Free-space photonics is the transmission and manipulation of light beams through open space (air, vacuum) rather than through guided waveguides or fibre. It spans a wide application set united by the absence of a physical medium between source and receiver: free-space optical communications (FSO), LiDAR, directed-energy sensing, inter-satellite laser links, and — at the intersection with computation — diffractive free-space neural networks where optical diffraction layers perform matrix–vector multiply in transit.
The engineering fundamentals differ sharply from guided photonics: beam divergence, pointing–acquisition–tracking (PAT), atmospheric turbulence (the primary impairment for terrestrial FSO), and background radiation dominate the link budget. Detectors must handle shot-noise-limited reception at high sensitivity — APDs and SPADs are standard receivers for terrestrial FSO; Mercury Cadmium Telluride FPAs serve MWIR atmospheric windows. Wavelengths are chosen for eye safety (1550 nm) and atmospheric transmission.
Commercial relevance concentrates in: (1) inter-satellite optical links (SpaceX Starlink, Mynaric, Tesat), where FSO offers 10–100× higher throughput than RF at competitive SWaP; (2) terrestrial last-mile in urban dense deployments; (3) LiDAR as a ranging modality (see LiDAR). Zigzag Networks applies free-space optical beam-forming in its network architecture. Diffractive Deep Neural Network descends from this space as a compute modality.
Frontier
- Can atmospheric-turbulence adaptive-optics correction (AO-FSO) reach the latency and power budget needed for untethered ground-to-ground links above 10 km?
- Will inter-satellite optical links (ISOL) become commodity infrastructure, and does that commoditise the terminal or the steering/pointing sub-system?
- What is the sustainable data-rate ceiling for a diffractive free-space compute link (light through air as a multiply-accumulate layer)?