Revealing hidden periodicity in momentum-encoded metasurfaces Authors: Seokwoo Kim, Jongha Yoon, Minsu Jeong, Joohoon Kim, Jihae Lee, Minseok Choi, Dongmin Jeon, Geon Lee, Kyungtae Kim, Junsuk Rho Published: 2026 ABSTRACT AbstractStructured planar materials known as metasurfaces enable versatile manipulation of electromagnetic waves at subwavelength scales. Especially, metasurfaces can serve as spatial sampling lattices that encode prescribed in-plane momentum profiles, as commonly employed in phase-gradient and Pancharatnam–Berry metasurfaces. However, such momentum-encoded metasurfaces typically lose global periodicity, complicating rigorous analyses. Here, we introduce a universal geometric framework that interprets momentum-encoded metasurfaces as directional Moiré lattices formed by imposing one-dimensional periodic perturbations onto two-dimensional Bravais lattices. We derive explicit commensurability conditions that reveal hidden periodicity in otherwise aperiodic structures and enable their representation as compact periodic superlattices. This supercell representation allows the discretization of radiation channels into defined diffraction orders, enabling clear, intuitive performance characterization. We validate the framework by applying it to local metasurfaces, obtaining diffraction results consistent with two-dimensional sampling theory, and to non-local metasurfaces with spin-dependent Rashba momentum, precisely identifying their superlattice conditions and their valley-addressable character. We further experimentally demonstrate the predicted valley addressability using fabricated momentum-encoded photonic crystal slabs. Our method thus provides an efficient and generalizable basis for analyzing complex metasurface architectures and exploring diverse lattice phenomena in wave physics.Subject terms: Nanophotonics and plasmonics, Metamaterials, Photonic devices, Topological matter, Computational methods Introduction Metasurfaces are planar components based on engineered lattices that mediate light–matter interactions at the subwavelength scale. The role of the lattice in metasurfaces can be broadly classified into two regimes—local and non-local—each supporting distinct manners of light control. First, the lattice functions as a local information storage platform, where each site, i.e., meta-atom, equidistantly samples the desired complex amplitude1. In this viewpoint, the metasurface lattice may be interpreted as a two-dimensional (2D) sampling grid2, encoding a target complex amplitude profile via geometric perturbations on scatterers, e.g., size scaling or in-plane rotation, at each lattice point. Through this process, an arbitrary spatial spectrum can be encoded on the surface with sub-wavelength resolution, thus enabling a broad range of functionalities, such as focusing3–6, holography7–10, and beam shaping11–14. Second, the lattice can serve as a coupling configuration between meta-atoms, where their spatial arrangement defines the crystalline symmetry. This symmetry underpins non-local phenomena, such as Bragg scattering15,16, band topology17–19, bound state in the continuum20–22, and valley-selective phenomena23–27. Notably, these two functions—local and non-local—are not mutually exclusive; the metasurface lattice can simultaneously function as both a local sampler and a non-local coupling template. This dual functionality is exemplified by non-local metasurfaces28–30 or the photonic Rashba effect26,27,31–33, where spatially varying phase or Berry-phase distributions encode an effective in-plane momentum onto the non-local lattice, giving rise to spectrally selective momentum shift of an incident wave. However, such momentum-encoded lattices generally lose their global periodicity since the periodicity and orientation of the encoded phase profile typically do not match the underlying lattice symmetry1. This forces us to adopt local periodic approximation34,35 and limit the scope of analysis to extremely simplified scenarios, i.e., supercells composed of integer multiples of the original lattice period, such as period doubling21 or tripling33 in order to use periodic boundary conditions. Here, we show that momentum-encoded metasurfaces can be collapsed into a single supercell by investigating a universal commensurability criterion of momentum-encoded lattices. We begin by introducing a geometric reinterpretation, where momentum-encoded metasurfaces can be understood as a distinct class of moiré lattices, which we term directional moiré lattices (Fig. 1a). Such lattices emerge naturally from the interference between a 2D host Bravais lattice and one-dimensional (1D) periodic perturbations. The formation of the moiré lattice can be explained by an overlapping of two different continuous scalar potentials36–38. In Fig. 1a, a base potential V0(r) constructs