Material platforms for optical metasurfaces

4.1 Refractory plasmonic materials

As discussed previously, plasmonic and nanophotonic metasurfaces have grown continuously and found application in a wide variety of optical elements. The two most typically used materials for plasmonic metasurfaces are gold and silver. These materials suffer from many limitations when considered for practical devices. They are incompatible with operations that require temperature stability, robustness to high-intensity lasers, chemical inertness, ultrathin films, and CMOS compatibility. In this application space of plasmonics, the emerging transition metal nitrides become essential. Here, we discuss their relevance to metasurface-based devices. In 2011–2012, titanium nitride (TiN) and zirconium nitride (ZrN) were first reported for their usefulness in plasmonics [80], [116]. Visually, TiN and ZrN resemble the appearance of gold (Au), suggesting a substantial similarity in their optical properties. Furthermore, these materials show chemical stability at temperatures beyond 2000°C, which classifies them as ceramics [80]. The plasma frequencies of both TiN and ZrN are in the visible range but typically exhibit a high Im(ε), i.e. relatively large optical losses in the visible and near-infrared (IR) spectrum as compared to commonly used noble metals (see Figure 2A,B). TiN has found ample uses as a cheap alternative to Au for metallic coatings and is already incorporated in standard CMOS processing [123], [124]. Large-area nanofabrication techniques for plasmonic applications have also been established for TiN [125]. Although initially reported to be nonstoichiometric, TiN is capable of being grown for single-crystalline ultrathin plasmonic metal films [126]. Their temperature-dependent optical properties have also been studied, which has opened up doors for their application in high-temperature sensing devices [127].

Figure 2:

Examples of TiN based metasurfaces.

(A)–(B) Measured optical properties [real (A) and imaginary (B) parts of the dielectric permittivity] for various different transition metal nitrides, e.g. TiN, ZrN, hafnium nitride (HfN), and tantalum nitride (TaN) [80]. (C) Simulated (dashed line) and measured (solid line) absorption spectra for the TiN MM absorber. (Inset – left) Schematic representation of a unit cell of the three-layer TiN MM absorber with dimensions of a=250 nm, w=50 nm, p=300 nm, h₁=30 nm, h₂=60 nm, and h₃=150 nm; (inset – right) SEM image of the fabricated TiN absorber [117]. (D) Experimental absorption spectra of 3D truncated TiN nanopillar structures with 160, 200, and 320 nm diameters at their maximum achievable structure heights of 800, 1000, and 1500 nm, respectively. The inset represents the cross-section of the nanostructures [118]. (E) Time-dependent temperature of the sapphire substrate heated by plasmonic nano-disk array when excited by an 800-nm laser illumination. The inset shows schematic [119]. (F) Reflected power distribution between circular copolarized (red) and cross-polarized (blue) components as obtained in simulations for ZrN nanoantenna on bilayer stack of ZrN/AlScN to create a gap plasmon resonance. The inset shows a SEM image of a fabricated metasurface [120]. (G) (Left) TiN hole array lattice schematic, (middle – right) TiN lattices with a single hole size can simultaneously focus light on multiple regions, (lower) SEM images of TiN lattice optics, and (upper) the corresponding confocal microscopy images (λ=800 nm). The lattices were optimized to produce (left) two focal points at x=±1 μm, y=0, and z=4, 6 μm and (middle) two focal points at x, y=0, and z=3, 7 μm [121]. (H) Colored SEM image of the 180-nm-long rectangular TiN nanoantenna; schematic of the SHG and the inset shows measured second harmonic (blue) and the corresponding extinction spectra (red) [122]. (A) and (B) reprinted with permission from [80], copyright 2013 by John Wiley and Sons; (C) reprinted with permission from [117], copyright 2014 by John Wiley and Sons; (D) reprinted with permission from [118], copyright 2017 by John Wiley and Sons; (E) reprinted (adapted) with permission from [119], copyright 2013 by the American Chemical Society; (F) reprinted (adapted) with permission from [120], copyright 2016 by The Optical Society; (G) reprinted (adapted) with permission from [121], copyright 2013 by the American Chemical Society; (H) reprinted (adapted) with permission from [122], copyright 2016 by the American Chemical Society.

In an early work, a TiN-based MM functioning as a broadband absorber was demonstrated [117]. Absorption higher than 87% is obtained for 400–600 nm spectral window for an array of square-ring-type TiN resonators on a bilayer stack of SiO₂ on TiN on a sapphire substrate (Figure 2C). This design utilized the lossy nature of the metal back layer at higher frequencies together with losses caused by LSPs in the square ring array toward the lower-frequency regime. In the same work, the authors also tested the structures for stability against high temperature and intense laser illumination. Recently, another demonstration of polarization insensitive, broadband absorption utilized plasmonic TiN coating on an array of 1.5-μm-tall Si pillar structures. Almost 94% absorption of the incident illumination is achieved in the wavelength range of 300–2300 nm (Figure 2D). Later, additive atomic layer deposition (ALD) grown hafnium oxide (HfO₂ ~30 nm) coating was used on the TiN pillars to protect it from ambient oxidation, and as a result, thermal stability for temperatures up to 1473 K was also demonstrated.

In another example, local heat generation was explored by utilizing the surface plasmon resonances in the near-IR [119]. An array of disk-shaped TiN nanoparticles was shown to be an efficient nanoscale heat source that outperformed a similar Au-based counterpart in the biological transparency window (Figure 2E).

Nanoparticles are vital components of many essential device applications of plasmonics. Single-crystalline colloidal plasmonic TiN nanoparticles have been examined early on, identifying them to be particularly useful in photothermal, photocatalytic, and nanoscale heating systems [128], [129]. Recently, researchers have also demonstrated efficient broadband hot electron generation and collection mechanism in TiN-nanoparticle-decorated TiO₂ nanowires [130]. Broad plasmonic resonances in TiN nanoparticles produced higher photocurrent enhancement in solar water splitting as compared to the Au nanoparticle counterparts.

The first demonstration of high-efficiency phase-gradient metasurface using refractory metal nitride used ZrN brick antenna structures on a bilayer plane of AlScN (aluminum scandium nitride) on ZrN [120]. This metasurface exhibits PSHE by creating a spatially separated mirror symmetric spectrum of the two opposite circular polarizations (right and left) of the incident light (Figure 2F). High working efficiency (~60%) similar to a previously reported Au-based design was achieved.

In yet another recent work, a metasurface device has been designed using subwavelength hole arrays in single-crystalline plasmonic TiN [121]. A powerful lattice evolution algorithm was implemented that utilized inverse form fitness function-based multiobjective optimization method. The resulting design of large area nanohole lattice in TiN could produce arbitrary far-field light patterns with balanced intensities at the wavelength of ~800 nm (Figure 2G).

Elevated damage threshold makes the refractory metal nitrides suitable for nanophotonic applications of nonlinear effects. Direct Z-scan measurements on TiN films have concluded the effective nonlinearities similar in magnitude to other standard metal films [131]. Low second harmonic intensity has also been previously measured from TiN thin films [132]. However, a significantly high laser power (5 GW/cm²) damage threshold has been experimentally verified for the thin films made of this metal nitride. More recently, in another work, second harmonic generation (SHG) was recorded from a rectangular array of TiN brick antenna structures with plasmonic resonances in the wavelength range of 950–1050 nm [122]. Compared to a similar Au-based design, TiN structures showed excellent stability and could endure laser intensities up to 15 GW/cm² without changing their optical properties or physical appearance (Figure 2H). Additionally, thin dielectric coated nanostructured Au has also been demonstrated useful for refractory plasmonics and nonlinear optics applications [77].

