Abstract
Microscopic structuring can change the effective properties of a material by several orders of magnitude. An example of this is animal bone, which has an effective elastic modulus that is more than 1,000 times larger than that of the constituent proteins. Here, we propose a broadbandenhancement principle of photonic nonlinearity that has a similar mathematical origin as the bone example. The proposed staggered array metamaterials violate the standard Miller’s rule in nonlinear optics and can enhance the thirdorder nonlinearity by more than a thousand to a billion times, depending on target operation frequencies. This metamaterial principle also enables manipulation of the individual components of the linear and nonlinear susceptibility tensors. Our biomimetic approach overcomes the fundamental speedefficiency tradeoff in current resonant enhancement schemes, making faster and more efficient alloptical devices possible for 1.55 μm wavelength. The principle is also applicable to ionic diffusion, heat conduction, or other transport problems.
Similar content being viewed by others
Introduction
Nonlinear photonic processes, such as harmonic generation, phase modulation, and multiphoton absorption, have diverse applications including laser sources and optical communications^{1,2,3}, threedimensional lithography^{4}, bioimaging^{5}, and many others^{6,7,8,9,10}. However, light–matter interaction is predominantly linear for common transparent photonic materials, thus necessitating the injection of highintensity artificial light sources along the optimal crystallographic directions^{11}. Therefore, numerous studies have been conducted with the aim to attain photonic nonlinearity enhancement. Most of these studies have focused on the introduction of tailored electromagnetic resonances near the operation wavelength to boost light–matter interactions^{12,13,14,15,16,17,18,19}. Both quantumelectrodynamic approaches involving the engineering of intersubband transition energy levels^{12,13}, and classical electrodynamic enhancement methods—mostly utilizing photonic cavities or resonant structures^{13,14,15,16,17,18,19}—have been proposed. However, these resonantenhancement schemes exhibit an inherent tradeoff relationship between the degree of enhancement and the frequency bandwidth, thereby hindering their practical applications. Even for applications that do not require large bandwidths, large resonanceinduced energy losses and a narrow fabrication error margin represent substantial obstacles. Moreover, only a few studies have investigated the controllability of the tensor form of nonlinear susceptibilities^{10,13,18}, despite the fact that it may have important practical consequences.
Here, we achieve both extreme enhancement and tensor form controllability of the linear and nonlinear susceptibilities of a nonlinear staggered array metamaterial (SAMM) over a broad bandwidth. The SAMM represents a periodic composite system of nonlinear materials and conductors with a deep subwavelength period. We analytically and numerically verify that the thirdorder nonlinear susceptibility can be enhanced by more than 10^{9} times at terahertz frequencies and by 10^{4} times at optical telecommunication frequencies. In addition, we show that 18 different crystal classes with secondorder nonlinearity can be realized based on simple variations of the spatial configuration of a given nonlinear material and a conductor. The proposed broadbandenhancement principle is mathematically analogous to that of naturally occurring mechanical SAMMs, such as animal bones and bonelike materials (e.g., nacre and sea shell), which have enhanced elastic modulus by more than three orders of magnitude compared to their constituent soft proteins^{20,21}.
Results
Bespoke nonlinear susceptibility tensor of the proposed SAMM
The proposed SAMM structure is composed of alternating metal and nonlinear dielectric layers (Fig. 1a). The metal layers are composed of insulated metal patches in a planar lattice, and the nearest metal layers are staggered with halfunitcell shifts in both the x and ydirections (denoted as “A” and “B” layers in Fig. 1a). The metal and nonlinear dielectric layers (‘I’) are stacked in the sequence A–I–B–I–. The lateral unitcell dimensions in the x and ydirections, the lateral gap sizes in the same directions, and the thicknesses of the metal and dielectric layers are denoted as a_{x} and a_{y}, g_{x} and g_{y}, and h_{m} and h_{d}, respectively. Figure 1b depicts one unit cell of the SAMM.
The proposed microscopic (subwavelength) structure closely resembles the mechanical microscopic structure of mineralized collagen fibrils in animal bone and other hard biological materials (Fig. 1c)^{20,21}. In a simplified model, the fibrils in the bones are an array of isolated hard mineral plates embedded in a soft protein matrix, and the mineral plates are in a staggered formation as in Fig. 1a. Figure 1d schematically shows the microscopic deformation of the mechanical SAMM under macroscopic tensile strain. The mathematical parallelism between the enhancement of mechanical and electromagnetic properties is presented in the discussion section.
From the point of view of materials, SAMMs can be considered homogeneous, crystalline photonic materials because their periods are much smaller than the wavelength—as those of natural crystals—and their effective material properties, such as the linear and nonlinear susceptibilities and refractive indices, are well defined and remain constant regardless of the thickness of the sample^{22,23,24}. For these metamaterials, their linear (n = 1) and nonlinear (n > 1) homogenized electric susceptibilities \({\tilde{\mathbf{\chi }}}_{{\mathrm{eff}}}^{(n)}\) can be calculated based on the reciprocity theorem^{13,19,25} using the electric susceptibility tensors, \({\tilde{\mathbf{\chi }}}_{\mathrm{d}}^{(n)}({\mathbf{r}})\), of the constituent materials themselves and pointwise fieldenhancement factors, \(F_{\alpha i}^{(\omega )}({\mathbf{r}}) = {\mathbf{\alpha }} \cdot {\mathbf{E}}_{{\mathrm{loc}}}^{(\omega )}({\mathbf{r}})/{\mathbf{E}}_{{\mathrm{ave}}}^{(\omega )}\) (with \({\mathbf{E}}_{{\mathrm{ave}}}^{(\omega )}{\mathbf{i}}\)), which quantifies how strong the αdirectional components of the local electric fields (\({\mathbf{E}}_{{\mathrm{loc}}}^{(\omega )}({\mathbf{r}})\)) are when the homogenized, macroscopic electric field (\({\mathbf{E}}_{{\mathrm{ave}}}^{(\omega )}\)) is along the idirection. α and i are the unit vectors in the α and idirections, respectively (see Methods and Supplementary Note 1 for detailed derivation).
