Metallic Metasurface Microcavities

• Metallic metasurface microcavities are subwavelength resonant structures defined by patterned metallic boundaries that tailor electromagnetic modes via geometry-driven confinement and radiation leakage.
• These structures span various architectures—from continuous-metal groove cavities and metal–insulator–metal absorbers to programmable active designs—each offering distinct spectral tunability and design trade-offs.
• Their applications include broadband absorbers, high-Q resonators, and dynamic optical elements, enabling advances in energy harvesting, sensing, and communication technologies.

Metallic metasurface microcavities are subwavelength resonant structures in which a patterned metallic boundary, often combined with a dielectric spacer, a continuous back reflector, or a perforated conducting film, confines electromagnetic energy in wavelength-scale or deeply subwavelength volumes and thereby tailors reflection, absorption, transmission, scattering, or intra-cavity field profiles. Reported realizations span continuous-metal groove cavities in single-crystal Au(111), metal–insulator–metal and metal–dielectric–metal absorbers, coaxial and slit-aperture arrays, anisotropic slit superlattices supporting Fano resonances and bound states in the continuum, active Huygens boundaries, chaotic-cavity-backed programmable surfaces, and electrically tunable plasmonic reflectarrays (Kiyumbi, 10 Mar 2026, Azad et al., 2015, Todorov et al., 2012, Díaz-Rubio et al., 2014, Lee et al., 2020, Wong et al., 2018, Faul et al., 2024, Mayoral-Astorga et al., 2024). Across these platforms, the defining feature is not a single geometry but the use of a metallic metasurface as the effective cavity boundary that sets mode confinement, radiation leakage, impedance matching, and spectral selectivity.

1. Architectural classes and defining characteristics

A first major class is the continuous-metal cavity metasurface. In this formulation, the cavity is carved directly into an optically thick metal rather than assembled from discrete antennas or multilayer stacks. The trifolium nanocavity arrays milled into single-crystal Au(111) microplates are an explicit example: three elongated V-groove lobes converge at a central junction, the cavities are arranged in square arrays of $15 \times 15$ with pitch $p \approx 1.35\,\mu\text{m}$, and the structured Au surfaces exhibit broad reflection bands and pronounced minima across the visible–near-infrared. The single-crystal Au(111) platform is used because it reduces grain-boundary scattering, sidewall roughness, and extrinsic damping relative to polycrystalline films (Kiyumbi, 10 Mar 2026).

A second major class is the metal-backed gap or slab microcavity. In metal–insulator–metal absorbers, patterned top metal resonators, a thin dielectric spacer, and a thick metallic ground plane form a cavity whose transmission is negligible and whose response is determined by destructive reflection and impedance matching. The broadband solar absorber based on a super-cell of sixteen Au resonant elements above a $60$ nm $\text{SiO}_2$ spacer and a $200$ nm Au ground plane is representative of this class. Closely related metal–dielectric–metal THz cavities employ a patterned Au top surface over a thin GaAs slab and a bottom Au mirror, so that TM-polarized energy is concentrated in the subwavelength metal–metal region (Azad et al., 2015, Todorov et al., 2012).

A third class uses apertures or perforations as the cavity itself. Square arrays of air-filled coaxial cavities covered by a thin FR4 dielectric sheet realize total absorption for p-polarized microwaves through a cavity-array impedance match. One-dimensional metallic slit arrays and slit superlattices in perfect-electric-conductor films support localized slit resonances, guided Bloch bands, quasi-guided leaky modes, and symmetry-protected BICs, depending on whether the structure consists of a single slit per period or multiple slits in a supercell (Díaz-Rubio et al., 2014, Deng et al., 2017, Lee et al., 2020).

A fourth class is explicitly programmable or active. In the Huygens’ box, a metallic enclosure populated with active Huygens meta-atoms synthesizes arbitrary intra-cavity fields by imposing electric and magnetic surface currents on the cavity boundary. In the chaotic-cavity-backed non-local programmable metasurface, a quasi-2D D-shaped metallic cavity with $N=14$ mechanically tunable shafts acts as a reverberant microcavity whose overlapping modes couple all meta-elements and ports non-locally. Electrically tunable plasmonic MOS metasurfaces introduce yet another variant, in which the metallic cavity is preserved but the local optical constants are tuned through carrier refraction in ITO (Wong et al., 2018, Faul et al., 2024, Mayoral-Astorga et al., 2024).