The studies discussed above have established the importance of TiN-based nanophotonic devices in numerous applications to harsh environment operating systems such as solar thermophotovoltaics, combustion thermophotovoltaics, waste heat harvesting in industries, as well as for durable optical elements for high power laser application, e.g. nonlinear signal converters, biological and chemical sensors, photocatalysis, HAMR, and so forth.

4.2 Epitaxial silver

As mentioned previously in the plasmonic metasurfaces section, silver has the highest negative permittivity in the visible frequency regime, which makes it suitable for plasmonic application. However, a silver thin film that is grown with electron beam evaporation often has poor optical properties as the film is not lattice matched to the substrate and is usually polycrystalline. For thin films that are subwavelength thick, the optical properties in the visible frequencies become less plasmonic and lossier due to the grain boundary defect scattering. An epitaxially grown silver thin film has the potential of having plasmonic properties similar to bulk silver films at a reduced thickness. Lattice matching of silver with substrates such as alumina (Al₂O₃), magnesium oxide (MgO), or mica is well known in the literature [133], [134]. The epitaxial silver film was explored for plasmonic application by Park et al. [135] on Mica substrate and patterned using focused ion beam (FIB). Wu et al. [136] demonstrated an epitaxially grown silver thin film on heavily doped Si substrate. Epitaxially grown silver thin film is shown to have an order of magnitude smaller root mean squared (RMS) surface roughness than that of equivalent thermally evaporated silver thin film. The 45-nm silver thin film had comparable optical properties of bulk silver films of Johnson and Christy [137]. Using a capping layer of Al₂O₃, they experimentally showed a surface plasmon polariton (SPP) propagation distance of 42 μm at 880 nm wavelength while theoretically calculating a 155 μm propagation distance for an ideal layer of film.

From measured optical constants of their thin film, Wu et al. numerically computed and showed that epitaxial silver nanostructures have the potential to exhibit a much sharper Fano-resonance than similar structures made from the evaporated silver film. Figure 3A shows experimentally obtained propagation length for epitaxial silver films, as well as numerical simulation results showing epitaxial silver outperforming evaporated silver with sharper Fano resonance and enhanced surface-enhanced Raman scattering (SERS). Numerical calculations show that for core-shell structures [136] with the gain medium, epitaxial silver can obtain normalized polarizability necessary for the lasing condition, where thermally evaporated silver cannot. For bow-tie antennas made of silver, they numerically predicted that epitaxial silver bow-ties would have 16 times SERS enhancement factor at resonance than evaporated silver-film-based structures.

Figure 3:

Epitaxial silver for metasurface applications.

(A) Epitaxial silver shows sharper Fano resonance or enhanced field confinement in bow-tie antenna compared to their evaporated counterparts [136]. Top-left panel shows experimental setup with SPP waves launched in the silver thin film. Top-right panel shows the propagation distance for different amplitudes obtained experimentally. Bottom-left and bottom-right panels show the theoretical calculation for Fano-resonant and SERS enhancement factor for epitaxial and thermally evaporated silver. (B) Epitaxial silver mirror reduces mode volume of lasing [138]. The top-left panel shows the GaN-InGaN core-shell structure, and top-right shows the field enhancement in SiO₂ spacer layer with the epitaxial silver thin film. Bottom-left shows the emission spectra with different pump fluences and bottom-right shows the lasing saturation. (C) Hyperbolic dispersion in 2D structure demonstrated with the help of epitaxial silver [139]. Top-left panel shows the SEM image of the device, and the top-middle panel shows the deflection angle as a function of wavelength. Below 550 nm wavelength, the device shows negative refraction and therefore operates in the hyperbolic regime. Top-right panel shows the propagation length in the hyperbolic metasurface. The bottom panel shows optical images of the refraction in three different wavelengths of 640 nm, 540 nm, and 490 nm. (D) Color holography with epitaxial silver metasurface [40]. Top-left shows a schematic of a single unit cell, while top-right shows a SEM image, with scale bar representing 500 nm. The bottom panel shows optical images with objective focused at different distances from the metasurface. At 5 μm distance, the image appears to be clearest. (A) Reprinted with permission from [136], copyright 2014 by John Wiley and Sons; (B) reprinted with permission from [138], copyright 2012 by the American Society for Advancement of Science; (C) reprinted with permission from [139], copyright 2015 by Springer Nature; (D) reprinted with permission from [40], copyright 2017 by John Wiley and Sons.

A direct comparison between evaporated and epitaxial silver was performed for lasing application by Lu et al. [138] that showed the superior performance of epitaxial silver. An epitaxial silver film is used as a back-reflector layer, and a 5-nm-thin spacer layer of SiO₂ is situated between InGaN@GaN core-shell structure and the silver back-mirror. FEM simulation shows strong field confinement in the spacer layer. With 405 nm CW semiconductor diode laser excitation, lasing at 510 and 522 nm was observed. In contrast, a similar InGaN@GaN core-shell structure with a similar thickness of SiO₂ but on a polycrystalline (evaporated) silver showed no lasing. Figure 3B shows high mode confinement using the epitaxial silver structure. It is evident that the epitaxial silver is critical to demonstrate lasing using the structure. Bimodal lasing with similar pumping threshold was experimentally observed.

It was predicted that hyperbolic metasurface could show anomalous dispersion [140] for propagating SPP similar to a bulk 3D metasurface. In their letter, Liu et al. showed that a metasurface, when engineered adequately with periodic gratings, can give rise to flat or hyperbolic isofrequency dispersion contours. The so-called “hyperbolic metasurface” was first demonstrated through epitaxial silver by High et al. [139]. Their fabricated metasurface showed in 2D some of the critical characteristics of bulk hyperbolic MMs, such as negative refraction and diffraction-free propagation. Figure 3C depicts the negative angle of refraction at wavelengths lower than 550 nm, denoting the hyperbolic regime, and the positive angle of refraction at the elliptical regime with wavelengths longer than 550 nm. These devices demonstrated polarization-dependent spin-orbit coupling of incoming photons, which could readily showcase the PSHE.

Epitaxial silver’s low loss and long SPP propagation length at the visible frequencies can be utilized to make metasurfaces for visible frequency applications. Since epitaxial silver is grain-boundary defect free, one can easily pattern the film using FIB, whereas for evaporated silver with grain boundaries, patterning causes rough side walls. Patterned rotated nanoslits can provide Pancharatnam-Berry geometric phase to an incoming electromagnetic wave. Again, for a hologram generation, one must use computed phase profile to reproduce an image to a virtual plane. By using the computed phase profile, rotated nanoslits are patterned onto an epitaxial silver film that gives incoming circularly polarized light the necessary phase shifts for creating the hologram. The resulting holograms are observed using confocal microscopy at a focal plane of 10 μm above the structures, as demonstrated in Figure 3 [40]. Figure 3D shows the observed holograms while the focal plane is shifted, showing the sharper image at the designed length of 5 μm.

Though silver itself is a conventional plasmonic material, epitaxial silver, having optical properties of bulk silver at nanoscale thicknesses, opens new dimensions for plasmonic metasurface design at visible wavelengths with strong field confinement and lower loss.

