Diffractive optical computing in free space

Abstract

Structured optical materials create new computing paradigms using photons, with transformative impact on various fields, including machine learning, computer vision, imaging, telecommunications, and sensing. This Perspective sheds light on the potential of free-space optical systems based on engineered surfaces for advancing optical computing. Manipulating light in unprecedented ways, emerging structured surfaces enable all-optical implementation of various mathematical functions and machine learning tasks. Diffractive networks, in particular, bring deep-learning principles into the design and operation of free-space optical systems to create new functionalities. Metasurfaces consisting of deeply subwavelength units are achieving exotic optical responses that provide independent control over different properties of light and can bring major advances in computational throughput and data-transfer bandwidth of free-space optical processors. Unlike integrated photonics-based optoelectronic systems that demand preprocessed inputs, free-space optical processors have direct access to all the optical degrees of freedom that carry information about an input scene/object without needing digital recovery or preprocessing of information. To realize the full potential of free-space optical computing architectures, diffractive surfaces and metasurfaces need to advance symbiotically and co-evolve in their designs, 3D fabrication/integration, cascadability, and computing accuracy to serve the needs of next-generation machine vision, computational imaging, mathematical computing, and telecommunication technologies.

Fig. 1. Overview of diffractive and metasurface systems for optical computing.

Diffractive surfaces consist of thickness- and/or index-tuned units (λ/2 or larger) while metasurfaces consist of substructured metallic and/or dielectric units below λ/2. These free-space processors can perform polarization processing, spatial processing, universal linear transformations, and spectral & temporal processing of waves.

Fig. 2. Timeline of diffractive^,– and metasurface^– based optical devices and systems.

Images are adapted from refs. ^,, with permissions from AAAS, refs. ^–,,, with permissions from © The Optical Society, refs. ^,,,, with permissions from Springer Nature, refs. ^, under CC BY 4.0, ref. with permissions from AIP publishing, and refs. ^,, with permissions from APS.

Fig. 3. Design of diffractive surfaces and metasurfaces.

Diffractive unit cell designs and their principles of wavefront modulation based on (a) thickness-tuning and (b) index-tuning^,. Metasurface unit cell designs using (c) plasmonic^,, and (d) dielectric materials^,,. b This is adapted with permission from ref. by CC BY 4.0. c This is adapted with permission from ref. by AAAS, refs. ^, by CC BY 4.0 and ref. by ACS. d This is adapted with permission from ref. by CC BY 4.0, ref. by John Wiley and Sons and ref. by ACS.

Fig. 4. Diffractive optical networks can perform universal linear transformations.

Diffractive surfaces trained to perform (a) linear operations including an arbitrary complex-valued transform, discrete Fourier transform (DFT) and (b) permutation operation. Extensions of this framework to perform universal linear transformations under spatially and/or temporally incoherent illumination were also demonstrated^,. a, b These are adapted with permission from ref. and ref. , respectively, by CC BY 4.0.

Fig. 5. Diffractive networks and metasurfaces for all-optical spatial processing of light.

All optical image recovery by diffractive networks from (a) computer-generated holograms and (b) OAM information channels. Metasurfaces for (c) single channel and (d) OAM-enabled multichannel holography. (e) Active electrically tuned spatial modulation by metasurfaces in liquid crystal cells. a This is adapted with permission from ref. by ACS. b This is adapted with permission from ref. by © The Optical Society. c This is adapted with permission from ref. by Springer Nature. d This is adapted with permission from ref. by Springer Nature. e This is adapted with permission from ref. by AAAS.

Fig. 6. Free-space polarization processors based on diffractive networks and metasurfaces.

a Metasurfaces for versatile polarization generation by nanorod units. b Polarization-dependent holographic images using anisotropic meta-units^,. c Diffractive networks enabled multichannel universal linear transformation by polarization multiplexing using a polarizer array. a This is adapted with permission from ref. by ACS. b This is adapted with permission from ref. by Springer Nature and ref. from APS. c This is adapted with permission from ref. by CC BY 4.0.

Fig. 7. Hybrid optical-electronic networks for intelligent free-space processing.

a Hybrid optoelectronic network for classification. b Digital encoder enables pixel super-resolution displaying by diffractive networks. c Deep-learning design of digital code for dynamic beamforming on a reconfigurable metasurface. d Robust, high-accuracy OAM measurement by a hybrid network. a This is adapted with permission from ref. by CC BY 4.0. b This is adapted with permission from ref. by AAAS. c This is adapted with permission from ref. by IEEE. d This is adapted with permission from ref. by CC BY 4.0.

Fig. 8. Tunable and reconfigurable responses from metasurfaces and diffractive networks.

Tuning of device responses by (a) Lego-like swapping of the layer designs, (b) electromechanical modulation that deforms the metasurface substrate, (c) MEMS systems that adjust distances between two metasurfaces. Unit-level metasurface reconfiguration by (d) electrical tuning of field-programmable gate arrays and (e) reversible laser writing on phase-change materials. a This is adapted with permission from ref. by Springer Nature. b This is adapted with permission from ref. by AAAS. c–e These are adapted with permission from ref. , ref. , and ref. , respectively, by Springer Nature.

Fig. 9. Roadmap and future outlook of diffractive surface- and metasurface-enabled computing platforms.

These free-space optical processors can have transformative impact on various technologies ranging from robotics to biomedical imaging and telecommunications, among others.