Using this approach, one can analytically derive the homogenized nonlinear susceptibility tensor of the proposed structure for secondharmonic generation (SHG). In this section, we assume for the sake of simplicity, the use of a metal with perfect electric conductivity and an insulator with frequencyindependent nonlinear coefficients, thus resulting in the Kleinman symmetry condition^{11,26} formulated according to \(F_{\beta \gamma }^{(\omega )} = F_{\beta \gamma }^{(n\omega )} = F_{\beta \gamma }\). These restrictions are eliminated in the rigorous analysis of the actual enhancements and numerical demonstrations presented in the following sections. For the proposed structure, the local electric field is mainly oriented in the zdirection inside the dielectric layers owing to the parallelplate capacitorlike geometry of the metal plates above and below the dielectric. Subsequently, the relevant fieldenhancement factors are only F_{zx}, F_{zy}, and F_{zz}, and other factors are being negligibly small. The zdirectional local electric fieldenhancement factors under an xdirectional macroscopic electric field, F_{zx}, of a specific structure (a_{x} = 4800 nm, a_{y} = 800 nm, g_{x,y} = 0.2a_{x,y}, and h_{d} = h_{m} = 20 nm) is plotted in Fig. 1e–f (see Methods). The actual profiles of F_{zx} and F_{zy} are very simple with a large and almost spatially uniform value (around 120 in the aforementioned case) throughout the space between the top and bottom metal plates, and with a negligible value elsewhere. These maximum fieldenhancement values are approximately F_{zx}  = 0.5a_{x}/h_{d} and F_{zy}  = 0.5a_{y}/h_{d}^{27}. For the nonlinear dielectric, we assume that the microscopic pole directions are either the +z or −zdirection or unpoled, thus resulting in the intrinsic secondorder nonlinear susceptibility of the form \(\chi _{33}^{(2)}({\mathbf{r}}) = \chi _{33}^{(2)}p({\mathbf{r}})\) with p(r) = +1, −1, or 0, respectively. As a representative example, we consider a specific spatial pole configuration of p(r) = sign[F_{zy}(r)]. Figure 1g–h show zdirectional local secondorder nonlinear enhancement factor under an xdirectional macroscopic electric field, \(\chi _{33}^{(2)}F_{zx}^2/\chi _{33}^{(2)}\), of this example. The nonlinear enhancement factor is very large (around 14,000) and spatially nearly uniform throughout the aforementioned highfield region between plates. The contracted notation^{11,26,28} of the effective nonlinear susceptibility simplifies to
if g/a_{x} and g/a_{y} are small (see Supplementary Note 2 for detailed derivation).
The tensor form in Eq. (1) corresponds to the monoclinic system with a single twofold rotational symmetry. In a similar manner, one can realize various other nonlinear crystal systems by appropriately designing the spatial symmetry of p(r). If we also include the crystal designs in which the sides of the metal plates are intentionally arranged in a nonparallel orientation to any of the lattice vectors to break mirror symmetries, all of the noncentrosymmetric triclinic, monoclinic, orthorhombic, and tetragonal crystal classes can be realized (Fig. 2). If we consider the use of hexagonal plates in a hexagonal lattice, all of the noncentrosymmetric trigonal and hexagonal crystal classes are also covered (see Supplementary Fig. 1). Note that by tuning both the microscopic structure and the pole configuration in this simple manner, all noncentrosymmetric crystal classes—with the exception of the cubic classes—can be realized (a total of 18 point groups) and that only \(\chi _{zzz}^{(2)}\) (\(\chi _{33}^{(2)}\) in the contracted notation) of the constituent dielectric is utilized in all cases.
Extreme enhancement of nonlinear susceptibilities at various frequency ranges
The local electric fieldenhancement factors of the proposed SAMM show three noticeable characteristics. First, the magnitude of the enhancement factors can become very large for highaspectratio structures (Fig. 1e–f, for example). The resulting nonlinear enhancements are even larger because they are products of multiple fieldenhancement factors, depending on the order of the nonlinearity (Fig. 1g–h, compared to Fig. 1e–f). Additionally, these enhancement factors are nearly frequencyindependent in the longwavelength regime. A simple model, assuming perfect electric conductors and dispersionless dielectrics, exhibits no frequency dispersion at all (F_{zx} = 0.5a_{x}/h_{d} and F_{zy} = 0.5a_{y}/h_{d}), and a rigorous model with realistic materials also exhibits a small frequency dependence, except for the cases with extremely high aspect ratios. Third, the enhanced local field appears over a significant portion (50% or even more) of the total volume of the effective medium, as opposed to small “hot spots”, as in previous tipenhanced nonlinear structures. In structures with a hot spot near a sharp tip, or a narrow gap between metallic inclusions, the volume of the hot spots becomes smaller if one increases the degree of field enhancement by reducing the tip’s radius of curvature or the gap size. Subsequently, after homogenization with volume integration, the effective nonlinear susceptibility of these structures is not as drastically enhanced as their fields. These are the key factors that differentiate this nonresonant, polarization accumulation scheme of extreme nonlinearity enhancement via spacefilling of local electric dipoles from many previous approaches. The conceptual explanation regarding to the polarization accumulation and the large susceptibility enhancement are in Supplementary Note 3 and Supplementary Fig. 2.