2. Resonance physics and analytical frameworks

The underlying physics is architecture dependent, but several analytical descriptions recur. In plasmonic groove cavities and metal-backed nanogaps, surface plasmon polaritons at a flat metal–dielectric interface follow

$$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$

with $k_0=\omega/c$. In V-grooves, gap-surface-plasmon confinement is frequently parameterized by an effective index $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$, and a useful approximation for moderately wide grooves is

$$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$

The long-wavelength resonance of the trifolium cavity is interpreted as a hybridized cavity resonance of such confined modes, with a Fabry–Pérot-like condition

$p \approx 1.35\,\mu\text{m}$0

and, in a simplified form,

$p \approx 1.35\,\mu\text{m}$1

Increasing groove depth therefore increases the optical path and redshifts the resonance (Kiyumbi, 10 Mar 2026).

For MIM absorbers, the central description is destructive reflection under blocked transmission. With $p \approx 1.35\,\mu\text{m}$2, one has

$p \approx 1.35\,\mu\text{m}$3

and maximum absorption occurs when the effective surface impedance matches free space,

$p \approx 1.35\,\mu\text{m}$4

In cavity language, the same structure satisfies

$p \approx 1.35\,\mu\text{m}$5

with $p \approx 1.35\,\mu\text{m}$6. This framework is used explicitly for broadband Au/$p \approx 1.35\,\mu\text{m}$7/Au absorbers in the visible–near-infrared (Azad et al., 2015).

In THz metal–dielectric–metal stripe microcavities, the patterned metal behaves as a short double-metal waveguide. The localized standing-wave resonances are described by

$p \approx 1.35\,\mu\text{m}$8

where $p \approx 1.35\,\mu\text{m}$9 is the mode order, $60$0 the stripe width, and $60$1 an effective modal index including end-reflection phase correction. For 2D patches, the corresponding relation is

$60$2

In the coaxial cavity array, the mono-mode resonance condition is written as

$60$3

showing directly that cavity length $60$4 is the primary frequency-tuning parameter (Todorov et al., 2012, Díaz-Rubio et al., 2014).

Perforated PEC slit superlattices introduce yet another regime. There, guided and quasi-guided Bloch modes are phase matched to free space through

$60$5

and their interference with a nonresonant background produces Fano line shapes,

$60$6

At symmetry-protected points, the radiative channel can vanish altogether, yielding a BIC embedded in the continuum (Lee et al., 2020).

3. Passive metallic microcavities as absorbers, reflectors, and converters

A large body of work treats metallic metasurface microcavities as absorptive structures. The broadband solar absorber based on eight pairs of Au nanoresonators above a thin $60$7 spacer and Au ground plane demonstrates $60$8 absorptance approximately in $60$9 in simulation, and absorptance $\text{SiO}_2$0 in $\text{SiO}_2$1 experimentally at $\text{SiO}_2$2 incidence, while measured absorptance becomes $\text{SiO}_2$3 for $\text{SiO}_2$4 and negligible $\text{SiO}_2$5 above $\text{SiO}_2$6. The resonances are intentionally overlapped, so the device operates in a low-$\text{SiO}_2$7 regime with broad $\text{SiO}_2$8 rather than as a narrowband cavity (Azad et al., 2015).

Other absorptive implementations are much more spectrally selective. The coaxial-cavity-on-metal platform covered by FR4 exhibits a low-frequency peak with near-total absorption for p-polarized waves, with the resonance shifting from $\text{SiO}_2$9 GHz at $200$0 mm to $200$1 GHz at $200$2 mm and $200$3 GHz at $200$4 mm for $200$5 mm and $200$6. Increasing dielectric thickness from $200$7 mm to $200$8 mm lowers the resonance from $200$9 GHz to $N=14$0 GHz at fixed $N=14$1 mm. A defining feature is that the peak frequency is essentially independent of $N=14$2 from $N=14$3 to $N=14$4, whereas the absorption amplitude peaks near $N=14$5 and collapses near normal incidence because the impinging wave cannot efficiently excite the coaxial resonant mode without a lateral phase variation across the array (Díaz-Rubio et al., 2014).

THz metal–dielectric–metal stripe and patch cavities demonstrate a related far-field-to-near-field conversion. When the dielectric thickness is very small compared with wavelength, the structure supports strongly localized modes concentrated in the subwavelength metal–metal regions; numerics and experiment show that absorption can approach $N=14$6 for 2D patches around $N=14$7 and $N=14$8, with coupling mediated by evanescent diffraction harmonics rather than by a propagating higher-order diffraction spectrum (Todorov et al., 2012).

