Technical Summary: Transmission Eigenvalues and Non-scattering

1. Problem Statement and Context

This paper surveys recent developments in the theory of scattering and non-scattering for the linear Helmholtz equation in the presence of inhomogeneous media with nontrivial contrast. The central inquiry is whether there exist time-harmonic incident waves at specific frequencies (wave numbers $k$) that are not scattered by a given inhomogeneity $(A_D, n_D, D)$. Such frequencies are termed non-scattering wave numbers, and the corresponding media are non-scattering inhomogeneities.

The problem is framed within the context of inverse scattering, where the scattered field carries information about the inhomogeneity. The existence of non-scattering waves implies a failure of injectivity for the far-field operator, potentially leading to the failure of certain reconstruction methods.

The authors distinguish between two primary scenarios:

1. Isotropic/Scalar Contrast ($A \equiv I$): The inhomogeneity differs from the background only in the refractive index $n$.
2. Anisotropic/Vector Contrast ($A \not\equiv I$): The inhomogeneity possesses a distinct conductivity matrix $A$ and refractive index $n$.

A fundamental necessary condition for non-scattering is that the wave number $k$ must be a transmission eigenvalue. However, the paper emphasizes that being a transmission eigenvalue is not sufficient for non-scattering; additional regularity conditions on the inhomogeneity and the incident wave are required.

2. Mathematical Formulation

The model considers a bounded domain $D \subset \mathbb{R}^d$ ($d=2,3$) with Lipschitz boundary $\partial D$. The medium is characterized by a matrix $A$ and scalar $n$, where $A=I$ and $n=1$ outside $D$. An incident wave $u_{in}$ satisfies the Helmholtz equation $\Delta u_{in} + k^2 u_{in} = 0$.

Non-scattering occurs if the scattered field $u_{sc}$ vanishes outside $D$. This leads to an overdetermined boundary value problem inside $D$:
$$\begin{cases} \nabla \cdot (A_D \nabla u) + k^2 n_D u = 0 & \text{in } D \\ \Delta v + k^2 v = 0 & \text{in } D \\ u = v, \quad \nu \cdot A_D \nabla u = \nu \cdot \nabla v & \text{on } \partial D \end{cases}$$
where $v = u_{in}|_D$ and $u = v + u_{sc}$. The values of $k$ for which this system admits a non-trivial solution $(u, v)$ are the transmission eigenvalues.

3. Methodology and Key Results

3.1 The Case $A \equiv I$ (Scalar Contrast)

In this section, the authors analyze media where only the refractive index $n_D$ differs from 1.

• Spherically Symmetric Inhomogeneities:

  • For spherically symmetric $n(r)$, the transmission eigenvalue problem reduces to a system of ordinary differential equations involving spherical Bessel functions.
  • Result: If $n(1) \neq 1$ or the integral mean $M_n \neq 1$, there exist infinitely many real transmission eigenvalues. Crucially, for these specific geometries, every real transmission eigenvalue corresponds to a non-scattering wave number. The corresponding incident waves are Herglotz wave functions (superpositions of plane waves).
  • Complex Eigenvalues: Under specific smoothness conditions ($n \in C^3$), the paper notes the existence of infinitely many truly complex transmission eigenvalues, confirming the non-selfadjoint nature of the problem.

• General Inhomogeneities and Existence of Real Eigenvalues:

  • The authors utilize a variational formulation involving fourth-order operators to prove the existence of real transmission eigenvalues.
  • Theorem 2.2: If $\partial D$ is Lipschitz and $n$ is positive with a uniform sign of contrast ($n-1$) in a boundary neighborhood $D_\delta$, there exists a countable infinite set of real transmission eigenvalues accumulating only at $+\infty$. This result holds even without assuming the contrast has a uniform sign in the entire domain, provided it holds near the boundary.