4.3 Silicon and its oxides and nitrides

Compared to plasmonic metasurfaces, dielectric metasurfaces has a clear advantage for low loss application. Section 3 discusses the principles of dielectric metasurfaces; in this subsection, many of the important works on silicon-based metasurfaces are listed. The primary advantage for silicon-based metasurface is the ability to use existing fabrication techniques and tools to manufacture the devices rapidly. Silicon has been the primary material of choice for the CMOS industry for over five decades. Since silicon manufacturing technology is robust, many of the proof-of-concept designs for dielectric metasurface chose silicon platform to demonstrate the concepts. Silicon is lossier than transparent dielectric materials in the visible-IR regime, and most devices that utilize silicon show operation in the near-IR regime. For an excellent overview of this topic and a possible future roadmap, please refer to Staude and Schilling’s review [83].

High-efficiency dielectric Huygens’ surfaces were proposed by Decker et al. [105], by spectrally overlapping the electric and magnetic resonances for Si nanodisks. By engineering the height and diameter of Si nanodisks, they overlapped the two resonances of the nanodisks. Figure 4A inset shows the array of Si nanodisks. The simulation results in Figure 4A illustrate that by adjusting the disk radius, the phase control of 2π can be achieved. For near-IR operation, the experimental results are shown in Figure 4B, which shows near-unity transmission with phase dependent on the incoming frequency.

Figure 4:

Silicon as a metasurface material platform.

(A) Computed phase profile of Si nanodisk array at different wavelengths and diameters of disks show full 2π phase control (inset: Si nanodisks engineered to have simultaneous electric and magnetic resonance). (B) Near-unity transmittance for the disk array shown both with simulation and experimental results [105]. (C) Optical image (inset: SEM image) of a Si vortex plate [53]. (D) Si-metasurface-based perfect reflector [89]. (E) Refractive index sensing using Si-based Fano-resonance metasurface [87]. (F) Enhanced focusing with the Si-based flat lens (inset: SEM image of the Si-based flat lens) [141]. (G) Si-metasurface-based broadband holography; subpanels show SEM image and hologram in different illumination wavelengths. (H) Top-left: false-colored SEM image of chiral metasurface made of Si; purple and green colors represent two different chiralities. Right panels: R and L image represent right-hand circularly polarized (RCP) and left-hand circularly polarized (LCP) illumination response. The bottom-left panel shows unpolarized light response [112]. (I) Left panel shows optical and SEM images of Si nano-post-based metasurface; right panel shows beam conversion output, both simulation and experimental results [142]. (J) Metasurface grating embedded into a silicon waveguide show mode conversion. TE0 mode is converted into TE1 mode [143]. (A)–(B) reprinted with permission from [105], copyright 2015 by John Wiley and Sons; (C) reprinted with permission from [53], copyright 2017 by the American Chemical Society; (D) reprinted with permission from [89], copyright 2014 by the American Physical Society; (E) reprinted with permission from [87], copyright 2014 by Springer Nature; (F) reprinted with permission from [141], copyright 2014 by the Optical Society of America; (G)–(H) reprinted from [112] through Creative Commons Attribution-NonCommercial license from Science Advance; (I) reprinted with permission from [142], copyright 2015 by Springer Nature; (J) reprinted with permission from [143], copyright 2016 by the American Chemical Society.

Dielectric resonators with silicon were demonstrated as a material platform by Yang et al. [53]. They designed the metasurface by maximizing the cross-polarized reflection and choosing the dimensions of the silicon nanoresonator to obtain proper phase shift. By patterning silicon nanoantennas backed by a silver reflector, vortex beam generation at a conversion efficiency of 98% at 200 nm bandwidth is shown. Figure 4C shows an optical image of the vortex plate. Yang et al. [87] also showed a cloaking effect using the all-dielectric silicon-based metasurface. Figure 4D shows near-unity transmission in telecommunication wavelength for the designed all-dielectric metasurface. They showed that the EIT device could easily outperform its plasmonic counterpart and could be used for refractive-index sensing with a higher sensitivity and figure-of-merit than its plasmonic counterparts. Figure 4E shows refractive index sensing using a Fano resonance Si metasurface. Moitra et al. [89] demonstrated high-efficiency perfect reflector using Si material platform by utilizing the Mie resonance. The array of Si nanodisks shows 98% reflectivity in 200 nm bandwidth at the near-IR wavelength.

Silicon nanoposts have been utilized to create an on-chip focusing lens by West et al. [141] where the nanopillars were fabricated using reactive-Ion etching on a silicon wafer. Figure 4F shows a SEM image of the structures of the lens and focusing of a Gaussian beam. Full phase control was achieved at telecommunication wavelength of 1.55 μm. Designed silicon lens showed subwavelength thin micro-lens effect. Such lenses could be used for reducing insertion loss by wafer-level integration of the lens. Kim et al. [144] demonstrated phase control and focusing by patterning nano-apertures in a silicon substrate. They showed the achromatic performance of the designed lens at operating wavelength of 1500–1900 nm. Khorasaninejad et al. showed aberration-free imaging by metasurface lens by patterning coupled dielectric resonator array. Invariance of focal length is demonstrated experimentally from 1300 nm to 1800 nm wavelength for their lens. By utilizing their Si platform, Khorasaninejad et al. demonstrated dispersion-free broad-band and chiral holograms with blazed-binary silicon grating structures and rotated nanoantennas [112]. Figure 4G demonstrates the broad-band hologram operating at 480, 530, 585, and 630 nm. Figure 4H shows the false-colored SEM image of the chiral hologram. Magenta and green represent the two chiralities. Figure 4H shows the output of the metasurface, while R is displayed when RCP light is illuminated on the device, and L is displayed for LCP light.

Arbabi et al. [142] showed high-efficiency metasurface with silicon by utilizing high contrast Si nanoposts. Each nanopost acts as a low-Q Farby-Perot resonator, while the orientation of ellipse determines the polarization-dependent phase shift of the structure. Experientially, up to 97% efficiency was shown for such metasurfaces. Figure 4I shows the optical microscopy and SEM image of such metasurface and illustrates both experimental and simulation results for an axial and radial polarization converter with the metasurface.

Recently, a high-efficiency Si-metasurface with crystalline Si was reported by Zhou et al. [145]. By transferring thin-film c-silicon onto a quartz substrate and patterning it with electron-beam lithography into nanoposts, they created Si metasurface on transparent substrates. They showed that the crystalline Si metasurface could obtain transmission efficiency of 71% and 95% diffraction efficiency with full 2π phase control at 532 nm wavelength.

The primary advantage of Si metasurfaces is that they can augment or modify the existing devices and bring in new applications. All dielectric chiral metasurface with CMOS compatible process was demonstrated with silicon with a quality factor of up to 100 [146]. Vortex converter with 45% efficiency at telecommunication wavelength is shown using Si [107]. Mode conversion was demonstrated with the help of metasurfaces embedded on a silicon waveguide [143], [147]. Figure 4J shows the SEM image and input and the output for such a device. The fabricated SOI structure can readily convert a TE0 mode to the TE1 mode using graded refractive index profile. The measured mode purity is shown to be 95% for the transformed mode and transmission of 88% is achieved. Enhanced nonperturbative high-harmonic emission from is demonstrated with a Si metasurface designed with Fano-resonance [148].

Silicon compounds such as silica (SiO₂) and silicon nitride (Si₃N₄) are both transparent in the visible and has lower losses and can be suitable for metasurface fabrication. Si₃N₄ has a higher refractive index of 2.032 compared to around 1.5 for SiO₂ and therefore can give better index contrast. Both SiO₂ and Si₃N₄ have lower indices than TiO₂, which has been explored more for metasurface fabrication and is further discussed in Section 4.4. Si₃N₄ has been utilized to make low-contrast metasurface with high transmission efficiency and focusing efficiency [149]. Park et al. [150] showed subtractive color filter with cyan, magenta, and yellow color with hydrogenated amorphous silicon (a-Si:H) disks, which shows potential to generate full-color pallet with such devices. Unpolarized high-numerical-aperture imaging with Si₃N₄ lens has also been experimentally demonstrated [151]. Planar a-Si metasurface embedded into SiO₂ was made by local oxidation of silicon [152] and reflects the viability of combining the silicon platform with its nitride and oxides.

