betaFeedback

WorksPhysics and Astronomy

A Passive Intelligent Metasurface: A Network Model

Abstract

Intelligent metasurfaces enable advanced wave manipulation, including negative refraction, focusing, beam splitting and beam steering, with significant potential for practical applications. However, conventional approaches often rely on active metamaterials with gain, or demand computationally intensive solutions of inverse problems to find the optimal configuration. In this work, we use the scattering framework of quantum graph theory to derive an exact, analytic expression for configuration of a passive metasurface that produces a prescribed reflected wave field. The metasurface terminates a periodic lattice of waveguides and consists of an array of channels, each a tube of length ℓm\ell_m ending in a vertex whose scattering parameter λmλ_m sets how much of the wave is lost to the wall. For any prescribed incident and reflected field, the exact lengths and scattering parameters follow channel by channel, in closed form, from the ratio of the wave amplitudes returning to and leaving the lattice, and the metasurface is passive whenever this ratio does not exceed unity in modulus. We demonstrate the construction for both a one-dimensional and a two-dimensional metasurface. The one-dimensional metasurface terminates a two-dimensional lattice and steers, splits or perfectly absorbs Gaussian beams. The two-dimensional metasurface terminates a three-dimensional lattice and generates a prescribed image in a plane inside the lattice. In each case, we obtain the exact boundary properties that produce the effect.

Introduction

Wave manipulation and wave control has emerged as an important research field, with applications in optics, wireless communication and acoustics, among other fields [1, 2, 3, 4]. The core question is, given a source and a (possibly complex or disordered) medium, how much control can be exerted over the resulting wave field by acting on a finite, structured set of degrees of freedom? Over the past two decades this question has been answered using linear wave theory and ideas from metamaterial design based on the following principles: a linear wave operator; a scattering or transfer matrix connecting inputs to outputs; and an optimisation problem over a constrained control space.

The seminal demonstrations of wavefront shaping through and around complex scattering media showed that even strongly multiple-scattering systems — normally scrambling coherent information irrecoverably — admit a well-defined input–output linear map, known as the transmission matrix, whose singular value decomposition reveals high-transmission “open channels” [5, 6, 7, 8]. Shaping the incident wavefront via spatial light modulators (SLM) [9, 10], spatial microwave modulators (SMM) [11] or an acoustic analogon [12, 13], allows one to send wave energy through opaque scattering layers thereby synthesising prescribed output patterns [1, 14]. This body of work, originally developed for optical speckle, has since been transposed to the microwave [15] and acoustic regimes [16], where the number of controllable degrees of freedom is more modest.

In a reverberant setting, such as wave fields in a room or cavity, multiple scattering typically leads to randomness of the wave field [17]. Here a different approach is tried and tested. Instead of shaping the incoming wave field to suit a given environment, the environment is itself transformed to provide a specified cavity wave field for a given source distribution [18, 19, 20, 21]. It has been shown to lead to controlled enhancement of the diversity and capacity of the communication channels both for microwave [22] and acoustic [23, 24] communication. The environment shaping effect is typically achieved by installing subwavelength-structured interfaces or panels — metasurfaces — capable of imposing an essentially arbitrary local phase, amplitude, or polarization response on an incident wave. Reconfigurability is achieved by embedding active elements (varactors, PIN diodes, MEMS switches, or their acoustic analogues) into each meta-atom, which in the simplest and most robust implementation yields a binary response, in which each unit cell can only be switched between two discrete states [25]. This restriction, while attractive from a fabrication and control standpoint, converts the continuous wavefront-shaping optimization into a combinatorial one.

An engineering counterpart of this program has developed, in the electromagnetic domain, around reconfigurable intelligent surfaces (RIS): large, low-cost, nearly-passive arrays of tuneable scattering elements deployed to shape the radio propagation environment for wireless communication, effectively turning the wireless channel itself into an optimisation variable [26, 27, 28, 29]. Practical RIS hardware is almost universally quantised to a small number of phase states, and the resulting discrete beamforming problem inherits the same combinatorial structure as binary coding metasurfaces [25, 27]. The acoustic analogue — reconfigurable acoustic metasurfaces used to control sound fields in rooms, ducts, or underwater channels — has been demonstrated to provide multichannel, cross-talk-free acoustic communication in reverberant rooms using binary-controlled acoustic reconfigurable metasurfaces [24].

The wave shaping scenarios mentioned above have one thing in common: In order to achieve the desired waveforms, optimisation algorithms have to be employed to find the best metasurface configuration for the task at hand. In this paper, we will use a description of tuneable metasurfaces in terms of quantum graphs. We will demonstrate that an exact map between incoming and outgoing reflected and transmitted waves can be obtained in this setting. Further, we will show that the process is a solvable inverse problem, that is, we can uniquely determine the parameters of the metasurface knowing the incoming and desired outgoing field, thus sidestepping an optimisation algorithm.

A quantum graph is a network of one-dimensional edges (waveguides, pipes, or transmission lines) meeting at vertices; free one-dimensional wave propagation along edges is considered with graph vertices acting as low-dimensional, unitary scattering centres [30, 31, 32]. Quantum graphs have served in mesoscopic and wave-chaotic physics as a computationally cheap model for wave-scattering dynamics. The response of a quantum graph, including its Green’s function for a forcing, can be assembled through relatively small matrix problems rather than through full-wave simulations [33]. Building on this correspondence, Lawrie et al. [34] showed that a periodic quantum graph, with finite sub-graphs attached at each vertex, mimicking subwavelength resonant elements, reproduces the band structure and exotic dispersion features of genuine electromagnetic metamaterials while remaining tractable analytically. The same framework was subsequently extended to acoustic metasurfaces built from networks of interconnected pipes, where tuning the graph topology and edge lengths was shown to engineer non-resonant negative refraction of surface waves, confirmed both in full-wave simulations and in experiments on acoustic pipe networks [35]. Choosing the edge lengths alone can even engineer a Doppler effect in a static material, producing vortices and black-, white- and worm-hole-like wave trajectories [36]. Modelling interfaces with graph methods has shown that structures placed between interfaces can be used to engineer their reflection and transmission coefficients [37], and has led to the discovery of angular filtering, which has been validated by numerical simulations [38] and by microwave experiments on networks of coaxial cables, which showed unit transmission at a designed angle [39]. Extending the formulation to a gauge-shifted wave equation, the magnetic Schrödinger equation of a charged particle, makes the transmission angle tunable by a magnetic flux [40], and the transmission profile of a waveguide network can be prescribed exactly, with the network constructed analytically [41]. A detailed account of the formalism behind these works is given in Ref. [42]. In the present work we adopt this quantum graph picture as the underlying model for our tunable metasurface interface which acts as a reflecting wall: each meta-atom is represented as a finite graph whose vertex-scattering data can be tuned a priori to provide the desired transmission matrix.

Physically, the wall we have in mind is an array of resonant channels, each a tube of variable length terminated by a tunable scatterer, see Figure 1. It is close to the experimental metasurfaces of acoustic cavity resonators that have been built to dissipate electric-motor noise in several frequency bands [43]; the tubes of our wall play the role of those cavities, and their lengths and terminations are the quantities we design. Arrays of resonators with different, graded geometries are known to produce rich scattering responses [44, 45]; here the grading is chosen exactly to produce a prescribed field.

Unit-cell strip and channel fields

Figure 1. (a) A unit-cell strip of the semi-infinite square lattice, with edges of length ℓ\ell and coordinates x=nℓx=n\ell, y=mℓy=m\ell, terminated at n=0n=0 by a wall of M=5M=5 channels. Each channel is a tube of length ℓm\ell_m ending in a vertex with parameter λm\lambda_m, beyond which a semi-infinite lead (dotted) carries waves away from the lattice; the arrows indicate an incident and a reflected wave. (b)–(d) The field Re⁡{ψ0,m,r}\operatorname{Re}\{\psi_{0,m,r}\} along a single channel, z0,m,r∈[0,∞)z_{0,m,r}\in[0,\infty), excited with α0,m,rout=1\alpha^{\mathrm{out}}_{0,m,r}=1 and β0,m,rin=0\beta^{\mathrm{in}}_{0,m,r}=0, for (b) λm=0\lambda_m=0, (c) an intermediate value of λm\lambda_m and (d) λm→∞\lambda_m\to\infty; the channel vertex sits at z0,m,r=ℓmz_{0,m,r}=\ell_m. Panel (b) labels the amplitudes travelling towards and away from the wall vertex and the channel vertex.

