Computational Design and Selection of Optimal Organic Photovoltaic Materials

TLDR

Conjugated organic polymers are key building blocks of low‑cost photovoltaic materials. The study uses computational predictions to perform inverse design of over 90,000 copolymers, aiming to identify structures with optimal optical and excited‑state energies for highly efficient solar cells. A genetic‑algorithm search over synthetically accessible six‑ and eight‑unit copolymers produced hundreds of candidates predicted to exceed 8 % efficiency, with many over 10 %, and the authors derived sequence‑motif trends and design rules while applying additional filtering to exclude polymers with poor hole mobility. The approach rapidly identifies copolymers with superior electronic and optical properties, outperforming serial experiments and computational studies, and is generalizable to other materials‑science challenges.

Abstract

Conjugated organic polymers are key building blocks of low-cost photovoltaic materials. We have examined over 90 000 copolymers using computational predictions to solve the "inverse design" of molecular structures with optimum properties for highly efficient solar cells (specifically matching optical excitation energies and excited-state energies). Our approach, which uses a genetic algorithm to search the space of synthetically accessible copolymers of six or eight monomer units, yields hundreds of candidate copolymers with predicted efficiencies over 8% (the current experimental record), including many predicted to be over 10% efficient. We discuss trends in polymer sequences and motifs found in the most frequent monomers and dimers in these highly efficient targets and derive design rules for the selection of appropriate donor and acceptor molecules. We show how additional computationally intensive filtering steps can be used, for example, to eliminate targets likely to have poor hole mobilities. Our method effectively targets optimum electronic structure and optical properties far more efficiently than time-consuming serial experiments or computational studies and can be applied to similar problems in other areas of materials science.