. The authors consider a problem where the solid occupies an ellipsoidal region initially and evolves approaching a circular equilibrium in time. Figure 1 demonstrates improved stability of the proposed method over an (conditionally stable) immersed boundary method for various mesh and time step sizes. Furthermore, convergence results in time for different time discretization schemes using a fine mesh are presented in Table 1. It appears that only the BDF1 scheme achieves its full rate of convergence in this case. Overall, this paper provides a very good overview the authors' results for the simulation of fluid-structure interaction problems based on a fictitious domain approach.