Non-Diffracting Resonant Angular Filter

• The paper introduces a quantum-graph formulation that achieves discrete angular selectivity with perfect transmission at selected incidence angles via resonant beyond-nearest-neighbor coupling.
• Finite-element simulations and acoustic meta-grating realizations verify that the filter confines energy to a single propagating channel below the diffraction limit.
• The work extends to flux-tunable configurations, demonstrating that magnetic phase shifts and δ-type vertex conditions can continuously control transmission angle and amplitude.

A non-diffracting resonant angular filter is a periodic interface which, at a given frequency, only transmits waves incident from a discrete set of angles, with unit transmission at those angles and effectively zero transmission at all others. In the formulation introduced for infinitely periodic quantum graphs and realized numerically as an acoustic meta-grating, angular selectivity is produced by resonant, beyond-nearest-neighbour coupling along the interface, while “non-diffracting” denotes operation below the grating diffraction threshold so that only a single transmitted propagating channel is available (Lawrie et al., 2024, Lawrie et al., 7 Oct 2025).

1. Definition and physical principle

The central object is a one-dimensionally periodic interface, with period $\ell$ along the interface direction, separating two half-spaces or waveguides. Unlike an ordinary diffraction grating, the interface is not designed to redistribute incident power among multiple propagating diffraction orders. Instead, the operating frequency is chosen so that higher orders are evanescent, and the interface switches the only propagating transmitted channel between almost total reflection and perfect transmission at selected tangential wavenumbers (Lawrie et al., 2024).

In the acoustic realization summarized later in the same research line, the filter consists of a periodic array of openings whose internal channels are interconnected beyond nearest neighbours. The earlier acoustic filter was interpreted in terms of internal harmonic modes that decouple from the environment; off resonance these modes impose effective Dirichlet conditions at the junctions, while at specific incidence angles the tangential wavenumber matches that of the internal bound mode and the effective boundary condition becomes periodic, enabling unit transmission (Lawrie et al., 7 Oct 2025).

This mechanism produces a discrete angular filter in $k$-space. For a plane wave in a homogeneous surrounding medium, the tangential Bloch wavenumber is $\kappa_y = k\sin\theta$, so a discrete set of admissible $\kappa_y$ values corresponds directly to a discrete set of incidence angles. The interface can therefore be designed to yield perfect transmission at customizable angles of incidence, without diffraction, and with transmission confined to arbitrarily narrow wavenumber windows (Lawrie et al., 2024).

2. Quantum-graph formulation

The original theory is developed on a discrete, infinitely periodic quantum graph. Vertices are indexed by $m\in\mathbb{Z}$ and located at $y=m\ell$. Each vertex is connected to a left lead, a right lead, and two internal bonds, denoted $d$ and $u$, that connect the vertex to its $\mu$-th neighbours along the interface. The non-locality is encoded by $\mu>1$, so the internal bonds implement beyond-nearest-neighbour coupling (Lawrie et al., 2024).

On each edge $k$0, the stationary field satisfies the one-dimensional Helmholtz equation

$k$1

and a Bloch-form solution is written as

$k$2

Here $k$3 is the operating wavenumber, $k$4 is the edge coordinate, and $k$5 is the Bloch quasi-momentum along the interface (Lawrie et al., 2024).

At every vertex, Kirchhoff–Neumann conditions are imposed. Continuity requires

$k$6

and current conservation requires

$k$7

These conditions yield a local vertex scattering matrix with entries

$k$8

for $k$9. The internal bonds add a second layer of phase matching: propagation along a bond of length $\kappa_y = k\sin\theta$0 contributes the phase $\kappa_y = k\sin\theta$1, while periodicity along the interface contributes $\kappa_y = k\sin\theta$2 (Lawrie et al., 2024).

The graph description is analytically advantageous because it separates the interface design problem into a local scattering law at each junction and a non-local propagation law along the resonant bonds. This is the structural source of the filter’s discrete angular response.

3. Resonance law and discrete pass directions

Eliminating the internal bond amplitudes yields an effective $\kappa_y = k\sin\theta$3 scattering matrix for the left and right leads,

$\kappa_y = k\sin\theta$4

with transmission coefficient

$\kappa_y = k\sin\theta$5

This closed-form expression contains the complete angle-frequency response of the ideal interface (Lawrie et al., 2024).

The resonance condition is

$\kappa_y = k\sin\theta$6

which means that the beyond-nearest-neighbour bonds have lengths equal to half-integer multiples of the wavelength. Under this condition, the perfect-transmission angles are determined by

$\kappa_y = k\sin\theta$7

or, equivalently, by the discrete set

$\kappa_y = k\sin\theta$8

At resonance, the transmission becomes binary: $\kappa_y = k\sin\theta$9 The filter therefore acts as a discrete angular selector in tangential wavenumber space (Lawrie et al., 2024).

Away from the discrete pass values, the resonant bonds impose effective Dirichlet conditions at the interface, so the structure behaves as a hard reflector. At the pass values, the effective boundary condition becomes periodic, and the interface opens a perfectly transmitting channel. This sharp switching is the basis for the claim that the angular passband can be made arbitrarily narrow: detuning is controlled by the resonant bond length $\kappa_y$0 and the operating proximity to $\kappa_y$1 (Lawrie et al., 2024).

