Two-level Schwarz method for solving variational inequality with nonlinear source terms (Q2475379): Difference between revisions

The numerical solution of some variational inequality with nonlinear source terms is analyzed. The proposed problems have many applications, e.g., in the diffusion involving Michaelis--Menten or second-order irreversible reactions. The authors extend the two-level Schwarz method for solving the variational inequality problem with nonlinear terms: Find \(u \in K = \{v \in H^1_0 (\Omega): v \geq 0 \text{ a.e. on } \Omega \}\) such that \[ a(u, v-u)+(f(u,\mathbf{x}),v-u) \geq 0\text{ for any }v \in K\tag{1} \] where \(a(u,v)=\int_\Omega \mathbf{grad}\, u . \mathbf{grad}\, v \,d\Omega, f(u,\mathbf{x})\) is continuously differentiable and \(\partial f/\partial u \geq 0 \text{ on } \widetilde{\Omega} \times \{u:u \geq0\}\). The lumped mass discretization method to deal with nonlinear source terms is used. Algebraically, the finite element approximation of the problem (1) (Find \(u \in K^n = \{ v\in\mathbb R^n:v \geq 0 \} \), such that: \((Au+F(u), v-u)\geq0\) , for any \(v \in K^n, A=(a_{ij})_{n \times m}, a_{ij}= a(\varphi_i, \varphi_j)\) is an M-matrix, \(F(u)=(F_i(u_i))^n_{i=1}, F_i(u_i)=\int_{w_i} f(u_i,x)\,dx\), \(w_i\) denotes the lumped mass discretization region) is equivalent to the nonlinear complementarity problem: find \(u \in K^n\), such that \[ \min(Au+F(u),u)=0\tag{2} \] If \(A\) is an \(M\)-matrix the problem (2) has a unique solution, which is the minimal on the set \(S=\{v \in K^n:Av+F(v) \geq 0\}\) (the supper solution of (2)). Here the solution \(u\) of the problem (2) divides the set \(N=\{1,2, \dots n\}\) into two non-overlapping subsets \([N^{\diamondsuit}, N^+]\), where \[ N^{\diamondsuit}=\{i \in N:u_i=0\}, N^+ = N \backslash N^{\diamondsuit}=\{i \in N:u_i>0\}.\tag{3} \] The authors propose the two-level Schwarz algorithm based on the partition (3). By using a specific monotonicity result, the computational domain can be partitioned into subdomains with linear and nonlinear subproblems. One of the advantages of this two-level Schwarz method is that a fast solver for the system of linear equations may be used. Main result: this paper extends the two-level Schwarz method to solve the variational inequality problems with nonlinear source terms. The precise proof of the convergence theorem is proposed. The method converges within finite steps with an appropriate initial point. Finally numerical results which show that the methods are efficient, are also presented.