• Regularity and Non-scattering (The Free Boundary Connection):

  • The paper establishes a deep connection between non-scattering and free boundary regularity.
  • Theorem 2.3: If an inhomogeneity is non-scattering for a non-degenerate incident wave (where $(n-1)v \neq 0$), the boundary $\partial D$ must possess high regularity. Specifically:
    • If $n$ is $C^{m,\mu}$, then $\partial D$ must be $C^{m+1,\mu}$.
    • If $n$ is real analytic, then $\partial D$ must be real analytic.
  • Corollary: This implies that inhomogeneities with corners, edges, or conical singularities always scatter for generic incident waves. The non-scattering property forces the boundary to be smooth.

• Singularities:

  • Using Complex Geometric Optics (CGO) solutions, the paper cites results showing that inhomogeneities with corners or edges scatter all incident waves, provided the contrast does not vanish at the singularity.

• Analytic Inhomogeneities:

  • For real analytic $n_D$, the paper discusses conditions under which non-scattering fails. For instance, strictly convex domains with specific width ratios or domains with inward cusps are shown to scatter plane waves.

3.2 The Case $A \not\equiv I$ (Anisotropic Contrast)

This section addresses media with contrast in both the principal part ($A$) and the lower-order term ($n$).

• Existence of Eigenvalues:

  • Similar to the scalar case, under conditions where the contrast in $A-I$ and $n-1$ has a uniform sign (or specific relationships), there exists an infinite set of real transmission eigenvalues.
  • Spectral analysis (discreteness, Weyl asymptotics) is conditional on ellipticity assumptions at the boundary (Agmon-Douglis-Nirenberg complementing condition).

• Regularity of the Boundary:

  • Theorem 3.1: Analogous to the scalar case, if a non-scattering solution exists for a wave number $k$ and an incident wave satisfying a non-degeneracy condition ($\nu \cdot (A-I)\nabla u_{in} \neq 0$), then the boundary $\partial D$ inherits the regularity of the coefficients $A$ and $n$.
  • Methodology: The proof employs the Hodograph transform, mapping the unknown curved boundary to a flat one. The regularity of the inverse transform determines the regularity of $\partial D$.

• Counter-examples and Non-degeneracy:

  • The paper highlights that the non-degeneracy condition is essential. It provides an example of a non-smooth domain (a square) with constant anisotropic coefficients where non-scattering occurs for specific incident waves (e.g., $\cos(k_1 \pi x_1)\cos(k_2 \pi x_2)$). In this case, the gradient of the incident wave vanishes at the corners, violating the non-degeneracy condition required for the regularity theorems.
  • Transformation Optics: The authors discuss a construction using diffeomorphisms ($\Phi$) to create anisotropic media that are non-scattering at all frequencies. However, they clarify that such constructions do not contradict the regularity theorems because the resulting boundary $\partial D$ must already possess the regularity asserted by the theorem if the non-degeneracy condition is met.

4. Significance and Claims

The paper positions itself as a survey of recent theoretical advances, emphasizing the following points:

1. Distinction between Eigenvalues and Non-scattering: The existence of transmission eigenvalues is a necessary but not sufficient condition for non-scattering. The paper clarifies that non-scattering imposes strict geometric constraints (smoothness) on the inhomogeneity.
2. Free Boundary Perspective: A major contribution is framing the non-scattering problem as a free boundary problem. This perspective allows the application of regularity theory (Kinderlehrer-Nirenberg, Caffarelli) to prove that "almost all" singularities (corners, edges) cause scattering.
3. Limitations of Non-scattering: The results suggest that non-scattering inhomogeneities are rare and highly structured. They typically require smooth boundaries and specific relationships between the material properties and the incident wave.
4. Inverse Scattering Implications: The non-injectivity of the scattering operator at non-scattering wave numbers is identified as a fundamental obstacle in inverse scattering reconstruction methods. Understanding the conditions for non-scattering is therefore critical for the stability and uniqueness of inverse problems.

The authors explicitly state that the existence of non-scattering inhomogeneities with corners or edges is generally impossible for non-degenerate waves, and that for analytic media, non-scattering is restricted to very specific geometries (like spheres) or requires the incident wave to vanish at singular points. The paper does not propose new experimental setups but rather consolidates the mathematical understanding of when and why invisibility (non-scattering) can or cannot occur.