4.4 Titanium dioxide

One major potential application of metasurface is to construct flat metalenses to realize focusing and imaging functionalities. Such lenses can be made with wavelength-comparable thicknesses, as opposed to their refractive counterparts, which require large thicknesses to realize phase accumulation through light propagation in the material. To shape the wavefront into the desired form, metalenses are implemented by discretizing the flat surface into sub-wavelength pixels (optical antennas), each imparting the locally required phase shift by either resonant or waveguiding effect [82], [153]. As discussed above in the Plasmonic Metasurface and Si Metasurface sections, although metals and silicon are widely used in proof-of-concept demonstrations for metasurface researches, they are not suitable for making optical metalenses due to the substantial optical losses in the visible frequencies. Instead, great research endeavor toward high-efficiency metalenses has been focused on titanium dioxide (titania, TiO₂) due to its minimal optical loss and high refractive index (~2.4) across the visible spectrum. The high refractive index results in large optical confinement inside the optical elements, which diminishes the optical crosstalk between neighboring elements when the spatial sampling period (i.e. pixel size) is reduced for higher efficiency. In fact, TiO₂ metalenses have been realized in as early as the 1990s. In a first effort known to date, Lalanne et al. [102], [103], [104] experimentally demonstrated 78–89% first-order diffraction efficiency at 633–860 nm wavelengths in a TiO₂-based binary blazed grating structure (Figure 5A). The beam diffraction was realized by patterning TiO₂ with variable fill fraction, thus local refractive index in each pixel, so that a distribution of effective refractive index could be engineered to achieve the desired beam diffraction. The pixel sizes were carefully chosen to optimize the performance while also accounting for fabrication limitations. The resulted diffraction efficiency was ~12% higher than that of conventional echelette lenses (also known as the Fresnel lenses) because the adverse shadowing effect (abnormal light refraction close to the vertical edge of the blazed gratings) inherent to echelette lenses was amended with waveguiding effect in the TiO₂ metalens [104].

Figure 5:

Titanium dioxide (TiO₂) as a metasurface material platform.

(A) Scanning electron micrograph (SEM) of a blazed binary diffractive lens made of top-down etched TiO₂ nanopillars; the local effective refractive index is controlled by the width of the pillars. (B) SEM of TiO₂ metalens fabricated using the bottom-up atomic layer deposition (ALD) method. Inset: Cross-section view of the nanofins exhibiting vertical sidewalls with a height of ~600 nm. (C) Simulated focusing efficiency of three metalenses designed for wavelengths 405 (blue curve), 532 (green curve), and 660 (red curve) nm and SEM of fabricated metalens. Inset: measured focal spot intensity at designed wavelength 405 nm (NA=0.8). (D) Simulated complex transmission coefficients of cylindrical nanopillars with various diameters at three design wavelengths: 405, 532, 660 nm. Each point represents the amplitude and phase of the transmission of a nanopillar, and the entire 2π phase coverage is realized by varying the nanopillar diameter. (E) Top: different holograms are imaged with LCP and RCP light illuminating on the same TiO₂ metasurface, which decouples the propagation phase design and geometric (Pancharatnam-Berry) phase design by controlling the nanofin dimensions and rotation angles, respectively. Bottom: images formed with RCP and LCP light, respectively. (F) Left: schematic of the off-axis meta-lens that focuses light of different wavelengths to different angles. Right: measured spectra from a supercontinuum laser with a planar ultracompact visible chiral spectrometer designed based on the off-axis focusing lens; the performance is comparable to a commercial handheld spectrometer. (A) reprinted with permission from [103], copyright 1999 by The Optical Society; (B) reprinted with permission from [154], the authors; (C) reprinted with permission from [27], copyright by the American Association for the Advancement of Science; (D) reprinted with permission from [110], copyright 2016 by the American Chemical Society; (E) reprinted with permission from [155], copyright 2017 by American Physical Society; (F) reprinted with permission from [156], using the Creative Commons Attribution (CC BY) license.

This field was recently revisited by a group of researchers at Harvard University, who used a novel ALD-based technique to fabricate TiO₂ nanopillars (referred to as “nanofins” in their publications) with aspect ratio as high as 15, nearly 90° vertical sidewalls, and roughness RMS as small as 0.738 nm [154]. As is shown in Figure 5B, the fabrication quality from this bottom-up etching-free approach surpasses that from the previously used top-down etching approach, and this new fabrication technique initiated a growing field of TiO₂-based metalens researches. In a pioneering work, Khorasaninejad et al. [27] designed, fabricated, and characterized a flat focusing lens with a high free space numerical aperture (NA) of 0.8, as well as diffraction-limited focusing with efficiencies of 86, 73, and 66% at 405, 532, and 660 nm wavelengths, respectively, outperforming the state-of-the-art commercial refractive lenses in both the focal spot size and profile symmetry; the imaging ability of the proposed flat metalens was also tested to be comparable to the bulk refractive lens (Figure 5C). In a later work, the NA was further increased from 0.8 to 1.1 by immersion method while maintaining a focusing efficiency of 50% [157]. These two metalenses rely on the rotation angles of the nanofins to locally impart the required phase shifts and work only for circularly polarized light. To overcome this constraint, Khorasaninejad et al. [110] demonstrated a polarization-insensitive metalens in the visible with similar NA and focusing abilities, by replacing birefringent nanofins with axisymmetric nanocylinders with varying diameters to attain the appropriate phase profile (Figure 5D). Apart from focusing metalenses, it has been pointed out that arbitrary phase profiles can be imposed on any two orthogonal polarization states by a single metasurface combining propagation phase design and geometric phase design (realized through varying the x-y dimensions and rotation angles of TiO₂ nanofins, respectively). Using the principle of geometric phase, a multiplexed hologram metasurface was shown that displays different images when the helicity of the illuminating circularly polarized light changes [155] (Figure 5E). Furthermore, a single metalens was designed for highly resolved chiral spectroscopy where the circularly polarized light of different helicities are focused to different spots upon passing through the metalens, eliminating the need for stacked and bulky optical systems to do the same job [158]. TiO₂-based metasurfaces were also used to demonstrate ultracompact spectrometers. Zhu et al. [156] designed a metasurface that strongly disperses focused light of different wavelengths to different angles, so that high spectral resolution can be realized in a single flat metasurface, in contrast to bulky conventional spectrometers, which depend on long propagation distances to separate the insufficiently dispersed light (Figure 5F). Specially designed TiO₂ metasurface can convert circularly polarized light into vortex beams, and researchers showed that the high versatility of the metasurface design enables the generation of vortex beams with arbitrary and even fractional topological charges (number of helical surfaces that constitutes the final helical mode), while exhibiting an absolute efficiency of 60%, which significantly surpasses that of metallic metasurfaces (8.4%) with similar functionalities [159].

