Diffractive Nonlocal Metasurfaces

Abstract

Metasurfaces are ushering in an era of multifunctional control over optical wavefronts realized with ultrathin planarized devices. Recent advances have been enabling unprecedented control over the frequency response of these surfaces, suggesting that the future of flat optics may tailor both spectral and spatial degrees of freedom in highly multispectral and multifunctional devices. Diffractive nonlocal metasurfaces are opening new opportunities in this direction: they leverage symmetry-protected scattering from quasi-bound states in the continuum and, by spatially manipulating controlled geometric perturbations, they support ultrasharp optical responses with wavefront-manipulating features. Encoded in nonlocal (i.e., spatially extended) resonant modes, the resulting response is observed exclusively within the bandwidth of the resulting Fano resonance, affording ideal features for a wide range of applications. In this perspective, this novel class of metasurfaces are discussed in the broader context of flat optics, highlighting their peculiar operation in contrast to relevant predecessors, and highlighting the opportunities for future advancement and applications. In particular, it is emphasized that nonlocality and selectivity are inherently related, but that spectral and spatial selectivity can be independently tuned in suitably tailored metasurfaces. In turn, this freedom allows the design and implementation of both wavefront-shaping and wavefront-selective devices. The novel optical responses, combined with the compatibility with rational design, herald new prospects for active, nonlinear and quantum metasurfaces, ultrathin devices for augmented reality, and compact tailored optical sources.

Why optics needs thickness

“…In space-invariant optics (8), where the behavior depends only on the relative separation of input and output points, we are then convolving with a fixed kernel. Then, D simplifies (8) to…”

“…With coordinates x and y on the input face and u and v on the output face (Fig. 1), as in the formalism of ( 8 ), generallyΦu,v=∬Du,v;x,yΨx,ydxdywhere Du,v;x,y is the kernel or the device operator ( 15 , 16 ), relating the output function Φu,v to the input function Ψx,y.…”

“…In space-invariant optics ( 8 ), where the behavior depends only on the relative separation of input and output points, we are then convolving with a fixed kernel. Then, D simplifies ( 8 ) toDu,v;x,y→Dx−v,y−vBecause the absolute position no longer matters, we simply choose one specific position for the calculations—e.g., for the output, such as u = 0, v = 0—and evaluate the matrices as required.…”

“…First, can we shrink the distance between a lens and the output plane in an imager—i.e., “squeezing space” ( 1 ), possibly with a spaceplate ( 2 – 5 )? Second, what kinds of mathematical operations could we perform—for example, on an image—using some metasurface structure with some thickness ( 6 – 8 )? This approach gives meaningful answers to these questions and others.…”

Emerging Planar Nanostructures Involving Both Local and Nonlocal Modes

“…One direct solution is seeking compensation. For example, the rippled PB phases that result from the neighboring couplings can be compensated by resonant phases through tiny geometric adjustments of meta-atoms. − , However, this approach still focuses on the local construction of metasurfaces and relies on time-consuming simulations to build structural libraries of meta-atoms, especially for nonlinear metasurfaces. The high demands on the nanostructure geometry limit its efficiency and practical applications as well.…”