Machine learning identifies scale-free properties in disordered materials

Q-Factor Optimization of Modes in Ordered and Disordered Photonic Systems Using Non-Hermitian Perturbation Theory

The quality factor, Q, of photonic resonators permeates most figures of merit in applications that rely on cavity-enhanced light-matter interaction such as all-optical information processing, high-resolution sensing, or ultralow-threshold lasing. As a consequence, large-scale efforts have been devoted to understanding and efficiently computing and optimizing the Q of optical resonators in the design stage. This has generated large know-how on the relation between physical quantities of the cavity, e.g., Q, and controllable parameters, e.g., hole positions, for engineered cavities in gaped photonic crystals. However, such a correspondence is much less intuitive in the case of modes in disordered photonic media, e.g., Anderson-localized modes. Here, we demonstrate that the theoretical framework of quasinormal modes (QNMs), a non-Hermitian perturbation theory for shifting material boundaries, and a finite-element complex eigensolver provide an ideal toolbox for the automated shape optimization of Q of a single photonic mode in both ordered and disordered environments. We benchmark the non-Hermitian perturbation formula and employ it to optimize the Q-factor of a photonic mode relative to the position of vertically etched holes in a dielectric slab for two different settings: first, for the fundamental mode of L3 cavities with various footprints, demonstrating that the approach simultaneously takes in-plane and out-of-plane losses into account and leads to minor modal structure modifications; and second, for an Anderson-localized mode with an initial Q of 200, which evolves into a completely different mode, displaying a threefold reduction in the mode volume, a different overall spatial location, and, notably, a 3 order of magnitude increase in Q.

Artificial Intelligence and Advanced Materials

Artificial intelligence (AI) is gaining strength, and materials science can both contribute to and profit from it. In a simultaneous progress race, new materials, systems, and processes can be devised and optimized thanks to machine learning (ML) techniques, and such progress can be turned into innovative computing platforms. Future materials scientists will profit from understanding how ML can boost the conception of advanced materials. This review covers aspects of computation from the fundamentals to directions taken and repercussions produced by computation to account for the origins, procedures, and applications of AI. ML and its methods are reviewed to provide basic knowledge of its implementation and its potential. The materials and systems used to implement AI with electric charges are finding serious competition from other information-carrying and processing agents. The impact these techniques have on the inception of new advanced materials is so deep that a new paradigm is developing where implicit knowledge is being mined to conceive materials and systems for functions instead of finding applications to found materials. How far this trend can be carried is hard to fathom, as exemplified by the power to discover unheard of materials or physical laws buried in data.

Heavy tails and pruning in programmable photonic circuits for universal unitaries

Developing hardware for high-dimensional unitary operators plays a vital role in implementing quantum computations and deep learning accelerations. Programmable photonic circuits are singularly promising candidates for universal unitaries owing to intrinsic unitarity, ultrafast tunability and energy efficiency of photonic platforms. Nonetheless, when the scale of a photonic circuit increases, the effects of noise on the fidelity of quantum operators and deep learning weight matrices become more severe. Here we demonstrate a nontrivial stochastic nature of large-scale programmable photonic circuits-heavy-tailed distributions of rotation operators-that enables the development of high-fidelity universal unitaries through designed pruning of superfluous rotations. The power law and the Pareto principle for the conventional architecture of programmable photonic circuits are revealed with the presence of hub phase shifters, allowing for the application of network pruning to the design of photonic hardware. For the Clements design of programmable photonic circuits, we extract a universal architecture for pruning random unitary matrices and prove that "the bad is sometimes better to be removed" to achieve high fidelity and energy efficiency. This result lowers the hurdle for high fidelity in large-scale quantum computing and photonic deep learning accelerators.

Predicting strongly localized resonant modes of light in disordered arrays of dielectric scatterers: a machine learning approach

In this work, we predict the most strongly confined resonant mode of light in strongly disordered systems of dielectric scatterers employing the data-driven approach of machine learning. For training, validation, and test purposes of the proposed regression architecture-based deep neural network (DNN), a dataset containing resonant characteristics of light in 8,400 random arrays of dielectric scatterers is generated employing finite difference time domain (FDTD) analysis technique. To enhance the convergence and accuracy of the overall model, an auto-encoder is utilized as the weight initializer of the regression model, which contains three convolutional layers and three fully connected layers. Given the refractive index profile of the disordered system, the trained model can instantaneously predict the Anderson localized resonant wavelength of light with a minimum error of 0.0037%. A correlation coefficient of 0.95 or higher is obtained between the FDTD simulation results and DNN predictions. Such a high level of accuracy is maintained in inhomogeneous disordered media containing Gaussian distribution of diameter of the scattering particles. Moreover, the prediction scheme is found to be robust against any combination of diameters and fill factors of the disordered medium. The proposed model thereby leverages the benefits of machine learning for predicting the complex behavior of light in strongly disordered systems.