Direct sampling methods for inverse medium scattering problems of elastic waves

Abstract: This paper concerns the inverse elastic scattering problem of determining an unknown penetrable obstacle from far-field data. Using the Helmholtz decomposition, the coupled boundary value problem is reformulated as a coupled scalar Helmholtz system. We prove the uniqueness of the associated coupled Helmholtz boundary value problem and of the corresponding boundary integral equation system, which is then discretized by a Nyström method for efficient numerical computation. Leveraging the relation between the compressional and shear far-field patterns of the Navier system and those of the coupled Helmholtz system, we employ three indicators to reconstruct the location and shape of the obstacle. Furthermore, we analyze the decay properties of these indicators and establish corresponding stability estimates. Numerical experiments are presented to illustrate the effectiveness and robustness of the proposed method, even for limited-aperture data.

• The paper introduces a novel direct sampling strategy that decouples elastic wave scattering into coupled scalar Helmholtz problems.
• It employs a high-order Nyström discretization with boundary integral formulations to achieve exceptional numerical accuracy and efficiency.
• The study demonstrates robust reconstruction of penetrable obstacles under noisy and limited-aperture conditions via specialized indicator functions.

Direct Sampling Methods for Inverse Medium Scattering of Elastic Waves: A Technical Analysis

Problem Setting and Reformulation

The paper "Direct sampling methods for inverse medium scattering problems of elastic waves" (2607.03745) addresses the reconstruction of penetrable obstacles embedded in a homogeneous elastic background from far-field data of time-harmonic elastic waves. The forward (direct) problem is governed by the 2D Navier equations for linear isotropic elasticity, coupled by transmission conditions of displacement and traction on the interface $\partial D$. The inverse problem is to determine the obstacle's location and geometry from measured compressional and shear far-field patterns.

A central innovation is the application of Helmholtz decomposition to recast the vector Navier equations into a coupled system of scalar Helmholtz equations for the potential fields. This reformulation enables the use of scalar boundary integral operator theory and associated numerical schemes, leading to improved computational efficiency and analytical tractability relative to direct treatments of the vector system involving hypersingular operators.

The Forward Problem: Uniqueness and Numerical Scheme

Boundary Integral Reduction

By representing the interior and scattered fields with potential functions—$\phi$ (compressional) and $\psi$ (shear)—the coupled system is reduced to scalar Helmholtz equations in each region, with intricate transmission conditions coupling normal and tangential derivatives and second-order boundary traces. The well-posedness of the forward problem is rigorously established: the authors prove uniqueness for the coupled boundary value problem through Rellich-type arguments and the properties of scalar Helmholtz solutions, ultimately showing that the only solution under homogeneous data is trivial.

Boundary Integral Formulation and Nyström Discretization

The boundary value problem is reformulated into a coupled four-density boundary integral equation (BIE) system using single-layer potentials for each scalar field. This leads to integral operators that require singular quadrature, for which the authors employ a high-order Nyström discretization leveraging kernel decompositions and custom quadrature rules (as in [Drule], [Kress]).

The integral system admits a unique solution provided the scalar wavenumbers are not interior Dirichlet eigenvalues. Detailed parameterizations for arbitrary analytic obstacle boundaries are specified, allowing for robust numerical evaluation of the BIE system. The implementation achieves high accuracy, with relative errors on the order of $10^{-13}$ to $10^{-14}$ in representative low-frequency cases with sufficient quadrature nodes, and demonstrates scalability to high-frequency regimes with increased discretization.

The Inverse Problem: Direct Sampling Indicators

Indicator Definitions

Building on the direct sampling methodology for elastic scattering ([DSM]), the authors extend the framework to penetrable (transmission) obstacles. Three sampling indicators are proposed:

• $\mathbf{I}_F$: utilizes the full matrix of compressional and shear far-field data for all incident and observation directions.
• $\mathbf{I}_P$: uses only compressional (P-wave) components of the far field.
• $\mathbf{I}_S$: uses only shear (S-wave) components.