the original 2D lattice. When we introduce a 1D periodic perturbation potential, Vpr=∑nϵZδVneinkp⋅r, a total potential V(r)=V0(r)+Vp(r) gives rise to a new class of structures, directional moiré lattices. This perturbation constitutes a weak modulation with wavenumber kp=2πλp, resulting in encoded momentum kp in a base potential. In metasurfaces, these perturbations may stem from geometric variations, such as scaling or rotating, that lead to a momentum shift in the radiation or a dispersion coupling procedure (Fig. 1b). Concept of directional moiré superlattices. a Directional moiré lattice. A base potential V0 characterizes the 2D Bravais lattice. A 1D periodic perturbation Vp, applied along a lattice-derived direction, modulates this structure to form a directional moiré lattice. b Momentum encoding process in metasurfaces. Geometric perturbation, which is realized by size scaling or rotating to a meta-atom, introduces an additional momentum kp in the coupling process. c Phase diagram illustrating the two-step commensurability process that reveals hidden periodicity in directional moiré lattices. Unlike typical moiré systems formed by twisted bilayers36–41, directional moiré lattices exhibit hidden translational periodicity when sequential angular and length commensurability conditions are satisfied. (Fig. 1c). These conditions are precisely governed by both the magnitude and direction of the encoded momentum kp. We provide closed-form commensurability criteria to determine when initially aperiodic momentum-encoded structures can be represented as moiré quasi-superlattices and, subsequently, fully periodic superlattices. By uncovering hidden periodicities within seemingly complex momentum-encoded metasurfaces, our geometric framework enables rigorous, approximation-free analyses using compact superlattice representations. Leveraging this framework, we apply our methodology to analyze both local and non-local momentum-encoded metasurfaces. For local metasurfaces, we obtain diffraction results consistent with the previously unexplored 2D sampling theory. Additionally, for non-local metasurfaces encoded with various spin-dependent momenta, we systematically explore and identify their corresponding superlattice conditions. Finally, we experimentally demonstrate the predicted valley addressability using fabricated momentum-encoded photonic crystal slabs, confirming that the hidden-periodicity-based supercell description captures the momentum-space relocation of valley modes. Our results demonstrate the broad applicability and accuracy of our approach, opening new pathways toward robust and computationally efficient design of complex metasurface architectures. Results First commensurability condition: angular alignment As in Fig. 2a, let us consider a scenario where cylinder meta-atoms arranged in a square lattice are perturbed by scaling. As illustrated in Fig. 2b, the first step in moiré superlattice formation occurs when the applied perturbation becomes angularly commensurate with the underlying 2D Bravais lattice. In other words, the perturbation wavevector kp is parallel to a specific linear combination of the primitive lattice vectors a₁ and a₂, which is expressed as where m1(2)\inZ (Fig. 2b). This angular matching acts as a key symmetry constraint, allowing the lattice periodicity to coexist with the imposed modulation. Once angular commensurability is satisfied, an extended unit cell (EUC) can be defined—an enlarged unit cell that reflects the modified periodicity introduced by the perturbation. This EUC is constructed by two basis vectors a∥=m1a1+m2a2, and a⊥=n1a1+n2a2 where a∥∥kp, a⊥⊥kp, and m12,n1(2)∈Z. Note that the basis vectors of EUC decompose the perturbed and non-perturbed direction while preserving the original lattice symmetry of the host lattice. At this stage, the system has strict periodicity along a⊥ while remaining aperiodic along a∥ (Fig. 2c). This intermediate state of moiré lattice defines a moiré quasi-superlattice, characterized by long-range correlations along the commensurate direction but lacking global periodicity. Directional moiré superlattice formation. a Moiré lattice formed by applying a 1D perturbation to a host Bravais lattice. For clarity, the coordinate axes are rotated such that the perturbation direction is aligned horizontally. b First commensurability condition for the moiré lattice. When the perturbation direction is parallel to a lattice vector, an EUC emerges, defined by basis vectors a∥ and a⊥, which are parallel and orthogonal to the perturbation direction, respectively. c Representation of a 1D moiré quasi-superlattice satisfying the first commensurability condition. The structure consists of a linear array of EUCs. d Second commensurability condition. e Moiré superlattice representation illustrates that the