The effective thirdorder nonlinear electric susceptibility tensor, \({\tilde{\mathbf{\chi }}}_{{\mathrm{eff}}}^{(3)}\), has an overall enhancement factor which is proportional to the fourth power of F_{zx} or F_{zy}. The longwavelength value of \({\tilde{\mathbf{\chi }}}_{{\mathrm{eff}}}^{(3)}\) for a SAMM with a_{x} = a_{y} = a becomes
based on the simple model, and it satisfies the tetragonal symmetry (4mm) (see Supplementary Note 2 for detailed derivation). In the following section, we focus on \(\chi _{{\mathrm{eff}},11}^{(3)}\), which leads to an extremely large thirdorder nonlinear polarization along the xdirection under an incident xdirectional electric field.
For more accurate calculations in the rest of the paper, we now adopt a rigorous analytical model, considering the actual complex permittivity of both the metal and the dielectric. This model is used to calculate the effective refractive index and the nonlinear susceptibilities (see Supplementary Note 4 for detailed derivation). Fig. 3 shows the analytically calculated and numerically simulated \(\chi _{{\mathrm{eff}},11}^{(3)}\), over several decades of frequencies (see Methods). In Fig. 3a, the calculated enhancement factor \(\chi _{{\mathrm{eff}},11}^{(3)}/\chi _{33}^{(3)}\) from the rigorous model is depicted as a function of both the aspect ratio of the SAMM, a/h_{d}, and the frequency, with the structural parameters of h_{d} = h_{m} = 10 nm and g = 0.2a. The lateral unitcell dimension, a, is changed to vary the aspect ratio. An almost uniform, extreme enhancement is achieved from nearzero frequency up to the lowest resonance frequency. For low frequencies, the simple estimation of \(\chi _{{\mathrm{eff}},11}^{(3)}\) in Eq. (2) matches the rigorous model well. As the frequency increases, materialinduced resonances occur when \(a^2/(4h_{\mathrm{m}}h_{\mathrm{d}})\sim \varepsilon _{\mathrm{m}}^{(3\omega )}/\varepsilon _{\mathrm{d}}^{(3\omega )}\) (gray dashed line) and when \(a^2/(4h_{\mathrm{m}}h_{\mathrm{d}})\sim \varepsilon _{\mathrm{m}}^{(1\omega )}/\varepsilon _{\mathrm{d}}^{(1\omega )}\) (black dashed line) (see Supplementary Note 4 for details about the resonance conditions). These resonances can be classified as electrostatic resonances, with a resonance condition determined by quasistatic boundary conditions^{29,30,31,32,33}. Because the magnitude of the metal permittivity is reduced as the frequency increases, the resonance frequency can be increased by lowering the aspect ratio. Therefore, one should design the structural parameters by considering the target frequency range, the amount of enhancement desired, and the acceptable amount of dispersion. To reduce the computational load, we consider structures that are uniform in the ydirection (g_{y} = 0) in the following calculations. \(\chi _{{\mathrm{eff}},11}^{(3)}\) for nonzero g_{y} is also of the same order of magnitude as long as g_{y}/a_{y} ≪ 1.
Figure 3b shows the enhancement factor (\(\chi _{{\mathrm{eff}},11}^{(3)}/\chi _{33}^{(3)}\)) for two different aspect ratios calculated by a nonlinear transfermatrixbased parameter extraction method using nonlinear finitedifference timedomain (FDTD) simulation results. The values are compared to the analytical model predictions. For a moderate aspect ratio (a = 900 nm, g = a/5, h_{d} = h_{m} = 10 nm, i.e. a/h_{d} = 90), we observed a millionfold enhancement of the effective nonlinear susceptibility compared to that of the constituent nonlinear dielectric (Fig. 3b, blue line), which is already two orders of magnitude higher than that of a previously proposed theoretical resonant structure^{17}. Moreover, the enhancement is almost uniform from a nearzero frequency up to 2 THz, which is much lower than the electrostatic resonance frequency (5 THz). A SAMM with a larger aspect ratio (a = 6000 nm, g = a/5, h_{d} = h_{m} = 10 nm, i.e. a/h_{d} = 600) shows more than a billionfold enhancement (Fig. 3b, green line). Although there is an electrostatic resonance at a subterahertz frequency for the structure with this high aspect ratio, the enhancement remains extremely large from the zero to the terahertz frequency range. The enhancement factors from the FDTD simulations and the rigorous analytical model are in excellent agreement. The retrieved \(\chi _{{\mathrm{eff}},11}^{(3)}\) value remained constant regardless of the thickness and the input intensity (see Supplementary Fig. 3). Therefore, the extracted \(\chi _{{\mathrm{eff}},11}^{(3)}\) is well defined as an effective bulk material property of the SAMMs.
