Andrew T. Barker,Susanne C. Brenner,Eun-Hee Park,Li-Yeng Sung
Abstract:Discontinuous Petrov-Galerkin (DPG) methods are new discontinuous Galerkin methods with interesting properties. In this article we consider a domain decomposition preconditioner for a DPG method for the Poisson problem.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
The paper aims to develop a one - level additive Schwarz preconditioner for the discontinuous Petrov - Galerkin (DPG) method to improve the efficiency and stability of solving the Poisson problem.
#### Specific background and problem description
1. **Ultra - weak form of the Poisson problem**:
- Let \(\Omega\) be a polyhedral domain in \(R^d\) (\(d = 2,3\)), and \(\Omega_h\) be a simplicial subdivision of \(\Omega\).
- The Poisson problem in the ultra - weak form can be expressed as: find \(U\in U\) such that
\[
b(U,V)=l(V),\quad\forall V\in V,
\]
where
\[
U = [L^2(\Omega)]^d\times L^2(\Omega)\times H^{1/2}_0(\partial\Omega_h)\times H^{-1/2}(\partial\Omega_h),
\]
\[
V = H(\text{div};\Omega_h)\times H^1(\Omega_h),
\]
\[
b(U,V)=\int_{\Omega}\sigma\cdot\tau\,dx-\sum_{K\in\Omega_h}\int_K u\,\text{div}\tau\,dx+\sum_{K\in\Omega_h}\int_{\partial K}\hat{u}\tau\cdot n\,ds-\sum_{K\in\Omega_h}\int_K\sigma\cdot\text{grad}v\,dx+\sum_{K\in\Omega_h}\int_{\partial K}v\hat{\sigma}\cdot n\,ds,
\]
\[
l(V)=\int_{\Omega}f v\,dx.
\]
2. **Application of the DPG method**:
- The DPG method computes \(U_h\in U_h\) such that
\[
b(U_h, V)=l(V),\quad\forall V\in V_h,
\]
where \(U_h\) and \(V_h\) are the trial space and the test space respectively.
3. **Requirement for the preconditioner**:
- In order to solve the above problems effectively, it is necessary to develop a suitable preconditioner to accelerate the iterative solution process. The paper proposes a one - level additive Schwarz preconditioner \(B_h\) to handle the linear system \(A_hU = F\) generated by the DPG method.
4. **Theoretical analysis and numerical results**:
- Theoretically, the effectiveness of this preconditioner is proved, and an estimate of the condition number is given.
- Numerical experiments show that after using this preconditioner, the number of iterations of the conjugate gradient method is significantly reduced, thus improving the solution efficiency.
By introducing this preconditioner, the paper solves the problems of slow convergence speed and high computational cost faced by the DPG method in solving the Poisson problem, and provides a more efficient numerical solution strategy.