Machine learning in materials science

Machine learning holds great potential to accelerate materials research. Many domains in materials science are benefiting from its application, but several challenges persist, and it remains to be seen whether the field will live up to the hype that surrounds it.

The design of new functional polymers depends on the successful navigation of their structure-function landscapes. Advances in combinatorial polymer chemistry and machine learning provide exciting opportunities for the engineering of fit-for-purpose polymeric materials.

Computational sustainability harnesses computing and artificial intelligence for human well-being and the protection of our planet. Materials science is central to many sustainability challenges. Exploiting synergies between computational sustainability and materials science advances both fields, furthering the ultimate goal of establishing a sustainable future.

Intense efforts are underway to produce circuits that integrate a technologically relevant number of qubits. Although qubit control in most material systems is by now mature, device variability is one of the main bottlenecks in qubit scalability. How do we characterize and tune millions of qubits? Machine learning might hold the answer.

The rapidly expanding biomaterials data are challenging to organize. Text mining systems are powerful tools that automatically extract and integrate information in large textual collections. As text mining leaps forward by leveraging deep-learning approaches, it is time to address the most pressing biomaterials information and data processing needs.

Google Applied Science is a division of Google Research that applies computational methods, and in particular machine learning, to a broad range of scientific problems. Patrick Riley, until recently one of their software engineers and now head of artificial intelligence at Relay Therapeutics, talks to Nature Reviews Materials about his experience working on machine-learning projects in an industrial setting.

The materials research landscape is being transformed by the infusion of approaches based on machine learning. This Review discusses the emerging materials intelligence ecosystems and the potential of human–machine partnerships for fast and efficient virtual materials screening, development and discovery.

Neural networks can capture nonlinear relationships in high-dimensional spaces and are powerful tools for photonic-system modelling. This Review discusses how deep neural networks can serve as surrogate electromagnetic solvers, inverse modelling tools and global device optimizers.

Machine learning can be applied for the controlled synthesis of nanoparticles with precise properties. This Review discusses different machine learning approaches for the synthesis of semiconductor, metal, carbon-based and polymeric nanoparticles, and highlights key approaches for the collection of large datasets.

High-throughput experimental technologies can generate large data sets of cell-type-specific information, allowing the study of multicellular complexity. This Review discusses machine learning approaches, in particular, deep learning and network-based models, which can be applied to analyse, interpret and model these data sets.

Machine learning is enabling a metallurgical renaissance. This Review discusses recent progress in representations, descriptors and interatomic potentials, overviewing metallic glasses, high-entropy alloys, superalloys and shape-memory alloys, magnets and catalysts, and the prediction of mechanical and thermal properties.

Batteries, as complex materials systems, pose unique challenges for the application of machine learning. Although a shift to data-driven, machine learning-based battery research has started, new initiatives in academia and industry are needed to fully exploit its potential.

Several key industries routinely use metal printing to make complex parts that are difficult to produce by conventional manufacturing. Here, we show that a synergistic combination of metallurgy, mechanistic models and machine learning is driving the continued growth of metal printing.

Artificial intelligence can be used to facilitate the 3D printing of functional materials and devices directly on target surfaces, such as human bodies. This Review surveys ex situ and in situ artificial-intelligence-assisted 3D printing of multifunctional materials and its combination with surgical robots to enable autonomous medical care and smart biomanufacturing.

The discovery of new thermoelectric materials is challenging owing to the diversity of the chemical space and to the serial nature of experimental work. This Review highlights the recent progress in computationally guided discovery of thermoelectric materials and identifies the key outstanding challenges.

The discovery of nanoporous materials is now being propelled by the analysis of big data combined with traditional computational thermodynamics calculations. In this Review, we analyse the current state of the art, with a focus on the generation of computational databases and results from large-scale screening for gas separations.

An article in Nature Communications reports a method for the rapid detection of SARS-CoV-2 in saliva samples using nanopores and a machine learning algorithm.

An article in npj Computational Materials reports a deep-learning workflow for the analysis of microscopy images that can adapt to changing conditions.

An article in Science Robotics presents an autonomous excavator system able to recognize and load a range of materials in different environments.

An article in Nature Chemistry uses the knowledge gathered in the Cambridge Structural Database to build a machine-learning model that predicts the oxidation states of metal–organic frameworks.

An article in npj Computational Materials reports a machine learning-based solver that predicts local stresses in mechanically heterogeneous solids.