entire structure can be tiled using an SUC. f Example illustrating the hierarchical relationship among the fundamental unit cell, EUC, and SUC. g, h Calculated EUC and SUC sizes as functions of the perturbation orientation. The plotted points correspond to configurations satisfying the commensurability conditions. The condition for EUC existence, i.e., angular commensurability, is obtained from the orthogonality between a⊥ and a∥, and can be expressed as a Diophantine equation: where both s=∣a2∣/∣a1∣ and θ=cos−1a1⋅a2∣a1∣∣a2∣ are parameters defining the host Bravais lattice, which is an unperturbed lattice. This Diophantine equation implies the geometric constraints that determine valid EUC configurations. When angular commensurability holds, the twist angle ϕ (see Fig. 2b), defined as the angle between the basis vector (a1) of the host lattice and the perturbation direction, can be expressed as follows: The relationship between the host lattice basis (a₁, a₂) and EUC basis (a∥, a⊥) is described by a linear transformation matrix TEUC: where TEUC=m1n1m2n2 is the explicit form of the transformation matrix. Eigenvalue analysis of this matrix yields complex eigenvalues with magnitudes exceeding one, indicating that the linear transformation involves both rotation and scaling. This structure reflects the fundamental geometric relationship between the primitive unit cell (PUC) and EUC. The area of the EUC is expressed as AEUC=∣detTEUC∣⋅APUC, where APUC=∣a1×a2∣ denotes the area of the PUC. This area scaling factor quantifies the extent of lattice expansion induced by the perturbation and serves as a key parameter for estimating the magnitude of moiré-induced effects in perturbed lattices. Second commensurability condition: length matching Although the first commensurability condition establishes angular commensurability, the resulting quasi-1D array remains incommensurate with the perturbation period λp. Creating a fully periodic moiré superlattice requires an additional condition: length commensurability, wherein the EUC periodicity and perturbation wavelength are rationally related (Fig. 2d). This second commensurability condition is met when integer multiples of the EUC length p∣a∥∣ along the perturbation direction are commensurate with integer multiples of the perturbation period qλp: where p,q∈Z>0. When this rational relationship holds, the system transitions from a 1D quasi-periodic arrangement to a 2D periodic superlattice, which can be fully represented by a single superlattice unit cell (SUC, Fig. 2e). From length commensurability, the basis vectors of SUC are constructed by integer multiples of the EUC basis set, i.e., pa∥,a⊥. This ensures that the SUC contains exactly p consecutive EUCs along the perturbation direction while preserving the original perpendicular periodicity. The resulting superlattice defines the minimal periodic unit that captures the full moiré pattern. The transformation matrix transforming the PUC basis to the SUC basis is described by the following hierarchical matrix product: where all matrix elements are integers. This matrix encapsulates the full geometric transformation from the original host lattice to the final superlattice, providing a unified description of both angular and length commensurability. The area of the SUC is expressed as ASUC=∣det(TSUC)∣AUC=pAEUC. This scaling relationship confirms that the SUC area is exactly p times greater than that of the EUC, reflecting the rational multiplication factor required for length commensurability. The hierarchical relationship among the PUC of the unperturbed Bravais lattice, EUC, and SUC constitutes a key architectural feature of moiré superlattices (see Fig. 2f). The sizes of EUC and SUC vary with the twist angle ϕ and periodicity λp (Fig. 2g, h). Furthermore, in square lattices, the EUC size diverges near high-symmetry directions (e.g., 45°), reflecting discrete commensurate matching conditions. This divergence is a hallmark of moiré lattices42, underscoring their extreme sensitivity to angular alignment and the discrete symmetries of the host lattice. Both the divergence of the EUC and the discrete growth of the SUC originate from the intrinsic incompatibility between continuous geometric parameters and discrete integer commensurability, which forces increasingly large integer solutions near incommensurate conditions, a fundamental mechanism underlying moiré physics. Likewise, the SUC size exhibits a complex dependence on the perturbation period (Fig. 2h), with discrete jumps corresponding to different rational commensurability ratios. The plotted configurations represent solutions to the two Diophantine equations, with integer parameters constrained to m1,m2,p,q∈[0,50]. Such sensitivity necessitates the careful selection of rational commensurate conditions that ensure manageable EUC and SUC