Complex nanophotonic structures hold the potential to deliver exquisitely tailored optical responses for a range of applications. Metal-insulator-metal (MIM) metasurfaces arranged in supercells, instance, can be by geometry and material choice exhibit variety absorption properties resonant wavelengths. With this flexibility, however, comes vast space design possibilities that classical paradigms struggle effectively navigate. To overcome challenge, here we demonstrate tandem residual network...

Understanding how nano- or micro-scale structures and material properties can be optimally configured to attain specific functionalities remains a fundamental challenge. Photonic metasurfaces, for instance, spectrally tuned through choice structural geometry achieve unique optical responses. However, existing numerical design methods require prior identification of material-structure combinations, device classes, as the starting point optimization. As such, unified solution that...

From higher computational efficiency to enabling the discovery of novel and complex structures, deep learning has emerged as a powerful framework for design optimization nanophotonic circuits components. However, both data-driven exploration-based machine strategies have limitations in their effectiveness inverse design. Supervised approaches require large quantities training data produce high-performance models difficulty generalizing beyond given complexity space. Unsupervised...

A fundamental challenge in the design of photonic devices, and electromagnetic structures more generally, is optimization their overall architecture to achieve a desired response. To this end, topology or shape optimizers based on adjoint variable method have been widely adopted due high computational efficiency ability create complex freeform geometries. However, functional understanding such remains black box. Moreover, unless space high-performance devices known advance, gradient-based...

In recent years, hybrid design strategies combining machine learning (ML) with electromagnetic optimization algorithms have emerged as a new paradigm for the inverse of photonic structures and devices. While trained, data-driven neural network can rapidly identify solutions near global optimum given data set's space, an iterative algorithm further refine solution overcome set limitations. Furthermore, such ML-optimization methodologies reduce computational costs expedite discovery novel...

In article number 2100548, Christopher Yeung, Aaswath P. Raman, and co-workers propose a global photonics materials design framework, based on generative adversarial networks, which simultaneously optimizes photonic system's device class, material properties, geometric structuring. This framework is demonstrated in the context of metasurface design, where unique combinations structures are generated that yield significantly more variation achievable optical responses than conventional deep...

Abstract The growth kinetics of localized corrosion, e.g. pits, in corrosive environments often controls the service life metallic components. Yet, our understanding these is largely based on coupon-level, mass-loss, studies which provide limited insights into evolution individual damage events. It critical to relate observed cumulative loss trends, such as links between changing humidity and mass rates, pits. Towards this goal, we leverage in-situ X-ray computed tomography measure rates...

We present a mission concept for long-duration solar-powered aircraft to explore the surface and cloud-deck of Venus. In this concept, an flies against planet's rotation prevailing winds remain roughly near sub-solar point indefinitely while performing related measurements atmosphere in 50-80 km altitude range. The spends significant time at 50 altitude, where it can measure upwelling radiance from surface. Since power constraints prohibit indefinite operations we identify "porpoising"...

A fundamental challenge in the design of nanophotonic devices is optimization subwavelength structures to achieve tailored and high-performance electromagnetic responses. To this end, topology or shape optimizers based on adjoint variables method have been widely adopted push performance limits systems. However, understanding such freeform remain obscure, gradient-based can get trapped low-performance local minima. Accordingly, elucidate relationships between device nanoscale structuring,...

From higher computational efficiency to enabling the discovery of novel and complex structures, deep learning has emerged as a powerful framework for design optimization nanophotonic circuits components. However, both data-driven exploration-based machine strategies have limitations in their effectiveness inverse design. Supervised approaches require large quantities training data produce high-performance models difficulty generalizing beyond given complexity space. Unsupervised...

In recent years, hybrid design strategies combining machine learning (ML) with electromagnetic optimization algorithms have emerged as a new paradigm for the inverse of photonic structures and devices. While trained, data-driven neural network can rapidly identify solutions near global optimum given dataset's space, an iterative algorithm further refine solution overcome dataset limitations. Furthermore, such ML-optimization methodologies reduce computational costs expedite discovery novel...

We present a machine learning-based photonics design strategy centered on encoding image colors with material and structural data. Given input target spectra, our model can accurately determine the optimal metasurface class, materials, structure.