At higher frequencies, including the infrared and visible, the highest aspect ratio that one can use for broadband enhancement becomes lower owing to the smaller magnitude of the metal permittivities. Figure 3c shows the enhanced thirdorder nonlinearity of the proposed structure with a = 75 nm, g = 15 nm, and h_{d} = h_{m} = 5 nm with realistic materials at 200 THz, which is close to an important opticalfiber communication band. The electrostatic resonances occur at 67 THz and 200 THz. The \(\chi _{{\mathrm{eff}},11}^{(3)}\) enhancement is larger than 4×10^{4} at 200 THz. In addition, the enhancement in the longwavelength limit is larger than 10^{3}, which is still a very high enhancement.
Thin and efficient nonlinear reflectors utilizing SAMMs incorporating nonlinear quantum wells
The proposed microscopic structure of the staggered conductors functions as a universal magnifier and symmetrymodifier of the intrinsic nonlinear susceptibility tensor of the constituent nonlinear material of choice, regardless of the latter. Accordingly, the quantitative enhancement factors shown in Eqs. (1) and (2) as well as the symmetrychanging capabilities demonstrated in Fig. 2 remain the same when we change the nonlinear materials. Thus, one can utilize quantum wells (whose subband energy levels are carefully designed to possess resonantenhanced nonlinear susceptibilities that are several orders of magnitude larger than those of typical nonlinear materials in their offresonance regime) as the nonlinear material between the conductors and the microscopic structure will further enhance these large nonlinearities, by another several orders of magnitude. This allows the use of metasurface types of nonlinear reflectors for massively parallel, pixelbypixel nonlinear optical operations (Fig. 4), rather than using waveguides or crystaltype devices with long propagation paths.
The nonlinear reflectors in Fig. 4a comprise a thin SAMM slab with GaN/AlN quantum well^{34}, dielectric impedancematching layers above and below^{35,36}, and a metallic mirror (see Methods). The electromagnetic properties of the GaN/AlN quantum well is summarized in Supplementary Fig. 4.
Figure 3b shows the switching performance of the optimized nonlinear reflector (a = 57 nm, g = 20 nm, h_{d} = 5.1 nm, h_{m}/2 = 10.2 nm, h_{u} = 357 nm, and h_{l} = 57 nm) for pulses which are 30 fs with a center wavelength of 1.55 μm. The thickness of the SAMM and the total thickness including dielectric layers are only of the order of 1/60 and 1/3 of the vacuum wavelength, respectively. While resonantcavitybased alloptical switching approaches sacrifice speed and bandwidth in exchange for lower switching power[2], our nonresonantenhancement scheme does not suffer from this limitation, which is apparent from the 26 fs output pulses (red curve in Fig. 4b inset). The recovery speed is ultimately limited by the relaxation time of the intersubband transitions in quantum wells, which can be of the order of 100 fs or faster^{34,37}. The system exhibits a modulation in excess of 10 dB with a pulse energy utilization of only 380 fJ, while the reflectance is more than 50% at the onstate mode of operation. This 10 dB modulation pulse energy is a recordlow value for alloptical switches with similar modulation speed[2], and is 250 times smaller than previous quantumwellbased alloptical switches^{37}. We visually compare switching performance of the result in Fig. 4 with several stateoftheart alloptical switches^{37,38,39,40,41,42,43} in Supplementary Fig. 5 using a Pareto optimality plot based on the switching time and the required switching energy. The maximum local electric field magnitudes are of the order of 5 × 10^{8} V m^{−1} for the 30 fs, 380 fJ pulses, which are below the damage threshold for many solid dielectric materials^{44}. Although the effective susceptibilities are now dispersive owing to the resonant nature of the quantum wells, the bandwidth is still large enough to cover a significant part of the optical communication bands or to allow femtosecondpulse operations. In addition, we verify that the designed device maintains more than 5 dB modulation depth with same pulse energy (380 fJ) under up to 30% (±10 nm) lateral missalignment (zero missalignment being the perfectly staggered case) between two metal layers of the SAMM slab, relative to the lateral unit cell size (Fig. 4c).