Each indicator is defined as a double integral of the measured far-field pattern multiplied by suitably constructed test functions (plane waves at each sampling point with compressional or shear wavenumber), without explicit solution of an auxiliary "background" problem. These indicators are designed to peak near the boundary of the obstacle, decaying rapidly away from it.

Analytical Decay and Stability

Rigorous analysis demonstrates that the indicators decay as Bessel functions of increasing sampling-point distance from the boundary, following the oscillatory and decaying nature of scalar Helmholtz solutions. The decay mechanism is analytically tracked using integral representations and properties of the Helmholtz Green's function and far-field patterns. The stability of the indicators is established in $L^2$ norm: the perturbation in the indicator due to noise in the far-field data is linearly controlled by the noise amplitude, independent of the sampling location. This holds for all indicators by direct application of Cauchy–Schwarz.

Numerical Results

A wide suite of numerical experiments confirms the theoretical findings. Reconstructions are performed for a variety of boundary shapes (including pear, flower, leaf, kite, apple, peanut, circles, rectangles, and multi-component configurations) under both full and limited aperture and with up to $30\%$ relative noise in the data.

Key empirical outcomes:

• All three indicators ($\phi$0, $\phi$1, $\phi$2) accurately localize the boundary and shape of the obstacle in full aperture and low to moderate noise regimes.
• Resolution is robust to moderate noise and somewhat tolerant to limited-aperture data, with boundary details recovered best near available incident/observation angles.
• Multi-obstacle and multi-component cases are qualitatively resolved, with performance degrading only when obstacles approach or overlap closely.
• High accuracy is achieved with moderate computational cost due to the direct, non-iterative nature of the method.

Practical and Theoretical Implications

The direct sampling strategy realized here offers a simple, efficient, and robust approach for qualitative identification of penetrable elastic inhomogeneities from minimal a priori information. The scalar reduction not only streamlines theoretical analysis (uniqueness, decay, stability) but also enables high-order numerics for complex geometries and high frequencies, which is generally challenging in vector elasticity.

On the theoretical side, the equivalence of the elastic far-field patterns with those from the scalar Helmholtz system (via the Helmholtz decomposition) is shown to underlie the efficacy of the indicators. This links the inverse elastic scattering problem to better-understood scalar inverse problems, permitting the transfer of theory and methods.

In applied settings (nondestructive testing, medical elastography, seismic imaging), the direct sampling scheme can serve as a rapid initial diagnosis or as an efficient module in a multi-stage inverse solver. Furthermore, the framework is directly extensible to multi-component and multi-physics settings (e.g., hybrid acoustic-elastic media, or fluid-structure interaction), as intimated in the appended multi-obstacle system.

Prospects for Further Development

Potential directions for extending this work include:

• 3D extension: While the current theory and numerics are 2D, extension to 3D vector elasticity will likely require further advances in the analysis and computational treatment of vector Helmholtz decompositions and boundary integral solvers.
• Anisotropic and heterogeneous media: The current framework assumes isotropy and homogeneity outside the obstacle; real-world applications may require accommodating more complex material models.
• Integration with data-driven methods: Direct sampling indicators can be integrated within hybrid frameworks (e.g., learning-based regularization or parameter estimation modules as suggested in [dp1]), potentially enhancing performance in challenging inverse problems with strongly limited data.
• Adaptive multi-frequency and multi-aperture strategies: Adaptive selection of frequencies or incident/observation apertures based on current reconstructions could improve resolution and robustness in practice.

Conclusion

This paper demonstrates a mathematically rigorous and numerically precise methodology for the inverse reconstruction of penetrable elastic scatterers, leveraging a potent combination of Helmholtz decomposition, boundary integral formulation, high-order discretization, and direct indicator sampling. The analytical results (uniqueness, decay rate, stability) and comprehensive numerical experiments establish this as a competitive tool for both theoretical investigation and applied inverse elasticity imaging (2607.03745).