Boundary-Sensitive Localization Physics

Updated 24 June 2026

Boundary-sensitive localization physics is the study of how specific boundary conditions, like Robin, Dirichlet, and Neumann, directly generate or suppress localized states in classical and quantum models.
It examines the influence of boundary geometry and disorder on localization metrics, such as eigenfunction decay rates and participation ratios, across diverse systems such as photonics and non-Hermitian chains.
The field drives practical insights for optimizing sensor designs, enhancing inverse problem techniques, and controlling dissipative and stochastic processes in engineered systems.

Boundary-sensitive localization physics concerns the emergence, suppression, or character of spatially localized states—classical or quantum—that arise specifically due to the presence, nature, or geometry of system boundaries, as opposed to localization induced by bulk disorder, bulk interactions, or global symmetries alone. Boundary conditions—either explicit (Dirichlet, Neumann, Robin), dynamically induced (through openness or coupling), or arising from topological or geometric singularities—govern spectral features, eigenfunction localization length, response functions, and observable quantities across quantum field theory, condensed matter, photonics, stochastic dynamics, and inverse problems.

1. Boundary Conditions as Localization Mechanisms

Boundary-sensitive localization arises when the imposition or tuning of boundary conditions defines or dramatically modifies the structure of localized modes. In scalar field theory on Minkowski spacetime, excising a point (the spatial origin) and imposing Robin boundary conditions at the puncture yields a discrete bound state exponentially localized near the excised region, with frequency $\omega_b = \sqrt{m_0^2 - 1/\beta^2}$ and radial profile $R_b(r) \propto e^{-r/\beta}/r$ . The explicit condition

$\lim_{r\to 0^+} \left[ r R + \beta (rR)' \right] = 0$

parametrized by $\beta\in\mathbb{R}$ , selects self-adjoint extensions and generates this localized sector directly from field-theoretic principles—without ad hoc confining potentials or external trapping fields (Ramos et al., 24 Sep 2025). This approach generalizes to spacetimes with genuine singularities (e.g., conical or global monopole backgrounds), where the same family of Robin-like extensions governs the existence of boundary-localized bound states.

Similarly, in quantum and optical lattice models, boundary interpolation and impurity strength can interpolate between Bloch-like extended states, scale-free boundary-localized states, and exponentially localized skin modes. The boundary parameter (e.g., link strength $\eta$ ) appears directly in self-consistency conditions (Lyapunov exponent or secular equations) that distinguish between scale-free localization (localization length diverging linearly with system size) and Anderson-like or skin-effect phases (finite correlation length, characteristic exponential decay) (Zhang et al., 20 Feb 2025, Li et al., 2020, Li et al., 24 Mar 2026).

2. Boundary-induced Anderson and Scale-free Localization

In both photonic and electronic disordered lattices, spatially confined or disordered boundaries can serve as the principal scattering centers, inducing bulk localization even without bulk disorder. In two-dimensional "ribbon" photonic arrays, disorder limited to the lateral boundaries leads to backscattering and localization along the longitudinal direction. The inverse participation ratio (IPR)—a quantitative measure of localization—remains elevated for moderate ribbon width and moderate disorder, with the localization length $\xi$ scaling as $\xi \sim W^{-2}$ with disorder amplitude and $\xi\sim M$ with ribbon width. The effect weakens as the system width increases, reflecting the boundary origin of the localization (Molina, 2010).

In sharp contrast, one-dimensional semi-infinite arrays reveal that localization at the edge is generally weaker than in the bulk: disorder-induced trapping is less efficient near the boundary, necessitating higher levels of disorder to achieve comparable decay rates as in the interior. The boundary acts as a repulsive defect, reducing mode overlap at the edge, as confirmed experimentally and in numerical simulations (Szameit et al., 2010).

Non-Hermitian chains with a tunable boundary impurity illustrate even richer boundary sensitivity: as the link parameter is varied, the system transitions between periodic boundary, scale-free accumulation, reversed scale-free modes, and non-Hermitian skin-effect regimes. The scale-free localization length scales inversely with system size ( $\xi\propto N$ ), and infinitesimal changes in boundary parameters can dramatically reorganize the spectrum and eigenmode profiles (Li et al., 2020, Zhang et al., 20 Feb 2025, Li et al., 24 Mar 2026).

