New Opportunity: Machine Learning for Polymer Materials Design and Discovery

Abstract

Under the guidance of the material genome initiative (MGI), the use of data‐driven methods to discover new materials has become an innovation of materials science. The polymer materials have been one of the most important parts in materials science for the excellent physical and chemical properties as well as corresponding complex structures. Machine learning, as the core of data‐driven methods, has taken an important place in polymer materials design and discovery. In this review, the authors have introduced the applications of machine learning in the design and discovery of polymer materials. The development tendency of published papers about machine learning in polymer materials, the commonly used algorithms, the polymer descriptors, the workflow of machine learning in polymer materials, and recent progresses of machine learning in materials are summarized. Then, the detail of how to use machine learning to assist design and discovery of polymer materials is fully discussed combined with two cases. Finally, the opportunities and challenges on the future development prospects of machine learning in the field of polymer materials are proposed.

Metal-related electrocatalysts for Li–CO2batteries: an overview of the fundamentals to explore future-oriented strategies

“…Second, driven by increased computing power, advances in algorithm development, and the availability of massive amounts of data, modern society has seen the application of machine learning expand into numerous research areas of materials. [142][143][144][145] In traditional materials science research, researchers only rely on the experience of material testing and application in the experimental process to accumulate material properties. Obviously, the development of high-performance new materials by traditional methods has a long period of time, and low efficiency, and is limited by experimental conditions.…”

Data-Driven Design of Polymer-Based Biomaterials: High-throughput Simulation, Experimentation, and Machine Learning

“…On the basis of our experiences, we highlight key aspects of successful representation strategies, considerations for choosing one algorithm over another, and facets of a data set that can limit the performance of surrogate models. For fuller discussion about featurization schemes and algorithms, largely for homopolymers, we refer readers to some other recent works and review articles. − …”

Accelerating Discovery of High Fractional Free Volume Polymers from a Data-Driven Approach

“…In the past few years, we have witnessed data-driven methods being increasingly used in prediction and design of new polymers. , For example, experimental and computational data of polymer properties (band gap, dielectric constant, refractive index, atomization energy, glass transition temperature, solubility parameter, and density) were accumulated to develop machine learning (ML) models . From a chemically diverse library of poorly soluble drugs and drug-polymer micellar solubilization data, predictive models were proposed to discover polymeric micelle formulations for poorly soluble drugs .…”