The paper is organised as follows. Section II introduces the model and Sec. III inverts the boundary condition of a single channel. Section IV describes the Bloch waves of the lattice and the scattering orders of a periodic wall, and Sec. V gives the design rule for an arbitrary prescribed field. We apply it to Gaussian beams that are steered, absorbed and split by the wall in Sec. VI, and to image formation by a two-dimensional metasurface in Sec. VIB, before concluding in Sec. VII. All derivations are collected in the appendices.

Formulation

We consider a semi-infinite square lattice of thin waveguides (edges) of length ℓ\ell, with vertices labelled (n,m)(n,m), n≤0n \le0, m∈Zm \in\mathbb{Z}, at the positions x=nℓx=n\ell and y=mℓy=m\ell, see Figure 1(a). The four edges meeting at a vertex are labelled by the direction in which they point, e∈{l,r,d,u}e \in\{l,r,d,u\} (left, right, down, up); on edge ee of vertex (n,m)(n,m), we introduce a local coordinate zn,m,e∈[0,ℓ]z_{n,m,e} \in[0,\ell] measured from the vertex. On every edge, the wave field satisfies the one-dimensional Helmholtz equation

d2ψn,m,edzn,m,e2+k2ψn,m,e=0.(1)\frac{\mathrm{d}^2\psi_{n,m,e}}{\mathrm{d}z_{n,m,e}^2}+k^2\psi_{n,m,e}=0. \tag*{(1)}

This equation describes acoustic waves in narrow pipes, transverse waves on strings, voltages and currents on transmission lines and in microwave networks, and electrons in quantum wires [31, 32, 35, 46, 47]. The time dependence is given by the phase e−iωt\mathrm{e}^{-\mathrm{i}\omega t}, where the relation between ω\omega and kk depends on the physical system being modelled.

At every vertex, the edges are coupled by two conditions: (1) the field is continuous, ψe(0)=ψe′(0)≡ϕ\psi_e(0)=\psi_{e'}(0)\equiv\phi for any two edges ee and e′e' meeting at the vertex; and (2) the outgoing derivatives satisfy ∑edψe/dze(0)=λϕ\sum_e d\psi_e/dz_e(0)=\lambda\phi, with a real vertex parameter λ\lambda [32]. Writing the field on each edge as a wave leaving and a wave arriving at the vertex, standard algebra turns these conditions into a vertex scattering matrix that maps the incoming onto the outgoing amplitudes (Appendix § 6.1 and Ref. [42]), such that

S(k;λ)=2ikvik−λE−I,(2)S(k;\lambda)=\frac{2ik}{vik-\lambda}\mathbf{E}-\mathbf{I}, \tag*{(2)}

where E\mathbf{E} is the v×vv\times v matrix of ones, I\mathbf{I} the identity and vv the valency of the vertex. In this paper, we have v=4v=4 or v=6v=6 for a 2D or 3D lattice and we generally assume Neumann–Kirchhoff junctions, that is, λ=0\lambda=0. General λ\lambda values can be interpreted, for example, as transverse waves on a network of strings under tension TT, a point mass μ\mu and a grounded spring of stiffness ss at the vertex. This then gives λ=(s−μω2)/T\lambda=(s-\mu\omega^2)/T, so that negative values of λ\lambda are mass-like and positive values spring-like.

The lattice is terminated at column n=0n=0 by the wall. The right-hand edge of every vertex (0,m)(0,m) is replaced by a tube of length ℓm\ell_m that ends in a vertex of valency two with parameter λm\lambda_m, beyond which a semi-infinite lead carries waves away from the structure Figure 1(a). The lead represents any mechanism that removes energy from the lattice, such as an anechoic termination, an absorbing material or a port into another system; λm\lambda_m determines how much of the wave is lost to the wall Figure 1(b)–(Figure 1d).

For λm=0\lambda_m=0 the vertex is transparent and the wave passes straight through, while as λm→∞\lambda_m\to\infty the vertex imposes a Dirichlet condition and the wave is fully reflected. We take the wall to be periodic with period L=MℓL=M\ell. One period thus contains MM channels with (ℓm+M,λm+M)=(ℓm,λm)(\ell_{m+M},\lambda_{m+M})=(\ell_m,\lambda_m).

Designing the Channels

Each channel leaving from a vertex (0,m)(0,m) consists of a tube of length ℓm\ell_m terminated by a vertex with v=2v=2 which in turn is connected to an infinite lead, see Figure 1(b). Let α0,m,rout\alpha^{\mathrm{out}}_{0,m,r} be the amplitude of the wave leaving the wall vertex (0,m)(0,m) along its right-hand edge rr, and α0,m,rin\alpha^{\mathrm{in}}_{0,m,r} the amplitude returning to it. For v=2v=2, the scattering matrix (2) at the channel vertex has reflection coefficient ρm=λm/(2ik−λm)\rho_m=\lambda_m/(2ik-\lambda_m). A wave entering the tube is reflected at the channel vertex and returns to the wall vertex. We require that the wall is passive, that is, no wave amplitude enters from the lead; we thus find that the two amplitudes α0,m,rin\alpha^{\mathrm{in}}_{0,m,r} are related by (Appendix § 6.1)

e2ikℓmρm=α0,m,rinα0,m,rout≡Γm.(3)e^{2ik\ell_m}\rho_m=\frac{\alpha^{\mathrm{in}}_{0,m,r}}{\alpha^{\mathrm{out}}_{0,m,r}}\equiv\Gamma_m. \tag*{(3)}

This is the key equation of this work. For any prescribed field in the lattice, the amplitudes on each tube, and hence Γm\Gamma_m, are known, and Eq. (3) is a single complex condition on the two real parameters ℓm\ell_m and λm\lambda_m. Its modulus fixes the vertex parameter and its phase fixes the tube length, which gives (Appendix A)

λm=σm2k∣Γm∣1−∣Γm∣2,(4)\lambda_m = \sigma_m \frac{2k|\Gamma_m|}{\sqrt{1-|\Gamma_m|^2}}, \tag*{(4)}
ℓm=12k[arg⁡Γm+σmarccos⁡(−∣Γm∣)]+qπk,(5)\ell_m = \frac{1}{2k}\left[\arg\Gamma_m+\sigma_m\arccos\left(-|\Gamma_m|\right)\right]+\frac{q\pi}{k}, \tag*{(5)}

with σm=±1\sigma_m=\pm1 and q∈Zq\in\mathbb{Z}. The vertex parameter is real, and the channel passive, if and only if ∣Γm∣≤1|\Gamma_m|\leq1, that is, if the channel absorbs rather than emits energy. A channel that must absorb everything (Γm=0\Gamma_m=0) is an open lead, λm=0\lambda_m=0, whose length is irrelevant, while a channel that must reflect everything (∣Γm∣=1|\Gamma_m|=1) is a Dirichlet channel, λm=±∞\lambda_m=\pm\infty, whose phase is set by the length alone. Throughout we take σm=−1\sigma_m=-1, so that every scatterer is mass-like; the integer qq allows any minimum tube length to be imposed.

Lattice solutions

To express the tube amplitudes through the lattice field we need the Bloch waves of the lattice. The solutions of the square periodic quantum graph are well known [34, 35, 38, 42], and we only state the key results here; details are given in the references and in Appendix B. Note that the lattice is simply a convenient way of representing the waves that scatter from the wall. Although the formalism is discrete, its solutions translate directly to the continuum: the discrete angular filter of Ref. [38], designed on a lattice, agreed excellently with full-wave simulations of the corresponding continuous structure. For this reason we use the discrete formalism throughout.

