Abstract:We introduce a new numerical method for solving time-harmonic acoustic scattering problems. The main focus is on plane waves scattered by smoothly varying material inhomogeneities. The proposed method works for any frequency $\omega$, but is especially efficient for high-frequency problems. It is based on a time-domain approach and consists of three steps: \emph{i)} computation of a suitable incoming plane wavelet with compact support in the propagation direction; \emph{ii)} solving a scattering problem in the time domain for the incoming plane wavelet; \emph{iii)} reconstruction of the time-harmonic solution from the time-domain solution via a Fourier transform in time. An essential ingredient of the new method is a front-tracking mesh adaptation algorithm for solving the problem in \emph{ii)}. By exploiting the limited support of the wave front, this allows us to make the number of the required degrees of freedom to reach a given accuracy significantly less dependent on the frequency $\omega$. We also present a new algorithm for computing the Fourier transform in \emph{iii)} that exploits the reduced number of degrees of freedom corresponding to the adapted meshes. Numerical examples demonstrate the advantages of the proposed method and the fact that the method can also be applied with external source terms such as point sources and sound-soft scatterers. The gained efficiency, however, is limited in the presence of trapping modes.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
This paper aims to solve the numerical calculation problems in high - frequency scattering problems. Specifically, the researchers are concerned with the problem of plane - wave scattering in media where the material properties change smoothly. Such problems become extremely difficult to solve under high - frequency conditions because traditional finite - difference or finite - element methods require a large number of degrees of freedom to achieve a given accuracy, and standard iterative solvers and multigrid methods fail or converge slowly at high frequencies.
### Main contributions
The paper proposes a new adaptive finite - element method, which is particularly suitable for high - frequency scattering problems. The main steps include:
1. **Calculate a suitable incident plane - wave packet**: Construct an incident plane - wave packet \(u_I(x,t)\) with compact support, in the form of \(u_I(x,t)=\omega\psi(\omega(t - (\hat{r}\cdot x)/c_0))\), where \(\psi\) is a real - valued smooth function, supported within \([-\xi_0,\xi_0]\), and its Fourier transform satisfies \(\mathcal{F}[\psi](- 1)=1\).
2. **Solve the time - domain scattering problem**: Use the above - mentioned incident plane - wave packet as a source term to solve the time - domain scattering problem:
\[
\partial_t^2(u_S + u_I)-\beta^{-1}\nabla\cdot(\alpha\nabla(u_S + u_I)) = 0\quad\text{in }\mathbb{R}^d\times(t_0,\infty)
\]
where \(t_0\) is the time when the incident plane - wave packet first enters the computational domain.
3. **Reconstruct the time - harmonic solution through Fourier transform**: Use the time - domain solution \(u_S(x,t)\) to reconstruct the time - harmonic solution \(U_S\) in time through Fourier transform:
\[
U_S=\mathcal{F}_t[u_S](\cdot,-\omega)=\lim_{t\rightarrow\infty}\int_{t_0}^{t}e^{i\omega\tau}u_S(\cdot,\tau)d\tau
\]
### Key techniques
- **Adaptive mesh generation**: By tracking the position of the wavefront, the mesh is adaptively adjusted so that a fine mesh is used near the wavefront and a coarse mesh is used in the area far from the wavefront. This significantly reduces the number of degrees of freedom required, thereby reducing the computational complexity.
- **Efficient Fourier transform algorithm**: Using the degrees of freedom reduced by the adaptive mesh, an efficient Fourier transform algorithm is designed to further improve the computational efficiency.
### Conclusion
This method shows significant advantages in dealing with high - frequency scattering problems, especially in the absence of trapped modes. Numerical experiments verify the effectiveness of this method and show its application potential under different external source terms (such as point sources and acoustically soft scatterers). However, in the presence of trapped modes, the efficiency of the method will be limited.