sizes for computational feasibility. Note that while a square lattice is used as the host lattice example here, the proposed framework is equally applicable to other types of Bravais lattices (Supplementary Note 1). Momentum-resolved analysis of momentum-encoded metasurfaces In this section, we apply the directional moiré superlattice framework to metasurfaces in which the surface phase is spatially encoded on each meta-atom to control radiation channels. As illustrated in Fig. 3a, phase encoding can be implemented by introducing geometric perturbations across a square-lattice metasurface. Here, we primarily consider cylindrical meta-atoms for clarity; however, the framework is not limited to meta-atom shape and remains applicable to any scatterer geometry, including anisotropic meta-atoms that rely on rotational perturbations to produce Pancharatnam–Berry (PB) geometric phase22,43; this generality is further supported by full-wave simulations and experimental measurements on PB-type momentum-encoded metasurfaces (Supplementary Note 2). Momentum-resolved analysis of directional moiré superlattices. a Conceptual illustration of a directional moiré superlattice formed by applying a one-dimensional periodic perturbation (periodicity λp) onto a two-dimensional host lattice, resulting in a compact SUC. b Schematic depiction of diffraction orders generated by the momentum-encoded metasurface under illumination. Target, aliasing, and stray diffraction orders are indicated. c Discretized radiation channels in reciprocal space, classifying diffraction into target, zeroth-order, aliasing, and stray orders. d Momentum-resolved diffraction efficiency maps showing how target-order diffraction performance varies with meta-atom refractive index. Each momentum-space point represents a target diffraction order defined by the directional moiré superlattice, and the color indicates the corresponding diffraction efficiency. e Influence of lattice periodicity (from 300 to 450 nm) on the emergence and evolution of aliasing diffraction orders. The dotted lines denote the sampling criterion obtained from multi-dimensional sampling theory. When engineering the far-field response, the design generally follows the principles of the generalized Snell’s law14, in which phase gradients are introduced to redirect light toward a target direction. Once the phase-encoded directional moiré metasurface is constructed, the resulting radiative channels include the zeroth-order non-diffracted beam, the target diffraction order, the aliasing orders that emerge when the spatial sampling condition is not satisfied1,2, and stray orders caused by unwanted near-field coupling and fabrication imperfections. Except for the target order, all remaining diffraction components can be interpreted as undesired diffractions. As discussed in the previous section, when the encoded momentum and host lattice satisfy both angular alignment and length commensurability, a superlattice can be constructed, allowing us to define a SUC with a well-defined 2D periodicity. This periodicity induces a discretized momentum basis, such that each radiative channel corresponds to a distinct diffraction order, enabling a physically intuitive classification of all scattered directions (Fig. 3c). The reciprocal lattice vectors of SUC are defined as (G∥,G⊥) where (G∥,G⊥)=2πpa∥,2πa⊥, and p is the integer scaling factor introduced by the second commensurability condition. In the far-field scattering process, all momentum transfers Δk between the incident wave and the metasurface can then be represented as linear combinations of these reciprocal basis vectors. Within this framework, the engineered radiation—the target diffraction order—is characterized by Δk=kp=qG∥, such that it maps to the coordinate (q,0) in the reciprocal momentum basis of SUC. If the spatial sampling condition is violated due to coarse meta-atom spacing, additional aliasing components appear, described by Δk=kp±Gi(i=1,2), where G1 and G2 are the reciprocal lattice vectors of the host lattice, since the host lattice acts as a sampling lattice, where the encoded spectrum in the Fourier space of the metasurface is replicated at intervals of G1 and G2. These aliasing orders correspond to diffraction directions indexed as (q±pmi,±ni), where detailed derivations are in the “Methods.” On the other hand, the zeroth-order transmission corresponds to the case Δk=0, which maps to the coordinate (0, 0). We further define stray orders as diffraction orders that do not fall into the target or aliasing directions. Through this order classification of discretized far-field components, it becomes possible to benchmark metasurfaces in a momentum-resolved manner, even for aperiodic or nonuniform phase profiles. Figure 3d, e show calculated diffraction orders—including target and aliasing