Advancing materials science through next-generation machine learning.

For over a decade, machine learning (ML) models have been making strides in computer vision and natural language processing (NLP), demonstrating high proficiency in specialized tasks. The emergence of large-scale language and generative image models, such as ChatGPT and Stable Diffusion, has significantly broadened the accessibility and application scope of these technologies. Traditional predictive models are typically constrained to mapping input data to numerical values or predefined categories, limiting their usefulness beyond their designated tasks. In contrast, contemporary models employ representation learning and generative modeling, enabling them to extract and encode key insights from a wide variety of data sources and decode them to create novel responses for desired goals. They can interpret queries phrased in natural language to deduce the intended output. In parallel, the application of ML techniques in materials science has advanced considerably, particularly in areas like inverse design, material prediction, and atomic modeling. Despite these advancements, the current models are overly specialized, hindering their potential to supplant established industrial processes. Materials science, therefore, necessitates the creation of a comprehensive, versatile model capable of interpreting human-readable inputs, intuiting a wide range of possible search directions, and delivering precise solutions. To realize such a model, the field must adopt cutting-edge representation, generative, and foundation model techniques tailored to materials science. A pivotal component in this endeavor is the establishment of an extensive, centralized dataset encompassing a broad spectrum of research topics. This dataset could be assembled by crowdsourcing global research contributions and developing models to extract data from existing literature and represent them in a homogenous format. A massive dataset can be used to train a central model that learns the underlying physics of the target areas, which can then be connected to a variety of specialized downstream tasks. Ultimately, the envisioned model would empower users to intuitively pose queries for a wide array of desired outcomes. It would facilitate the search for existing data that closely matches the sought-after solutions and leverage its understanding of physics and material-behavior relationships to innovate new solutions when pre-existing ones fall short.

High-Resolution Volumetric Imaging and Classification of Organisms with Standard Optical Microscopy.

Three-dimensional (3D) characterization of organisms is important for the study of cellular phenotypes, structural organization, and mechanotransduction. Existing optical techniques for 3D imaging rely on focus stacking or complex multiangle projection. Focus stacking has deleterious axial resolution due to the one-angle optical projection. Herein, we achieve high-resolution 3D imaging and classification of organisms based on standard optical microscopy coupled to optothermal rotation. Through a seamless fusion of optical trapping and rotation of organisms on a single platform, our technique is applicable to any organism suspended in clinical samples, enabling contact-free and biocompatible 3D imaging. Moreover, when applying deep learning to distinguish different types of biological cells with high similarity, we demonstrate that our platform improves the classification accuracy (96% vs 85%) while using one-tenth the number of training samples compared with conventional deep-learning-based classification.

Decoding Optical Data with Machine Learning.

Optical spectroscopy and imaging techniques play important roles in many fields such as disease diagnosis, biological study, information technology, optical science, and materials science. Over the past decade, machine learning (ML) has proved promising in decoding complex data, enabling rapid and accurate analysis of optical spectra and images. This review aims to shed light on various ML algorithms for optical data analysis with a focus on their applications in a wide range of fields. The goal of this work is to sketch the validity of ML-based optical data decoding. The review concludes with an outlook on unaddressed problems and opportunities in this emerging subject that interfaces optics, data science and ML.

A mixture-density-based tandem optimization network for on-demand inverse design of thin-film high reflectors.

Deep learning (DL) has emerged as a promising tool for photonic inverse design. Nevertheless, despite the initial success in retrieving spectra of modest complexity with nearly instantaneous readout, DL-assisted design methods often underperform in accuracy compared with advanced optimization techniques and have not proven competitive in handling spectra of practical usefulness. Here, we introduce a tandem optimization model that combines a mixture density network (MDN) and a fully connected (FC) network to inversely design practical thin-film high reflectors. The multimodal nature of the MDN gives access to infinite candidate designs described by probability distributions, which are iteratively sampled and evaluated by the FC network to allow for rapid optimization. We show that the proposed model can retrieve the reflectance spectra of 20-layer thin-film structures. More interestingly, it reproduces with high precision the periodic structures of high reflectors derived from physical principles, even though no such information is included in the training data. Improved designs with extended high-reflectance zones are also demonstrated. Our approach combines the high-efficiency advantage of DL with the optimization-enabled performance improvement, enabling efficient and on-demand inverse design for practical applications.

Deep Convolutional Mixture Density Network for Inverse Design of Layered Photonic Structures.

Machine learning (ML) techniques, such as neural networks, have emerged as powerful tools for the inverse design of nanophotonic structures. However, this innovative approach suffers some limitations. A primary one is the nonuniqueness problem, which can prevent ML algorithms from properly converging because vastly different designs produce nearly identical spectra. Here, we introduce a mixture density network (MDN) approach, which models the design parameters as multimodal probability distributions instead of discrete values, allowing the algorithms to converge in cases of nonuniqueness without sacrificing degenerate solutions. We apply our MDN technique to inversely design two types of multilayer photonic structures consisting of thin films of oxides, which present a significant challenge for conventional ML algorithms due to a high degree of nonuniqueness in their optical properties. In the 10-layer case, the MDN can handle transmission spectra with high complexity and under varying illumination conditions. The 4-layer case tends to show a stronger multimodal character, with secondary modes indicating alternative solutions for a target spectrum. The shape of the distributions gives valuable information for postprocessing and about the uncertainty in the predictions, which is not available with deterministic networks. Our approach provides an effective solution to the inverse design of photonic structures and yields more optimal searches for the structures with high degeneracy and spectral complexity.

Intelligent nanophotonics: merging photonics and artificial intelligence at the nanoscale.

Nanophotonics has been an active research field over the past two decades, triggered by the rising interests in exploring new physics and technologies with light at the nanoscale. As the demands of performance and integration level keep increasing, the design and optimization of nanophotonic devices become computationally expensive and time-inefficient. Advanced computational methods and artificial intelligence, especially its subfield of machine learning, have led to revolutionary development in many applications, such as web searches, computer vision, and speech/image recognition. The complex models and algorithms help to exploit the enormous parameter space in a highly efficient way. In this review, we summarize the recent advances on the emerging field where nanophotonics and machine learning blend. We provide an overview of different computational methods, with the focus on deep learning, for the nanophotonic inverse design. The implementation of deep neural networks with photonic platforms is also discussed. This review aims at sketching an illustration of the nanophotonic design with machine learning and giving a perspective on the future tasks.