A common misconception is to treat the device as a conventional grating. That interpretation is inaccurate in the operating regime of interest. The filter is explicitly designed so that only the specular transmitted order propagates; the angular response comes from resonant non-local coupling, not from distributing energy among propagating diffraction orders (Lawrie et al., 2024).

4. Acoustic realizations

The original theory was realized numerically as an acoustic meta-grating in several settings. In the discrete implementation, the surrounding medium is a square lattice of acoustic tubes, and the interface is formed by helices of length $\kappa_y$2 that connect lattice points separated by $\kappa_y$3 along the interface. In the reported example, the tube radius is $\kappa_y$4, the lattice period is $\kappa_y$5, the resonant bonds have length $\kappa_y$6, and $\kappa_y$7 (Lawrie et al., 2024).

The same filtering geometry was then transferred to continuous acoustic waveguides. One realization uses two rectangular waveguides joined by a periodic array of vertical tubes or helices at the interface. A second realization uses a single continuous waveguide whose top wall contains periodically spaced holes connected to the resonant filter. In both cases, the operating frequency is chosen well below the grating diffraction limit, so that only one transmitted mode exists and the response remains non-diffractive (Lawrie et al., 2024).

Finite-element simulations reproduce the analytical quantum-graph predictions. In the discrete lattice environment, point-source excitation generates transmitted beams only at the predicted discrete angles. In the continuous two-waveguide and single-waveguide realizations, the angle-frequency transmission maps show narrow resonant lobes that track the graph-theoretic pass conditions. The continuous environments modify the detailed dispersion and therefore the exact angular positions, but the discrete-angle transmission mechanism persists (Lawrie et al., 2024).

These realizations also clarify the physical meaning of the graph parameters. Graph edges correspond to acoustic pipes supporting the fundamental mode; vertices correspond to junctions satisfying continuity of pressure and conservation of acoustic flux; and the bond length $\kappa_y$8 sets the resonant frequencies

$\kappa_y$9

with $m\in\mathbb{Z}$0. The graph is therefore not merely an abstract model but a direct design language for the acoustic interface.

5. Tunability and later generalizations

A later development replaced the classical wave equation on the internal network with the magnetic Schrödinger equation and introduced tunable $m\in\mathbb{Z}$1-type vertex conditions. In that version, the same filtering geometry becomes flux-tunable: a magnetic phase shift generated by a solenoidal flux $m\in\mathbb{Z}$2 and a vertex parameter $m\in\mathbb{Z}$3 allow both the transmission angle and the transmission coefficient to be controlled continuously (Lawrie et al., 7 Oct 2025).

The corresponding transmission coefficient is

$m\in\mathbb{Z}$4

At resonance, the discrete pass directions become

$m\in\mathbb{Z}$5

and the transmission amplitude at those directions becomes

$m\in\mathbb{Z}$6

The original acoustic filter is recovered by setting $m\in\mathbb{Z}$7 and $m\in\mathbb{Z}$8, in which case the pass directions are topology-fixed and the transmission at those angles is unity (Lawrie et al., 7 Oct 2025).

This later formulation shows that the original non-diffracting resonant angular filter can be understood as the zero-flux, Kirchhoff-coupled limit of a broader class of discrete angular filters. A plausible implication is that topology, resonance, and tunable phase all enter the same scattering architecture through a small set of analytically tractable parameters.

6. Relation to broader angular-filter research

The discrete graph-based filter belongs to a wider research program on angularly selective interfaces, but it is distinguished by its combination of non-diffracting operation and discrete pass directions. In contrast, non-uniform metagratings have been used to synthesize continuous low-pass, high-pass, and all-pass angular transfer functions of the fundamental Floquet mode at $m\in\mathbb{Z}$9, with the design cast as an impedance-matrix optimization over reactive loads (Kim et al., 27 Jan 2026). Those structures engineer spatial dispersion over a continuous angular spectrum rather than producing topology-fixed discrete pass angles.

Optical nonlocal metasurfaces provide a second adjacent line of work. Non-Hermitian metasurfaces based on symmetry-protected bound states in the continuum have been proposed as compact spatial filters operating over an angular range of approximately $y=m\ell$0 around normal incidence, with low-pass and high-pass behavior determined by nonlocal guided-mode resonances (Chen et al., 26 Jun 2025). The “Metapinhole” architecture extends this idea to planar Fourier optics without lenses, using metagratings to realize low-pass and controllable high-pass filtering in transmission and band-pass filtering in reflection (Abouelatta et al., 6 Sep 2025). These are angular filters in the continuous Fourier-optics sense, not discrete graph filters.

A third related direction concerns non-diffracting beam generation by momentum-space ring selection. High-quality-factor nonlocal metasurfaces have been used to generate vortex Bessel beams, establishing a direct link between non-diffracting-beam generation and photonic band curvature, and demonstrating an order-of-magnitude enhancement in propagation distance compared to conventional Laguerre–Gaussian modes (Kim et al., 30 Sep 2025). This suggests a complementary viewpoint: the discrete graph filter suppresses unwanted angular channels, whereas nonlocal metasurfaces can shape a narrow angular spectrum into a Bessel-like transmitted or reflected beam.

Across these developments, the graph-based non-diffracting resonant angular filter remains notable for the clarity of its transmission law, the explicit role of beyond-nearest-neighbour coupling, and the fact that perfect angular selectivity is achieved without opening diffracted propagating orders (Lawrie et al., 2024).