Intrinsically nonlocal metamaterials Authors: A Ware, J LaMountain, R C White, S R Bank, E Narimanov, D Wasserman, V A Podolskiy Published: 2026 ABSTRACT AbstractThe electromagnetic response of materials serves as the foundation for a broad range of vital applications, from imaging, to sensing, to classical and quantum communications. Here we demonstrate, theoretically and experimentally, a fundamentally new regime of electromagnetic material response originating from inherent material nonlocality. We show that by structuring materials on the intrinsic scale of this nonlocal response, it becomes possible to alter the electromagnetics of the composite, revealing the inherent nonlocal behavior of the constituent components. These intrinsically nonlocal metamaterials exhibit strong intrinsic (as opposed to effective) nonlocality, easily detectable at room temperatures, in realistic (lossy), macroscopic materials. Intrinsically nonlocal metamaterials open a new design space for electromagnetic composites, beyond photonic crystals, metasurfaces, and effective medium composites. This allows the control of electromagnetic fields at a deep subwavelength scale, revealing a new dimension for control of light-matter interactions.Subject terms: Metamaterials, Nanophotonics and plasmonics Introduction Nonlocality1 typically reveals itself in exotic, difficult-to-achieve regimes, such as entangled quantum systems, highly controlled nanoscale structures, or ultra-low-temperature single-crystal materials2–13. The electromagnetic response of any material is inherently nonlocal, defined at the fundamental intrinsic nonlocality scale (linl) by the classical or quantum dynamics of its charges in its crystal lattice. While this scale can be as large as microns in low-temperature phononic materials, it is on the order of 100 nm in plasmonic and excitonic media, and is reduced to single nanometers for almost all dielectrics. As a result, the electromagnetic response of the vast majority of everyday materials is not affected by nonlocal corrections. Here we demonstrate that by spatially structuring nonlocal materials, it becomes possible to bring this inherent nonlocal electromagnetism back to the macroscopic scale. Notably, this new metamaterial regime signifies the unexpected breakdown of the local effective medium description14 of composites, and it is fundamentally different from the “structural nonlocality” often discussed in the context of homogenizing the response of metamaterials and metasurfaces with local components15–25 as well as from other common regimes of optical composites (Fig. 1). We realize intrinsically nonlocal metamaterials in an epitaxially-grown, all-semiconductor, room-temperature materials platform. Our research presents new opportunities for shaping the optical response of materials: we are able to design the distribution of electromagnetic fields within, and the propagation of light through, these composite nanostructured media by leveraging a heretofore difficult-to-realize mechanism for controlling light-matter interaction. Different regimes of optical composites. The behavior of optical composites was thought to be determined by the relationship between the typical inhomogeneity scale d and the operating wavelength λ0. When d ≫ λ0, the response of the laminates and large-scale optical assemblies is best described by the ray optics formalism; the photonic crystal d ~ λ0 regime results from interference-based phenomena; the response of finely-structured d ≪ λ0 composites is known to resemble a homogeneous effective medium, with granularity-induced deviations from effective-medium response often described by structural nonlocality15 - 24. The response of small, isolated particles is known to deviate from bulk-material models due to inherent nonlocality5,6,10,38; here we show that this intrinsically nonlocal response fundamentally alters the response of the composite as a whole when d ~ linl (with linl being the nonlocality scale), revealing the intrinsically nonlocal metamaterial regime that is dominated by the interference of nonlocal modes. Materials’ response to light can be described in terms of the polarization distribution, fundamentally tied to the dynamics of charges within the materials26,27. The finite mass of these charges yields a time lag between the driving external electric field and the resulting polarization response. This time-lag results in the dispersion (frequency-dependence) of polarizability, the function that relates the polarization within the material and the external electric field (see Fig. 2a). This behavior is generally local–polarization P at a given location inside the material depends only on the electric field E at that same point. However, at the nanoscale, where charge dynamics are non-trivial, materials’ responses in adjacent positions couple to each other. For example, in free-electron-dominated plasmonic materials, electron motion due to the incident electromagnetic field induces polarization at the electron’s current location—potentially shifted from the electron’s location at the time of field application. Intrinsic nonlocalities in bulk materials. a A material’s response to an excitation electric field E pulse is often delayed in time and can affect the regions neighboring the excitation point (e.g., due to charge motion). The time-delay yields a frequency dependence of polarizability P (and equivalently, permittivity) while the spatial nonlocality yields a dependence on wavenumber k~, introducing additional waves; b The dispersion of the two waves supported by a highly-doped nonlocal plasmonic semiconductor, as a function of vacuum wavelength λ0, modeled with local (symbols) and nonlocal (lines) permittivity (see Methods and SI for details; note the nonlinear scale of the vertical axis). Despite the existence of an additional wave, its large propagation constant prevents efficient coupling of diffraction limited light to this mode, leaving the reflectivity of the plasmonic layer unaffected by the intrinsic nonlocality; c illustrates reflectivity of the layer for light incident from vacuum at 60o, modeled via local (symbols) and nonlocal (line) descriptions. Electromagnetic nonlocality, originally introduced and described in Ref. 1, makes the