components—for each encoded momentum, assuming cylindrical meta-atoms with a full 2π phase coverage and known refractive index and lattice parameters. In these momentum-space maps, each point corresponds to a diffraction order defined by the directional moiré superlattice, while the color scale represents the calculated diffraction efficiency coupled into that order. We note that, unlike PB-type meta-atoms whose phase varies linearly with rotation angle, the phase response of cylindrical meta-atoms does not follow a linear dependence on diameter; therefore, all encoded perturbations were implemented using precomputed nonlinear diameter–phase lookup tables rather than assuming linear scaling (look-up tables are in Supplementary Note 3). In Fig. 3d, as the refractive index of the meta-atoms decreases, diffraction efficiency tends to drop. This occurs because low-index meta-atoms are unable to sufficiently confine incident light, leading to enhanced interactions between adjacent meta-atoms44 and, consequently, a breakdown of the periodic approximation34,35 commonly assumed in meta-atom design. Since most phase maps are extracted under periodic assumptions, performance degradation becomes more severe for higher momentum encodings and lower refractive indices. The supercell-based simulation framework successfully captures these effects, offering a physically realistic assessment of diffraction performance. Interestingly, the behavior of aliasing orders further illustrates the predictive power of this framework. In Fig. 3e, we calculate the coupling efficiency of aliasing orders for varying lattice periods under different encoded momenta. The dashed boundaries in each graph are analytically derived from multi-dimensional sampling theory1,2 and correspond to the onset of aliasing. The simulation results conform precisely to these theoretical boundaries, validating the accuracy of the momentum-resolved analysis. As the lattice period increases, higher-order components in the spatial frequency spectrum are more readily coupled out, and the aliasing behavior exhibits a strong dependence on the direction of the encoded momentum. This directional dependence highlights the need for a fully momentum-resolved, multi-dimensional treatment of sampling in metasurface design. To our knowledge, this presents the first comprehensive calculation of sampling performance in metasurfaces that is fully resolved in momentum space. Without leveraging the supercell framework, evaluating each momentum encoding would require full-wave simulations of finite-sized metasurfaces—an approach that is computationally demanding and less generalizable. Directional moiré superlattices in non-local metasurfaces In the previous section, we focused on the role of the lattice as a local regime, i.e., sampling lattice, and analyzed its behavior within the moiré superlattice framework. In this section, we extend our discussion to demonstrate that the same geometric framework remains valid even when the lattice operates both locally and non-locally in tandem. Notably, recent work33 has demonstrated that introducing the gradient of Berry phase in a non-local metasurface (or photonic crystal slab) can shift Dirac cones at high symmetric points in the Brillouin zone into the light cone, enabling efficient coupling of guided modes to free-space radiation. Here, the moiré superlattice framework can systematically describe and control the directional moiré patterns arising from Berry-phase-encoded photonic crystals. We consider a hole-based honeycomb photonic crystal (Fig. 4a). This structure features a lattice constant of 260 nm and a complex refractive index of n = 2 + 0.005i, with the imaginary part introduced to account for optical losses in the calculation. The lattice comprises two distinct sublattices: anisotropic elliptical holes (sublattice A, minor and major axes of 58 and 98 nm, respectively) and isotropic circular holes (sublattice B, diameter of 120 nm). This structural asymmetry breaks the inversion symmetry between the two sublattices, thereby opening Dirac gaps at the K and K′ points of the Brillouin zone (Fig. 4a, b). Throughout this work, we use the term “Dirac cone” to refer to these gapped valley dispersions at K and K′, following the convention widely adopted in the valley photonics literature. Demonstration of valley addressability by directional moiré metasurfaces. a Real-space design flow: starting from a C6-symmetric honeycomb lattice, inversion-symmetry breaking reduces the symmetry to C3 and defines inequivalent sublattices A and B, opening a gap at the Dirac point. A 1D Berry-phase gradient with wavevector kp is then encoded selectively on sublattice A, forming a directional moiré superlattice supercell composed of multiple EUCs. b Momentum-space picture: inversion-symmetry breaking gaps the Dirac