It is worth noting that besides dry-etching and ALD-based fabrication methods, an e-beam evaporation and lift-off-based technique has also been investigated to fabricate TiO₂ metasurfaces. Using this method, a tunable TiO₂ metasurface on a stretchable substrate was fabricated and shown to exhibit 5.08% (0.96%) resonance red (blue) shift in the visible frequencies when 6% strain is applied transverse (parallel) to the polarization of incident light [160]; TiO₂ color filters have also been realized with full coverage of the red, green, blue colors in the reflection mode, and the relatively small pixel sizes are promising for bright, high-contrast, and high-resolution structural color generation [161].

The recent years have seen tremendous progress in TiO₂-based metasurfaces in terms of focusing with high efficiencies and large NA, multiplexed holograms, beam splitters for elliptically polarized light, ultracompact spectrometers, etc. To make more applicable devices based on this platform, future efforts should be spent on the following aspects:

Expanding the effective bandwidth: Currently, three different metalenses are needed to cover the entire visible spectrum; in real applications, however, a single lens should focus all visible light with high enough efficiencies.

Broadening the field of view: Images obtained with TiO₂ metalenses still suffer from a geometrical aberration, even at a single wavelength, probably due to the large NA.

Developing the mass-production technology: The current fabrication technique requires e-beam lithography and ALD deposition of TiO₂, which impose significant hurdle on the manufacturing cost and throughput. New technologies such as the nanoimprint need to be explored for mass production without compromising device quality.

4.5 Transparent conducting oxides and other metal oxides

Metal oxides (MOs) are a diverse set of heavily doped, wide-bandgap semiconductors that have shown great promise as alternative materials for metal-based metasurface and plasmonic applications [80], [162], [163], [164]. In contrast to many doped semiconductors, MOs support large charge donor densities – metal dopants and/or point defects – and have corresponding carrier concentrations as large as 10²¹ cm⁻³. Their large bandgaps restrict interband absorption primarily within the ultraviolet and visible spectrum, making intraband absorption the dominant loss mechanism at IR wavelengths. When combined, these two properties make MOs behave as a low-loss Drude metal at IR frequencies and a potential alternative material candidate for plasmonic and metasurface application and devices. Furthermore, MOs are exceptionally tunable materials. The past few years have seen an explosion of research into the nonlinear optical properties of MOs driven by optical, electrical, and thermal sources [51], [165], [166], [167], [168]. Here, we review recent and concurrent research and applications of MO materials for realizing low-loss and tunable plasmonic metasurfaces.

In general, MOs are well described using the Drude model. Within this description, the optical response of a MO transitions from dielectric to metallic at the crossover wavelength. To achieve near-IR crossover wavelengths, large carrier concentrations with simultaneous low effective electron masses are necessary. Indium tin oxide (ITO) was among the first semiconductors to achieve carrier concentrations of 10²¹ cm⁻³ and a plasmonic response at near-IR frequencies [169]. Gregory et al. [170] demonstrated one of the first MO MMs using both directly patterned and complementary ITO split ring resonator antennas (Figure 6A). Subsequent studies into other doped semiconductors quickly expanded the list of available oxides capable of supporting large carrier concentrations [80]. Kim et al. [172] demonstrated near-IR localized and surface plasmon resonances in periodically patterned aluminum- and gallium-doped zinc oxide (Al:ZnO and Ga:ZnO, respectively) films. In more advanced designs, the same authors fabricated Ga:ZnO/ZnO lamellar nanodisks for a highly sensitive mid-IR chemical sensor and also fabricated a Ga:ZnO quarter-waveplate operating at near-IR frequencies (Figure 6B) [51], [173].

Figure 6:

Transparent conducting oxides (TCO) as a metasurface material platform.

(A) IR transmission spectrum of direct (left) and complimentary (right) arrays of ITO nanoantennas showing plasmonic resonance dips. Insets show scanning electron micrographs of the fabricated ITO antennas. (B) Fabricated Ga:ZnO metasurface for near-IR (1.75–2.5 μm) quarter waveplate operation. Polar plot shows the intensity of reflected light after passing through a polarizer, demonstrating the linear to circular polarization conversion at the design wavelengths. (C) Ultrafast and broadband modulation of ITO nanorod arrays. The modulation amplitude exceeds 60% and recovers in less than a picosecond. (D) Ultrafast nonlinear wavelength shift of Al:ZnO thin films subject to both interband (UV) and intraband (NIR) pump. The two pump excitations induce opposite changes to the wavelength shift of the films and, when simultaneously combined, can completely cancel. (E) Enhance nonlinear effective refractive index of an ITO film at the ENZ wavelength of 1240 nm. Wavelength-dependent Kerr coefficient is measured using Z-scan and shows a clear enhancement at the ENZ wavelength, which increases drastically for increasing angles of incidence. (F) Electrically tunable metasurface arranged from plasmonic gap resonators. The voltage applied to the electrical contacts of the gold backing layer and gold antenna strips confines electrons at the ITO/Al₂O₃ interface. The accumulated electron density shifts the plasma frequency of the ITO layer and provides the necessary modulation functionality. Here, the authors were able to induce a dynamic 180° shift to the phase of reflected light. (A) Reprinted with permission from [170], copyright 2015 by the American Chemical Society; (B) reprinted with permission from [51], copyright 2016 by the American Chemical Society; (C) reprinted with permission from [171], copyright 2016 by Springer Nature; (D) reprinted with permission from [165], under terms of the Creative Commons CC BY license; (E) reprinted with permission from [166], copyright 2016 by the American Society for Advancement of Science; (F) reprinted with permission from [167], copyright 2017 by the American Chemical Society.

Although MOs can themselves be designed as the metallic constituents of a metasurface, noble metals continue to outperform MOs for plasmonics because of their outstanding optical properties. However, MOs exhibit very large optical tunability, a quality that is not found in noble metals. The exceptional tunability of MOs was first highlighted and pioneered by the group of Otto Muskens. In 2011, Abb et al. [174] showed that the optical properties of gold antennas on an ITO substrate could be significantly modified on an ultrafast timescale by optically pumping at the resonance wavelength. This effect was attributed to the excitation and injection of free carriers into the ITO layer via the large electric fields of the antenna, which then modified the dielectric constant of ITO and shifted the antenna resonance. The authors extended their results by patterning multifrequency crossed antenna arrays onto an ITO substrate and demonstrated control over the fast Kerr nonlinearities and the slow thermal nonlinear contributions [175]. Metzger et al. [176] utilized the large nonlinearities of ITO nanocrystals placed in the gap of a gold dimer plasmonic antenna to enhance significantly the third-harmonic signal of the hybrid antenna structure. The group of Robert Chang showed large modulation amplitudes and picosecond recombination rates in 2D arrays of ITO nanorods [177], [178]; in 2016, this same group broke their previous record and reached subpicosecond recombination rates in their report of arrays of ITO nanorods operating at visible wavelengths (Figure 6C) [179]. Conformal coating of phase-change material was used to achieve broadband all-optical switching [180]. Clerici et al. [165] used films of aluminum-doped zinc oxide (Al:ZnO) to show that the ultrafast response in their MO films could be achieved with both intraband and interband pump energies and, furthermore, that the combined nonlinear response could be additively combined to achieve switching speeds up to a few terahertz (Figure 6D). Several groups have also demonstrated remarkable modulation enhancement of MOs near their epsilon-near-zero (ENZ) wavelength – the frequency at which the permittivity of the media is approximately zero [181], [182], [183]. In 2016, the group of Robert Boyd demonstrated enhancement factors of ~2000 for the Kerr nonlinear coefficients of ITO films by utilizing the field discontinuity property of ENZ materials (Figure 6E).

