Prediction of Electronic Properties of Radical-Containing Polymers at Coarse-Grained Resolutions

The properties of soft electronicmaterials depend onthe couplingof electronic and conformational degrees of freedom over a wide rangeof spatiotemporal scales. The description of such properties requiresmultiscale approaches capable of, at the same time, accessing electronicproperties and sampling the conformational space of soft materials.This could in principle be realized by connecting the coarse-grained(CG) methodologies required for adequate conformational sampling toconformationally averaged electronic property distributions via backmappingto atomistic-resolution level models and repeated quantum-chemicalcalculations. Computational demands of such approaches, however, havehindered their application in high-throughput computer-aided softmaterials discovery. Here, we present a method that, combining machinelearning and CG techniques, can replace traditional backmapping-basedapproaches without sacrificing accuracy. We illustrate the methodfor an emerging class of soft electronic materials, namely, nonconjugated,radical-containing polymers, promising materials for all-organic energystorage. Supervised machine learning models are trained to learn thedependence of electronic properties on polymer conformation at CGresolutions. We then parametrize CG models that retain electronicstructure information, simulate CG condensed phases, and predict theelectronic properties of such phases solely from the CG degrees offreedom. We validate our method by comparing it against a full backmapping-basedapproach and find good agreement between both methods. This work demonstratesthe potential of the proposed method to accelerate multiscale workflowsand provides a framework for the development of CG models that retainelectronic structure information.

Automated summary

The properties of soft electronic materials depend on the coupling of electronic and conformational degrees of freedom. Traditional approaches for describing these properties require multiscale methods, which can access electronic properties and sample the conformational space of soft materials. In this study, a machine learning method combined with coarse-grained techniques is proposed to replace the traditional backmapping-based approaches without sacrificing accuracy.