Abstract:When solving the Poisson equation by the finite element method, we use one degree of freedom for interpolation by the given Laplacian - the right hand side function in the partial differential equation. The finite element solution is the Galerkin projection in a smaller vector space. The idea is similar to that of interpolating the boundary condition in the standard finite element method. Due to the pointwise interpolation, our method yields a smaller system of equations and a better condition number. The number of unknowns on each element is reduced significantly from $(k^2+3k+2)/2$ to $3k$ for the $P_k$ ($k\ge 3$) finite element. We construct 2D $P_2$ conforming and nonconforming, and $P_k$ ($k\ge3$) conforming interpolated Galerkin finite elements on triangular grids. This interpolated Galerkin finite element method is proved to converge at the optimal order. Numerical tests and comparisons with the standard finite elements are presented, verifying the theory and showing advantages of the interpolated Galerkin finite element method.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: when solving the Poisson equation, reduce the number of unknowns and improve the condition number by using a new interpolated Galerkin finite - element method in the finite - element method (FEM). Specifically, the paper proposes the following points:
1. **Reduce the number of unknowns**: The traditional finite - element method has more degrees of freedom on each element, resulting in higher computational complexity. The method proposed in this paper significantly reduces the number of unknowns on each element by interpolating the given Laplace operator (i.e., the function \( f \) on the right - hand side of the partial differential equation) and replacing the internal Lagrange node values with internal Laplace values.
2. **Improve the condition number**: Due to the application of point interpolation, the linear equation system generated by the new method is smaller in scale and has a better condition number, which helps to improve numerical stability.
3. **Maintain the optimal convergence order**: Although the number of unknowns is reduced and the condition number is improved, the new method can still guarantee the same optimal convergence order as the standard finite - element method, that is, the \( H^1 \)-norm convergence order of \( O(h^k) \).
4. **Applicable to high - order polynomial spaces**: For the \( P_k (k\geq3) \) finite - element space, the new method redefines the basis functions and constructs conforming and non - conforming interpolated Galerkin finite - elements on triangular meshes.
In summary, the main objective of this paper is to develop an improved finite - element method to solve the Poisson equation more efficiently while maintaining or improving the accuracy and stability of the numerical solution.
### Formula representation
- Poisson equation:
\[
-\Delta u = f \quad \text{in } \Omega,
\]
\[
u = 0 \quad \text{on } \partial \Omega.
\]
- Form of the interpolated Galerkin finite - element solution:
\[
u_h=\sum_{x_i\in\partial K\setminus\partial\Omega}u_i\varphi_i+\sum_{x_j\in K_0}c_j\varphi_j+\sum_{x_k\in\partial\Omega}c_k\varphi_k,
\]
where \( c_j \) is the interpolated value of the right - hand side function \( f \) at the internal node \( x_j \), \( c_k \) is the interpolated value on the boundary, and only \( u_i \) is obtained through Galerkin projection.
- Convergence estimate:
\[
\|u - u_h\|_0+h|u - u_h|_1\leq C h^{k + 1}|u|_{k+1},
\]
where \( |\cdot|_k \) represents the Sobolev seminorm \( H^k(\Omega) \).