Broadband absorption can also be obtained through continuous self-similarity rather than through a discrete superposition of resonators. The logarithmic spiral metasurface, backed by a metallic surface, combines a no-cutoff fundamental TM mode in a coiled tapered channel with the scale-invariant geometry $N=14$9. In the optimized graded-index case, the device absorbs $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$0 of incident microwave energy from $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$1 GHz to $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$2 GHz and lies within $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$3 mm of the Rozanov thickness–bandwidth limit; the 2D orthogonal version remains above $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$4 from $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$5 GHz to $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$6 GHz for both TM and TE polarizations (Wang et al., 2018).

4. Symmetry breaking, channel selection, and spectral tunability

Symmetry is one of the most consequential control parameters in metallic metasurface microcavities. In the trifolium Au(111) nanocavity, the three-lobed profile breaks full rotational symmetry and introduces preferred in-plane directions for mode excitation. This produces a measurable azimuth-dependent response under sample rotation: for groove depths $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$7 nm and $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$8 nm, the dominant long-wavelength reflection minimum shifts from $$k_{\mathrm{spp}}(\omega)=k_0\sqrt{\frac{\varepsilon_m(\omega)\varepsilon_d}{\varepsilon_m(\omega)+\varepsilon_d}},$$9 nm to $k_0=\omega/c$0 nm, a net $k_0=\omega/c$1 nm for $k_0=\omega/c$2 nm, corresponding to $k_0=\omega/c$3, while azimuthal rotation induces spectral displacements on the order of $k_0=\omega/c$4–$k_0=\omega/c$5 nm in the $k_0=\omega/c$6–$k_0=\omega/c$7 nm band. This explicitly contrasts with the azimuthally invariant behaviour often associated with circular groove cavities (Kiyumbi, 10 Mar 2026).

Angle can function as an equally direct channel selector. In the asymmetric slit-array metasurface with $k_0=\omega/c$8, the same localized slit microcavity resonance decays predominantly into the $k_0=\omega/c$9th transmission order at small incident angles, yielding extraordinary optical transmission, or into the $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$0st reflection order at larger angles, yielding extraordinary optical diffraction. For the representative case $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$1, $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$2, $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$3, and $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$4, the transition proceeds through three regimes: $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$5 for the EOT zone, approximately $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$6 for the total-internal-reflection mirror zone, and $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$7 for the EOD zone, with Littrow retroreflection at $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$8. The analytical treatment uses a PEC model and therefore attributes the response to the slit cavity resonance rather than to grating-assisted SPP excitation (Deng et al., 2017).

Slit superlattices show how additional symmetry and supercell degrees of freedom generate a richer modal spectrum. A superlattice with $\eta_{\mathrm{eff}}=k_{\mathrm{gsp}}/k_0$9 slits per period supports $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$0 Bloch bands $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$1 and $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$2 band gaps $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$3; among these, higher-order $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$4 bands can enter the radiation continuum and produce Fano resonances, while mirror symmetry at $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$5 can protect a band-edge mode from radiation, forming a BIC. For the $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$6 example, finite-element simulations report $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$7 at the symmetry-protected BIC and $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$8 for small $$k_{\mathrm{gsp}} \approx k_0 \sqrt{\varepsilon_d + \frac{2 \varepsilon_d \sqrt{\varepsilon_d-\varepsilon_m}}{k_0 w(-\varepsilon_m)}}.$$9, while leaky $p \approx 1.35\,\mu\text{m}$00 modes in the continuum have $p \approx 1.35\,\mu\text{m}$01. The resulting picture makes clear that metallic microcavities can be either low-$p \approx 1.35\,\mu\text{m}$02 absorbers or extremely high-$p \approx 1.35\,\mu\text{m}$03 quasi-BIC resonators, depending on symmetry and radiation leakage (Lee et al., 2020).

At visible and near-infrared wavelengths, strong coupling provides another route to spectral restructuring. In the metallo-dielectric hybrid metasurface composed of Si nanodisks above an Al film separated by $p \approx 1.35\,\mu\text{m}$04, the nanodisk array supports an anapole state while the lattice period phase matches an SPP at the Al–$p \approx 1.35\,\mu\text{m}$05 interface. When the spacer is near $p \approx 1.35\,\mu\text{m}$06 nm and the two resonances are tuned to zero detuning, the reflectance exhibits an avoided crossing with vacuum Rabi splitting up to $p \approx 1.35\,\mu\text{m}$07 meV; the diameter sweep yields $p \approx 1.35\,\mu\text{m}$08 meV, $p \approx 1.35\,\mu\text{m}$09 meV, $p \approx 1.35\,\mu\text{m}$10 meV, and $p \approx 1.35\,\mu\text{m}$11 meV (Ravishankar et al., 2021).

