Weyl–Titchmarsh Theory in Spectral Analysis

• Weyl–Titchmarsh Theory is a framework for analyzing spectral properties of self-adjoint operators using the m-function and Dirichlet-to-Neumann map.
• It provides an analytic tool to characterize eigenvalues and distinguish between absolutely continuous, singular continuous, and discrete spectra.
• The theory unifies boundary value formulations and inverse problems via Kreĭn-type resolvent formulas, applicable to multidimensional and operator-valued settings.

Weyl–Titchmarsh Theory

The Weyl–Titchmarsh theory generalizes the spectral analysis of self-adjoint differential operators, originally formulated for the one-dimensional Sturm–Liouville problem, to higher-dimensional, matrix-valued, operator-valued, and discrete settings. Central to the framework is the Weyl–Titchmarsh function (or m-function), an analytic (typically Herglotz) function or operator-valued map that encodes the full spectral information of the operator—eigenvalue structure, spectral type, and multiplicities—through its analytic structure, limiting behavior, and associated boundary value problems. Modern developments recognize the Dirichlet-to-Neumann map as a natural higher-dimensional or operator-theoretic generalization of the classical m-function.

1. Fundamental Setup and Core Definitions

Consider a self-adjoint Schrödinger operator on an unbounded exterior domain $\Omega \subset \mathbb R^n$, $n\ge2$, with $\mathbb R^n \setminus \Omega$ nonempty, bounded, and $C^2$ boundary $\partial\Omega$. For a real, bounded potential $q\in L^\infty(\Omega)$, define

$$A u = (-\Delta + q)u, \qquad \operatorname{dom}(A) = \{ u\in H^2(\Omega):\ u|_{\partial\Omega}=0 \}.$$

$A$ is self-adjoint, bounded below, and its spectrum accumulates only at $+\infty$ (Behrndt et al., 2012).

The multidimensional Weyl–Titchmarsh theory associates to $A$ the Dirichlet-to-Neumann (DtN) map: $$M(\lambda)g := \partial_\nu u_\lambda|_{\partial\Omega}, \qquad u_\lambda \in H^2(\Omega)\ \text{solving}\ (-\Delta + q - \lambda)u_\lambda=0,\ u_\lambda|_{\partial\Omega}=g.$$ Here $$M(\lambda):H^{3/2}(\partial\Omega) \to H^{1/2}(\partial\Omega)$$ is an operator-valued holomorphic function, viewed as an analogue of the scalar m-function in one dimension. The Poisson operator $$\gamma(\lambda):H^{3/2}(\partial\Omega)\to H^2(\Omega)$$, $\gamma(\lambda)g = u_\lambda$, maps boundary data to interior solutions.

For operator-valued, vector-valued, or discrete settings, the construction follows the same paradigm: identify a boundary or initial point, construct canonical solutions (matrix-, operator-, or vector-valued), and define the m-function as the boundary data-to-derivative map that singles out the square-integrable or "Weyl" solution (Gesztesy et al., 2011, Acharya, 2017, Anderson, 2010).

2. Spectral Characterization via the M-Function

The Dirichlet-to-Neumann map $M(\lambda)$ encodes all spectral properties of $A$. For real $x\notin \sigma_p(A)$, the strong limit $\lim_{\eta\to0+} M(x + i\eta)g$ exists for each $g$. The principal relations are:

• $x\in\rho(A) \iff M(\lambda)$ extends analytically (as an operator-valued function) to $\lambda=x$.
• $$x\in\sigma_p(A) \iff \text{``s-lim''}_{\eta\to0+} \eta M(x + i\eta) \neq 0$$.
• Simple poles of $M(\lambda)$ at $\lambda_0$ correspond to isolated eigenvalues; the residue $\operatorname{Res}_{\lambda_0}M$ gives normal derivatives of eigenfunctions.
• $$x\in\sigma_c(A) \iff \text{``s-lim''}_{\eta\to0+} \eta M(x + i\eta) = 0$$ but $M(\lambda)$ cannot be analytically continued. These statements generalize the classical one-dimensional correspondence between poles/jumps of the m-function and spectral data (Behrndt et al., 2012).

The spectrum type is further refined:

• The absolutely continuous spectrum is characterized by growth properties of $\operatorname{Im} M(x + i0)$:

$$\sigma_{\mathrm{ac}}(A) = \bigcup_{g \in H^{3/2}(\partial\Omega)} \operatorname{cl}_{\mathrm{ac}} \{ x \in \mathbb R: 0 < -\operatorname{Im}(M(x+i0)g, g) < \infty \},$$

where $\operatorname{cl}_{\mathrm{ac}}$ denotes essential closure.

• Absence of singular continuous spectrum in an interval $(a,b)$ follows if, for all $g$, the set where $\operatorname{Im}(M(x+i0)g, g) = +\infty$ and $\lim_{\eta\to0+} \eta(M(x + i\eta)g, g)=0$ is at most countable (Behrndt et al., 2012).

3. Analytic Structure, Kreĭn-Type Resolvent Formulas, and Operator Theory

The operator-valued M-function admits rich analytic and algebraic structure:

• $$\gamma(\lambda)^*:L^2(\Omega)\to L^2(\partial\Omega)$$ is given by the boundary normal derivative of the $(A-\bar\lambda)^{-1}$ applied to $u$.
• The resolvent admits a Kreĭn-type formula:

$$(A-\lambda)^{-1}-(A-\mu)^{-1} = \gamma(\lambda) [M(\lambda) - M(\mu)] \gamma(\mu)^* = (\lambda - \mu) \gamma(\lambda) \gamma(\mu)^*.$$

• The difference $$M(\lambda) - M(\mu)^* = (\lambda - \mu)\gamma(\mu)^*(A-\lambda)^{-1}\gamma(\mu)$$ (Behrndt et al., 2012).
• The real part $\operatorname{Re} M(\lambda)$ is explicitly controlled via the resolvent and $\gamma(\mu)$.
• For fixed boundary data $g$, the strong limit $\lim_{\eta\to0+} M(x + i\eta)g$ exists in $L^2(\partial\Omega)$ for real $x\notin\sigma_p(A)$.