At every vertex the amplitudes leaving along the edges (l,r,d,ul,r,d,u) follow from those arriving through aout=Sain\mathbf{a}^{\mathrm{out}}=\mathbf{S}\mathbf{a}^{\mathrm{in}}, with S\mathbf{S} from Eq. (2) for λ=0\lambda=0 and v=4v=4. By writing the incoming amplitudes at a vertex in terms of the outgoing amplitudes of its neighbours, ain=Paout\mathbf{a}^{\mathrm{in}}=\mathbf{P}\mathbf{a}^{\mathrm{out}}, via the phase matrix P\mathbf{P}, which contains the phase eikℓe^{ik\ell} accumulated along each edge and the Bloch phases e±iκxℓe^{\pm i\kappa_x\ell} and e±iκyℓe^{\pm i\kappa_y\ell} between neighbouring vertices (Appendix B), we construct the eigenvalue problem

[I−P(k,κx,κy)S]ain=0.(6)\left[\mathbf{I}-\mathbf{P}(k,\kappa_x,\kappa_y)\mathbf{S}\right]\mathbf{a}^{\mathrm{in}}=0. \tag*{(6)}

Nontrivial solutions exist when det⁡[I−PS]=0\det[\mathbf{I}-\mathbf{P}\mathbf{S}]=0, which gives the dispersion relation Figure 2(a)

Band diagram, iso-frequency contour, and neighbourhood of a lattice vertex

Figure 2. (a) Band diagram kℓ(κx,κy)k\ell(\kappa_x,\kappa_y) of the square periodic lattice, Eq. (7). (b) Iso-frequency contour at k=1/ℓk=1/\ell, showing the right-travelling solutions for three values of κy\kappa_y: (i) κy=1/ℓ\kappa_y=1/\ell, (ii) κy=0\kappa_y=0 and (iii) κy=−1/ℓ\kappa_y=-1/\ell; the insets show the corresponding Bloch wave fields on the lattice. (c) Neighbourhood of a vertex (n,m)(n,m) of the lattice in inset (ii), with the directions e∈{l,r,d,u}e\in\{l,r,d,u\} and the edge coordinates zn,m,ez_{n,m,e} marked. The real part of the wave function, Re⁡{ψn,m,e(zn,m,e)}\operatorname{Re}\{\psi_{n,m,e}(z_{n,m,e})\}, is plotted on each edge.

cos⁡(κxℓ)+cos⁡(κyℓ)=2cos⁡(kℓ).(7)\cos(\kappa_x\ell)+\cos(\kappa_y\ell)=2\cos(k\ell). \tag*{(7)}

In the long-wavelength regime, kℓ≪1k\ell\ll1, expanding the cosines gives κx2+κy2=2k2\kappa_x^2+\kappa_y^2=2k^2, so that up to a factor of two the lattice reproduces the free-space dispersion relation. For a given kk and κy\kappa_y, Eq. (7) gives a pair ±κx\pm\kappa_x, corresponding to a wave travelling towards the wall and a wave travelling away from it, or to a pair of evanescent waves Figure 2(b). The eigenvectors of Eq. (6) give the field on every edge Figure 2(c).

A wall of period L=MℓL=M\ell preserves the Bloch phase of the incident wave modulo 2π/L2\pi/L. An incident wave with tangential wavenumber κy(0)\kappa_y^{(0)} is, therefore, scattered into the MM orders

κy(j)=κy(0)−2πjL,j∈{0,1,…,M−1},(8)\kappa_y^{(j)}=\kappa_y^{(0)}-\frac{2\pi j}{L},\qquad j\in\{0,1,\ldots,M-1\}, \tag*{(8)}

each of which propagates or decays into the lattice depending on whether Eq. (7) gives a real or a complex κx(j)\kappa_x^{(j)}. The field in the lattice is the incident Bloch wave, with amplitude AA, plus a sum over the reflected orders with amplitudes B(j)B^{(j)}, and the choice of the B(j)B^{(j)} defines the reflected field we wish to produce.

Designing the wall

Evaluating the prescribed field on the tube attached to each wall vertex gives α0,m,rout\alpha_{0,m,r}^{\mathrm{out}} and α0,m,rin\alpha_{0,m,r}^{\mathrm{in}}, and, therefore, Γm\Gamma_m. This is most transparent in terms of the field at the wall vertex, Φm\Phi_m, and its continuation Φ~m\widetilde{\Phi}_m to a fictitious column n=1n=1, the value the missing lattice vertex would have had. Using Eq. (8), both are discrete Fourier sums over the scattering orders,

Φm=eiκy(0)mℓ[A+∑jB(j)e−2πijm/M],(9)\Phi_m=e^{i\kappa_y^{(0)}m\ell}\left[A+\sum_j B^{(j)}e^{-2\pi i jm/M}\right], \tag*{(9)}
Φ~m=eiκy(0)mℓ[Aeiκx(0)ℓ+∑jB(j)e−iκx(j)ℓe−2πijm/M],(10)\widetilde{\Phi}_m=e^{i\kappa_y^{(0)}m\ell}\left[Ae^{i\kappa_x^{(0)}\ell}+\sum_j B^{(j)}e^{-i\kappa_x^{(j)}\ell}e^{-2\pi i jm/M}\right], \tag*{(10)}

with jj running over the MM orders (8). Expressing the tube amplitudes through these sums (Appendix C) gives

Γm=eikℓΦm−Φ~mΦ~m−e−ikℓΦm.(11)\Gamma_m=\frac{e^{ik\ell}\Phi_m-\widetilde{\Phi}_m}{\widetilde{\Phi}_m-e^{-ik\ell}\Phi_m}. \tag*{(11)}

Equations (4)–(11) form the complete design rule, such that we can choose the reflected amplitudes B(j)B^{(j)}, evaluate two discrete Fourier transforms, and read off λm\lambda_m and ℓm\ell_m for every channel. Each channel then presents to its wall vertex exactly the response that the missing lattice column would have presented to the prescribed field, so that the prescribed field is the exact response of the designed wall.

The amplitudes B(j)B^{(j)} are free parameters. Writing B(j)=BB^(j)B^{(j)}=B\widehat{B}^{(j)} for a chosen profile B^(j)\widehat{B}^{(j)}, the overall amplitude BB sets how much energy is reflected. By tuning BB down we ensure that the reflected field carries less energy than the incident field and that no channel has to emit energy, which results in a purely passive reflecting boundary (Fig. 3). The largest such amplitude, B′B', is reached when the first channel reflects perfectly, ∣Γm∣=1|\Gamma_m|=1, and in the examples below we take B=B′B=B'.

Figure 3. ∣Γm∣2|\Gamma_m|^2 of every channel as a function of the amplitude BB of the reflected field, for (a) the steering and (b) the splitting walls of Figure 4. At B=0B=0 all channels take the perfect-absorber value ∣ρ′∣2|\rho'|^2. The wall is passive while every channel lies below ∣Γm∣2=1|\Gamma_m|^2=1; the first channel to reach the line sets B′B', beyond which (shaded) the wall would require gain.

Results

Reflections from the wall in 2D

See Figure 4, where we consider a non-trivial reflection angle, complete absorption and beam splitting for an incident beam at normal incidence. The beams are superpositions of Bloch waves with a Gaussian weight in the tangential wavenumber [34]; the wall is designed for the central component and held fixed, and every other component is reflected according to the scattering problem of that wall (Appendix D).