5. Programmable and active metallic microcavities

Passive microcavity operation is only one branch of the field. The active Huygens’ box reformulates the microcavity boundary as a programmable source surface. On a closed surface $p \approx 1.35\,\mu\text{m}$12 with outward normal $p \approx 1.35\,\mu\text{m}$13, the tangential jumps obey

$p \approx 1.35\,\mu\text{m}$14

For interior-only synthesis, one sets $p \approx 1.35\,\mu\text{m}$15, giving $p \approx 1.35\,\mu\text{m}$16 and $p \approx 1.35\,\mu\text{m}$17. Implemented in a parallel-plate metallic cavity at $p \approx 1.35\,\mu\text{m}$18 GHz with $p \approx 1.35\,\mu\text{m}$19 independent channels, this boundary was used to generate traveling plane waves at $p \approx 1.35\,\mu\text{m}$20, $p \approx 1.35\,\mu\text{m}$21, and $p \approx 1.35\,\mu\text{m}$22, a standing plane wave, a cylindrical Bessel $p \approx 1.35\,\mu\text{m}$23 profile, and a subwavelength superoscillation focal spot without evanescent components; experiments indicate faithful synthesis over an $p \approx 1.35\,\mu\text{m}$24 bandwidth limited by components (Wong et al., 2018).

The chaotic-cavity-backed non-local programmable metasurface is programmable in a different sense: the boundary remains passive during operation, but its geometry is reconfigured in situ. The device uses a compact quasi-2D D-shaped metallic cavity of area $p \approx 1.35\,\mu\text{m}$25, height $p \approx 1.35\,\mu\text{m}$26 cm, three ports, and $p \approx 1.35\,\mu\text{m}$27 mechanically tunable metallic shafts forming the top metasurface. Over $p \approx 1.35\,\mu\text{m}$28 random configurations, transmission magnitudes $p \approx 1.35\,\mu\text{m}$29, $p \approx 1.35\,\mu\text{m}$30, and $p \approx 1.35\,\mu\text{m}$31 reach maxima $p \approx 1.35\,\mu\text{m}$32 at $p \approx 1.35\,\mu\text{m}$33 GHz and $p \approx 1.35\,\mu\text{m}$34 near $p \approx 1.35\,\mu\text{m}$35 GHz, while minima approach $p \approx 1.35\,\mu\text{m}$36 throughout $p \approx 1.35\,\mu\text{m}$37–$p \approx 1.35\,\mu\text{m}$38 GHz; the composite $p \approx 1.35\,\mu\text{m}$39 rises from $p \approx 1.35\,\mu\text{m}$40 at $p \approx 1.35\,\mu\text{m}$41 GHz to $p \approx 1.35\,\mu\text{m}$42 at $p \approx 1.35\,\mu\text{m}$43 GHz. Closed-loop optimization then imposes reflectionless scattering modes, transmissionless scattering modes, approximate transmissionless exceptional points near $p \approx 1.35\,\mu\text{m}$44 and $p \approx 1.35\,\mu\text{m}$45 GHz, low-loss routing with undesired outputs suppressed by at least $p \approx 1.35\,\mu\text{m}$46 dB and desired transmission attenuated by at most $p \approx 1.35\,\mu\text{m}$47 dB, and reprogrammable multi-band filters with reject bands below $p \approx 1.35\,\mu\text{m}$48 dB and pass bands above $p \approx 1.35\,\mu\text{m}$49 dB (Faul et al., 2024).

Electrical tunability at optical frequencies is illustrated by the gold-dipole MOS reflectarray. Each subwavelength Au dipole sits above conformal HfO$p \approx 1.35\,\mu\text{m}$50, ITO, and a backside metal mirror, so the nanoantenna and mirror simultaneously form a MOS capacitor and a reflective MIM microcavity. Biasing from $p \approx 1.35\,\mu\text{m}$51 V to $p \approx 1.35\,\mu\text{m}$52 V changes the carrier density in a $p \approx 1.35\,\mu\text{m}$53 nm perturbed ITO layer and yields a measured reflection phase excursion of $p \approx 1.35\,\mu\text{m}$54 at $p \approx 1.35\,\mu\text{m}$55 nm, with experimental $p \approx 1.35\,\mu\text{m}$56 on average and $p \approx 1.35\,\mu\text{m}$57 flatness at $p \approx 1.35\,\mu\text{m}$58 nm, plus the absence of secondary lobes because the pixel periodicities satisfy $p \approx 1.35\,\mu\text{m}$59 in glass. The measured time constants are $p \approx 1.35\,\mu\text{m}$60–$p \approx 1.35\,\mu\text{m}$61 ns, implying a maximum operation frequency of $p \approx 1.35\,\mu\text{m}$62 MHz (Mayoral-Astorga et al., 2024).

