Resonant Nanophotonics

Elementary emitters of photons, like molecules, quantum dots and color centers, are by themselves deeply subwavelength in size. As consequence individual quantum emitter are not very bright, and their emission is not very directional. In applications like LEDs where many emitters are distributed in an extended device, a main challenge is to make emission brighter, more efficient and directional. A rich playground to enhance emission and impart spatial coherence is through diffractive metasurfaces

The invention of the blue light-emitting diodes (LEDs) enabled the solid-state lighting (SSL) revolution. SSL typically achieves white light via phosphor-based down-conversion, where blue or UV LEDs excite phosphors that re-emit at longer wavelengths. In remote phosphor systems, nanophotonic metasurfaces leverage waveguide local density of optical states (LDOS) to induce spatial coherence and direct spontaneous emission into specific modes, which are then outcoupled via engineered periodic arrays.

In projects with Philips, Signify and Lumileds we have worked on these strategies, and on  translating them to devices like compact phosphor-converted (pc) micro-LEDs for high brightness displays. Another exciting new  direction is on the interface of chirality and light-matter interaction. How do you control the local density of states for chiral fluorescent transitions,  and how do you transfer polarization (and phase) properties at will to ensembles of emitters?  Circularly polarized emission is important, for instance, for 3D display technology.

How much information about the nanoscale geometry of an object can you retrieve from its diffraction pattern? This question is relevant in metrology – for instance, the semiconductor lithography industry uses diffraction measurements on bespoke scattering targets that are placed on the wafer in between actual devices to measure the accuracy of lithography in terms of feature size, alignment, andsoforth.  This raises the question how you can optimally use a finite footprint for a maximally  informative scattering target,  what the best illumination wavefront is, and what features in a diffraction pattern are optimal for reading out geometrical perturbations.

We tackle these questions in the AMOLF-ARCNL collaborative research program with Lyuba Amitonova, Peter Kraus, and Arie den Boef, and in industry co-funded projects. We take inspiration from the fields of resonant metasurfaces for design, wavefront shaping for illumination, and from the concept of Fisher information to measure information content. We explore metasurface motifs for metrology. For instance, in one recent experiment we show that Pancharatnam-Berry phase metasurfaces can transduce nanoscale alignment information in background free polarimetric signatures.

In other experiments we focus on the use of Fano resonances and Bound States in the Continuum. We combine Fourier microscopy to measure diffraction patterns of sub-wavelength sized scattering targets, wavefront shaping with spatial light modulators to program the incident wave,  and information theory to design maximum information illuminations, read outs and targets. Beyond linear scattering we also study metasurfaces as nonlinear light generation devices for metrology, and do theory hand in hand with experiments.

The field of metasurfaces seeks to implement analog mathematical operations, as well as tasks like inference directly by light diffraction. We perform research on the information processing capacity of metasurfaces, as well as seeking strategies to reconfigure and rapidly modulate metasurface based processing.

Light-based interpretation of the environment is ubiquitous in many applications, from imaging, augmented and virtual reality, to automotive (e.g., LIDAR), and diagnostics (e.g., handheld spectroscopy, sensing and microscopy). Light encodes information in many degrees of freedom – aside from the ~ 10⁶ DOFs in a typical 2D image, this also includes spectral, polarization, phase, and 3D depth-degrees of freedom.

A main bottleneck in “edge” applications is that conversion of light to electronics for processing, classification is energy-inefficient:  the optical data sets are large, and computing tasks such as classification and inference are highly CPU intensive. Electronics-based solutions hence either require to deal with large energy consumption locally, or when outsourcing the computing to datacenters, cause large data traffic.

In this spirit we imagine that future energy-efficient computing strategies should be hybrid (e.g., hybrid photonic-electronic), and should avoid unnecessary DOF-conversion. Thus, we envision that large optical datasets should be (pre)processed and compressed as much as possible with light-based technologies, leaving the conversion to integrated circuits and electronics to the last.