Abstract Battery technology presents a major economic opportunity, spanning diverse metal-ion battery chemistries with distinct material, performance and supply chain characteristics. How knowledge is shared across these chemistries is essential for industrial strategies informed by techno-economic forecasting but understudied. Here we report the use of advanced large language models for automated patent classification to map knowledge flows within and across lithium-ion and sodium-ion battery chemistries, based on patent citation networks encompassing over 15,000 patent families classified at the chemical structure level. We find substantial, persistent knowledge flows within and across chemistries for product and process innovation, with continuous directional knowledge flows from mature lithium-ion to emerging sodium-ion chemistries. For some lithium-ion chemistries, cross-chemistry knowledge flows even exceed within-chemistry flows. These results indicate that industrial strategies aiming to leap-frog to new chemistries without design and manufacturing experience in existing chemistries are unlikely to succeed, and battery chemistries should therefore not be treated as fully independent technologies in forecasting models.
Hemmelder et al. (Tue,) studied this question.