Molecular Weight Transferable Electronic Structure Prediction for Coarse-Grained Polymers

Soft materials modeling for non-crystalline systems involves the integration of multiscale classical and quantum mechanical techniques. Recently, we introduced a methodology known as electronic coarse-graining that relies upon techniques from supervised machine learning to make accurate quantum-chemical predictions directly from coarse-grained degrees of freedom, thereby rapidly accelerating quantum-mechanical predictions for soft materials. In this talk, I will outline the extension of this methodology to polymers, specifically addressing the problem of the molecular weight transferability of electronic coarse-graining predictions. Focus will be paid to the integration of machine learning techniques with phenomenological quantum-mechanical Hamiltonians, enhancing transferability and interpretability of electronic structure simulations at a coarse-grained resolution.