Accelerated Discovery of High-Refractive-Index Polyimides via First-Principles Molecular Modeling, Virtual High-Throughput Screening, and Data Mining

Abstract

<div>We present a high-throughput computational study to identify novel polyimides (PIs) with exceptional refractive index (RI) values for use as optic or optoelectronic materials. Our study utilizes an RI prediction protocol based on a combination of first-principles and data modeling developed in previous work, which we employ on a large-scale PI candidate library generated with the ChemLG code. We deploy the virtual screening software ChemHTPS to automate the assessment of this extensive pool of PI structures in order to determine the performance potential of each candidate. This rapid and efficient approach yields a number of highly promising leads compounds. Using the data mining and machine learning program package ChemML, we analyze the top candidates with respect to prevalent structural features and feature combinations that distinguish them from less promising ones. In particular, we explore the utility of various strategies that introduce highly polarizable moieties into the PI backbone to increase its RI yield. The derived insights provide a foundation for rational and targeted design that goes beyond traditional trial-and-error searches.</div>

A deep neural network model for packing density predictions and its application in the study of 1.5 million organic molecules

“…In this work, we develop a DNN prediction model for the packing density of small organic molecules in an amorphous bulk phase and conduct a hyperscreening of 1.5 million candidate compounds. Our interest in this target property originates from our ongoing in silico discovery and design efforts for polymers with high refractive index (RI) [32][33][34] to be used in optic and optoelectronic applications. 35,36 We previously established an RI modeling protocol based on the Lorentz-Lorenz equation and parametrized with the polarizability and number density.…”

ChemML: A Machine Learning and Informatics Program Package for the Analysis, Mining, and Modeling of Chemical and Materials Data

“…ChemML features other methodological innovation, e.g., in the areas of physics-infused neural network architectures, learned features, local domain models, training set design, on-the-fly assessment of learning curves [31], chemical pattern recognition [32], etc., that we will describe elsewhere.…”

“…We have been employing ChemML in a number of realworld application studies, both for the creation of dataderived prediction models and chemical pattern recognition. These studies include discovery and design projects for new high-refractive-index polymers for optical applications [30][31][32][33][34][35], deep eutectic solvents for supercapacitors [36], and organic semiconductors for photovoltaics and other applications [37,38] (using data of the Harvard Clean Energy Project [39][40][41][42][43]).…”

Pareto Optimization of Oligomer Polarizability and Dipole Moment using a Genetic Algorithm

“…[17][18][19] Similar computational discovery in the related field of polarizable materials has been conducted through the lens of discovery of materials with high refractive indices through similar inverse design approaches, including machine learning 20 and high throughput virtual screening. 21 Other computational studies searching specifically for novel OPV conjugated polymer structures tend to focus structural motifs optimizing properties such as energy levels and band gaps. [22][23][24] In this study, we searched for novel high dielectric constant materials, focusing on π conjugated materials because of their high polarizability and proven utility for this application.…”