cones at K and K′, while the subsequent Berry-phase encoding imparts opposite momentum shifts, +kp and −kp, to the valley states. This displacement folds the valley dispersions into the light cone, enabling radiative coupling and valley-selective addressing. c Distribution of kp vectors satisfying the directional moiré superlattice condition. d SUCs corresponding to the four superlattice configurations marked in (c). e Calculated angle-resolved CD distributions at f = 520 THz for the SUCs shown in (d). Blue and red contours indicate opposite CD signs, originating from spin-valley selective coupling of the Dirac cones at K and K′. Each encoded Rashba momentum is illustrated in the Brillouin zone. Then, applying an in-plane rotational perturbation to sublattice A introduces spin-dependent momentum shifts, analogous to the Rashba effect31–33 (Fig. 4a, b). Consequently, the Dirac cones located at the K and K′ valleys undergo momentum displacements, which move them from outside to inside the light cone at the target frequency. Specifically, the Dirac cone at K shifts by +kp, while its time-symmetry conjugate at K’ shifts by −kp (Fig. 4b). We next identify perturbation momenta kp, that satisfy the angular and length commensurability conditions required for superlattice formation. Figure 4c presents candidate perturbation momentum vectors for encoding, represented as the post-encoding positions of the Dirac cone at the K point. From these, we select four representative vectors (highlighted by red rectangles in Fig. 4c) that ensure the Dirac cones lie within the light cone (shown as a dotted circle) while also limiting the SUC size to a manageable computational scale, defined as det(TSUC) < 30. Note that these vectors are selected for demonstrative purposes. Figure 4d illustrates the resulting SUCs, each annotated with their respective commensurability indices [m1, m2; p, q]. To confirm the directional relocation of Dirac cones into the light cone, we perform angle-resolved simulations under periodic boundary conditions at a frequency of 520 THz using circularly polarized light. We adopt the rigorous coupled-wave analysis for simulation. Figure 4e presents the resulting circular dichroism (CD) distributions, defined as the absorption difference between right- and left-circularly polarized incident light. These reveal Dirac cone cross-sections shifted into the light cone according to the encoded perturbation momenta. Owing to intrinsic spin-valley locking at the K and K′ valleys, each shifted Dirac cone exhibits distinct valley-selective circular polarization, which appears as red or blue contours corresponding to opposite CD signs. The calculated band structures for each photonic crystal slab are in Supplementary Note 4. Having established the superlattice conditions and the predicted momentum-space relocation of the valley modes, we next experimentally test whether these commensurability-defined shifts can be directly observed. To experimentally validate the proposed directional moiré superlattice framework in the context of valley addressing, we fabricate momentum-encoded photonic crystal slabs using a silicon nitride (Si₃N₄) thin film with a thickness of 70 nm. The structures are defined by electron-beam lithography to create a resist mask, followed by dry etching to transfer the designed patterns into the Si₃N₄ layer. (see “Methods” section) The fabricated lattice consists of a hexagonal arrangement with a period of 310 nm. The anisotropic sublattice A is composed of elliptical holes with major and minor axes of 130 and 65 nm, respectively, while sublattice B consists of circular holes with a diameter of 140 nm. Due to fabrication constraints, the geometric parameters of the fabricated samples are slightly modified from those used in the numerical calculations in Fig. 4. In particular, the lattice period and feature sizes are adjusted to ensure reliable pattern transfer and structural fidelity. Nevertheless, these modifications do not affect the essential physics governed by geometric commensurability. To probe the momentum-resolved photonic response, angle-resolved measurements are performed along a 1D trajectory in momentum space (Fig. 5). Since the valley modes originate from Dirac cones that are selectively shifted into the light cone along specific directions, it is necessary to align the measurement axis with the encoded momentum direction. To achieve this, each sample is physically rotated such that the momentum-shift direction of the valley modes coincides with the fixed scanning axis of the measurement system. This alignment enables direct access to the relevant band dispersion within the measurable angular range, effectively capturing the photonic band structure along a 1D momentum cut intersecting the shifted valleys. Experimental validation of valley addressing