Over the last few years, the chances of creating new conducting polymers with the help of machine learning have caught the attention of many researchers in the field of chemistry. Now, a team of researchers has discovered a new kind of polymer which contains high thermal conductivity and can be beneficial to the 5G mobile communication technologies.
Researcher Ryo Yoshida said that many aspects remain to be explored, such as “training” computational systems to work with limited data by adding more suitable descriptors. He added, “ML for polymer or soft material design is a challenging but promising field as these materials have properties that differ from metals and ceramics, and are not yet fully predicted by the existing theories.”
This research was conducted by Statistical Mathematics (ISM), Research Organization of Information and Systems, Tokyo Institute of Technology and Center for Materials Research by Information Integration, Research and Services Division of Materials Data and Integrated System (MaDIS).
Methodology Used
The goal behind this computational molecular design is to identify new molecules whose physicochemical properties meet the given arbitrary requirements. The researchers were successful in designing new polymers with high thermal conductivity with the help of a machine learning algorithm that is referred to as Bayesian molecular design.
In the Bayesian molecular design process which generated a library of virtual chemical structures, the researchers specified a higher region of glass transition temperatures and melting temperatures as alternative design targets, for which sufficient data were given to obtain reliable prediction models.
The model not only identifies new molecules but also helps to mitigate the issue of limited data. In addition to that, transfer learning which is an ML framework is also applied in order to obtain a thermal conductivity model with the given small data set.
Dataset used
The study was performed on a dataset of polymeric properties from PoLyInfo which is the largest database of polymers in the world at NIMS. The database contains a limited amount of data on the heat transfer properties of polymers. In order to predict the heat transfer properties from the given amount of data, machine learning models on proxy properties were pre-trained and then these pre-trained models extracted the common features which are relevant to the related tasks. The pre-trained model is merged with a specially designed ML algorithm for computational molecular design which is known as the iQSPR algorithm.
Other Use Cases
Last year, a group of researchers from the department of chemical engineering at Virginia Tech developed a temperature-independent computational model for a particular polymer which is sensitive to temperature. The simulation trajectories of this computational model were analysed with the help of a data-driven machine learning method. The computational model is known as the coarse-grained model which utilises a specific data-driven ML approach, known as non-metric multidimensional scaling method to analyse the molecular dynamic simulation trajectories of a coarse-grained model of a temperature-sensitive polymer.
Outlook
Traditionally, materials and polymers are created with a trial-and-error approach, but with ML, researchers can build and create new materials in a cost-efficient manner, in less time. In this modern era, databases and computation models are the keys to create resistive and efficient materials.
The researchers at the Tokyo Institute of Technology are striving to create ML-driven high-throughput computational systems in order to design next-generation soft materials for applications going beyond the 5G era. The polymers with high thermal conductivity would be the key to heat management in the fifth-generation (5G) mobile communication technologies.
(You can read the full paper here.)
