Metasurfaces are subwavelength arrays of engineered resonators that impose amplitude, phase and polarisation control on a wave in a layer thinner than the wavelength, and the 2026 literature shows the field splitting into two active branches: resonance-engineered optical components (sensing, absorption, nonlinear conversion, analogue image processing) and programmable radio-frequency surfaces proposed for 6G MIMO.
Summary
A metasurface is a planar array of scatterers whose spacing and size are below the operating wavelength, so the array does not diffract in the usual sense but instead acts as an engineered boundary condition that rewrites the amplitude, phase and polarisation of transmitted or reflected waves. “Flat optics” is the application framing: replacing curved, bulk elements with patterned films. The physical machinery is resonance. Dielectric elements use Mie-type resonances, metallic ones use plasmonic modes, and both can be pushed into interference-driven regimes such as bound states in the continuum, lattice resonances and anapoles. A recent perspective argues these platform-specific vocabularies (photonic crystals, plasmonics, dielectric metasurfaces) describe the same underlying object, the eigenmodes of an open system, seen through different scattering features.
The parameters that decide a metasurface are quality factor, loss, field overlap and fabrication tolerance, and they trade against each other. High Q concentrates the field and sharpens the spectral response but is fragile to dissipation and geometric error: an all-metallic design that balances a Fabry-Perot BIC dark mode against a Rayleigh-anomaly lattice resonance reached a measured Q of 2180 (2800 in theory) with nearly 99% absorption, and held that behaviour across pillar heights of 70 to 120 nm and radii of 210 to 280 nm. The opposite strategy also works: deliberately low-Q all-dielectric surfaces with engineered evanescent modes can beat the sensing performance normally expected of low-Q structures by maximising near-field overlap with the analyte. Quasi-BIC resonances are usually created by breaking geometric symmetry; one 2026 result instead breaks it with material anisotropy, inserting a 20 nm BeS layer with permittivity anisotropy of about 0.27 into a symmetric TiO2 nanobar pair to obtain Q of roughly 181 while leaving the geometry untouched, and uses the resulting transfer function for optical edge detection.
Three other threads matter. Tunability: phase-change chalcogenides offer refractive index changes above 0.7 on switching, and a two-photon printing route now writes them as freeform 3D structures rather than patterned thin films; hydrostatic pressure up to 5 GPa shifts a silicon nitride Fabry-Perot resonance by up to 30 nm and changes index by up to 4%, enabling a polarisation-converting surface intended for extreme environments. Active and nonlinear function: a thin-film lithium niobate metasurface performs wavelength-selective second-harmonic wavefront shaping, converting a Gaussian pump near 1100 nm into a first-order Hermite-Gaussian mode at 550 nm, and Er-implanted Si3N4 nanocylinder arrays enhance room-temperature telecom-band photoluminescence by a factor of about 18 with a tenfold lifetime reduction via the Purcell effect, on a CMOS-compatible platform. Design tooling: differentiable multiple-scattering solvers now return exact gradients with respect to geometry and position for both finite clusters and infinite metasurfaces, motivated explicitly by sub-10 nm fabrication expanding the design space, and simplified multipolar models reduce design complexity by exploiting the origin dependence of multipole moments.
Evidence base
Open questions
Assessment drafted 2026-08-31 from up to 18 KB sources using the technology-scorecard framework; scores are a draft read pending review.