Predicting Electronic Properties of Radical Polymers at Coarse-Grained Resolutions

Radical polymers are conducting polymers comprised of non-conjugated backbones and radical-containing side chains. Radical polymers are promising materials for applications in all-organic energy or memory storage devices. A fundamental understanding of how the chemical structure impacts charge transport in these materials is still lacking. To address this, computational approaches encompassing both mesoscale morphological features and electronic properties are required.

Here we use machine learning (ML) to predict electronic-structure information pertaining to charge transport at coarse-grained (CG) resolutions. ML models are trained on data obtained from quantum chemical calculations on conformations sampled from condensed-phase simulations at all-atom resolution. ML models then enable electronic property predictions directly from CG polymer morphologies. Electronic predictions that depend on either a single radical site (e.g. energy levels) or a pair of sites (e.g. electronic couplings) are investigated as a function of CG resolution.

The approach has the potential to drastically accelerate computational workflows, hence opening the way for high-throughput exploration of the chemical space of radical polymers while taking into account both mesoscale and electronic properties.