It has long remained challenging to predict the properties of complex chemical systems, such as polymer-based materials and their composites. We have constructed the largest database of lithium-conducting solid polymer electrolytes (104 entries) and employed a transfer-learned graph neural network to accurately predict their conductivity (mean absolute error of less than 1 on a logarithmic scale). The bias-free prediction by the network helped us to find superionic conductors composed of charge-transfer complexes of aromatic polymers (ionic conductivity of around 10-3 S/cm at room temperature). The glassy design was contrary to the traditional concept of rubbery polymer electrolytes, but it was found to be appropriate to achieve fast, decoupled motion of ionic species from polymer chains and to enhance thermal and mechanical stability. The unbiased suggestions generated by machine learning models can help researches to discover unexpected chemical phenomena, which could also induce a paradigm shift of energy-related functional materials.
ASJC Scopus subject areas
- Colloid and Surface Chemistry