Figure 4. Wall designs for a Gaussian beam at normal incidence (kℓ=π/3k\ell= \pi/3, wall period L=6ℓL = 6\ell, beam width σ=0.09/ℓ\sigma= 0.09/\ell): (a)–(c) reflection at +45∘+45^\circ, (d)–(f) complete absorption and (g)–(i) splitting into ±45∘\pm45^\circ. (a), (d), (g) Total field Re⁡ϕn,m\operatorname{Re}\phi_{n,m}; arrows mark the incident and reflected directions. (b), (e), (h) One period of the wall, boxed in (a), (d), (g), with the field on the lattice edges and vertices, in the tubes and on the truncated leads; large circles mark the channel vertices. (c), (f), (i) Tube lengths ℓm/ℓ\ell_m/\ell (blue) and vertex parameters λmℓ\lambda_m\ell (orange) from Eqs. (4)–(11); hatched bars mark Dirichlet channels, λm=−∞\lambda_m = -\infty. Lengths are defined modulo π/k\pi/k. Colour scales are linear.

Wave scattering in 3D and image forming

The same construction applies to a cubic lattice terminated by a two-dimensional metasurface. The vertices (n,m,p)(n,m,p) sit at x=nℓx=n\ell, y=mℓy=m\ell and z=pℓz=p\ell, the lattice occupies p≥0p\geq0, and the wall lies in the plane p=0p=0. Each vertex has six edges, e∈{l,r,d,u,i,o}e\in\{l,r,d,u,i,o\}, pointing along ∓x\mp x, ∓y\mp y and ±z\pm z [Fig. 5(a)(i)], and each wall vertex (n,m,0)(n,m,0) is connected through its edge oo to a tube of length ℓn,m\ell_{n,m} ending in a vertex with parameter λn,m\lambda_{n,m} [Fig. 5(a)(ii)]. The eigenvalue problem (6) now gives cos⁡(κxℓ)+cos⁡(κyℓ)+cos⁡(κzℓ)=3cos⁡(kℓ)\cos(\kappa_x\ell)+\cos(\kappa_y\ell)+\cos(\kappa_z\ell)=3\cos(k\ell), a wall with a period of N×MN\times M channels scatters an incident wave into N×MN\times M orders, and Γn,m\Gamma_{n,m} is again given by Eq. (11), now with two-dimensional Fourier sums (Appendix E).

To form an image in the plane p=pfp=p_f, we take the desired image Un,mU_{n,m} on one period, compute its discrete Fourier coefficients U~η,μ\widetilde{U}_{\eta,\mu}, and propagate each component back from the image plane to the wall, to give

B(η,μ)=BU~η,μe−iκz(η,μ)pfℓHη,μ,(12)B^{(\eta,\mu)}=B\widetilde{U}_{\eta,\mu}e^{-i\kappa_z^{(\eta,\mu)}p_f\ell}H_{\eta,\mu}, \tag*{(12)}

where Hη,μ=1H_{\eta,\mu}=1 for propagating orders and Hη,μ=0H_{\eta,\mu}=0 for evanescent ones. Figure 5 shows the result for the letters “ANA” (for the Adelaide–Nottingham Alliance), together with the tube lengths and vertex parameters of every channel.

Figure 5. Image formation by a two-dimensional metasurface terminating a cubic lattice at kℓ=1k\ell= 1, with a period of 64ℓ×64ℓ64\ell\times64\ell (4096 channels). (a) A plane wave incident in the xx-zz plane is reflected to form the letters “ANA” as thanks to our research funding the Adelaide-Nottingham Alliance; (∣Un,m∣=1|U_{n,m}| = 1 on the letters and 0 elsewhere. The reflected field Re⁡ϕref\operatorname{Re}\phi^{\mathrm{ref}} is shown on the plane of incidence and ∣ϕref∣2|\phi^{\mathrm{ref}}|^2 on one period of the image plane; the tubes and truncated leads face down along two edges of the metasurface. (i) A lattice vertex and (ii) a vertex of the metasurface, marked by the small unit cells on the right-hand edge of the lattice, with their edges ll, rr, dd, uu, ii, oo and the total field Re⁡ψ\operatorname{Re}\psi plotted inside the edges; in (ii) the edge oo is the tube, which ends in the channel vertex λn,m\lambda_{n,m}. (b) Tube lengths ℓn,m/ℓ\ell_{n,m}/\ell and (c) vertex parameters λn,mℓ\lambda_{n,m}\ell (Appendix E); lengths are defined modulo π/k\pi/k. Colour scales are linear; (c) saturates below λℓ=−2.85\lambda\ell= -2.85, and the black square marks the single Dirichlet channel.

Discussion and conclusions

We have shown that the boundary of a quantum graph lattice can be designed exactly to produce a prescribed reflected field, without optimisation. The required boundary condition at each channel reduces to a single complex ratio, Γm\Gamma_m, of returning to outgoing amplitudes, Eq. (11), which follows from the incident and desired reflected

fields through two discrete Fourier sums. This ratio is inverted in closed form for a tube length and a vertex parameter, Eqs. (4) and (5). The wall is passive whenever ∣Γm∣≤1|\Gamma_m| \le1, that is, whenever every channel absorbs rather than emits energy. We demonstrated the construction for a one-dimensional metasurface terminating a two-dimensional lattice, which steers, splits or perfectly absorbs Gaussian beams, and for a two-dimensional metasurface terminating a three-dimensional lattice, which forms a prescribed image.

The model maps directly onto physical waveguide networks. Quantum graphs describe thin acoustic, elastic and electromagnetic waveguides [46] and can be built as acoustic pipe networks [35] and as microwave networks of coaxial cables [39, 47]. The tube lengths can be folded into a compact wall by coiling [48], and walls of acoustic cavities of this kind have already been built and tested [43]. The vertex parameter, λm\lambda_m, corresponds to a local mass or spring, or more generally to any lumped scatterer, and the semi-infinite leads correspond to anechoic terminations or absorbing material. In such realisations, λm\lambda_m usually depends on frequency (for a mass, λ∝−ω2\lambda\propto-\omega^2), which does not affect the design at the operating frequency. Because a single real parameter per channel controls loss and a single length controls phase, the wall is naturally reconfigurable. Tuning λm\lambda_m, for instance with adjustable masses, springs or resonators, switches

between absorbing, steering and splitting states while the structure remains passive throughout. In contrast to the binary and phase-quantised surfaces discussed in Sec. I, the required settings are obtained in closed form rather than by a combinatorial search.

Several extensions follow directly. Other lattices and vertex conditions only change the Bloch amplitudes of Appendix B. Adding further tubes or vertices to each channel provides more parameters per channel, which can be used to prescribe richer fields, for several frequencies or several incident waves at once. The same reasoning applies to transmission through a structured layer [37, 41], and the semi-infinite leads could be replaced by a second lattice into which the wall transmits.

Acknowledgments

Any data and code generated for this paper can be made available upon reasonable request. This research was funded by the Universities of Adelaide and Nottingham through the Adelaide-Nottingham Alliance (ANA). Tristan Lawrie would like to acknowledge the financial support of the META4D program grant (EP/Y015673/1), as well as many useful discussion with Simon Horsley.

The channel condition and its inversion

Vertex scattering matrix

On an edge ee leaving a vertex we write the solution of Eq. (1) as ψe(z)=aeouteikz+aeine−ikz\psi_e(z)=a_e^{\mathrm{out}}e^{ikz}+a_e^{\mathrm{in}}e^{-ikz}, with zz measured from the vertex. Continuity at the vertex gives aeout+aein=ϕa_e^{\mathrm{out}}+a_e^{\mathrm{in}}=\phi for every edge, where ϕ\phi is the vertex value, and the condition ∑eψe′(0)=λϕ\sum_e\psi'_e(0)=\lambda\phi gives ik∑e(aeout−aein)=λϕik\sum_e(a_e^{\mathrm{out}}-a_e^{\mathrm{in}})=\lambda\phi. Eliminating ∑eaeout=vϕ−∑eaein\sum_ea_e^{\mathrm{out}}=v\phi-\sum_ea_e^{\mathrm{in}} gives

ϕ=2ikvik−λ∑eaein,aeout=ϕ−aein,(13)\phi=\frac{2ik}{vik-\lambda}\sum_ea_e^{\mathrm{in}},\qquad a_e^{\mathrm{out}}=\phi-a_e^{\mathrm{in}}, \tag*{(13)}

which is Eq. (2); see also Ref. [42]. For the channel vertices, v=2v=2, and the scattering matrix has the form

Sm=(ρmτmτmρm),ρm=λm2ik−λm,τm=2ik2ik−λm,(14)S_m=\begin{pmatrix}\rho_m&\tau_m\\\tau_m&\rho_m\end{pmatrix},\qquad\rho_m=\frac{\lambda_m}{2ik-\lambda_m},\qquad\tau_m=\frac{2ik}{2ik-\lambda_m}, \tag*{(14)}

with ∣ρm∣2+∣τm∣2=1|\rho_m|^2+|\tau_m|^2=1.