polarizability dependent on wavenumber in addition to frequency. Nonlocal behavior qualitatively changes materials’ optical response by introducing additional electromagnetic waves. We illustrate this phenomenon using a highly doped (plasmonic) semiconductor as a representative material. Figure 2b illustrates the dispersion of two waves in this material (characterized via dimensionless propagation constant k~=kc/ω, with ω being the angular frequency of light and c being the speed of light in vacuum; see Methods and supplementary information (SI) for details of analytical derivations and numerical calculations). Note that the dispersion of one of the modes (primary wave) is nearly identical to the prediction of local electromagnetism, while the second (additional) wave can only be described by a nonlocal formalism. At shorter wavelengths, both waves propagate (k~2>0) while at longer wavelengths, both waves exhibit evanescent decay (k~2<0). While this behavior somewhat resembles the structural nonlocality often seen in metamaterials and metasurfaces15,17–25,28–36—in that both cases feature wavevector-dependent permittivity and are associated with excitation of additional waves—the two regimes rely on fundamentally different physics and are therefore distinct from each other (see SI for a more detailed discussion). In the conventional regime, generally referred to as structural nonlocality, the actual response of the underlying materials remains fundamentally local, and nonlocality represents the mathematical formalism used to simplify the modeling of multi-scale composites15,28. The appearance of effective additional waves in these composites can often be related to components’ resonances and/or to diffractive phenomena originating from the granularity of the inherently local composites. Furthermore, in these composites, the scale of spatial modulation of electromagnetic fields is defined by the size of the inclusions d, with structural nonlocality corrections vanishing in the limit d ≪ λ017,37. In contrast, intrinsic nonlocality represents a novel channel for light-matter interaction at the materials’ level associated with the intrinsic nonlocality lengthscale linl and manifested by additional waves that cannot be modeled by local-electromagnetic frameworks. For the free-carrier response linl ~ λ0vF/c with vF being Fermi velocity. The existence of inherent nonlocality in isolated plasmonic nanostructures has been widely recognized by the plasmonics community5,6,12,28,38, and signatures of resonant excitations of additional waves have been predicted in numerical analysis of plasmonic-local dielectric multilayers and dielectric-patterned plasmonic films39,40. As we show below, structuring a nonlocal composite on the length scale of the order of inherent nonlocality (d ≲ linl) fundamentally alters the electromagnetic response of the composite, often making the effective medium description impossible even when d ≪ λ0. Notably (see Fig. 2b), the magnitude of the propagation constant of the additional wave in commonly used materials is significantly higher than that of its primary counterpart, virtually preventing the coupling of free-space radiation to nonlocality-driven additional modes. As a result, corrections due to nonlocality in the majority of previously explored situations are vanishingly small. As an example, Fig. 2c illustrates reflection from a plasmonic doped semiconductor, calculated using both local and nonlocal polarizability descriptions (see SI). The local and nonlocal far-field optical responses of this representative system are essentially identical. Up until now, inherent nonlocality has only been observed in a few exotic situations, such as single-crystal phononic materials operating in deep cryogenic environments at epsilon-near-zero frequencies3,4 or in coupled plasmonic nano-particles5,6,10,38. As a result, analysis of nonlocal materials has been limited to isolated nonlocal layers or to bi-component nanostructures. In this work, for the first time, we consider macroscopic composite materials comprising multiple adjacent intrinsically nonlocal components and present an approach to bring the underlying nonlocal response of materials to the forefront of their electromagnetic response. Results Emergence of intrinsic nonlocality The fundamental limitation to accessing and leveraging the nonlocality-related additional modes in any optical material lies in these modes’ weak coupling to free-space radiation. Homogeneous materials exhibit measurable nonlocality only when their permittivity is vanishingly small, such that propagation constants of nonlocal waves are small enough to interact with incoming light27. Alternatively, as demonstrated in this work, composite structures may utilize coupling between additional waves in their components, thereby engineering the behavior of the resultant coupled modes and enabling their coupling with free-space light. The origin of the intrinsically nonlocal metamaterial behavior can be uncovered if we follow the evolution of the electromagnetic response of a (weakly) nonlocal composite as the size of its components is gradually reduced from the macroscopic wavelength scale to the deeply subwavelength nonlocality scale. While the transition to the nonlocal response is a universal property of any composite material, we illustrate this transition with the example of stratified media comprising nonlocal components (Fig. 3a), focusing on the most relevant subset of regimes illustrated in Fig. 1. Field distributions across composites operating in different electromagnetic regimes. a Periodically stratified medium with plasmonic and dielectric components, used here as an example. When the size of the components d = 1.5 μm is comparable with the operating wavelength b the composite operates in the photonic crystal regime, with highly oscillatory electromagnetic fields resulting from the interference of counter-propagating waves inside the components (By component of the magnetic field is shown). As the components’ size is reduced d = 250 nm ≪ λ0 the electromagnetic fields become increasingly homogeneous, with the metamaterial transitioning to the effective medium regime c that is best described by averaged (effective) permittivities; up until this point, the effect of underlying nonlocality appears negligible, with local calculations adequately reflecting the response of the composite. However, when the size of the components is reduced further (d = 10 nm) and becomes comparable to the nonlocality length scale (here, linl ~ 50 nm), the field of the fundamental mode in the composite becomes strongly inhomogeneous, in sharp contrast to the predictions of local calculations (d). This clear signature of the intrinsically nonlocal metamaterial regime results from the interference of counter-propagating additional waves inside the components. In (b–d), the top bar represents a 3-μm section of the composite; the permittivity of the dielectric layers is fixed at ϵ=10+(10−4+3⋅10−5i)k~z2 with the nonlocal portion ignored in local calculations; plasmonic materials are modeled using the nonlocal Drude model, see Methods and SI. We follow the formalism originally developed to analyze the dispersion of modes in periodic arrays of electromagnetically-local layers41 and develop a nonlocal transfer matrix method (see Methods for details). We then use the developed approach to analyze the propagating properties of the waves in periodic electromagnetically nonlocal arrays. Our results are summarized in Figs. 3, Fig. 4 that show the field distribution in the fundamental (largest k~z2) mode across the unit cell of the composite, for fixed composition and operating wavelength (λ0 = 5.05 μm), and the dispersion of modes in periodic bi-layer structures of fixed composition as a function of operating wavelength, respectively. Dispersion of electromagnetic modes in composites operating in different electromagnetic regimes, illustrated on the example of periodically stratified materials with layer size d (see Fig. 3a), as a function of vacuum wavelength λ0. a Composites operating in the photonic crystal regime (d = 1.5 μm) exhibit a series of transparent bands and reflective bandgaps with propagating and evanescent modes, respectively; note that the nonlocality-induced additional wave rapidly decays inside the composite (k~2≪−1). As the layer size is reduced (d = 250 nm), the response of the composite converges to that of an effective medium b with the properties of the main wave well-described by averaged permittivities; deviations from the local effective medium response can be described in terms of a structural nonlocality, resulting in corrections ∝ d/λ0. Notably, the inherent-nonlocality-induced additional mode, once again, exponentially decays inside the composite (k~2≪−1). However, when the size of the composite is reduced further (d = 10 nm) and becomes comparable with the nonlocality scale linl, the main and additional waves in neighboring layers couple to each other (c), dramatically changing the overall dispersion of the modes, with multiple propagating modes existing at certain frequencies. This intrinsically nonlocal metamaterial regime requires some nonlocality in each component of the composite and cannot be described by effective medium parameters. Insets represent a 3-μm long slab of the composites. When the layer thickness d is comparable to the operating wavelength λ0, the structure operates as a one-dimensional photonic crystal, with the fields resulting from the interference of two counter-propagating waves within each layer of the composite (Fig. 3b) and the modes exhibiting a series of transparent propagation bands and opaque photonic band-gaps41 (Fig. 4a). Incorporating nonlocality into the numerical calculations does not measurably alter these responses while revealing an additional mode that exponentially decays into the composite. Therefore, as expected, the overall behavior of the periodic layered material is not affected by the nonlocality. As the layer thickness is reduced and the individual layers become sub-wavelength, the optical response of the composite transitions to the behavior of a homogeneous material whose dielectric permittivity is well-described by effective medium theory14,42. In this regime, the multilayer composite behaves as a uniaxial material, with an optical axis perpendicular to the layers42. The granularity of the composite can often be incorporated into this effective-medium description, yielding a structural nonlocality with nonlocal corrections ∝ d/λ0 (see SI)17,18,25. For substantially thin layers, the structural nonlocality vanishes, and the optical properties of the composite converge to the predictions of effective medium theory that are independent of d (and depend only on the ratio of layer thicknesses)14,42. The electromagnetic fields in this regime are almost constant throughout the unit cell (Fig. 3c)17. The dispersion of the layered materials in the effective medium regime exhibits a well-studied elliptic-to-hyperbolic transition42,43, Fig. 4b. Including the intrinsic nonlocality, once again, does not fundamentally alter these results, with the exception of introducing a rapidly decaying additional wave. As the layer thickness is reduced further, the predictions of local electromagnetism no longer change. However, the actual material response—that includes inherent nonlocality—keeps evolving (see SI for a detailed analysis of this evolution). Eventually, when the layer size becomes comparable to the scale of its components’ intrinsic nonlocality linl, the electromagnetic response of the composite undergoes a qualitative change. A composite operating in this regime is inherently nonlocal; its optical response is not adequately described by local permittivity, and the composite as a whole is dominated by the interference of nonlocality-induced additional waves and, as a result, does not behave as a quasi-homogeneous effective-medium material. This point is illustrated in Fig. 3d, where we show that despite the small layer thickness, the electromagnetic fields of primordial modes vary on the scale of the layer, opening the door to novel mechanisms for molding light at the nanoscale. Furthermore, the nonlocal theory now predicts the existence of the additional propagating mode and a significant deviation of the dispersion of the main wave from the (local) effective medium response, Fig. 4c. The presence of multiple interacting nonlocal components in the material is of vital importance for the existence of observable, propagating, additional waves. For example, as the nonlocality of one of the layers is decreased, the behavior of the bi-layer composite converges back to the predictions of local effective medium theory (evolution of photonic band structure of composites with different magnitudes of nonlocality is illustrated in SI). As mentioned above, the emergence of the nonlocal response, illustrated here with the example of a bi-layer periodic material, is a universal property of composite media whose components are structured on a length scale commensurate with the scale of their inherent nonlocal response (linl). As a result, intrinsic nonlocality can be realized in other coupled nonlocal environments (such as two- and three-dimensional periodic materials, nonlocal nano-cavities comprising nonlocal core and cladding layers, and interface regions). This inherently nonlocal electromagnetism, introduced here, is fundamentally different from previously reported phenomena, such as local hyperbolic response42, strong coupling-metamaterials44, and ballistic resonances45,46, all of which are adequately described by local material parameters. Indeed, the ballistic resonance originates from the confinement of free electron motion and manifests itself at scales smaller than the nonlocality scale, while strong coupling-induced phenomena44,47 are observed in the long-wavelength hyperbolic regime and for large angles of incidence. In all these previous studies, ultra-thin multilayer composites behaved as homogeneous slabs of materials with some effective permittivity. In contrast, intrinsically nonlocal composites feature fast oscillations of electromagnetic fields on the scale of their components and cannot be described by effective medium parameters. (A homogenized description of inherently nonlocal media may be possible in the regime d ≪ lint ≪ λ0. However, this regime is not achievable in the material platform considered in this work as it would require layer thicknesses that are smaller than the electron mean free path). Furthermore, another essential property of these metamaterials is that they are also fundamentally distinct from the quantum-well composites whose optical response is strongly affected by the energy level quantization related to charge confinement48–50. Experimental demonstration of intrinsically nonlocal metamaterials Experimental realization of intrinsically nonlocal metamaterials requires a composite with components that have (i) macroscopic (multi-atom, ~ nm) nonlocality scale, (ii) high-quality interfaces (with typical roughness smaller than the nonlocality scale), and (iii) uniformity of component size across the composite. These requirements can be satisfied using an epitaxially grown semiconductor platform. Semiconductor plasmonic designer metals whose nonlocality originates from their free-electron dynamics (see Methods) can be epitaxially grown, with excellent control of layer thickness, free-electron concentration, and interface quality, and thus present the material of choice for our study. Depending on the concentration of free electrons, such materials can exhibit transparent dielectric, reflective (plasmonic) metal, or even epsilon-near-zero response51,52. The nonlocal metamaterial platform realized in our work is schematically shown in Fig. 5a. The highly doped InAs (n++-InAs) layers are designed to exhibit a plasmonic transition at λp ≃ 6.3 μm, while their lightly doped “dielectric” counterparts (n-InAs) exhibit a similar transition at approximately 17 μm. These nonlocal plasmonic layers are separated by thin undoped AlAsSb barriers, used to implement the “hard” interfaces that exist in our theoretical description. The doping of the n-InAs dielectric layer was chosen to bring its permittivity in line with the permittivity of the AlAsSb (ϵ = 10) barriers at the operating wavelength λ0 ~ λp. Therefore, from the standpoint of electromagnetism, the lightly doped InAs and the AlAsSb barrier should exhibit almost identical behavior, differing only in their nonlocal response. Experimental Demonstration of Intrinsically Nonlocal Response. a Illustrates the metamaterial layer stack realized in this work. The remaining panels show the numerically simulated (b–d) and experimentally measured (e–g) TM-polarized transmission spectra for composites with varying barrier thickness but constant period. The black dashed lines in (b–d) represent 60∘ transmission simulated using purely local permittivities. For thick t = 30 nm barriers (b,e) the composite operates in a local effective medium regime, exhibiting the transmission dip that is associated with an elliptic-to-hyperbolic topological transition43. However, when the barrier thickness is reduced to t = 20 nm (c,f), the single resonance splits into two, with the composite entering an intrinsically nonlocal metamaterial regime that can only be described by nonlocal calculations. Reducing barrier thickness to t = 5 nm (d,g) further increases effective nonlocality of the barriers, changing the splitting of the resonance. Molecular beam epitaxy (MBE) was used to grow several series of metamaterials consisting of five periods of the structures described above. In the first set of samples, we varied the size of the AlAsSb barriers between 5 and 30 nm (keeping the total thickness of the AlAsSb/n-InAs/AlAsSb structure at 80 nm, equal to the thickness of the highly doped InAs layers) aiming to vary the strength of the nonlocal response within the composite. The angle- and wavelength-resolved transmission of this set of composites is shown in Fig. 5. The composite with the thickest AlAsSb barrier exhibits behavior that is almost identical to the previously reported optical response of all-semiconductor-based, strongly anisotropic (hyperbolic) metamaterials42, with a strong angle-dependent transmission feature at the wavelength where the permittivity of the highly doped layers (and the effective permittivity of the composite in the direction of its optical axis) is close to zero. However, as the thickness of the AlAsSb barrier is reduced, the optical behavior undergoes both quantitative and qualitative changes: the single angle-dependent transmission dip spectrally broadens and splits into two. Notably, such splitting in transmission (or reflection) resonances has been previously identified as a clear signature of a strongly nonlocal response in low-temperature homogeneous media3,4. The optical response of the composites was modeled with a nonlocal transfer matrix method, where the permittivity of the doped layers (both n-InAs and n++-InAs) was approximated as that of (nonlocal) degenerate plasmas53,54 and the AlAsSb layers were assumed to have frequency-independent permittivity with layer-thickness-dependent nonlocality α (see Methods and SI), with α being the single frequency-independent fitting parameter in our theoretical description. Notably, the results of our theoretical calculations closely match the experimental data. In the second set of experiments, we kept the thickness of the AlAsSb layers constant (thereby fixing the nonlocal response of the individual components of the metamaterial), but changed the thickness of the doped layers of the structure, varying the period of the metamaterial from 80 to 240 nm. Results of these studies are illustrated in Fig. 6d–f, along with results of theoretical modeling, Fig. 6a–c. The effective medium descriptions of the metamaterials with varying (deep subwavelength) period should be identical (and thus overall transmission should scale with overall composite thickness, dashed lines in a–c). Instead, a dramatic change of the optical response of the composite is observed as the metamaterial unit cell period is decreased by a factor of three, from 240 nm ( ~ λ0/30) to 80 nm ( ~ λ0/80). This drastic modulation of the overall electromagnetic response illustrates one of the central properties of the intrinsically nonlocal metamaterial response: even in the limit when individual layers are deeply subwavelength (here, layer thickness ~ λ0/100), the composite does not behave as a homogeneous effective material. Experimental control of Nonlocal Response. TM-polarized transmission spectra were simulated (a–c) and measured, (d–f), for HMMs of varying period and constant barrier thickness. The black dashed lines in (a–c) represent 60∘ transmission simulated using purely local permittivities. The red dashed lines in (d–f) indicate the spectral position of the nonlocal feature, controlled by adjusting the period of the nonlocal metamaterial. Figure 6 illustrates another unique property of intrinsically nonlocal metamaterials: by changing the scale of their components, it becomes possible to control nonlocality-induced optical signatures outside the previously explored epsilon-near-zero frequency range. This is clearly observed in the shortest period structure of Fig. 6c, f, where the signature of the nonlocal mode appears at wavelengths well below the epsilon-near-zero wavelength of 6.3 μm. Notably, optical features at λ0 ≲ λp cannot be explained by conventional plasmon polariton modes that only exist for λ0 > λp26. The behavior reported closely follows the universal properties of composites, introduced above. The as-grown composite comprises three components with different nonlocality scales in each component. Our analysis suggests that in n++-InAs linl ~ 50 nm, in n-InAs linl ~ 20 nm, while the nonlocality of AlAsSb layers varies between linl ~ 10 nm for 30-nm-thick barriers and linl ~ 30 nm for 5-nm-thick ones. Therefore, thick t = 30 nm barriers operate in linl ≪ d regime, reducing the response of the composite to the previously studied local effective medium regime (Fig. 5b, e). In contrast, when the thickness of the barrier approaches its nonlocality scale, the collective interference of the additional waves across the composite is clearly reflected in the transmission spectra (Fig. 5d, g). Increasing the period of the composite makes InAs components (and the composite overall) appear more local (Fig. 6a, d), while decreasing the period makes nonlocality-driven phenomena more prominent (Fig. 6c, f). Discussion Intrinsically nonlocal metamaterials represent a previously unexplored regime of optical composites that is dominated by the inherent nonlocal response of the components and results from the interference of nonlocality-induced additional waves. As such, intrinsically nonlocal metamaterials are expected to offer new opportunities in engineering light-matter interaction. These opportunities can be broadly grouped into (i) fundamentally new phenomena that are inherent to nonlocal media and (ii) exciting quantitatively novel regimes where nonlocal materials can provide alternatives to existing materials platforms. A broad class of new optical phenomena stems from the fact that nonlocal materials can react to gradients of electromagnetic fields and that they support additional propagating modes. The ability of material to react to field inhomogeneities–the core of nonlocal electromagnetism–is expected to fundamentally alter light-matter interactions. For example, light emission in local media is dominated by the dipolar response, resulting in suppression of dipole-forbidden transitions of molecules, quantum dots, and other quantum objects, and limiting the performance of classical non-dipolar antennae. In contrast, nonlocality provides direct coupling of non-dipolar emitters to the waves propagating in the material, fundamentally changing emission, absorption, and scattering processes. Separately, the existence of multiple propagating waves increases the density of information that can be carried by the structure, with implications for lifetime engineering, sensing, and communications. On the quantitative level, inherently nonlocal composites enable nanophotonic analogs of phenomena previously thought to be limited to wavelength-scale structures: photonic band engineering and molding of the distribution of electromagnetic fields. These phenomena are illustrated above (see Figs. 3 and 6). The effect of the nonlocal response on bulk optical properties of the composites is further illustrated in Fig. 7, which shows two additional potential applications of such metamaterials: compact platforms for accurate pulse shaping technologies and designer nonlinear response. Intrinsically nonlocal metamaterials for pulse shaping and nonlinear applications. While local materials provide some ability to modulate the shape of ultra-fast pulses (by adjusting group velocity, vg (a)), any significant modulation of group velocity in these materials generally coincides with strong absorption and is pinned at the plasma wavelength. In contrast, intrinsically nonlocal materials offer strong modulation of group velocity—including negative group velocity—in relatively low-loss regimes (b); parameter κ in (a, b) represents the imaginary part of the wavevector at the given wavelength λ0. In a related matter, any plasmonic material is known to provide a nonlinear change in transmission in response to a strong ultrafast pump pulse. In local materials (c), such modulation is limited to the epsilon-near-zero (plasma wavelength) regime; intrinsically nonlocal materials decouple strong nonlinearities from the epsilon-near-zero frequency (d). Control over group velocity is important for communication and optical computation applications. It has been widely accepted that strong group velocity modulation in macroscopic materials is only possible as a result of strong material dispersion—the regime that is inherently affected by either absorption (in passive media) or gain (in active materials)55–57. Our analysis of light propagation in local composites (Fig. 7a) fully confirms this expected behavior: the short-wavelength response of the material is essentially identical to a bulk semiconductor with refractive index n ~ 3, with a single resonance of group velocity at the plasma wavelength, and small group velocity in the type II hyperbolic regime—the latter two both associated with strong absorption. In contrast, the dispersion of intrinsically nonlocal metamaterials (Fig. 7b) features resonances of group velocity that are not associated with the resonances of the metamaterial components. In addition, the vg < 0 response exhibits relatively low loss. Furthermore, these materials allow for an unprecedentedly broad frequency band of negative group velocity ( ~ 1.2 THz according to our analysis). Figure 7c, d illustrates another benefit of intrinsically nonlocal composites: their ability to modulate nonlinear response (here caused by “hot” electrons in doped materials) outside of the well-studied epsilon-near-zero regime29,58,59. Here, we assume that the hot electrons in highly doped semiconductors result from the strong ultrafast pump pulse, resulting in the change of effective mass of these charges59, which is being detected as the change in overall transparency of the sample analyzed by the probe pulse. In local materials, the maximum of transmission modulation coincides with the plasma frequency of the highly doped component of the optical stack. In contrast, the response of the nonlocal metamaterial reflects the interference of the additional modes in its components. Future studies will explore the implications of the inherently nonlocal response on other optical phenomena. The intrinsically nonlocal metamaterial response, introduced in the present work, is a universal regime of optical composites that are dominated by the nonlocality of their components. Our work is therefore relevant to the nonlocality of (isolated) nanostructured plasmonic inclusions5,6,28, as well as the nonlocality of macroscopic and structured phononic media3,4,60,61. Notably, theoretical descriptions of these phenomena relied on material-specific simulation tools (e.g., hydrodynamic model for plasmonic nanostructures38 or ionic continuity models for phononic systems61). In contrast, our formalism is materials-agnostic and fully incorporates the inherent nonlocality of the components into macroscopic material response (see Methods). In doing so, our formalism avoids the fitting parameters and the topic of additional boundary conditions that are often associated with the nonlocal response of isolated inclusions, consistent with ideas put forth in ref. 62 for a single planar slab of nonlocal media. Since the functional form of nonlocality used in our work is different from that used by hydrodynamic models, more research is needed to analyze the interplay between these different theoretical descriptions. We note, however, that the approach presented in this manuscript relies on a single frequency-independent fitting parameter to fully describe multiple experimental results. The inherently nonlocal response is expected to manifest itself in a variety of plasmonic, phononic, and excitonic media as long as (i) the metamaterials comprise multiple nonlocal components in direct contact with each other and (ii) the composites are structured at the spatial scale commensurate with the scale of the inherent nonlocality of their components. Potential material platforms include noble metals and photonic media (CdS4) at visible frequencies, transparent oxides at visible and near IR, doped semiconductors and photonic materials at mid-IR, doped 2D materials at IR and THz, as well as excitonic media throughout visible and IR ranges. To conclude, we, for the first time, proposed and realized a macroscopic material that integrates multiple coupled nonlocal components and supports additional nonlocal waves. We have demonstrated, theoretically and experimentally, that structuring material at the deep subwavelength scale comparable with the scale of its components’ nonlocal response can dramatically change the electromagnetic response of the material as a whole. The electromagnetic properties of the resulting media are strongly influenced by the intrinsically nonlocal interactions of the components, regardless of whether the components by themselves or the material as a whole exhibit epsilon near zero response—a prerequisite to all previous realizations of nonlocal materials4,12,27. Importantly, intrinsically nonlocal metamaterials exhibit strong nonlocality even at room temperature and with realistic (lossy) materials, while their optical properties cannot be adequately described by homogenized bulk material parameters even when the individual components are much smaller than the operating wavelength. Intrinsically nonlocal metamaterials, illustrated here with a mid-infrared semiconductor designer metal material system, provide a novel platform for controlling electromagnetic fields at the scale of individual deeply subwavelength components, affecting all aspects of light-matter interactions—from bulk properties (demonstrated in this work) to light emission, detection, pulse shaping, and nonlinear light interactions. Methods Nonlocal response of composites in real- and wavenumber- spaces We assume that electromagnetic radiation propagates in the xz plane of the Cartesian coordinate system, that material properties change only along the z direction, and that materials exhibit nonlocality primarily along the same direction. Under these circumstances, the electric displacement field in the material is related to the electric field via with ϵ and α describing local and nonlocal contributions to material permittivity, respectively, with parameter α related to the intrinsic nonlocality scale linl via linl=λ0α. Any monochromatic field propagating in the medium can then be represented as a linear combination of transverse electric (TE, E∥y^) and transverse magnetic (TM, B∥y^) modes, with the latter affected by the nonlocal response and being the main focus of this study. Propagation of TM-polarized light in non-magnetic materials (B = H) in this geometry can be derived63 by substituting B=y^By(z)exp(−iωt+ikxx) into Maxwell’s equations written in Cartesian coordinates, expressing the components of electric field in terms of By, with Ex=−icωϵ∂By∂z, and Ez=−1kxi∂Ex∂z+ωcBy, and simplifying Eq. (1) along with the remaining component of Maxwell equations, Dz=−kxcωBy, to : As can be explicitly verified, when Eq. (2) is applied to homogeneous materials, its solutions yield63 a set of plane waves with E,B∝exp(−iωt+ik⋅r) whose dispersion is given by and whose constitutive relationship reduces to Dβ=∑γϵδβγ+δβγαkz2c2ω2δγzEγ with subscripts corresponding to Cartesian coordinates and δβγ being the Kroenecker delta symbol. When Eq.(2) is applied to an interface between two homogeneous materials, the requirement for arguments of spatial derivatives to be differentiable yields a set of boundary conditions63 requiring continuity of Ex, By, Ez, and α∂Ez∂z across the interface. Note that the two former conditions represent the conventional boundary conditions of local electromagnetism, while the two latter relationships represent the additional boundary conditions that are required to fully solve for coupling between primary and additional waves at the interface between two nonlocal media. Continuity of Ez then becomes redundant at the local/nonlocal interface. Nonlocal material parameters of epitaxially grown layers The permittivity of the doped semiconductors is approximated by the permittivity of a degenerate collisionless plasma53, corrected for the contribution from the background permittivity of the semiconductor. Explicitly, For the doped InAs materials, we use background permittivity ϵ∞ = 12.3, ωp corresponding to plasma wavelengths of 6.3 μm and 17.2 μm, respectively, and calculate Fermi velocity based on Fermi energy and electron effective mass vF=2EF/m*(EF). The latter parameters are estimated using fully quantum mechanical calculations, previously described in Refs. 45,46 that derive the optical response of highly doped semiconductor layers based on discrete quantum well transitions and that incorporate mass non-parabolicity. Importantly, these calculations demonstrate that the layer thicknesses in our materials are large enough for the Drude approach to adequately describe the response of the highly doped layers. Inelastic damping parameters are set at γ = 8 THz, γnl = 28.3 THz. Figure 8 illustrates the spectral response of the highly doped layers where we introduced a dimensionless nonlocality parameter αpl=−35ωpl2vF2ω(ω+iγnl)c2 (see Supplementary Notes). Relative permittivity and nonlocality of highly-doped plasmonic layers. a, b The dependence of permittivity (ϵpl) and nonlocality (αpl) on the operating wavelength λ0, respectively; line colors correspond to real (blue) and imaginary (orange) parts of the material parameters. To obtain the value of αeff we fixed ϵb = 10 and minimized the deviation between the experimentally measured transmission and the predictions of the nonlocal transfer matrix formalism described below (sample code is provided). Our analysis shows that while the exact value of the effective nonlocality depends on the spectral region, the wavelength/frequency spacing, and the sample used for fitting (multiple nominally identical samples have been analyzed), the variation in the effective nonlocality due to the changes above does not significantly affect the resulting optical spectra. After multiple fitting runs that cover λ0 ∈ [4…8] μm and different (nominally identical) samples, we choose αeff = (2.75 + 1.5i) × 10−5 for 5-nm barriers; nonlocality of 20- and 30-nm-thick barriers is reduced by 10% and 75%, respectively. Nonlocal transfer matrix formalism The transfer matrix formalism is a powerful linear algebra technique for calculating the propagation of light through an arbitrary array of planar layers41. In our implementation, the (column vector) of the fields at a given point inside a layer l is given by the product of the layer-specific structure matrix Nl^, a diagonal phase matrix Fl^(z), and a vector of mode-specific amplitudes al. Explicitly, with ϵl, αl representing material parameters of the layer l, kzl,m describing the propagation constant of the mode m = 1, 2 in the layer l [see Eq.(3)], and k~zl,m=ckzl,m/ω. The rows of the field matrix N^l represent distributions of Ex, By, α∂Ez∂z, and Ez fields, respectively, while matrix F^l describes the phase of individual modes. The boundary conditions at the interface between the layers l and l + 1, located at the point z0 can then be written as that allows one to relate the amplitudes of the modes in layer l + 1 to the amplitudes of the modes in layer l via the transfer matrix T^l: al+1=T^lal, with T^l=(N^l+1F^l+1(z0))−1Nl^Fl^(z0). Similar to the (local) transfer matrix method (TMM), originally introduced in ref. 41, the nonlocal TMM can be used to calculate transmission and reflection of light by layered optical stacks with arbitrary nonlocal layer configuration. Alternatively, Eq.(6) can be used as a basis to formulate the Scattering-Matrix Method implementation63,64 that, in turn, can be utilized to calculate the optical properties of multilayered structures that combine local and nonlocal components. Sample code that implements the boundary conditions above is provided in ref. 65 and is included with the manuscript The nonlocal TMM can also be used to calculate dispersion and field profiles of the modes propagating in periodically stratified materials. In the case of bi-layer periodic materials with thicknesses d1 and d2, the Bloch-periodic condition reduces to: with q being the wavenumber of the mode, a1 describing the magnitudes of the fields in the first layer, and I^ being the identity matrix. Dispersions of the modes propagating in periodically stratified nonlocal bilayer media are shown in Fig. 4 and are further illustrated in Supplementary Notes. Group velocity calculations Once the dispersion of the modes in periodically stratified materials is known (see above), group velocity can be calculated as with k=kxx^+qz^ being the (quasi-)wavevector of the mode. In our studies, we focus on the z-component of the group velocity, vgz=∂q/∂ω−1, with q(ω) relationship calculated using the transfer-matrix formalism described above. The real part of group velocity is shown in Fig. 7. Analysis of optical nonlinearities Here we focus on ultrafast optical nonlinearities that are commonly seen in pump-probe studies where the intense “pump” pulse of light is primarily absorbed by the electronic subsystem of the materials, raising the electronic temperature and modulating the optical response for a short period of time until the energy is re-distributed into atomic/phononic states of the material (see, e.g., ref. 58 and references therein). The optical response of doped semiconductors under these conditions can be adequately modeled by increasing the effective mass of the electrons59, the approach used in our work. To analyze the expected nonlinear behavior, we used the formalism described above to calculate the expected change in transmission through the sample resulting from variation of effective electron mass by 0.1%. The results of these calculations are shown in Fig. 7. Metamaterial growth All metamaterial samples presented in this work were grown via molecular beam epitaxy using a Varian Gen-II system on 1/4 2" unintentionally doped InAs substrates. All samples were subjected to a load chamber bake at 500 ∘C for 2.5 h after initial loading, then transferred to a buffer chamber and baked individually at 300 ∘C for 1 h. Before growth, each sample was transferred to the substrate manipulator of the growth chamber and baked at 300 ∘C for 15 min. After this initial bake, the substrate manipulator was rotated to the growth position, and a final bake at 510 ∘C was performed under Arsenic flux to remove the surface oxide from the substrate. Before the growth of the epitaxial stacks shown in this work, a 200 nm InAs buffer layer was grown at 475 ∘C to reduce defect densities in the subsequent layers66. The remainder of the growth was maintained at 475 ∘C. The lightly n-type doped InAs layers (n-InAs) were doped using a GaTe source with an intended doping concentration of ND = 2.09 × 1018 cm-3, corresponding to a plasma wavelength λp = 17.2 μm. The degenerately-doped InAs layers (n++-InAs) were doped with a Silicon source. The approximate doping concentrations extracted from the transmission spectra were ND = 3.00 × 1019 cm-3, corresponding to a plasma wavelength λp = 6.3 μm. Following the growth of the metamaterial stack, the samples were capped with 10 nm of InAs to prevent oxidation of the aluminum-containing layers. Metamaterial characterization All samples grown for this work were characterized using polarization and angle-dependent Fourier Transform Infrared (FTIR) transmission spectroscopy. The experimental system used is shown in Fig. 9. Broadband MIR light from an internal globar source was passed through the internal interferometer of the FTIR. After exiting the FTIR, this light was passed through a series of collimating apertures as well as a ZnSe wire-grid polarizer. Samples were mounted vertically on a rotating stage, aligned to rotate about the point where light is incident on the sample, allowing variation of the incidence angle. Light transmitted through the samples was collected using a 2” focal length parabolic gold mirror and focused onto a Mercury Cadmium Telluride detector (MCT). All transmission measurements shown in this work are normalized to transmission through air using the same experimental system, with the sample removed. Experimental transmission setup used in sample characterization. MCT and FTIR stand for Mercury Cadmium Telluride (detector) and Fourier Transform Infrared interferometer (here, used as a source of infrared radiation). Supplementary information Supplementary Information Transparent Peer Review file Acknowledgements This research was supported by NSF DMREF program (award #2118787 (VP), award #2119157 (EN), award #2119302 (DW&SB)). Author contributions A.W. and R.C.W. were responsible for the growth and characterization of samples. J.L.M. led the implementation of the nonlocal transfer matrix and other theoretical simulations. E.N. was responsible for analytical calculations of nonlocality parameters. D.W. and S.B. supervised the experimental part of the project, while V.P. and E.N. supervised theory and numerical efforts. All authors discussed the data and contributed to the manuscript. Peer review Peer review information Nature Communications thanks Adria Canos Valero and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available. Data availability All data are available in the manuscript or the supplementary materials. All data will be provided upon request to the authors. Code availability Source codes are publicly available through a Code Ocean capsule (10.24433/CO.4510424.v1) and on GitHub (https://github.com/viktor-podolskiy/primordial_metamaterials)65. 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. These authors contributed equally: A. Ware, J. LaMountain. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74396-z. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review file Data Availability Statement All data are available in the manuscript or the supplementary materials. All data will be provided upon request to the authors. Source codes are publicly available through a Code Ocean capsule (10.24433/CO.4510424.v1) and on GitHub (https://github.com/viktor-podolskiy/primordial_metamaterials)65.