ITO thin films have also been demonstrated as efficient modulators [168], [184], [185], [186]. Huang et al. [186] showed a gate-tunable metasurface capable of generating a 180° phase shift and ~30% change in reflectance with a 2.5 V gate bias. Here, the authors fabricated 2D arrays of gold stripes on an aluminum oxide/ITO/gold layered substrate and applied voltage between sets of three antenna stripes and the gold back contact. The field effect at the aluminum oxide/ITO layer generated a substantial change in the refractive index of the ultrathin ITO layer and allowed for control of the metasurface reflectance amplitude and phase. Park et al. [167] also utilized a very similar design and reported an electrically controlled metasurface capable of 180° phase control. Here, the voltage was applied to the ITO layer and the gold back contact, eliminating the need to apply a voltage to individual or sets of antennas (Figure 6F).

4.6 Graphene and 2D materials

Since its successful cleavage from bulk graphite in 2004, graphene – a monolayer of carbon atoms arranged in a 2D hexagonal lattice [187], [188] – has been intensively investigated as a plasmonic material. While the high optical losses and lack of dynamic tunability hinder conventional noble metal plasmonics from practical applications, graphene is perceived to be the ideal candidate to overcome these hurdles owing to its unique electrical and optical properties. Before experimental demonstrations of graphene plasmons were carried out, theoretical calculations conducted by multiple groups suggested that graphene plasmons exhibit two enticing features: ultrahigh optical confinement on the order of several hundred (λ_air/λ_plasmon~100) and long SPP propagation lengths (more than 100 SPP wavelengths in the mid-IR frequencies) [189], [190], [191]. These two features, together with graphene’s dynamic tunability in optical properties, form the basis of the unique graphene plasmon platform with merits unparalleled by conventional noble metal plasmons. However, the substantial difference between free space wavelength and plasmon wavelength imposes a significant challenge to launch surface plasmons in graphene (Figure 7A). Launching plasmons requires sharp tips (with a radius of a few tens of nanometers) to excite SPPs [197], [198], [199] or nano-patterning graphene into metasurface to excite LSPs [192], [193], [195], [200], [201], [202], [203], [204], [205], [206]. Several early theoretical works on graphene metasurface applications have suggested that LSPs in graphene nanodisks can lead to complete optical absorption in mid-IR in a Salisbury screen configuration [207], and graphene ribbons can be made into electrically modifiable waveguides with longer propagation distances and higher edge field confinement than conventional bulk metallic waveguides [205], etc. Notably, a recent theoretical work pointed out that ultrafast (~200 fs) radiative heat transfer happens in two graphene nanodisks vertically separated by 1 nm via plasmon interaction, resulting from the extraordinarily large plasmonic field concentration and low electronic heat capacity in graphene [208]. These theoretical works provided necessary tools to guide later experimental research on graphene metasurfaces.

Figure 7:

Graphene as a metasurface material platform.

(A) Calculated plasmon dispersion relation in graphene with Fermi level of E_F=1 eV as a function of in-plane wavelength 2π/k_‖. The light line (dotted curve) and the plasmon dispersion relations in the Drude (dashed curve) and local random phase approximation (local-RPA, solid curve) models are shown for comparison. The density plots are obtained using the full RPA conductivity for graphene with mobility μ=2000 cm²/(V⋅s). (B) Measured extinction spectra of 50-nm-wide graphene nanoribbons at varying Fermi levels. The dotted vertical line indicates the energy of the in-plane optical phonon of graphene. Greater than 100× wavelength confinement (λ_p≤λ₀/100) is achieved in 20-nm-wide nanoribbons with E_F=0.21 eV (shown in a separate graph from the reference). Inset shows the schematic of FTIR spectroscopy measurement. (C) Measured extinction spectra of protein sensor consisting of graphene nanoribbons array before (dashed) and after (solid) protein bilayer formation. Gray vertical strips indicate amide I and II vibrational bands of the protein. This graphene protein sensor exhibits 6× sensitivity in a spectral shift and 3× sensitivity in spectral modulation (dips at the two vertical strips) compared to conventional gold plasmonic antenna sensors (shown in a separate graph from the reference). Inset shows schematic of graphene nanoribbon graphene sensor. (D) Measured extinction spectra for nanorings with various outer diameters with a fixed inner diameter (60 nm), as shown by labels. Note that the antibonding mode plasmonic resonance energy reaches up to ~0.34 eV in 100 nm outer diameter graphene nanoring. Inset shows schematic of the device. (E) Measured reflectance spectra from a graphene loaded gold plasmonic antenna array for different gating voltages (V_G−V_CNP; V_CNP is the gate voltage when the concentrations of electrons and holes in the graphene sheet are equal), showcasing the considerable dynamic tuning range (1100 nm) of plasmonic resonance caused by graphene. All spectra are normalized to the reflection spectrum from a 300 nm Au film evaporated on the same substrate. Inset: SEM of the gold plasmonic structure, the graphene sheet is underneath the gold antennas and is in the background of the SEM. (F) Measured photovoltage generation enhancement on fractal metasurface compared to plain source-drain contact. A rather broadband enhancement is achieved with fractal metasurface, and the enhancement is insensitive to incident light polarization due to the hexagonal symmetry of the fractal metasurface (shown in a separate graph from the reference). Top inset: artistic rendering of fractal metasurface enhanced graphene photodetector; bottom inset: geometry of fractal metasurface used to enhance photovoltage generation in graphene. (A) Reprinted with permission from [192], copyright 2014 by the American Chemical Society; (B) reprinted with permission from [193], copyright 2013 by the American Chemical Society; (C) reprinted with permission from [194], copyright 2015 by the American Association for the Advancement of Science; (D) reprinted with permission from [195], copyright 2013 by the American Chemical Society; (E) reprinted with permission from [196], copyright 2014 by the American Chemical Society; (F) reprinted with permission from [66], copyright 2017 by American Chemical Society.

Thanks to the development of nanofabrication technology, graphene metasurfaces with elements small enough to excite graphene plasmons were successfully fabricated, and the experimental works [193], [195], [202], [203] that followed explored in great detail the nature of graphene plasmons. The metasurface configurations used include graphene nanoribbons, nanodisks, and nanorings (Figure 7B–D). As the theory predicts, optical confinement more than 100, as well as plasmonic resonance tunability of 0.1 eV, has been experimentally achieved with LSP in graphene nanoribbons. It is worth noting that the confinement is much stronger than that obtained with SPP in unpatterned graphene sheet (λ_air/λ_plasmon~40–60). The damping mechanisms of LSP in graphene nanoribbons via an interaction with graphene optical phonon and substrate phonon have also been investigated by several groups, revealing ~20 fs plasmon lifetime at energies above the optical phonon energy (~0.2 eV) of graphene [203]. Quite remarkably, resonance energies up to 335 meV (3.7 μm free-space wavelength) have been achieved in graphene nanorings [195]. However, it remains a significant challenge to further push the intrinsic graphene plasmonic resonance to near-IR and visible frequencies for telecom and detection applications. To accomplish this, nanostructures smaller than 5 nm and doping levels higher than 1 eV are required, and efforts are underway toward this goal. Theoretical calculations were also performed to support experimental results, where graphene was modeled either as a material with finite thickness (~0.3 nm) [193], [204] or as a surface current with zero thickness, which dramatically reduces computational cost [202]. Both numerical methods and analytical calculation [195], [204] achieved excellent agreement with experiments.

