Mapping Chemical Structure–Glass Transition Temperature Relationship through Artificial Intelligence

Abstract

Artificial neural networks (ANNs) have been successfully used in the past to predict different properties of polymers based on their chemical structure and to localize and quantify the intramonomer contributions to these properties. In this work, we propose to move forward in order to use the mathematical framework of the ANN for embedding the chemical structure of monomers into a high-dimensional abstract space. This approach allows us not only to accurately predict the glass transition temperature (T g) of polymers but, even more important, also to encode their chemical structure as m-dimensional vectors in a mathematical space. For this aim, we employed a fully connected neural network trained with a set of more than 200 atactic acrylates that provide the coordinates of the vectorized chemical structures into the m-dimensional space. These data points were then treated with a hierarchical nonparametric clusterization method in order to automatically group similar chemical structures into clusters with alike properties. These clusters were then projected into a human-readable three-dimensional space using principal component analysis. This approach allows us to deal with chemical structures as if they were mathematical entities and therefore to perform quantitative operations, so far hardly imaginable, being essential for both the design of new materials and the understanding of the structure–property relationships.

Benchmarking Machine Learning Models for Polymer Informatics: An Example of Glass Transition Temperature

“…Besides, the use of large and diverse data sets (22 polymer classes in total) is critical for pattern recognition. Otherwise, the trained ML model can only apply to certain classes of polymers in the local chemical space. ,, …”

“…As an exciting alternative, utilizing machine learning (ML) techniques and the increasing amount of polymer data sets offer a new opportunity to tackle the challenge in the polymer field. Successful polymer informatics attempts have touched upon the a number of property predictions like polymers’ electronic bandgap, , dielectric constant, refractive index, etc., but a lot more attention has been paid to the prediction of polymers’ glass transition temperatures. ,− This is primarily reflective of the facts that (1) the glass transition temperature is an important property controlling the phase transition and therefore the application of polymers and (2) the glass transition temperature T g is the most reported experimental measurement in publicly accessible databases like PoLyInfo, the Polymer Property Predictor and Database (NIST), and the CROW Polymer Properties Database …”

“…Extensive studies have been done for the T g prediction of polymers, and researchers have made different choices for Steps 1–3. For example, Miccio et al ,, (1) focused on the monomer structure of polymers (2) converted the monomer’s SMILES into binary matrices with the one-hot encoding algorithm; and (3) used the FFNN or CNN method to build ML models. Their ML models showed a relative error of about 6% and observed a reasonable generalization ability on about 300 polymers from different classes.…”

Lignin-Inspired Polymers with High Glass Transition Temperature and Solvent Resistance from 4-Hydroxybenzonitrile, Vanillonitrile, and Syringonitrile Methacrylates

“…For example, poly­(styrene- co -acrylonitrile) (SAN) has a T g above 100 °C, depending on the acrylonitrile content . Moreover, benzonitrile methacrylate polymers have recently been predicted by artificial neural network to possess high T g . From an industrial point of view, nitrilation is readily achieved by ammoxidation, using ammonia, oxygen, and a vanadium or molybdenum oxide catalyst. , With the aim to develop biobased high-performance polymethacrylates with high T g values resulting from both high macromolecular chain rigidity and polarity, we have in the present work prepared three different nitrile-containing methacrylate monomers.…”

An Enhanced Scheme for Multiscale Modeling of Thermomechanical Properties of Polymer Bulks

“…So far, many empirical methods − have been proposed to quickly predict various properties of polymers from the structures of their repeating units or monomers. In addition, quantitative structure–property relationships (QSPRs) based on machine learning (ML) have recently attracted renewed interests. , However, these methods cannot directly consider the effects of complex environments (solutions, blends, composites, confinements, cross-links, and so forth.) and at the same time disclose the underlying molecular mechanism.…”