Abstract:We introduce a new Petrov-Galerkin multiscale method for the numerical approximation of the Helmholtz equation with large wave number $\kappa$ in bounded domains in $\mathbb{R}^d$. The discrete trial and test spaces are generated from standard mesh-based finite elements by local subscale corrections in the spirit of numerical homogenization. The precomputation of the corrections involves the solution of coercive cell problems on localized subdomains of size $\ell H$; $H$ being the mesh size and $\ell$ being the oversampling parameter. If the mesh size and the oversampling parameter are such that $H\kappa$ and $\log(\kappa)/\ell$ fall below some generic constants and if the cell problems are solved sufficiently accurate on some finer scale of discretization, then the method is stable and its error is proportional to $H$; pollution effects are eliminated in this regime.
What problem does this paper attempt to address?
This paper attempts to solve the pollution effect problem encountered in the numerical solution of the Helmholtz equation, especially in the case of large wave number (large wave number κ). Specifically, the goal of the paper is to eliminate this pollution effect through local sub - scale correction.
### Background of the Paper
The numerical solution of the Helmholtz equation is very challenging in the case of large wave numbers. Due to the highly oscillatory nature of the solution and the wave - number - dependent pollution effect, the requirements for grid size are very strict. Usually, in order to meaningfully represent highly oscillatory functions, at least 5 to 10 degrees of freedom per wavelength are required. However, in practical applications, this condition is often more stringent than the theoretical minimum requirement.
### Pollution Effect
The pollution effect refers to a non - asymptotic effect related to the wave number that appears in the numerical solution, which will lead to a significant decline in the quality of the numerical solution. Many studies have attempted to overcome or at least reduce this effect, but in two - dimensional or three - dimensional standard methods, this is usually not possible. The work of Melenk and Sauter shows that by coupling the polynomial order \( p \) of the Galerkin finite - element space with the wave number \( \kappa \) (i.e., \( p\approx\log\kappa \)), the pollution effect can be suppressed in some model Helmholtz problems.
### Solution
This paper proposes a new Petrov - Galerkin multi - scale method for the numerical approximation of the Helmholtz equation under large wave numbers. The core idea of this method is to generate discrete trial function and test function spaces through local sub - scale correction. The specific steps include:
1. **Standard Grid Finite Element**: Generate the initial trial function and test function spaces from the standard grid - based finite elements.
2. **Local Sub - scale Correction**: Perform correction by pre - computing some coercive cell problems on local sub - domains. The size of these sub - domains is \( \ell H \), where \( H \) is the grid size and \( \ell \) is the oversampling parameter.
3. **Stability Condition**: If the grid size and the oversampling parameter satisfy \( H\kappa\lesssim1 \) and \( \ell\gtrsim\log(\kappa) \), and the cell problems are solved sufficiently accurately on a finer discretization scale, then the method is stable and the error is proportional to \( H \).
### Main Conclusions
This method can eliminate the pollution effect under the quasi - minimum resolution condition (i.e., \( H\kappa\leq O(1) \)), at the cost of a moderate increase in the amount of inter - element communication, that is, logarithmic - level oversampling. This means that by moderately expanding the stencil, the pollution effect can be eliminated.
### Practical Applications
Although this method is constructive in theory, its relevance in actual computation is not yet clear. This multi - scale method needs to pre - compute local cell problems at a finer numerical resolution, and the appropriate fine - scale selection depends on the stability characteristics of the problem. However, in some cases, for example, when multiple different force terms or incident wave directions of problems with the same geometric setting and wave number need to be solved, the pre - computation can be reused multiple times, thereby significantly reducing the computational complexity.
In conclusion, this paper proposes an effective method to solve the pollution effect problem of the Helmholtz equation in the case of large wave numbers, providing new ideas and tools for numerical solution.