Channel condition

Let β0,m,lin\beta_{0,m,l}^{\mathrm{in}} and β0,m,rin\beta_{0,m,r}^{\mathrm{in}} be the amplitudes arriving at the channel vertex from the tube and from the lead, and β0,m,lout\beta_{0,m,l}^{\mathrm{out}}, β0,m,rout\beta_{0,m,r}^{\mathrm{out}} those leaving it [Figure 1(b)], so that (β0,m,lout,β0,m,rout)T=Sm(β0,m,lin,β0,m,rin)T(\beta_{0,m,l}^{\mathrm{out}},\beta_{0,m,r}^{\mathrm{out}})^{T}=S_m(\beta_{0,m,l}^{\mathrm{in}},\beta_{0,m,r}^{\mathrm{in}})^{T}. Propagation along the tube gives β0,m,lin=eikℓmα0,m,rout\beta_{0,m,l}^{\mathrm{in}}=e^{ik\ell_m}\alpha_{0,m,r}^{\mathrm{out}} and α0,m,rin=eikℓmβ0,m,lout\alpha_{0,m,r}^{\mathrm{in}}=e^{ik\ell_m}\beta_{0,m,l}^{\mathrm{out}}, so that

α0,m,rin=e2ikℓmρmα0,m,rout+eikℓmτmβ0,m,rin.(15)\alpha_{0,m,r}^{\mathrm{in}}=e^{2ik\ell_m}\rho_m\alpha_{0,m,r}^{\mathrm{out}}+e^{ik\ell_m}\tau_m\beta_{0,m,r}^{\mathrm{in}}. \tag*{(15)}

A nonzero β0,m,rin\beta_{0,m,r}^{\mathrm{in}} would inject energy into the lattice from the lead. For a passive wall β0,m,rin=0\beta_{0,m,r}^{\mathrm{in}}=0, and Eq. (15) reduces to Eq. (3). The wave leaving through the lead, β0,m,rout=eikℓmτmα0,m,rout\beta_{0,m,r}^{\mathrm{out}}=e^{ik\ell_m}\tau_m\alpha_{0,m,r}^{\mathrm{out}}, carries the energy flux k∣τm∣2∣α0,m,rout∣2=k(∣α0,m,rout∣2−∣α0,m,rin∣2)k|\tau_m|^2|\alpha_{0,m,r}^{\mathrm{out}}|^2=k(|\alpha_{0,m,r}^{\mathrm{out}}|^2-|\alpha_{0,m,r}^{\mathrm{in}}|^2) out of the lattice.

Inversion

Since kk, ℓm\ell_m and λm\lambda_m are real, the modulus of Eq. (3) gives ∣ρm∣2=λm2/(λm2+4k2)=∣Γm∣2|\rho_m|^2=\lambda_m^2/(\lambda_m^2+4k^2)=|\Gamma_m|^2, which is solved by Eq. (4). The phase of Eq. (3) gives e2ikℓm=Γm/ρm=Γm(2ik/λm−1)e^{2ik\ell_m}=\Gamma_m/\rho_m=\Gamma_m(2ik/\lambda_m-1). From Eq. (14), cos⁡(arg⁡ρm)=−∣ρm∣\cos(\arg\rho_m)=-|\rho_m| and sgn⁡(arg⁡ρm)=−σm\operatorname{sgn}(\arg\rho_m)=-\sigma_m, so that arg⁡ρm=−σmarccos⁡(−∣Γm∣)\arg\rho_m=-\sigma_m\arccos(-|\Gamma_m|), which gives Eq. (5). Equivalently, ℓm=(2ik)−1ln⁡[Γm(2ik/λm−1)]+qπ/k\ell_m=(2ik)^{-1}\ln[\Gamma_m(2ik/\lambda_m-1)]+q\pi/k, where the argument of the logarithm has unit modulus, so that ℓm\ell_m is real.

Bloch waves of the square lattice

With the amplitudes around a vertex ordered as (l,r,d,u)(l,r,d,u) and Bloch conditions ϕn+1,m=eiκxℓϕn,m\phi_{n+1,m}=e^{i\kappa_x\ell}\phi_{n,m} and ϕn,m+1=eiκyℓϕn,m\phi_{n,m+1}=e^{i\kappa_y\ell}\phi_{n,m} on the vertex values, the phase matrix in Eq. (6) is P=diag⁡{px,py}P=\operatorname{diag}\{p_x,p_y\}, with

pj=eikℓ(0e−iκjℓeiκjℓ0).(16)p_j=e^{ik\ell}\begin{pmatrix}0&e^{-i\kappa_j\ell}\\e^{i\kappa_j\ell}&0\end{pmatrix}. \tag*{(16)}

The dispersion relation (7) gives

κx(k,κy)=1ℓarccos⁡(2cos⁡(kℓ)−cos⁡(κyℓ)),(17)\kappa_x(k,\kappa_y)=\frac{1}{\ell}\arccos\left(2\cos(k\ell)-\cos(\kappa_y\ell)\right), \tag*{(17)}

and we take the branch with Im⁡κx≥0\operatorname{Im}\kappa_x\geq0 and, for propagating waves with 0<kℓ<π0<k\ell<\pi, 0<κxℓ<π0<\kappa_x\ell<\pi. Then e+iκxnℓe^{+i\kappa_xn\ell} either carries energy towards the wall or decays in that direction, and e−iκxnℓe^{-i\kappa_xn\ell} either carries energy away from the wall or decays into the lattice.

Continuity at the vertex as specified in Appendix A requires aeout+aein=ϕa_e^{\mathrm{out}}+a_e^{\mathrm{in}}=\phi with ϕ\phi given in (A1). Equation (6) is homogeneous and fixes aina^{\mathrm{in}} only up to a constant factor, which we choose such that ϕ=1\phi=1. With this normalisation, we have ψe(ℓ)=aeouteikℓ+aeine−ikℓ=eiθeψ(0)=eiθe\psi_e(\ell)=a_e^{\mathrm{out}}e^{ik\ell}+a_e^{\mathrm{in}}e^{-ik\ell}=e^{i\theta_e}\psi(0)=e^{i\theta_e}, which yields

aeout=eiθe−e−ikℓ2isin⁡kℓ,aein=eikℓ−eiθe2isin⁡kℓ,(18)a_e^{\mathrm{out}}=\frac{e^{i\theta_e}-e^{-ik\ell}}{2i\sin k\ell},\qquad a_e^{\mathrm{in}}=\frac{e^{ik\ell}-e^{i\theta_e}}{2i\sin k\ell}, \tag*{(18)}

where θe=(−κxℓ,κxℓ,−κyℓ,κyℓ)\theta_e=(-\kappa_x\ell,\kappa_x\ell,-\kappa_y\ell,\kappa_y\ell) for e=(l,r,d,u)e=(l,r,d,u) is the Bloch phase of the neighbouring vertex. These amplitudes satisfy aeout+aein=1a_e^{\mathrm{out}}+a_e^{\mathrm{in}}=1 for every ee, and the vertex condition ∑e(aeout−aein)=0\sum_e(a_e^{\mathrm{out}}-a_e^{\mathrm{in}})=0 reduces to ∑eeiθe=4cos⁡kℓ\sum_e e^{i\theta_e}=4\cos k\ell, which is the dispersion relation (7). The amplitudes beb_e of the waves travelling away from the wall follow with κx→−κx\kappa_x\to-\kappa_x. The field on an edge from a vertex position (n,m)(n,m) in direction ee is obtained as

