Number Density Descriptor on Extended-Connectivity Fingerprints Combined with Machine Learning Approaches for Predicting
- PDF / 1,081,362 Bytes
- 6 Pages / 432 x 648 pts Page_size
- 85 Downloads / 162 Views
MRS Advances © 2018 Materials Research Society DOI: 10.1557/adv.2018.454
Number Density Descriptor on ExtendedConnectivity Fingerprints Combined with Machine Learning Approaches for Predicting Polymer Properties Takuya Minami1, Yoshishige Okuno1 1
Research Association of High-Throughput Design and Development for Advanced Functional Materials, Ibaraki, Japan
Abstract
We developed a new type of polymer descriptor based on Extended Connectivity Fingerprints. The number densities, that are substructure numbers divided by the number of atoms in a polymer model, were employed. We found that this approach is superior in accurately predicting linear polymer properties, compared to the conventional approach, where just the substructure numbers are used as descriptors. In addition, dimension reduction and multiple replication of repeat unit were found to improve prediction accuracy. As a result, the novel descriptor based on the Extended Connectivity Fingerprints with machine learning approaches was found to achieve accurate prediction of the refractive indices of linear polymers, which is comparable to that by ab initio density functional theory. Although process-dependent properties such as mechanical properties were difficult to predict, the present approach was found to be applicable to prediction of substructure-dependent properties, for example, optical properties, thermal stabilities.
INTRODUCTION: Application of informatics or machine learning to material science has attracted researchers, since it accelerates the development of novel functional materials [1][2]. This is a data-driven approach, which predicts structures and/or properties of unknown materials by learning correlations between descriptors and properties of already-known materials. In recent years, researchers have succeeded in applying materials informatics to several functional materials such as thermoelectric materials [3], molecular organic light-emitting diodes [4], and low-thermal-conductivity compounds [5]. In addition, process-structure-property (PSP) linkages [6], Integrated Computational Materials Engineering (ICME) [7], as well as catalyst informatics have been reported recently [8]. One of the key for applying informatics to material science is selecting appropriate descriptors. To make computers understand a material, we need to convert features of materials into descriptors, that are computer-friendly data such as digital vectors. Since prediction accuracy depends on the quality of descriptors, the development of descriptor is very important.
Downloaded from https://www.cambridge.org/core. YBP Library Services, on 19 Aug 2018 at 08:14:30, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1557/adv.2018.454
In case of functional polymers, Mannodi-Kanakkithodi et al., have reported a pioneering work on designing polymer dielectrics [9]. In their approach, a polymer is represented by the chain of blocks composed of several atoms. The descriptors are the numbers of single b
Data Loading...