3. Geometric and Irregular Boundary Enhancement

Boundary geometry, specifically local curvature and roughness, strongly governs the localization of Laplacian eigenfunctions under Neumann conditions. In domains with conical spikes, narrow slits, or fractal boundaries, enhanced return probabilities of Brownian motion near the boundary lead to the concentration of low-lying eigenfunctions at high-curvature or trapping features. Quantitatively, Neumann eigenfunctions satisfy the inequality

$\phi_k(x)^2 \gtrsim \text{(geometry-dependent lower bound)}$

where, for instance, at the tip of a cone of opening $R_b(r) \propto e^{-r/\beta}/r$ 0, the mode amplitude is amplified by $R_b(r) \propto e^{-r/\beta}/r$ 1 compared to flat regions. In acoustics, this implies order-of-magnitude enhancements in local pressure fields and, consequently, in sound attenuation for walls engineered with spikes or fractal edges (Jones et al., 2018).

In inverse problems, the local mean curvature $R_b(r) \propto e^{-r/\beta}/r$ 2 of an inclusion boundary explicitly amplifies high-frequency components in shape reconstruction from generalized polarization tensors (GPTs) or scattering coefficients (SCs). The principal symbol of the relevant microlocal operators contains a $R_b(r) \propto e^{-r/\beta}/r$ 3 weight, indicating that at high curvature points, finer spatial details (higher-frequency Fourier modes) are more robustly encoded and stably recovered from far-field measurements—a rigorous mechanism for boundary-induced super-resolution (Ammari et al., 2019).

4. Boundary-sensitive Scaling Laws in Stochastic and Dissipative Systems

In stochastic lattice models with dual conservation laws (mass and energy), nonequilibrium steady states (NESS) induced by distinct boundary baths can exhibit spatial condensation (localization of a finite fraction of the conserved quantity), even when both boundaries are tuned to subcritical parameters. The criterion for condensation in NESS is geometric: if the line connecting boundary parameters $R_b(r) \propto e^{-r/\beta}/r$ 4 in the space of density and energy crosses the equilibrium condensation curve $R_b(r) \propto e^{-r/\beta}/r$ 5, a condensed, spatially localized profile emerges in the bulk. The localization is maintained dynamically by the balance of peak-creation and boundary absorption, rather than by static microcanonical constraints (Giusfredi et al., 2024).

In open quantum or dissipative (Lindblad) systems, boundary conditions radically alter the relaxation spectrum and the localization of Liouvillian eigenmodes (Liouvillian skin effect). Switching from periodic to open boundaries in non-Hermitian Lindbladian models can render all low-lying eigenmodes exponentially localized (skin-modes) or, depending on microscopic feedback mechanisms, alter only the excited-mode structure. The scaling of the relaxation time to steady state transitions from diffusive ( $R_b(r) \propto e^{-r/\beta}/r$ 6) to ballistic ( $R_b(r) \propto e^{-r/\beta}/r$ 7) due to boundary-induced mode localization and biorthogonality effects between left and right Liouvillian eigenmodes (Feng et al., 2023).

5. Information, Observability, and Boundary-aware Design

Boundary sensitivity arises not only at the physical level but also in the deployment of sensors and algorithms for localization tasks. In magnetic dipole localization, planar sensor arrays suffer geometric loss of observability near the boundary (e.g., close approaches or in the far field), signaled mathematically by the Fisher Information Matrix (FIM) developing small minimum eigenvalues and ill-conditioned inversion. Optimizing the sensor geometry—for instance, via staggered split-array topologies—substantially enhances the observability across the workspace, suppresses catastrophic boundary outliers, and pushes the error floor towards its Cramér-Rao bound (Xie et al., 24 Apr 2026).