Discussion
Nonlinear susceptibilities of many natural nonlinear crystals can be estimated using Miller’s rule, which states that the nthorder nonlinear susceptibilities (e.g., \(\chi ^{(2)}(2\omega ;\omega ,\omega )\)) can be approximated to an (n + 1)thorder linear susceptibilities (e.g., \(\chi ^{(1)}(2\omega )\chi ^{(1)}(\omega )\chi ^{(1)}(\omega )\)) with a constant factor, δ, known as the Miller’s delta^{11,19,45}. However, according to O’Brien et al.^{19}, the nonlinear response predicted by Miller’s rule can be several times larger or smaller than the measured values for composite media such as metamaterials. In case of the proposed SAMMs, the actual nonlinear susceptibilities (Eqs. (1) and (2)) can differ from those predicted by Miller’s rule, by several orders of magnitude. Furthermore, the factor of discrepancy can be expressed as [M_{d}]^{−n}[M_{s}]^{−1}, where M_{d} and M_{s} are the enhancement factors of the effective oscillator density and strength, respectively, in a homogenized oscillator model of SAMMs (see Supplementary Note 5 for detailed derivation). Given that M_{d} and M_{s} are a/(4h_{d}) and a/(2h_{d}), respectively, for the proposed SAMMs^{27} and can be larger than 1000, the predictions using the original Miller’s rule without these correction factors may be 10^{9} and 10^{12} times different from the true values for \(\chi _{{\mathrm{eff}}}^{(2)}\) and \(\chi _{{\mathrm{eff}}}^{(3)}\), respectively. In Fig. 5, we compare the 3rd order nonlinear susceptibility enhancement factors (\(\chi _{{\mathrm{eff}}}^{(3)}/\chi _{\mathrm{d}}^{(3)}\)) of the SAMMs in this paper (blue crosses) with the estimated values based on the conventional Miller’s rule (black dots) and the modified Miller’s rule developed in this study (red dots) (see Methods). It is confirmed that the modified Miller’s rule with \(\delta^{\prime} = \delta [M_{\mathrm{d}}]^{  3}[M_{\mathrm{s}}]^{  1}\) estimates thirdorder nonlinear susceptibilities of SAMMs well. While large effective oscillator strength enhancement factor M_{s} can be found in many previous studies^{16,17,31,33,46,47}, large effective oscillator density enhancement factor M_{d} is a unique characteristic to SAMMs and induces a dramatic deviation from the conventional Miller’s rule. In fact, our SAMMs can be considered as an example of a new category of electromagnetic media with designable oscillator densities (see Supplementary Note 5 for details).
The enhancement of the electric susceptibilities demonstrated herein can be viewed as a special case of a general enhancement principle of effective material properties that occur in twophase composite materials. For example, the enhancement of the elastic modulus of the mineralized collagen fibrils in bone (Fig. 1c, d) can be conceptually explained as follows: First, the macroscopic tensile strain induces a shear strain enhancement by a factor of 0.5a/h_{d} inside the protein layer, which results in an enhanced shear stress by a factor of 0.5a/h_{d}. This enhanced shear stress accumulates over the distance of 0.5a and appears as tensional stress in the mineral plates with a thickness of h_{m}, thereby inducing another enhancement factor equal to 0.5a/h_{m}. Rigorous derivations of the effective mechanical properties can be found in previous reports^{20,21}, and the expression of the effective elastic modulus is similar to Eq. (3) in Methods with n = 1 (linear case) and becomes M_{eff} ≈ G_{p}(a/h_{d})^{2}/8, where G_{p} is the shear modulus of the protein matrix. This shows that the system can have an effective elastic modulus that is several orders of magnitude larger than that of the host material, for cases involving high aspect ratios (a/h_{d} ≫ 1). This mechanical enhancement principle has been previously adopted for biomimetic hard and tough materials made of ceramics and polymers^{48}. We note that this mechanism for enhanced elastic modulus is analogous to the explanation in Supplementary Note 3.
In general, we argue that the same principle can be encountered in various homogenization problems (including the examples already discussed) with a constituent equation of the form F = σE, where F is the relevant flux density (e.g., heat or mass), E is the conservative driving force (e.g., the temperature gradient or mass density gradient), and σ is the conductivity. E is an irrotational vector field owing to the conservative nature, and F becomes divergencefree in a steady state. Subject to these conditions, the effective conductivity for a twophase composite material with isolated highconductivity phases embedded in a continuous lowconductivity phase host material in a staggered formation, as in our SAMM, is σ_{eff} = σ(a/h_{d})^{2}/8 (see Supplementary Note 6 and Supplementary Fig. 6 for detailed derivation). Thus, the effective properties of the composite materials can be many orders of magnitude higher than those of the continuous host, which may occupy more than half of the total volume, while the enhancement factor is purely geometrically controlled. Moreover, the nonlinearity of the conductivity is enhanced by even higher orders of magnitude in a manner identical to that of the enhancement of the nonlinear susceptibility, which might find use in nonlinear control of heat and mass flow, for example.
It has been shown that the symmetry and strength of individual components of linear and nonlinear susceptibility tensors can be tuned by the appropriate design of microscopic composites of nonlinear dielectrics and metals and that 18 nonlinear crystal classes can be realized. A near dispersionless enhancement of χ^{(3)} of the order of 10^{6} was demonstrated for broad terahertz frequency ranges, and resonant enhancements of the orders of 10^{9} and 10^{4} were demonstrated for terahertz and nearinfrared frequencies, respectively. Ultrafast, lowpower, pixelbypixel alloptical modulation schemes at optical communication wavelengths was presented using a structure with a subwavelength thickness (λ/3). These SAMMs may serve as a useful platform for scientific studies of photonic nonlinearity with weak light or for highdensity integration of nonlinear photonic devices. Moreover, the enhancement principle may find applications in other branches of science, while synergistic enhancements of multiphysical interactions are also imaginable.