enabled by directional moiré superlattices. a–d SEM images of fabricated momentum-encoded photonic crystal slabs with different orientations of the encoded momentum relative to the underlying lattice. The white rectangles indicate the corresponding superlattice unit cells defined by the commensurability conditions. Scale bars, 500 nm. e–h Corresponding momentum-space representations of each configuration. The hexagonal Brillouin zone of the host lattice is shown with high-symmetry points K and K′. The dashed circle denotes the light cone, within which radiative channels can be accessed. Bottom panels, angle-resolved circular dichroism (CD) measurements as a function of incident angle and wavelength. Distinct CD features corresponding to momentum-shifted valley modes are observed and systematically evolve with the encoded momentum direction, consistent with the predicted relocation of the valley states in momentum space. As shown in Fig. 5, the angle-resolved CD measurements reveal distinct spectral features that systematically evolve with the orientation of the encoded momentum. Here, CD is defined as the transmission difference between right- and left-circularly polarized incident light. These features correspond to the momentum-shifted valley modes predicted by the directional moiré superlattice analysis. The observed agreement between the measured CD response and the theoretical momentum-space picture confirms that the proposed framework accurately captures the redistribution of radiative channels and enables deterministic valley addressing in momentum-encoded photonic crystal slabs. Discussion We have introduced a geometric framework that identifies hidden periodicity in momentum-encoded metasurfaces and represents otherwise aperiodic structures as fully periodic superlattices through a two-step commensurability process. The key strength of this framework lies in its generality. While our demonstration focused on linear phase perturbations derived from the generalized Snell’s law14, the underlying formulation is not limited to such cases. Arbitrary periodic perturbations—including nonlinear, non-analytic, or superposed functions—can be incorporated, as long as the commensurability conditions are satisfied (Supplementary Note 5). Moreover, the methodology is inherently modality-agnostic: relying solely on geometric symmetry and lattice matching, it is readily applicable to various wave-based systems, such as electronics, acoustics, and mechanical metamaterials. Moreover, the lattice does not have to be a discrete system; our formulation accommodates a spatially continuous potential lattice38. Beyond single-directional perturbations, our framework naturally extends to bi-directional configurations (Supplementary Note 6). Beyond far-field radiation and coupling-process engineering demonstrated in this work, the directional moiré superlattice framework may, in principle, be extended to explore other lattice-driven phenomena governed by commensurability and symmetry. For example, moiré-induced flat-band physics37–40, widely studied in electronic and photonic systems, naturally arises from long-period supercell formation. While such phenomena are beyond the scope of the present study, our framework provides a compact and deterministic geometric basis that could facilitate their systematic investigation in momentum-encoded metasurfaces. Nonetheless, several limitations remain. Notably, our current formulation assumes a uniform momentum encoding across the metasurface. In practical scenarios, e.g., metalenses3,5,6, the phase gradient varies continuously across space, locally spanning a spectrum of momenta. Under such conditions, the concept of a single supercell breaks down, and the advantages of periodic simplification are diminished. However, even in these cases, the directional moiré framework offers value as a momentum-resolved benchmarking tool without relying on the local periodic approximations34,35. By evaluating metasurface responses under individual momentum encodings, one can indirectly assess the performance of a meta-atom library, providing physical insights into the bandwidth, sampling fidelity, and diffraction efficiency. Finally, the directional moiré framework presents a fundamentally distinct design paradigm compared to traditional bilayer moiré systems36–41, which rely solely on global twist angles between lattices. By introducing two independent commensurability conditions—angular alignment and length matching—our framework offers greater geometric flexibility and design freedom. This dual-commensurate approach expands the accessible design space for realizing programmable supercell physics. Methods Diffraction order identification In the moiré superlattice, we recover a 2D discrete translation symmetry with primitive periods pa∥ and a⊥, where p∈Z>0. With