Machine learning estimators that encode physics-aware features (e.g., input saturation masks, geometry-aware attention) can exploit boundary-induced information anisotropy, further improving localization robustness. In environmental source localization, PDE surrogates constructed with precise enforcement of boundary conditions (Dirichlet, Neumann) and mesh adaptation allow mobile sensor teams to exploit boundary geometry (e.g., obstacles, inlets) for more informative sampling, demonstrating superior convergence over heuristic or variance-only strategies (Shaffer et al., 17 Sep 2025).

6. Boundary-induced Universality and Non-universal Fluctuations

Boundary conditions can select between distinct universality classes of fluctuations even in otherwise "universal" bulk disordered systems. In two-dimensional Anderson localization, the full distribution of conductance fluctuations changes from the Tracy–Widom GUE to GOE class depending on whether the leads are narrow–narrow (NN; mimicking droplet initial conditions in KPZ) or wide–narrow (WN; flat initial conditions). This selection is robust to the bulk symmetry class (orthogonal/unitary)—breaking time-reversal with a magnetic field rescales the localization length $R_b(r) \propto e^{-r/\beta}/r$ 8 but leaves the conductance distribution functional form unchanged. This boundary-induced selection contradicts the traditional Thouless criterion, establishing a new boundary-sensitive paradigm in disordered transport (Swain et al., 23 Apr 2025).

7. Localization via Boundary Localization in Field Theory

In supersymmetric quantum field theory, localization techniques applied to manifolds with boundary ("gluing") enable the reduction of infinite-dimensional path integrals to boundary integrals over rigid BPS data. The induced boundary theories inherit localization properties controlled by the choice of boundary polarization and compatible supercharge. In supersymmetric quantum mechanics, the gluing integral reduces to sums over zero-dimensional vacua; in 3D and 4D theories on spheres, boundary localization yields finite-dimensional Coulomb/Higgs branch integrals with explicit one-loop measures. This framework renders the computation of protected quantities (partition functions, indices, correlators) manifestly sensitive to boundary conditions and provides a powerful mechanism to study boundary observables, interfaces, and dualities (Dedushenko, 2018).

References:

Boundary-sensitive mechanism	Example domains / models	Key reference(s)
Robin/self-adjoint extension induces localization	Massive scalar field with punctured Minkowski origin; conical and global monopole spacetimes	(Ramos et al., 24 Sep 2025)
Boundary-disorder–induced bulk Anderson localization	Ribbons with disordered edges in 2D; open-boundary suppression and repulsion in 1D optical lattices	(Molina, 2010, Szameit et al., 2010)
Geometric enhancement: curvature/roughness	Neumann Laplacian in conical/slit/fractal domains; GPT/SC-based inverse problems: curvature boosts local resolution	(Jones et al., 2018, Ammari et al., 2019)
Tunable boundary-driven scale-free or skin effects	Non-Hermitian chains: impurity interpolation, Floquet-driven chains: boundary-induced PT breaking and scale-free states	(Li et al., 2020, Zhang et al., 20 Feb 2025, Li et al., 24 Mar 2026)
Sensor/geometric optimization for boundary observability	Magnetic tracking: FIM, staggered arrays, boundary-aware GAA and PIF neural nets	(Xie et al., 24 Apr 2026)
Nonequilibrium localization from boundary driving	Stochastic mass/energy-conserving lattices: NESS condensation via parametric path crossing of condensation line	(Giusfredi et al., 2024)
Liouvillian skin effect in dissipative dynamics	Relaxation scaling, eigenmode structure in open vs periodic Lindblad chains	(Feng et al., 2023)
Supersymmetric field theory localization (gluing)	Path integrals reduced to finite-dimensional boundary data; boundary wavefunctions, mirror symmetry	(Dedushenko, 2018)
Universality class selection by boundary condition	KPZ-GOE vs GUE conductance distributions in 2D Anderson via narrow/wide leads	(Swain et al., 23 Apr 2025)

Boundary-sensitive localization thus serves as a unifying principle: boundary conditions, geometry, and interface properties fundamentally organize, enable, and tune localization phenomena across a wide domain of physical, mathematical, and engineering systems.