In addition to investigating the intrinsic graphene plasmons in graphene metasurface, researchers also used graphene to tune the plasmonic resonance in graphene/metal metasurface hybrid structures [209]. The idea is to make use of the tunable optical properties of graphene to dynamically change the environment dielectric function for the noble metal metasurfaces, therefore actively tuning their plasmonic resonances. It has been demonstrated that graphene can actively tune both the resonance linewidths [210] and frequencies [196], [211], [212], [213], [214] in the mid-IR in back-gated graphene devices. The dynamic tuning range achieved reaches as large as 1.1 μm with a modulation frequency of 30 MHz [196] (Figure 7E), and 100% modulation depth was also shown in a tunable metasurface mid-IR perfect absorber [211]. For graphene to tune plasmonic resonances in near-IR and visible frequencies, gating voltages beyond typical dielectric breakdowns are needed to supply adequate carrier densities; therefore, alternative techniques to dielectric back gating are required. Emani et al. [215] demonstrated the electrical tunability of Fano-type resonance with graphene at ~2.4 μm with ionic gel top gating.

Based on the developments in the graphene-plasmon related works mentioned above, a number of applications utilizing dynamically tunable graphene plasmons with superior performance to conventional plasmonic material platforms have been realized. The following highlights some of the efforts.

Biosensing: Rodrigo et al. [194] showed that graphene nanoribbons exhibit 6× enhancement in sensitivity compared to gold dipole antennas when used to detect protein bilayers and that the resonance in graphene nanoribbons can be electrically tuned to sense difference vibrational bands (Figure 7C).

Photodetection: Noble metal metasurfaces [66], [216], [217], [218], [219], [220], as well as graphene nanoribbons [221], [222] have been used to enhance the sensitivity of graphene-based photodetectors by many research groups, and typically, the sensitivity can be improved by an order of magnitude (Figure 7F), even as high as by 200 times [217].

Lasing: Chakraborty et al. [223] showed that graphene could be used to dynamically tune the emission spectrum of quantum cascade lasers when placed on a metallic waveguide with aperiodic slits.

Optical absorber/modulator: Jang et al. [224] demonstrated tunable optical absorption (0–25%) with graphene nanoribbons in a Salisbury screen structure; Kim et al. [225] experimentally showed 28.6% transmission modulation efficiency by coupling graphene nanoribbons to metallic extraordinary optical transmission structure, whereas 95.7% modulation efficiency was predicted by calculations; graphene has also been demonstrated to be capable of light modulation in plasmonic slot waveguides, which are promising for ultracompact optoelectronic devices [226], [227], [228].

Since the successful exfoliation of monolayer graphene, the discovery of other 2D materials has surged. Among these 2D materials, phosphorene (black phosphorus) and transition metal dichalcogenides (MoS₂, WS₂, WSe₂, etc.) have been shown to exhibit outstanding quantum emission properties and plasmonic response [229], [230], [231], [232], which make them promising candidates for metasurface applications. In addition, noble metal plasmonic structures were integrated with various 2D materials to enhance optical absorption and photon emission properties, as is in graphene/metal hybrid structures [233], [234]. To date, the relatively small sizes of exfoliated flakes of these 2D materials pose technological challenges to pattern them into metasurfaces; however, this field remains an active area of research because of the large wave confinement and dynamic tunability exhibited by the surface plasmons in such 2D materials [235], [236], [237], [238], [239], [240].

4.7 Phase transition/change materials for tunable metasurfaces

Next-generation metasurface systems and devices will require efficient, dynamic, and programmable functionality to realize complete and versatile control of electromagnetic radiation. Phase transition and phase change materials (PTMs and PCMs) are a promising and emerging class of photonic materials that offer outstanding optical modulation performance across a wide breadth of control inputs, such as thermal, electrical, and optical excitations [241]. Their exceptional properties are well studied and have been utilized for a variety of applications well before their introduction into photonics [242]. Broadly speaking, the distinction between PTMs and PCMs is whether the process reverses naturally: PTMs are naturally reversible or, equivalently, volatile and will return to their initial, unperturbed state after a period of time if left unaltered; in contrast, PCMs are irreversible, that is, nonvolatile, and will remain in a fixed state of matter unless an input excitation “resets” the PCM back to its original state. Both volatile and nonvolatile processes have advantages and disadvantages for photonic applications. In this section, we discuss and review two prominent materials that have shown great promise for tunable metasurfaces, namely the volatile PTM vanadium dioxide (VO₂) and the nonvolatile PCM germanium-antimony-tellurium (GST).

At room temperatures, vanadium dioxide is in its insulating phase with a ~0.7 eV bandgap and a monoclinic crystal structure. If the temperature of VO₂ increases past the insulator-metal-transition (IMT) temperature – typically 340 to 345 K – the crystal structure becomes tetragonal rutile and the resistance decreases by nearly four orders of magnitude. The colossal change in resistance is associated with a corresponding modification of VO₂’s dielectric permittivity ε(ω). Upon cooling, VO₂ will return to its insulating phase, but at a temperature lower than the IMT temperature, i.e. the cyclic transition exhibits hysteresis. In addition to thermal sources, switching between insulator and metal phases can be triggered with electrical, magnetic, or optical external perturbations. For optical-induced IMTs, the transition time can occur on the femtosecond time scale. These properties are what make vanadium dioxide such an attractive material platform for realizing tunable functionality in metasurfaces.

Initial studies of VO₂-based metasurfaces utilized arrays of split-ring resonators (SRRs) operating at IR frequencies. Driscoll et al. [243] patterned gold SRRs on a 90-nm film of VO₂ and monitored the resonance wavelength as a function of applied temperature. For temperatures higher than the IMT of 340 K, the sharp reduction in electric resistance of the VO₂ shorted the SRR’s gap and shifted the 1.6 THz resonance frequency by up to 20%. In the following year, this same group demonstrated the very first “memory MM” [244] and circumnavigated the challenge using VO₂ for nonvolatile applications (Figure 8A). In addition to using a similar SRR+VO₂ metasurface design, Driscoll et al. attached electrodes to the VO₂ film to apply sequential 1-s electrical pulses and operated their device at maximum hysteresis to obtain persistent tuning via memory-capacitance. The resonance shift of the metasurface device could then be tuned in discrete, controllable steps.

Figure 8:

Phase-transition materials for metasurface applications.

(A) Memory metasurface consisting of gold SRR on VO₂ films. Electronic contacts provide heat to raise the temperature of the VO₂ layer and induced a memory capacitance via a hysteretic phase transition. For larger temperatures, the resonance of the device (see inset) is red-shifted. (B) Ultrafast modulation of SRR on VO₂ substrate with terahertz electric fields. The large electric fields in the SRR gaps induce an insulator-metal transition in the underlying VO₂ substrate and actively shift the resonance of the metasurface (see inset). (C) VO₂ films selectively irradiated using ion beams to alter the local defect concentrations; consequentially, the irradiated areas exhibit a much lower insulator-to-metal phase transitions temperature, and the film will exhibit a subwavelength variation in refractive index when heated. By judiciously controlling the ion pattern, intrinsic metasurfaces, such as waveplates, can be designed. (D) Schematic of an active metasurface utilizing GST for dynamic beam steering. Switching between the amorphous (red) and crystalline (blue) phase of the underlying GST layer dynamically switches the resonance of the metasurface from resonance A to resonance B and changes the direction of the transmitted beam. (E) Reconfigurable photonic metasurface using a thin film of GST sandwiched between ZnS-SiO₂ and a femtosecond laser source for writing. The top image shows the reflection map of the dielectric metasurface written into the GST film. The bottom image shows the transmission and reflection for horizontally polarized light. (F) V-shaped antenna wrote into a 30-nm-thick amorphous GST layer using a nanosecond laser and measured using near-IR nanoscopy. The bottom left and the right images show the dielectric contrast of the GST after being written and the generation of surface phonon-polaritons, respectively. (A) Reprinted with permission from [244], copyright 2009 by the American Association for the Advancement of Science; (B) reprinted with permission from [245], copyright 2012 by Springer Nature; (C) reprinted with permission from [246], copyright 2015 by the American Chemical Society; (D) reprinted from [247] under the Creative Commons (Attribution-Noncommercial) license; (E) reprinted with permission from [248], copyright 2015 by Springer Nature; (F) reprinted with permission from [249], copyright 2017 by Springer Nature.

