All-optical deep learning: Deep learning uses multilayered artificial neural networks to learn digitally from large datasets. It then performs advanced identification and classification tasks. To date, these multilayered neural networks have been implemented on a computer. Linet al.demonstrate all-optical machine learning that uses passive optical components that can be patterned and fabricated with 3D-printing. Their hardware approach comprises stacked layers of diffractive optical elements analogous to an artificial neural network that can be trained to execute complex functions at the speed of light.Science, this issue p.1004

Diffractive networks encode the spatial information of objects into power spectrum to classify images with a single-pixel detector.

Recent advances in deep learning have been providing non-intuitive solutions to various inverse problems in optics. At the intersection of machine learning and optics, diffractive networks merge wave-optics with deep learning to design task-specific elements to all-optically perform various tasks such as object classification and machine vision. Here, we present a diffractive network, which is used to shape an arbitrary broadband pulse into a desired optical waveform, forming a compact and passive pulse engineering system. We demonstrate the synthesis of various different pulses by designing diffractive layers that collectively engineer the temporal waveform of an input terahertz pulse. Our results demonstrate direct pulse shaping in terahertz spectrum, where the amplitude and phase of the input wavelengths are independently controlled through a passive diffractive device, without the need for an external pump. Furthermore, a physical transfer learning approach is presented to illustrate pulse-width tunability by replacing part of an existing network with newly trained diffractive layers, demonstrating its modularity. This learning-based diffractive pulse engineering framework can find broad applications in e.g., communications, ultra-fast imaging and spectroscopy.

Plasmonic photoconductive antennas have great promise for increasing responsivity and detection sensitivity of conventional photoconductive detectors in time-domain terahertz imaging and spectroscopy systems. However, operation bandwidth of previously demonstrated plasmonic photoconductive antennas has been limited by bandwidth constraints of their antennas and photoconductor parasitics. Here, we present a powerful technique for realizing broadband terahertz detectors through large-area plasmonic photoconductive nano-antenna arrays. A key novelty that makes the presented terahertz detector superior to the state-of-the art is a specific large-area device geometry that offers a strong interaction between the incident terahertz beam and optical pump at the nanoscale, while maintaining a broad operation bandwidth. The large device active area allows robust operation against optical and terahertz beam misalignments. We demonstrate broadband terahertz detection with signal-to-noise ratio levels as high as 107 dB.

The scope and potential uses of time-domain terahertz imaging and spectroscopy are mainly limited by the low optical-to-terahertz conversion efficiency of photoconductive terahertz sources. State-of-the-art photoconductive sources utilize short-carrier-lifetime semiconductors to recombine carriers that cannot contribute to efficient terahertz generation and cause additional thermal dissipation. Here, we present a novel photoconductive terahertz source that offers a significantly higher efficiency compared with terahertz sources fabricated on short-carrier-lifetime substrates. The key innovative feature of this source is the tight three-dimensional confinement of the optical pump beam around the terahertz nanoantennas that are used as radiating elements. This is achieved by means of a nanocavity formed by plasmonic structures and a distributed Bragg reflector. Consequently, almost all of the photo-generated carriers can be routed to the terahertz nanoantennas within a sub-picosecond time-scale. This results in a very strong, ultrafast current that drives the nanoantennas to produce broadband terahertz radiation. We experimentally demonstrate that this terahertz source can generate 4 mW pulsed terahertz radiation under an optical pump power of 720 mW over the 0.1–4 THz frequency range. This is the highest reported power level for terahertz radiation from a photoconductive terahertz source, representing more than an order of magnitude of enhancement in the optical-to-terahertz conversion efficiency compared with state-of-the-art photoconductive terahertz sources fabricated on short-carrier-lifetime substrates.