ψn,m,e(z)=ϕn,meiθesin⁡kz+sin⁡k(ℓ−z)sin⁡kℓ,(19)\psi_{n,m,e}(z)=\phi_{n,m}\frac{e^{i\theta_e}\sin kz+\sin k(\ell-z)}{\sin k\ell}, \tag*{(19)}

valid for kℓ∉πZk\ell\notin\pi\mathbb{Z} with ϕ(n,m)\phi_{(n,m)}, the amplitude at vertex (n,m)(n,m). A Bloch wave of vertex amplitude AA carries the energy flux k(∣aeout∣2−∣aein∣2)∣A∣2=k∣A∣2sin⁡(κxℓ)/sin⁡(kℓ)k(|a_e^{\mathrm{out}}|^2-|a_e^{\mathrm{in}}|^2)|A|^2=k|A|^2\sin(\kappa_x\ell)/\sin(k\ell) along each horizontal edge, and evanescent waves carry none. Energy flows along the group velocity, ∇κω∝(sin⁡κxℓ,sin⁡κyℓ)\nabla_{\boldsymbol{\kappa}}\omega\propto(\sin\kappa_x\ell,\sin\kappa_y\ell), normal to the iso-frequency contour, at the angle θ=arctan⁡(sin⁡κyℓ/sin⁡κxℓ)\theta=\arctan(\sin\kappa_y\ell/\sin\kappa_x\ell) to the xx axis; because of the anisotropy of the lattice this generally differs from the direction of (κx,κy)(\kappa_x,\kappa_y).

The design rule

The prescribed field at the vertices is the incident Bloch wave plus the reflected orders (8),

ϕn,m=Aei(κx(0)n+κy(0)m)ℓ+∑j=0M−1B(j)ei(−κx(j)n+κy(j)m)ℓ,(20)\phi_{n,m}=A e^{i(\kappa_x^{(0)}n+\kappa_y^{(0)}m)\ell}+\sum_{j=0}^{M-1}B^{(j)}e^{i(-\kappa_x^{(j)}n+\kappa_y^{(j)}m)\ell}, \tag*{(20)}

and the field on every edge follows from Eq. (19). On the tube attached to the wall vertex (0,m)(0,m), the outgoing and incoming amplitudes are the corresponding sums of the edge amplitudes ara_r and brb_r of Eq. (18),

α0,m,rout/in=Aarout/in,(0)eiκy(0)mℓ+∑jB(j)brout/in,(j)eiκy(j)mℓ,(21)\alpha_{0,m,r}^{\mathrm{out/in}}=A a_r^{\mathrm{out/in},(0)}e^{i\kappa_y^{(0)}m\ell}+\sum_jB^{(j)}b_r^{\mathrm{out/in},(j)}e^{i\kappa_y^{(j)}m\ell}, \tag*{(21)}

see Figure 1(b). Inserting Eq. (18) and using Eqs. (9) and (10), these become α0,m,rout=(Φ~m−e−ikℓΦm)/(2isin⁡kℓ)\alpha_{0,m,r}^{\mathrm{out}}=(\tilde{\Phi}_m-e^{-ik\ell}\Phi_m)/(2i\sin k\ell) and α0,m,rin=(eikℓΦm−Φ~m)/(2isin⁡kℓ)\alpha_{0,m,r}^{\mathrm{in}}=(e^{ik\ell}\Phi_m-\tilde{\Phi}_m)/(2i\sin k\ell), whose ratio is Eq. (11).

The energy flux absorbed by channel mm is

Pm=k(∣α0,m,rout∣2−∣α0,m,rin∣2)=ksin⁡kℓIm⁡(Φ~m∗Φ~m).(22)P_m=k\left(\lvert\alpha_{0,m,r}^{\mathrm{out}}\rvert^2-\lvert\alpha_{0,m,r}^{\mathrm{in}}\rvert^2\right)=\frac{k}{\sin k\ell}\operatorname{Im}\left(\widetilde{\Phi}_m^*\widetilde{\Phi}_m\right). \tag*{(22)}

so the wall is passive, ∣Γm∣≤1\lvert\Gamma_m\rvert\le1, if and only if Pm≥0P_m\ge0 for every mm. When the reflected field is prescribed up to an overall amplitude, B(j)=BB^(j)B^{(j)}=B\widehat{B}^{(j)}, max⁡m∣Γm∣\max_m\lvert\Gamma_m\rvert grows with BB from its value for the absorbing wall at B=0B=0, and reaches unity at the critical amplitude B′B'.

The prescribed field satisfies Eq. (1) on every edge and the vertex condition at every lattice vertex, and once Eq. (3) holds it also satisfies the condition at every channel vertex; being unique, it is the response of the designed wall. Conversely, for a given wall (ℓm,λm)(\ell_m,\lambda_m) and incident amplitude AA, imposing Eq. (3) at all MM channels gives a linear system for the MM amplitudes B(j)B^{(j)}. This forward problem returns the prescribed amplitudes for every designed wall and is used to construct the beams of Appendix D.

Gaussian beams and the perfect absorber

Following Ref. [34], a beam is constructed by integrating the Bloch solutions over the tangential wavenumber with a Gaussian weight centred on κy′\kappa'_y,

ϕn,min=∫dq g(q)ei[κx(q)n+qm]ℓ,g(q)=e−(q−κy′)2/2σ2,(23)\phi^{\mathrm{in}}_{n,m}=\int dq\,g(q)e^{i[\kappa_x(q)n+q m]\ell},\qquad g(q)=e^{-(q-\kappa'_y)^2/2\sigma^2}, \tag*{(23)}

with κx(q)\kappa_x(q) from Eq. (17). The wall is designed for the central component q=κy′q=\kappa'_y and held fixed. Every other component is scattered into the orders q−2πj/Lq-2\pi j/L, with amplitudes B(j)(q)B^{(j)}(q) given by the forward problem of Appendix C, and the reflected beam is

ϕn,mref=∫dq g(q)∑j=0M−1B(j)(q)ei[−κx(j)(q)nℓ+(q−2πj/L)mℓ].(24)\phi^{\mathrm{ref}}_{n,m}=\int dq\,g(q)\sum_{j=0}^{M-1}B^{(j)}(q)e^{i[-\kappa_x^{(j)}(q)n\ell+(q-2\pi j/L)m\ell]}. \tag*{(24)}

Incident and reflected beams together form an exact solution of the graph with the designed wall. For kℓ=π/3k\ell=\pi/3 and L=6ℓL=6\ell, the orders j=1j=1 and j=5j=5 of a normally incident wave have κx(j)=∣κy(j)∣=k\kappa_x^{(j)}=\lvert\kappa_y^{(j)}\rvert=k, so that they leave the wall at exactly ∓45∘\mp45^\circ, while all other nonspecular orders are evanescent.

For perfect absorption, B(j)=0B^{(j)}=0, we have Φ~m=eiκx(0)ℓΦm\widetilde{\Phi}_m=e^{i\kappa_x^{(0)}\ell}\Phi_m, and Eq. (11) gives the same value for every channel,

Γm=ρ′=eikℓsin⁡[(k−κx(0))ℓ/2]sin⁡[(k+κx(0))ℓ/2].(25)\Gamma_m=\rho'=e^{ik\ell}\frac{\sin\left[(k-\kappa_x^{(0)})\ell/2\right]}{\sin\left[(k+\kappa_x^{(0)})\ell/2\right]}. \tag*{(25)}

The absorbing wall is therefore uniform, and it is always passive, since ∣ρ′∣<1\lvert\rho'\rvert<1 is equivalent to sin⁡kℓsin⁡κx(0)ℓ>0\sin k\ell\sin\kappa_x^{(0)}\ell>0.