Methods
Linear and nonlinear effective electric susceptibility tensors of metamaterials
The general form of the effective electric susceptibility tensors for a threedimensional metamaterial can be formulated in a similar manner to that of the twodimensional case (metasurface), which was previously derived from the Lorentz reciprocity theorem^{13,19,25}. In general, the nthorder nonlinear susceptibilities are (n + 1)thorder tensors. However, for nthharmonic generation problems, the corresponding nonlinear susceptibilities can be represented with a more concise secondorder tensor form using a contracted notation^{11,26,28},
where V is the unitcell volume and \({\tilde{\mathbf{{\chi }}}}_{{\mathrm{eff}}}^{(n)}\) is a \(3 \times \left( {\left( \begin{array}{c}3\\ n\end{array} \right)} \right)\) homogenized nonlinear susceptibility tensor where \(\left( {\left( \begin{array}{c}3\\ n\end{array} \right)} \right) = (2 + n)!n!^{  1}2^{  1}\) is a combination with repetitions. \({\tilde{\mathbf{F}}}_n^{(\omega )}({\mathbf{r}})\) is a \(\left( {\left( \begin{array}{c}3\\ n\end{array} \right)} \right)\)by\(\left( {\left( \begin{array}{c}3\\ n\end{array} \right)} \right)\) tensor of field enhancements. Each component of \({\tilde{\mathbf{F}}}_n^{(\omega )}({\mathbf{r}})\) is a combinatorial summation of the products of n individual fieldenhancement factors, \({E}_{{\alpha}i}^{({{\omega})}}({\bf{r}})={\bf{\alpha}}\cdot{\bf{E}}_{{\mathrm{loc}}}^{({{\omega})}}({\bf{r}}) / {\bf{E}}_{{\mathrm{ave}}}^{({{\omega})}}\) (with \({\mathbf{E}}_{{\mathrm{ave}}}^{(\omega)}{\parallel} \, {\mathbf{i}}\)), which quantifies how strong the αdirectional components of the local electric fields (\({\mathbf{E}}_{{\mathrm{loc}}}^{(\omega )}({\mathbf{r}})\)) are when the macroscopic electric field (\({\mathbf{E}}_{{\mathrm{ave}}}^{(\omega )}\)) is in the idirection. \(\tilde{\chi}^{(n)}{(\bf{r})}\) is the local \(3 \times \left( {\left( \begin{array}{c}3\\ n\end{array} \right)} \right)\) nonlinear susceptibility tensor at location r for the nth harmonic generation (for n = 1, Eq. (3) is reduced to the homogenized linear susceptibility). The derivation and explicit, elementwise form of Eq. (3) is given in Supplementary Note 1.
Local electricfieldenhancement factor plot in Fig. 1
A lossy Drude model is assumed for gold, which is the metallic material (plasma frequency: 2.18 PHz, damping frequency: 6.45 THz)^{49}. Crystalline zinc oxide is assumed to be the dielectric^{50}. The enhancement factor values are obtained from FDTD simulations (We used commercial FDTD software of Lumerical Inc. for all FDTD simulations in this paper). A threeunitcellthick SAMM slab is used in the simulations. The enhancement factor is measured at 2 THz, which is a sufficiently low frequency and can be considered to be in the longwavelength regime.
Effective nonlinear electric susceptibilities calculation in Fig. 3
\(\chi _{{\mathrm{eff}},\,11}^{(3)}\) is retrieved from the simulations using a nonlinear transfer matrix method. The complex amplitude of the thirdharmonic wave is obtained based on nonlinear FDTD simulations, whereas the linear electromagnetic property of the SAMM is obtained from separate linear FDTD simulations. A dispersionless nonlinear dielectric (ε_{11} = ε_{33} = 6, \(\chi _{33}^{(3)}\) = 10^{−21} m^{2} V^{−2} for 10 GHz to 100 THz) and a lossy Drude model for gold^{49} are assumed as the constituent materials for Fig. 3a. Zinc oxide^{50} (ε_{11} = ε_{22} = 7.8, ε_{33} = 8.9, \(\chi _{33}^{(3)}\) = 10^{−18} m^{2} V^{−2} for 0.2 to 6 THz frequency) and the lossy Drude model for gold are assumed as the constituent materials, and a SAMM slab with twounit cell thickness is used for the FDTD simulation for Fig. 3b. Zinc oxide^{51} (ε_{11} = ε_{22} = ε_{33} = 4, \(\chi _{33}^{(3)}\) = 10^{−18} m^{2} V^{−2} for 20 to 750 THz frequency) and aluminum^{52} are assumed as the constituent materials, and a SAMM slab with a halfunit cell thickness is used for the FDTD simulations for Fig. 3c.
Alloptical switch demonstration in Fig. 4
Owing to the limitation of the computational load, we obtained the results based on twodimensional FDTD simulations (i.e. uniform in the ydirection in Fig. 4a). The pulse energy was calculated for a threedimensional beam whose electric field profile in the radial direction is the same as that of the twodimensional beam used in the simulations. The optical property of the GaN/AlN quantum well is calculated from twolevel electron model^{11,34} whose carrier density, excitation energy, transition dipole length, relaxation time, dephasing time, and background relative permittivity are 5 × 10^{19} cm^{−3}, 0.8 eV, 0.5 nm, 100 fs, 10 fs, 5.37, respectively^{34,37}. Calculated intensitydependent permittivity of the GaN/AlN quantum well is presented in Supplementary Fig. 4. The dynamics of the material property of the quantum well is treated as twolevel electron model which incorporate equation of motion of electron population in sublevels^{53} in nonlinear FDTD simulations. Aluminum^{52} and silver^{52} are assumed to be the materials used for the metallic part of the SAMM and for the backreflecting mirror, respectively. Silica is used for the upper and lower impedancematching dielectric layers (relative permittivity: 2.074). A rigorous analytical model for the electric susceptibilities of SAMMs, a transfer matrix method, and a particle swam optimization are used for the structural parameter optimization.