6. Applications, limitations, and design trade-offs

The application space is broad because different cavity regimes emphasize different observables. Continuous-metal groove cavities on Au(111) directly support reflective structural colour, compact colour filtering, frequency-selective reflective surfaces, and optical-variable anti-counterfeiting features, with depth and azimuth providing two independent optical variables (Kiyumbi, 10 Mar 2026). Broadband MIM absorbers are aligned with solar thermophotovoltaics, where the combination of $p \approx 1.35\,\mu\text{m}$63 absorptance over most of the solar band and $p \approx 1.35\,\mu\text{m}$64 absorptance above $p \approx 1.35\,\mu\text{m}$65 is advantageous for spectral selectivity (Azad et al., 2015). Coaxial, THz MDM, and logarithmic-spiral cavities are relevant to sensing, EMI/RFI shielding, stealth or RCS reduction, and compact absorbers; slit superlattices add high-$p \approx 1.35\,\mu\text{m}$66 filtering, angle-tunable spectral control, enhanced nonlinear or THz generation, and quasi-BIC lasing concepts (Díaz-Rubio et al., 2014, Todorov et al., 2012, Wang et al., 2018, Lee et al., 2020). Active and programmable metallic microcavities extend this to imaging, communication, medical therapy, cognitive radio, anti-jamming filtering, programmable routing, and analog signal processing (Wong et al., 2018, Faul et al., 2024).

The field is equally defined by trade-offs. Reported devices range from deliberately low-$p \approx 1.35\,\mu\text{m}$67 broadband absorbers to ultrahigh-$p \approx 1.35\,\mu\text{m}$68 symmetry-protected states, so $p \approx 1.35\,\mu\text{m}$69 is architecture dependent rather than a universal performance metric. Metals broaden resonances through ohmic loss; gold’s intrinsic loss broadens optical resonances, and in the trifolium platform FIB milling imposes tolerances on width and sidewall taper and can introduce surface damage (Kiyumbi, 10 Mar 2026). Polarization dependence is also common: the coaxial absorber requires p-polarized oblique incidence, the slit EOT/EOD platform operates under TM illumination, and TE coupling is weak in PEC slit superlattices (Díaz-Rubio et al., 2014, Deng et al., 2017, Lee et al., 2020).

Visible-wavelength cavity benchmarks show the cost of metallic loss particularly clearly. In the holographic microcavity with $p \approx 1.35\,\mu\text{m}$70 nm Au mirrors at $p \approx 1.35\,\mu\text{m}$71 nm and $p \approx 1.35\,\mu\text{m}$72, the measured linewidth is $p \approx 1.35\,\mu\text{m}$73 nm, the quality factor is $p \approx 1.35\,\mu\text{m}$74, the free spectral range is $p \approx 1.35\,\mu\text{m}$75 nm, the finesse is $p \approx 1.35\,\mu\text{m}$76, and the measured round-trip loss is $p \approx 1.35\,\mu\text{m}$77. Simulations replacing the metallic mirrors with $p \approx 1.35\,\mu\text{m}$78-pair TiO$p \approx 1.35\,\mu\text{m}$79/SiO$p \approx 1.35\,\mu\text{m}$80 DBRs narrow the linewidth to $p \approx 1.35\,\mu\text{m}$81 nm and raise $p \approx 1.35\,\mu\text{m}$82 to $p \approx 1.35\,\mu\text{m}$83, while also improving image fidelity. This does not negate metallic microcavities; rather, it establishes a quantitative boundary between metallic compactness and absorber-free high-$p \approx 1.35\,\mu\text{m}$84 performance at visible wavelengths (Mason et al., 2023).

Taken together, the literature shows that metallic metasurface microcavities are not a single device family but a unifying framework for confining and sculpting electromagnetic fields with patterned metallic boundaries. Depending on geometry, symmetry, and the balance between radiation leakage and internal loss, the same overarching concept yields broadband absorbers, narrowband reflectors, angle-switched diffraction devices, quasi-BIC resonators, programmable transfer-function synthesizers, active wave-generating cavities, and electrically tunable phased arrays.