Metasurface and metalens

A conventional lens shapes a wavefront by varying the thickness of glass, accumulating more phase where the glass is thicker. A metasurface imposes the phase abruptly instead, with a single layer of nanostructures, pillars, fins or holes a few hundred nanometers across and spaced below the wavelength, each acting as a small waveguide or resonator that delays the light passing through it by an amount set by its size or orientation. Arranging the elements so that the local delay varies across the surface produces any desired phase profile in one lithographic step. Because the elements are smaller than the wavelength, the surface does not diffract into unwanted orders the way a coarse grating does; it behaves as a continuous phase plate. Rotating anisotropic elements gives a phase that depends on circular polarization, and structuring them differently along two axes controls polarization directly.

A metalens applies the hyperbolic phase profile that focuses a plane wave to a point, $\phi(r) = -(2\pi/\lambda)\left(\sqrt{r^2 + f^2} - f\right)$, wrapped modulo $2\pi$ across the surface. Wrapping is what makes the lens flat, and it is also what makes the lens chromatic in the way of a diffractive optic rather than a refractive one: the focal length varies roughly inversely with wavelength, so a 5% change in wavelength shifts the focus by about 5%, many times the chromatic error of a glass lens. Dispersion engineering of the elements can correct this over limited bandwidths and apertures. High numerical apertures are reachable: a lens 2 mm across with a 1 mm focal length has an NA of 0.71.

Efficiency is the other constraint to state plainly. Each element must supply its phase with high transmission, and the phase must be sampled finely enough across the aperture; at high NA the local period of the $2\pi$ phase zones near the edge shrinks toward the element spacing, so each zone is sampled by only a few elements, and efficiency falls. Reported focusing efficiencies for visible metalenses are typically well below those of a refractive lens, and light not focused goes into background that lowers image contrast. Large apertures also require very large numbers of elements, patterned with the lithography of semiconductor fabs.

Where a flat optic helps, metasurfaces are in production: in compact sensing modules such as the projectors and receivers of depth sensors, where the wavelength is a single laser line and chromatic error does not matter, and where integration with the sensor wafer saves assembly. They are also studied for polarization cameras, beam shaping, augmented-reality waveguide couplers, and on-chip optics. For broadband imaging with high efficiency, refractive lenses remain the better choice.

References: N. Yu et al., Science 334, 333 (2011); M. Khorasaninejad et al., Science 352, 1190 (2016).