These properties parallel and generalize the classical scalar Herglotz property and resolvent identities of m-functions.

4. Algebraic and Analytic Dependencies: Examples and Comparison

Special cases illustrate the Weyl–Titchmarsh framework:

• On bounded domains, $\sigma(A)$ is purely discrete and $M(\lambda)$ is meromorphic; poles coincide with Dirichlet eigenvalues.
• For radially symmetric exterior domains, separation of variables reduces the analysis to the one-dimensional channel, and the classical m-function emerges in each partial wave.
• In one dimension, the scalar m-function is a Herglotz function whose poles correspond to eigenvalues and whose jumps/density correspond to absolutely continuous or singular continuous spectral components (Behrndt et al., 2012, Derkach et al., 2024).
• In matrix, vector, or operator-valued settings, the Weyl–Titchmarsh function is an operator or matrix-valued Nevanlinna–Pick function (e.g., maps the upper half-plane into the Siegel upper half-space) (Gesztesy et al., 2011, Acharya, 2017).

A table summarizing spectrum-mapping by $M(\lambda)$:

Spectral Component	M-function Behavior	Comments
Isolated eigenvalue	Simple pole of $M(\lambda)$	Residue encodes normal trace
Embedded or continuous	No analytic continuation/$\eta M(x+i\eta)\to 0$	Limiting absorption principle
Absolutely continuous	$0 < -\operatorname{Im}(M(x+i0)g,g) < \infty$	For $g\in H^{3/2}(\partial\Omega)$
Singular continuous (absence criterion)	Exceptional divergence of $\operatorname{Im} M$	See also conditions in (Behrndt et al., 2012)

5. Generalizations: Matrix-, Operator-, and Time Scale Theory

The Weyl–Titchmarsh paradigm admits extensive generalization:

• For Schrödinger operators with operator- or matrix-valued potentials, an operator- or matrix-valued m-function (Weyl–Titchmarsh function) is constructed, with Herglotz–Nevanlinna property and corresponding spectral representation (Gesztesy et al., 2011, Acharya, 2017).
• The framework applies to discrete operators (Jacobi and CMV matrices, including those with critical/double-root behavior or Wigner-von Neumann type perturbations), with spectral density formulae in terms of the boundary value of the m-function (Naboko et al., 2019, Janas et al., 2010, Clark et al., 2010, Simonov, 2010).
• Unified continuous/discrete analysis appears in dynamic Hamiltonian systems on Sturmian time scales, where Weyl disks converge to limiting sets encoding spectral subspaces (Anderson, 2010, Anderson, 2010).
• For first-order symmetric or more general systems, the m-function encodes extensions and spectral data even when minimal deficiency indices are unequal, via boundary triplets and Nevanlinna families (Albeverio et al., 2012).
• Singular and distributional potentials are encompassed by an appropriate extension of the singular Weyl–Titchmarsh function, which may reside in a generalized Nevanlinna class (Kostenko et al., 2010, Kostenko et al., 2010, Eckhardt et al., 2012).

6. Physical and Computational Interpretations

The Weyl–Titchmarsh theory underlies both direct spectral analysis and inverse problems. In computational and applied contexts:

• The m-function (or M-function) acts as an exact Dirichlet-to-Neumann (DtN) operator for the stationary or time-dependent Schrödinger equation, enabling the formulation of transparent and absorbing boundary conditions, rational approximation, and numerical schemes efficient across broad frequency ranges (Ehrhardt et al., 2024, Derkach et al., 2024).
• In multidimensional and PDE contexts, the Dirichlet-to-Neumann operator plays a central role in computational methods for open or exterior domain problems, including scattering and quantum dynamics (Behrndt et al., 2012, Ehrhardt et al., 2024).
• The abstract machinery also governs boundary control, system-theoretic identification, and boundary inverse problems, with the spectral and analytic properties of the M-function providing the theoretical foundation (Avdonin et al., 29 May 2025).

7. Structural Synthesis: Unifying Principles

The Weyl–Titchmarsh framework provides a transparent, structurally unifying apparatus for spectral theory:

• The (scalar, matrix, or operator-valued) M-function serves as a complete unitary invariant for the operator under appropriate extension and boundary parametrizations.
• The analytic nature, pole structure, and limiting absorption behavior of $M(\lambda)$ encode eigenvalues, embedded spectrum, and absolutely/the singular continuous spectrum on equal footing.
• The functional calculus, spectral measure, and direct/inverse transforms are reconstructed from the M-function or its Herglotz representation.
• All major analytical and algebraic features of the classical Sturm–Liouville theory—Kreĭn resolvent formula, residue calculus, boundary value characterization—are preserved and generalized, with the Dirichlet-to-Neumann map and Poisson operator as central conceptual tools (Behrndt et al., 2012, Gesztesy et al., 2011, Acharya, 2017).

This theory sets the foundation for multidimensional, non-scalar, and operator-theoretic spectral analysis, as well as for modern computational approaches in mathematical physics and engineering.