For a wave incident with κy(0)=k\kappa_y^{(0)}=k, Eq. (7) gives κx(0)=k\kappa_x^{(0)}=k, so that ρ′=0\rho'=0 and λ′=0\lambda'=0: a wave travelling along the lattice diagonal is absorbed completely by open leads attached to the boundary vertices.

The cubic lattice

For the cubic lattice the amplitudes around a vertex are ordered as (l,r,d,u,i,o)(l,r,d,u,i,o), with l,rl,r along ∓x\mp x, d,ud,u along ∓y\mp y and i,oi,o along +z+z and −z-z. The Neumann–Kirchhoff scattering matrix has elements Se,e′=1/3−δe,e′S_{e,e'}=1/3-\delta_{e,e'}, and the phase matrix is P=diag⁡{px,py,pz}\mathbf{P}=\operatorname{diag}\{p_x,p_y,p_z\} with pjp_j as in Eq. (16). The zeros of det⁡[I−PS]\det[\mathbf{I}-\mathbf{P}\mathbf{S}] give the dispersion relation

cos⁡(κxℓ)+cos⁡(κyℓ)+cos⁡(κzℓ)=3cos⁡(kℓ),(26)\cos(\kappa_x\ell)+\cos(\kappa_y\ell)+\cos(\kappa_z\ell)=3\cos(k\ell), \tag*{(26)}

and the eigenvectors are again given by Eq. (18), now with six Bloch phases θe\theta_e. A wall with a period N×MN\times M channels scatters an incident wave with tangential wavenumbers (κx(0),κy(0))(\kappa_x^{(0)},\kappa_y^{(0)}) into the orders κx(η)=κx(0)−2πη/(Nℓ)\kappa_x^{(\eta)}=\kappa_x^{(0)}-2\pi\eta/(N\ell) and κy(μ)=κy(0)−2πμ/(Mℓ)\kappa_y^{(\mu)}=\kappa_y^{(0)}-2\pi\mu/(M\ell), with κz(η,μ)\kappa_z^{(\eta,\mu)} from Eq. (26) and Im⁡κz≥0\operatorname{Im}\kappa_z\ge0. The prescribed field at the vertex (n,m,p)(n,m,p) is

ϕn,m,p=∑η,μEn,m(η,μ)[A(η,μ)e−iκz(η,μ)pℓ+B(η,μ)eiκz(η,μ)pℓ],(27)\phi_{n,m,p}=\sum_{\eta,\mu}E_{n,m}^{(\eta,\mu)}\left[A^{(\eta,\mu)}e^{-i\kappa_z^{(\eta,\mu)}p\ell}+B^{(\eta,\mu)}e^{i\kappa_z^{(\eta,\mu)}p\ell}\right], \tag*{(27)}

with En,m(η,μ)=ei(κx(η)n+κy(μ)m)ℓE_{n,m}^{(\eta,\mu)}=e^{i(\kappa_x^{(\eta)}n+\kappa_y^{(\mu)}m)\ell} and sums over 0≤η<N0\le\eta<N, 0≤μ<M0\le\mu<M. The incident field may itself contain several orders, and the field on every edge follows from Eq. (19). The tube of channel (n,m)(n,m) is the edge of the wall vertex (n,m,0)(n,m,0). Repeating the steps of Appendix C, Γn,m\Gamma_{n,m} is given by Eq. (11) with the two-dimensional Fourier sums

Φn,m=∑η,μEn,m(η,μ)[A(η,μ)+B(η,μ)],\Phi_{n,m}=\sum_{\eta,\mu}E_{n,m}^{(\eta,\mu)}\left[A^{(\eta,\mu)}+B^{(\eta,\mu)}\right],
Φ~n,m=∑η,μEn,m(η,μ)[A(η,μ)eiκz(η,μ)ℓ+B(η,μ)e−iκz(η,μ)ℓ].(28)\widetilde{\Phi}_{n,m}=\sum_{\eta,\mu}E_{n,m}^{(\eta,\mu)}\left[A^{(\eta,\mu)}e^{i\kappa_z^{(\eta,\mu)}\ell}+B^{(\eta,\mu)}e^{-i\kappa_z^{(\eta,\mu)}\ell}\right]. \tag*{(28)}

the field at the wall vertex and its continuation to the missing layer p=−1p=-1. Equations (4) and (5) then give λn,m\lambda_{n,m} and ℓn,m\ell_{n,m}.

For an incident plane wave and the choice (12), with Un,m=∑η,μU~η,μe−2πi(ηn/N+μm/M)U_{n,m}=\sum_{\eta,\mu}\widetilde{U}_{\eta,\mu}e^{-2\pi i(\eta n/N+\mu m/M)}, the reflected field in the image plane is Bei(κx(0)n+κy(0)m)ℓB e^{i(\kappa_x^{(0)}n+\kappa_y^{(0)}m)\ell} times the band-limited version of Un,mU_{n,m}, in which only the Fourier components inside the iso-frequency surface (E1) survive. Its intensity therefore reproduces ∣Un,m∣2\lvert U_{n,m}\rvert^2 for any angle of incidence. Because the wall and the incident wave are periodic, so is the image; Figure 5(a) shows one period of the image plane, directly above one period of the wall.

