Machine Learning-Guided Design of a Flexible Highly Conductive Additive-Free Polymer Cathode

Organic cathode materials (OCMs) are promising sustainable alternatives to inorganic counterparts for next-generation batteries, yet their widespread application is largely hindered by intrinsically low electrical conductivity (below 10^-6 S cm^-1) and material dissolution. The vast chemical space for exploration complicates the discovery of optimal OCMs. In this work, we utilized a machine learning (ML)-based discovery process with a pretrained transformer model in ZINC organic molecules database, yielding a couple of potential high-performance OCMs candidates, including isoindigo-type redox units. The output of such efficient screening inspires the design of poly-benzodifurandione (PBFO) as a free-standing cathode material for high-performance Li-ion and Na-ion storage. The flexible PBFO film exhibits a breakthrough conductivity of 5.9×10² S cm^-1, setting a new benchmark for additive-free organic cathodes. The neat PBFO cathodes achieve a reversible capacity of 262 mAh g^-1 averaging at 2.5 V versus Li⁺/Li at 25 mA g^-1, delivering a high electrode-level energy density of 655 Wh kg^-1, among the highest reported for OCMs. This work provides the first flexible, high-conductivity organic cathodes without conductive additives and binders, opening a new direction toward viable organic batteries.