the basis of the unperturbed lattice as APUC=[a1a2], the basis of superlattice ASUC is written with a∥=m1a1+m2a2 and a⊥=n1a1+n2a2 (mi,ni\inZ). For a square lattice as a host lattice, the orthogonality condition forces n1=−m2 and n2=m1; the following expressions, however, remain valid for any host Bravais lattice. The reciprocal basis of the superlattice is obtained from that of the unperturbed lattice, G=[G1G2]=2π(A−1)T, which gives where Δ=m1n2−m2n1≠0. Hence, the discretised far-field grid is spanned by the magnitudes G∥=2π/(pa∥) and G⊥=2π/a⊥. Any radiative channel emerging from the metasurface with in-plane momentum Δk is classified by the integer pair so that (N∥,N⊥) denotes the diffraction order in the superlattice basis. The metasurface is designed to impart an in-plane momentum of magnitude kp=2π/λp parallel to G∥. Enforcing the commensuration condition qλp=pa∥ in the second commensurability condition yields kp=qG∥ and therefore i.e., the intended far-field channel is the (q,0) order. Additional aliasing channels arise because sparse sampling distance of the unperturbed lattice, which results in the folded momentum: Expressing each primitive EUC reciprocal vector in the superlattice basis via ei′=TSUCTei (with e1=(1,0)T and e2=(0,1)T) gives Combining with kp we obtain the aliasing diffraction orders For the square lattice subset (n1,n2)=(m2,−m1), the dominant aliasing channel is (q−pm1,n1) for twist angles 0∘<φ<45∘ and (q−pm2,−n2) for 45∘<φ<90∘; the algebraic form above ensures continuity across these regimes. Fabrication of directional moiré metasurfaces Metasurfaces were fabricated on a glass substrate with a 500 μm thickness. As a first step, a 70 nm-thick layer of silicon nitride (Si3N4) was deposited using plasma-enhanced chemical vapor deposition (PECVD, BMR Technology HiDep-SC). A layer of positive tone photoresist (ZEP 520 A) was spin-coated at 5000 rpm onto the Si3N4 film and subsequently baked at 180 °C for 2 min. High-speed electron-beam lithography (ELS-BODEN, Ellionoix) was used to transfer the metasurface patterns onto the photoresist. The exposed photoresist patterns were developed by immersing the sample in ZEP developer (ZED-N50) at 0 °C for 1 min. The ZEP patterns were transferred onto the silicon nitride film using a dry etching process (DRM85DD, TEL). Finally, the remaining ZEP was removed by immersing it in the anisole solvent. Supplementary information Supplementary Information Transparent Peer Review File Acknowledgements This work was financially supported by the POSCO-POSTECH-RIST Convergence Research Center program funded by POSCO, and the National Research Foundation (NRF) grant (RS-2024-00356928) funded by the Ministry of Science and ICT (MSIT) of the Korean government. S.K., M.J., and J.K. acknowledge the Sejong Science fellowships (RS-2026-25484953, RS-2026-25489396, RS-2026-25496744), respectively, funded by the MSIT of the Korean government. M.C. and G.L. acknowledge the Presidential Science fellowship funded by the MSIT of the Korean government. G.L. and K.K. acknowledge the NRF Ph.D. fellowships (RS-2024-00406756, RS-2025-25436773), respectively, funded by the Ministry of Education (MOE) of the Korean government. Author contributions S.K. and J.R. conceived the idea and initiated the project. S.K. designed the overall research, developed the theoretical framework, and performed the main theoretical analyses, numerical simulations, and data interpretation. J.Y. assisted with theoretical calculations and numerical simulations. J.K. and J.L. assisted with the nanofabrication of the devices. M.J. assisted with the band structure and angle-resolved optical measurements. S.K., M.J., J.K., and J.Y. analyzed the experimental and numerical data. M.C., D.J., G.L., and K.K. contributed to discussions and provided comments that helped refine the concept and interpretation. S.K. mainly wrote the manuscript. All authors discussed the results and approved the final version of the manuscript. J.R. guided the entire work. Peer review Peer review information Nature Communications thanks Patrice Genevet, Baile Zhang, and the other anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Data availability The data generated in this study are provided in the manuscript, Supplementary Information. Additional data supporting the findings of this study are available from the corresponding author upon request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information The online version contains Supplementary material available at 10.1038/s41467-026-74150-5. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review File Data Availability Statement The data generated in this study are provided in the manuscript, Supplementary Information. Additional data supporting the findings of this study are available from the corresponding author upon request.