Modified Miller’s rule for the SAMMs in Fig. 5
The black dots in Fig. 5 are calculated using conventional Miller’s rule as
The red dots in the same figure are calculated using our modified Miller’s rule as
where
All \(\chi _{{\mathrm{eff}}}^{(1)}\) are obtained from the simulation results.
Data availability
The authors declare that the data supporting the findings of this study are available within the article (and its Supplementary Information files) and are available from the corresponding author upon reasonable request.
Code availability
The code to carry out simulations and analysis can be found at https://doi.org/10.6084/m9.figshare.12043308 and are available from the corresponding author upon reasonable request.
References
Kivshar, Y. S. & Agrawal, G. Optical Solitons: From Fibers to Photonic Crystals (Academic Press, Cambridge, MA, 2003).
Chai, Z. et al. Ultrafast alloptical switching. Adv. Opt. Mater. 5, 1600665 (2017).
Ou, J.Y., Plum, E., Zhang, J. & Zheludev, N. I. Giant nonlinearity of an optically reconfigurable plasmonic metamaterial. Adv. Mater. 28, 729–733 (2016).
Maruo, S., Nakamura, O. & Kawata, S. Threedimensional microfabrication with twophotonabsorbed photopolymerization. Opt. Lett. 22, 132–134 (1997).
Denk, W., Strickler, J. & Webb, W. Twophoton laser scanning fluorescence microscopy. Science 248, 73–76 (1990).
Cai, W., Vasudev, A. P. & Brongersma, M. L. Electrically controlled nonlinear generation of light with plasmonics. Science 333, 1720–1723 (2011).
Ren, M., Plum, E., Xu, J. & Zheludev, N. I. Giant nonlinear optical activity in a plasmonic metamaterial. Nat. Commun. 3, 833 (2012).
Li, G. et al. Continuous control of the nonlinearity phase for harmonic generations. Nat. Mater. 14, 607 (2015).
Tymchenko, M. et al. Gradient nonlinear pancharatnamberry metasurfaces. Phys. Rev. Lett. 115, 207403 (2015).
Chen, S. et al. Giant nonlinear optical activity of achiral origin in planar metasurfaces with quadratic and cubic nonlinearities. Adv. Mater. 28, 2992–2999 (2016).
Boyd, R. W. Nonlinear Optics (Academic press, Cambridge, MA, 2003).
Sirtori, C., Capasso, F., Sivco, D. L. & Cho, A. Y. Giant, triply resonant, thirdorder nonlinear susceptibility χ^{(3)}_{3ω} in coupled quantum wells. Phys. Rev. Lett. 68, 1010–1013 (1992).
Lee, J. et al. Giant nonlinear response from plasmonic metasurfaces coupled to intersubband transitions. Nature 511, 65–69 (2014).
Pendry, J. B., Holden, A. J., Robbins, D. & Stewart, W. Magnetism from conductors and enhanced nonlinear phenomena. IEEE Trans. Microw. Theory Tech. 47, 2075–2084 (1999).
Zharov, A. A., Shadrivov, I. V. & Kivshar, Y. S. Nonlinear properties of lefthanded metamaterials. Phys. Rev. Lett. 91, 037401 (2003).
Kim, S. et al. Highharmonic generation by resonant plasmon field enhancement. Nature 453, 757–760 (2008).
Rose, A., Larouche, S. & Smith, D. R. Quantitative study of the enhancement of bulk nonlinearities in metamaterials. Phys. Rev. A 84, 053805 (2011).
Lassiter, J. B. et al. Thirdharmonic generation enhancement by filmcoupled plasmonic stripe resonators. ACS Photonics 1, 1212–1217 (2014).
O’Brien, K. et al. Predicting nonlinear properties of metamaterials from the linear response. Nat. Mater. 14, 379–383 (2015).
Jäger, I. & Fratzl, P. Mineralized collagen fibrils: a mechanical model with a staggered arrangement of mineral particles. Biophys. J. 79, 1737–1746 (2000).
Gao, H. Application of fracture mechanics concepts to hierarchical biomechanics of bone and bonelike materials. Int. J. Fract. 138, 101–137 (2006).
Smith, D. R. & Pendry, J. B. Homogenization of metamaterials by field averaging. J. Opt. Soc. Am. B 23, 391–403 (2006).
Alù, A. Firstprinciples homogenization theory for periodic metamaterials. Phys. Rev. B 84, 075153 (2011).
Gorlach, M. A., Voytova, T. A., Lapine, M., Kivshar, Y. S. & Belov, P. A. Nonlocal homogenization for nonlinear metamaterials. Phys. Rev. B 93, 165125 (2016).
Roke, S., Bonn, M. & Petukhov, A. V. Nonlinear optical scattering: the concept of effective susceptibility. Phys. Rev. B 70, 115106 (2004).
Kleinman, D. A. Nonlinear dielectric polarization in optical media. Phys. Rev. 126, 1977–1979 (1962).
Chang, T. et al. Broadband giantrefractiveindex material based on mesoscopic spacefilling curves. Nat. Commun. 7, 12661 (2016).
Yang, X.l & Xie, S.w Expression of thirdorder effective nonlinear susceptibility for thirdharmonic generation in crystals. Appl. Opt. 34, 6130–6135 (1995).