References

  1. [1]Stefan Rotter and Sylvain Gigan. Light fields in complex media: mesoscopic scattering meets wave control. Reviews of Modern Physics, 89:015005, 2017.DOI
  2. [2]Alexander V. Kildishev, Alexandra Boltasseva, and Vladimir M. Shalaev. Planar photonics with metasurfaces. Science, 339:1232009, 2013.DOI
  3. [3]Badreddine Assouar, Bin Liang, Ying Wu, Yong Li, Jian-Chun Cheng, and Yun Jing. Acoustic metasurfaces. Nature Reviews Materials, 3:460–472, 2018.DOI
  4. [4]Ertugrul Basar, Marco Di Renzo, Julien De Rosny, Merouane Debbah, Mohamed-Slim Alouini, and Rui Zhang. Wireless communications through reconfigurable intelligent surfaces. IEEE Access, 7:116753–116773, 2019.
  5. [5]I. M. Vellekoop and A. P. M. Mosk. Focusing coherent light through opaque strongly scattering media. Optics Letters, 32:2309–2311, 2007.
  6. [6]I. M. Vellekoop and A. P. Mosk. Universal optimal transmission of light through disordered materials. Physical Review Letters, 101:120601, 2008.
  7. [7]S. M. Popoff, G. Lerosey, R. Carminati, M. Fink, A. C. Boccara, and S. Gigan. Measuring the transmission matrix in optics: an approach to the study and control of light propagation in disordered media. Physical Review Letters, 104:100601, 2010.
  8. [8]Allard P. Mosk, Ad Lagendijk, Geoffrey Lerosey, and Mathias Fink. Controlling waves in space and time for imaging and focusing in complex media. Nature Photonics, 6:283–292, 2012.DOI
  9. [9]Roberto Di Leonardo, Francesca Ianni, and Giancarlo Ruocco. Computer generation of optimal holograms for optical trap arrays. Optics Express, 15(4):1913–1922, 2007.
  10. [10][Geng2013]. Reference to be supplied, 2013.
  11. [11][Zhou2018]. Reference to be supplied, 2018.
  12. [12]Gianluca Memoli, Mihai Caleap, Michihiro Asakawa, Deepak R. Sahoo, Bruce W. Drinkwater, and Sriram Subramanian. Metamaterial bricks and quantization of meta-surfaces. Nature Communications, 8:14608, 2017.DOI
  13. [13][Memoli20]. Reference to be supplied, 2020.
  14. [14]Moonseok Kim, Youngwoon Choi, Changhyeong Yoon, Wonjun Choi, Jaisoon Kim, Q-Han Park, and Wonshik Choi. Maximal energy transport through disordered media with the implementation of transmission eigenchannels. Nature Photonics, 6:581–585, 2012.DOI
  15. [15]Philipp Ambichl, Andre Brandstötter, Julian Böhm, Matthias Kühnmayer, Ulrich Kuhl, and Stefan Rotter. Focusing inside disordered media with the generalized Wigner-Smith operator. Physical Review Letters, 119:033903, 2017.DOI
  16. [16]Etienne Rivet, Andre Brandstötter, Konstantinos G. Makris, Hervé Lissek, Stefan Rotter, and Romain Fleury. Constant-pressure bound waves in non-Hermitian disordered media. Nature Physics, 14:942–947, 2018.
  17. [17]C. W. J. Beenakker. Random-matrix theory of quantum transport. Reviews of Modern Physics, 69:731–808, 1997.DOI
  18. [18]Nadège Kaina, Matthieu Dupré, Geoffrey Lerosey, and Mathias Fink. Shaping complex microwave fields in reverberating media with binary tunable metasurfaces. Scientific Reports, 4:6693, 2014.DOI
  19. [19]Matthieu Dupré, Philipp del Hougne, Mathias Fink, Fabrice Lemoult, and Geoffrey Lerosey. Wave-field shaping in cavities: waves trapped in a box with controllable boundaries. Physical Review Letters, 115:017701, 2015.pubmed.ncbi.nlm.nih.gov/26182120
  20. [20]Jean-Baptiste Gros, Philipp del Hougne, and Geoffrey Lerosey. Tuning a regular cavity to wave chaos with metasurface-reconfigurable walls. Physical Review A, 101:061801(R), 2020.DOI
  21. [21]Benjamin W. Frazier, Thomas M. Antonsen, Steven M. Anlage, and Edward Ott. Wavefront shaping with a tunable metasurface: creating cold spots and coherent perfect absorption at arbitrary frequencies. Physical Review Research, 2:043422, 2020.DOI
  22. [22]Philipp del Hougne, Mathias Fink, and Geoffrey Lerosey. Optimally diverse communication channels in disordered environments with tuned randomness. Nature Electronics, 2:36–41, 2019.
  23. [23]Guangcong Ma, Xiying Fan, Ping Sheng, and Mathias Fink. Shaping reverberating sound fields with an actively tunable metasurface. Proceedings of the National Academy of Sciences, 115:6638–6643, 2018.DOI
  24. [24]Hongkuan Zhang, Qiyuan Wang, Mathias Fink, and Guangcong Ma. Optimizing multi-user indoor sound communications with acoustic reconfigurable metasurfaces. Nature Communications, 15, 2024. doi:10.1038/s41467-024-45435-4.DOI
  25. [25]Lei Zhang et al. A wireless communication scheme based on space- and frequency-division multiplexing using digital metasurfaces. Nature Electronics, 4:218–227, 2021.DOI
  26. [26]Marco Di Renzo et al. Smart radio environments empowered by reconfigurable AI meta-surfaces: an idea whose time has come. EURASIP Journal on Wireless Communications and Networking, 2019:129, 2019.
  27. [27]Qingqing Wu and Rui Zhang. Towards smart and re-configurable environment: intelligent reflecting surface aided wireless network. IEEE Communications Magazine, 58(1):106–112, 2020.arxiv.org/abs/1905.00152
  28. [28]Marco Di Renzo et al. Smart radio environments empowered by reconfigurable intelligent surfaces: how it works, state of research, and the road ahead. IEEE Journal on Selected Areas in Communications, 38:2450–2525, 2020.
  29. [29]George C. Alexandropoulos, Geoffrey Lerosey, Merouane Debbah, and Mathias Fink. Reconfigurable intelligent surfaces and metamaterials: the potential of wave propagation control for 6G wireless communications, 2020. arXiv:2006.11136.arxiv.org/abs/2006.11136
  30. [30]Tsampikos Kottos and Uzy Smilansky. Periodic orbit theory and spectral statistics for quantum graphs. Annals of Physics, 274:76–124, 1999.DOI
  31. [31]Sven Gnutzmann and Uzy Smilansky. Quantum graphs: applications to quantum chaos and universal spectral statistics. Advances in Physics, 55:527–625, 2006.DOI
  32. [32]Gregory Berkolaiko and Peter Kuchment. Introduction to Quantum Graphs, volume 186 of Mathematical Surveys and Monographs. American Mathematical Society, Providence, RI, 2013.DOI
  33. [33]Tristan Lawrie, Sven Gnutzmann, and Gregor Tanner. Closed form expressions for the Green’s function of a quantum graph—a scattering approach. Journal of Physics A: Mathematical and Theoretical, 56:475202, 2023.arxiv.org/abs/2309.11251
  34. [34]Tristan Lawrie, Gregor Tanner, and Dimitrios Chronopoulos. A quantum graph approach to metama- terial design. Scientific Reports, 12:18006, 2022.
  35. [35]T. M. Lawrie, T. A. Starkey, G. Tanner, D. B. Moore, P. Savage, and G. J. Chaplain. Application of quantum graph theory to metamaterial design: negative refraction of acoustic waveguide modes. Physical Review Materials, 8:105201, 2024.arxiv.org/abs/2409.07133
  36. [36]T. M. Lawrie et al. On the Doppler effect in static materials: vortices and black, white and worm-hole like wave trajectories, 2026. arXiv:2610.33141.arxiv.org/abs/2610.33141
  37. [37]T. M. Lawrie, G. Tanner, and G. J. Chaplain. Engineering metamaterial interface scattering coefficients via quantum graph theory. Acta Physica Polonica A, 144(6):486, 2023.DOI
  38. [38]T. M. Lawrie, G. Tanner, and G. J. Chaplain. Non-diffracting resonant angular filter. Physical Review Research, 7:023209, 2025.arxiv.org/abs/2410.17329
  39. [39]Michał Ławniczak, Tristan M. Lawrie, Szymon Bauch, Gregor Tanner, and Leszek Sirko. Experimental realization of a discrete k-space angular filter in microwave networks. Physical Review Applied, 25:064043, 2026.DOI
  40. [40]T. M. Lawrie and O. M. Brown. A flux-tunable discrete angular filter. Acta Physica Polonica A, 148(5):S25, 2025.DOI
  41. [41]T. M. Lawrie. Which waveguide network realizes a prescribed transmission profile? An exact forward construction, 2026. arXiv:2606.23645.arxiv.org/abs/2606.23645
  42. [42]Tristan M. Lawrie. A Quantum Graph Approach to Metamaterial Design. PhD thesis, University of Nottingham, 2024.DOI
  43. [43]Z. Tian, J. Bennett, J. Yang, T. Lawrie, W. Elmadih, A. Bardalai, C. Gerada, J. Zhu, and D. Chronopoulos. Experimental investigation of mechanical, acoustic and hybrid metamaterial designs for enhanced and multi-band electric motor noise dissipation. Engineering Structures, 271:114945, 2022.DOI
  44. [44]Luke G. Bennetts, Malte A. Peter, and Richard V. Craster. Graded resonator arrays for spatial frequency separation and amplification of water waves. Journal of Fluid Mechanics, 854:R4, 2018.DOI
  45. [45]Han Meng, Waiel Elmadih, Huan Jiang, Tristan Lawrie, Yanyu Chen, and Dimitrios Chronopoulos. Broadband vibration attenuation achieved by additively manufactured 3D rainbow hollow sphere foams. Applied Physics Letters, 119(18):181901, 2021.DOI
  46. [46]Peter Kuchment. Graph models for waves in thin structures. Waves in Random Media, 12:R1–R24, 2002.DOI
  47. [47]Oleh Hul, Szymon Bauch, Prot Pałoński, Nazar Savyt-skyy, Karol Życzkowski, and Leszek Sirko. Experimental simulation of quantum graphs by microwave networks. Physical Review E, 69:056205, 2004.DOI
  48. [48]Zixian Liang and Jensen Li. Extreme acoustic metamaterial by coiling up space. Physical Review Letters, 108:114301, 2012.pubmed.ncbi.nlm.nih.gov/22540476

Paper details

vlnc:W_bxrd68c2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents