Predicting physical response of an artificially structured material is of particular interest for scientific and engineering applications. Here we use deep learning to predict optical response of artificially engineered nanophotonic devices. In addition to predicting forward approximation of transmission response for any given topology, this approach allows us to inversely approximate designs for a targeted optical response. Our Deep Neural Network (DNN) could design compact (2.6 × 2.6 μm²) silicon-on-insulator (SOI)-based 1 × 2 power splitters with various target splitting ratios in a fraction of a second. This model is trained to minimize the reflection (to smaller than ~ −20 dB) while achieving maximum transmission efficiency above 90% and target splitting specifications. This approach paves the way for rapid design of integrated photonic components relying on complex nanostructures.

For decades, progress in the field of optical (including solar) energy conversion was dominated by advances in the conventional concentrating optics and materials design. In recent years, however, conceptual and technological breakthroughs in the fields of nanophotonics and plasmonics combined with a better understanding of the thermodynamics of the photon energy-conversion processes reshaped the landscape of energy-conversion schemes and devices. Nanostructured devices and materials that make use of size quantization effects to manipulate photon density of states offer a way to overcome the conventional light absorption limits. Novel optical spectrum splitting and photon-recycling schemes reduce the entropy production in the optical energy-conversion platforms and boost their efficiencies. Optical design concepts are rapidly expanding into the infrared energy band, offering new approaches to harvest waste heat, to reduce the thermal emission losses, and to achieve noncontact radiative cooling of solar cells as well as of optical and electronic circuitries. Light–matter interaction enabled by nanophotonics and plasmonics underlie the performance of the third- and fourth-generation energy-conversion devices, including up- and down-conversion of photon energy, near-field radiative energy transfer, and hot electron generation and harvesting. Finally, the increased market penetration of alternative solar energy-conversion technologies amplifies the role of cost-driven and environmental considerations. This roadmap on optical energy conversion provides a snapshot of the state of the art in optical energy conversion, remaining challenges, and most promising approaches to address these challenges. Leading experts authored 19 focused short sections of the roadmap where they share their vision on a specific aspect of this burgeoning research field. The roadmap opens up with a tutorial section, which introduces major concepts and terminology. It is our hope that the roadmap will serve as an important resource for the scientific community, new generations of researchers, funding agencies, industry experts, and investors.

The ability to modulate light using 2-dimensional (2D) materials is fundamentally challenged by their small optical cross-section leading to miniscule modal confinements in diffraction-limited photonics despite intrinsically high electro-optic absorption modulation (EAM) potential given by their strong exciton binding energies. However the inherent polarization anisotropy in 2D materials and device tradeoffs lead to additional requirements with respect to electric field directions and modal confinement. A detailed relationship between modal confinement factor and obtainable modulation strength including definitions on bounding limits are outstanding. Here, we show that the modal confinement factor is a key parameter determining both the modulation strength and the modulator extinction ratio-to-insertion loss metric. We show that the modal confinement and hence the modulation strength of a single-layer modulated 2D material in a plasmonically confined mode is able to improve by more than 10× compared to diffraction-limited modes. Combined with the strong-index modulation of graphene, the modulation strength can be more than 2-orders of magnitude higher compared to Silicon-based EAMs. Furthermore, modal confinement was found to be synergistic with performance optimization via enhanced light-matter-interactions. These results show that there is room for scaling 2D-material EAMs with respect to modal engineering toward realizing synergistic designs leading to high-performance modulators.

The class of transparent conductive oxides includes the material indium tin oxide (ITO) and has become a widely used material of modern every-day life such as in touch screens of smart phones and watches, but also used as an optically transparent low electrically-resistive contract in the photovoltaics industry. More recently ITO has shown epsilon-near-zero (ENZ) behavior in the telecommunication frequency band enabling both strong index modulation and other optically-exotic applications such as metatronics. However, the ability to precisely obtain targeted electrical and optical material properties in ITO is still challenging due to complex intrinsic effects in ITO and as such no integrated metatronic platform has been demonstrated to-date. Here we deliver an extensive and accurate description process parameter of RF-sputtering, showing a holistic control of the quality of ITO thin films in the visible and particularly near-infrared spectral region. We are able to custom-engineer the ENZ point across the telecommunication band by explicitly controlling the sputtering process conditions. Exploiting this control, we design a functional sub-wavelength-scale filter based on lumped circuit-elements, towards the realization of integrated metatronic devices and circuits.

Deep learning is now playing a major role in designing photonic devices, including nanostructured photonics. In this paper, we investigate three models for designing nanophonic power splitters with multiple splitting ratios. The first model is a forward regression model, wherein the trained deep neural network (DNN) is used within the optimization loop. The second is an inverse regression model, in which the trained DNN constructs a structure with the desired target performance given as input. The third model is a generative network, which can randomly produce a series of optimized designs for a target performance. Focusing on the nanophotonic power splitters, we show how the devices can be designed by these three types of DNN models.