Fredkin, D. R. & Mayergoyz, I. D. Resonant behavior of dielectric objects (electrostatic resonances). Phys. Rev. Lett. 91, 253902 (2003).
Quinten, M. Optical Properties of Nanoparticle Systems: Mie and Beyond (John Wiley & Sons, Hoboken, NJ, 2010).
Kim, J. Y. et al. Highly tunable refractive index visiblelight metasurface from block copolymer selfassembly. Nat. Commun. 7, 12911 (2016).
Chung, K., Kim, R., Chang, T. & Shin, J. Optical effective media with independent control of permittivity and permeability based on conductive particles. Appl. Phys. Lett. 109, 021114 (2016).
Kim, R. et al. Metal nanoparticle array as a tunable refractive index material over broad visible and infrared wavelengths. ACS Photonics 5, 1188–1195 (2018).
Suzuki, N. & Iizuka, N. Electron scattering rates in AlGaN/GaN quantum wells for 1.55µm intersubband transition. Jpn. J. Appl. Phys. 37, L369 (1998).
Cho, S.H. et al. A black metaldielectric thin film for highcontrast displays. J. Korean Phys. Soc. 55, 501–507 (2009).
Maniyara, R. A., Mkhitaryan, V. K., Chen, T. L., Ghosh, D. S. & Pruneri, V. An antireflection transparent conductor with ultralow optical loss (<2%) and electrical resistance (<6 Ω sq^{−1}). Nat. Commun. 7, 13771 (2016).
Iizuka, N., Kaneko, K. & Suzuki, N. Alloptical switch utilizing intersubband transition in GaN quantum wells. IEEE J. Quantum Electron. 42, 765–771 (2006).
Hu, X., Jiang, P., Ding, C., Yang, H. & Gong, Q. Picosecond and lowpower alloptical switching based on an organic photonicbandgap microcavity. Nat. Photonics 2, 185 (2008).
Koos, C. et al. Alloptical highspeed signal processing with silicon–organic hybrid slot waveguides. Nat. Photonics 3, 216 (2009).
Nozaki, K. et al. Subfemtojoule alloptical switching using a photoniccrystal nanocavity. Nat. Photonics 4, 477 (2010).
Martínez, A. et al. Ultrafast alloptical switching in a siliconnanocrystalbased silicon slot waveguide at telecom wavelengths. Nano Lett. 10, 1506–1511 (2010).
Ren, M. et al. Nanostructured plasmonic medium for terahertz bandwidth alloptical switching. Adv. Mater. 23, 5540–5544 (2011).
Shcherbakov, M. R. et al. Ultrafast alloptical switching with magnetic resonances in nonlinear dielectric nanostructures. Nano Lett. 15, 6985–6990 (2015).
Wood, R. M. LaserInduced Damage of Optical Materials (CRC Press, Cleveland, Ohio, 2003).
Ettoumi, W., Petit, Y., Kasparian, J. & Wolf, J. P. Generalized miller formulæ. Opt. Express 18, 6613–6620 (2010).
Shin, J., Shen, J.T. & Fan, S. Threedimensional metamaterials with an ultrahigh effective refractive index over a broad bandwidth. Phys. Rev. Lett. 102, 093903 (2009).
Choi, M. et al. A terahertz metamaterial with unnaturally high refractive index. Nature 470, 369–373 (2011).
Munch, E. et al. Tough, bioinspired hybrid materials. Science 322, 1516–1520 (2008).
Ordal, M. A., Bell, R. J., Alexander, R. Jr, Long, L. & Querry, M. Optical properties of fourteen metals in the infrared and far infrared: Al, Co, Cu, Au, Fe, Pb, Mo, Ni, Pd, Pt, Ag, Ti, V, and W. Appl. Opt. 24, 4493–4499 (1985).
Azad, A. K., Han, J. & Zhang, W. Terahertz dielectric properties of highresistivity singlecrystal ZnO. Appl. Phys. Lett. 88, 021103 (2006).
Bond, W. L. Measurement of the refractive indices of several crystals. J. Appl. Phys. 36, 1674–1677 (1965).
Palik, E. D. Handbook of Optical Constants of Solids. Vol. 3 (Academic press, Cambridge, MA, 1998).
Taflove, A. & Hagness, S. C. Computational Electrodynamics: The FiniteDifference TimeDomain Method. (Artech house, Norwood, MA, 2005).
Acknowledgements
This work was supported by the Samsung Science & Technology Foundation (grant SSTFBA140106) and an NRF (National Research Foundation of Korea) grant funded by the Korean Government (NRF2017Global Ph.D. Fellowship Program).
Author information
Authors and Affiliations
Contributions
J.S. conceived the idea and supervised the project. T.C. and J.S. conducted the theoretical analyses. T.C. performed the numerical simulations. T.C., S.J., and M.H. performed the matched dielectric layer design.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Chang, T., Jeon, S., Heo, M. et al. Mimicking biomechanical principles in photonic metamaterials for giant broadband nonlinearity. Commun Phys 3, 79 (2020). https://doi.org/10.1038/s4200502003520
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s4200502003520
This article is cited by

Verification of the Effect of Local Electric Field Resonance Behaviors on the Transmission Properties of the DualBand Terahertz Metasurface
Journal of Electronic Materials (2023)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.