Following the demonstration of SRR+VO₂ metasurface, several works expanded upon this design by extending the operating frequencies into the near- and mid-IR [250], [251] and increasing the ON/OFF extinction ratio by orders of magnitude [252]. Temperature-induced phase transitions were still the predominant method of control; however, in 2012, the first all-optical modulated VO₂ metasurface was demonstrated using the large field enhancement in SRR gaps and picosecond THz pulses (Figure 8B) [245]. The IMT was attributed to generation Poole-Frenkel excitations via the extremely high electric fields within the VO₂ layer. In addition to optical tuning, voltage switching of a VO₂ metasurface was also accomplished by using ionic gels covering a VO₂+SRR metasurface [253]. In 2016, it was shown that the IMT temperature of VO₂ could be selectively and locally controlled via ion beam irradiation (Figure 8C) [246].

The second class of materials that have generated a considerable degree of excitement for tunable metasurface devices is GST. Unlike VO₂, GST is a phase change material (i.e. nonvolatile), which exhibits a stable crystalline-to-amorphous phase change for temperatures above the melting point (~600 K) and an amorphous-to-crystalline change for temperatures above the crystallization temperature (~160 K). Each phase will persist until an external source – electric, optical, thermal – “resets” the material; in this way, GST is a natural memory material and does not rely on hysteresis effects like VO₂. The amorphous and crystalline phases have excellent optical contrast throughout the IR wavelengths, which make them ideal as a tunable material platform for MMs.

GST’s potential as a memory material has been known since the 1980s, when it was first considered for energy storage devices; however, it was not until three decades later that the first GST MM was demonstrated. In 2013, Gholipour et al. designed and fabricated a multilayer metasurface composed of a sandwiched GST film between two buffer layers of ZnS/SiO₂ on top of a gold metasurface. By switching from amorphous-to-crystalline states using a 660-nm short- and high-intensity pulse and crystalline-to-amorphous states using a longer and lower-intensity pulse, the authors were able to switch the resonance of the metasurface antennas and modulate the near- and mid-IR reflection by as much as 400% (Figure 8D) [254]. The metasurface was able to endure 50 amorphous-to-crystalline cycles before showing degradation. In the same year, Chen et al. [255] patterned gold nanodisks directly on a 20-nm-thick GST layer and demonstrated nonvolatile tuning across a 1.89–2.27 μm range via thermal heating. In addition to metallic metasurface designs, GST has been incorporated in all-dielectric metasurfaces for functional tuning at visible and near-IR wavelengths [256]. Here, Karvounis et al. patterned subwavelength-thick GST films into a grating metasurface and demonstrated switching contrast ratios of 7 dB via a 532 nm laser excitation.

The application of the high-contrast change in optical properties of GST was efficiently utilized in the mid-IR spectrum, where losses are exceptionally low, by Michel et al. who demonstrated effective tuning of aluminum nanoantennas directly fabricated on GST. Here, the authors operated at 3000 cm⁻¹, where the relative change in the real portion of permittivity (ε₁) between amorphous (a) and crystalline states (c) was ∆ε_1a→_c=27.2, which corresponded with an antenna resonance shift of nearly 20% [257]. Tittl et al. [258] used a similar strategy for controlling a metal antenna metasurface but instead used GST as a spacer layer between an aluminum backing layer and an array of aluminum square antennas. Their design performed as a temperature-dependent switchable plasmonic perfect absorber from 3 to 5 μm with multispectral thermal imaging capabilities. Just recently in 2015, the same group demonstrated a tunable and switchable chiral metasurface with a mid-IR circular dichroism response from 4.15–4.90 μm [259]. Furthermore, by applying a bias layer to their design, the authors were able to flip the circular dichroism signal’s sign using the large tunability of GST.

Of GST’s remarkable optical properties, the ability to locally control the phase of the material is perhaps its most novel. In 2015, Wang et al. [248] demonstrated an all-optical reconfigurable dielectric metasurface based on a 70-nm-thick layer of GST sandwiched between ZnS/SiO₂ layers capable of generating nearly arbitrary reflection and transmission responses at near-IR frequencies. The metasurface was “written” using a femtosecond laser for generating the amorphous-to-crystalline phase change, a spatial light modulator for controlling the laser intensity pattern, and a microscope to focus this pattern down onto the GST layer (Figure 8E). In addition to a metasurface, a number of other planar optical components were demonstrated, including a visible bichromatic, multifocal Fresnel lens, a greyscale hologram, and a super-oscillatory lens. A similar and promising approach for generating mid-IR reconfigurable metasurfaces was recently demonstrated using crystalline GST as a phonon-polariton material [249]. Here, it was shown that micron-sized patterns of crystalline GST could be written using a nanosecond 660 nm laser and erased using the same laser, albeit with a longer pulse duration. The crystalline GST patterns could then support highly confined surface phonon-polariton modes (Figure 8F).

The impressive tunablility of GST and VO₂ has driven the exploration and discovery of other phase change/transition materials. Recently, a class of phase change materials within the perovskite family has emerged as promising alternative materials for tunable meta-devices. Strongly correlated samarium nickelate (SmNiO₃) is known to exhibit metal-insulator phase transitions at temperatures of approximately 140°C, similar to VO₂ and GST [260]. However, SmNiO₃ exhibits a second phase transition via electron doping, which can drastically change the optical properties of the film across a vast spectral range (0.4–17 μm). The electron doping effectively increases the bandgap of SmNiO₃ and is accomplished through either gas phase, liquid phase, or solid-state dopant injection. Li et al. [261] utilized this remarkably broadband tuning in their demonstration of several platinum on SmNiO₃ plasmonic metasurface structures using a variety of control inputs, including lithium intercalation, thermal annealing, and electrical biasing. Their finding suggests that SmNiO₃ is an ideal candidate for modulating absorption at mid-IR frequencies and holds great promise for thermal applications.

As a closing remark, it should be noted that there exists another class of materials that exhibit a metal-to-insulator phase change when subjected to gaseous chemical reagents, namely hydrogen. Since the conceptual emergence of hydrogen as an energy source, a large body of research has concentrated on effectively storing and sensing hydrogen gas. Palladium is well known to absorb substantial concentrations of hydrogen within its atomic lattice and sequentially transition into the dielectric phase known as palladium hydride; the phase transition is fully reversible upon removal of the hydrogen environment [262]. Additionally, palladium exhibits an LSP resonance throughout the ultraviolet, visible, and near-IR spectrum [263], [264]. These qualities have made palladium an attractive option for tunable metasurface devices, particularly ones that are functionalized for hydrogen sensing applications. In 2011, Tittl et al. [265] demonstrated a perfect absorbing metasurface structure composed of palladium strips optimized to detect changes in hydrogen concentrations as low as 0.5% in a matter of seconds. Similar hydrogen-induced phase transitions have been observed in other metals, such as yttrium and magnesium [266], [267], [268].