In the dynamic landscape of battery technology, understanding the flow of innovation and technological knowledge among different chemistries is pivotal. Recent breakthroughs have demonstrated profound interdependencies between lithium-ion batteries (LIBs) and sodium-ion batteries (SIBs), two of the most promising monovalent metal-ion battery technologies. A ground-breaking study, soon to be published in Nature Energy, unpacks these knowledge exchanges by leveraging patent citation analysis, revealing how these advanced battery systems build upon and evolve through shared technical insights.
The research team’s approach centers around patent families as discrete units of technological innovation. Unlike individual patents that can be filed across multiple jurisdictions, patent families consolidate these filings, providing a clearer picture of unique inventions while eliminating redundancies caused by cross-country patents. This method captures the nuanced technological progress embodied in complete electrode assemblies—from positive electrode materials to intricate cell-level integrations—yielding unprecedented granularity on how knowledge flows traverse different battery chemistries.
Crucially, the investigation drills down into chemical structure-level knowledge flows, examining how similarities or differences in molecular frameworks influence the trajectory of innovation. This analysis is further enhanced by dissecting the innovation types, contrasting product innovations—such as novel materials or cell design breakthroughs—with process innovations that improve manufacturing techniques or production efficiency. Temporal trends are also mapped out, identifying whether evolving chemistries like SIBs gradually develop technological independence or remain tightly coupled with the mature LIB knowledge base.
The team draws from the extensive patent data housed within The Lens database, renowned for its comprehensive family-level citation data and English-language patent claim texts. This choice ensures high-quality, consistent data vital for sophisticated network analyses. Leveraging baseline validated query strategies from prior leading studies ensures the capture of core patents for LIB and SIB technologies across crucial jurisdictions including the US, Europe, Japan, and China, the latter being particularly critical for the nascent SIB domain.
Data curation involved stringent filters to guarantee patent validity and relevance. Only granted patent families with at least one forward citation and available English claims qualified, refining the dataset to over 33,000 LIB and more than 2,200 SIB patent families. To overcome language barriers, professional translations supplemented original texts, ensuring that insights into Chinese filing strategies were incorporated, especially given China’s pivotal role in SIB research and commercialization.
A pioneering element of the study lies in its use of advanced artificial intelligence tools. The researchers integrated GPT-4o, a state-of-the-art large language model, via an API to automate multi-level patent classification. This technological leap enabled precise categorization by battery type (LIB vs. SIB), electrode material specifications, and innovation typologies. Rigorous manual validation demonstrated that AI classification paralleled expert human annotation, providing consistent, reproducible results while scaling analysis across tens of thousands of records.
The classification framework is meticulously hierarchical. For LIBs, the study identified key positive electrode materials such as lithium cobalt oxide (LCO), nickel manganese cobalt (NMC), and lithium iron phosphate (LFP), alongside primary negative electrode materials like graphite and silicon-carbon composites. For SIBs, classes ranged from iron-based Prussian blue analogues (Fe-PBA) to various polyanion frameworks and hard carbon negative electrodes. Distinctions between product and process innovations also proved integral for understanding the different ways knowledge proliferates within and across these chemistries.
Network analysis represents the heart of the study’s methodology. By constructing directed citation networks, the researchers mapped how later patents reference prior art, effectively tracing the directional flow of technological knowledge backward in time. These networks differentiate between positive and negative electrode innovations and capture bidirectional interactions, spotlighting the inter-electrode material influences fundamental to battery development. The visualization employs material-year nodes, offering temporal context and highlighting fluctuations in innovation volume and knowledge exchange density.
Addressing patent citation volume disparities was vital for accurate interpretation. Established LIB chemistries accumulate patent volumes and citations over several decades, whereas SIBs, being emergent, are comparatively sparse in both volume and age. To navigate this, a normalization process was devised that quantifies knowledge flow as a percentage of each battery class’s total knowledge base, allowing comparability by contextualizing raw citation counts relative to each receiver’s existing citation habits.
Further refinement came from calculating relative knowledge flow intensities by benchmarking cross-class citation flows against within-class self-citations. This reference point treats citations within the same chemistry as a natural baseline, enabling the identification of when one battery chemistry draws significantly from another’s intellectual contributions. Remarkably, this metric reveals instances where emerging SIB chemistries disproportionately rely on established LIB innovations, offering insights into potential innovation leverage points or technological bottlenecks.
An intriguing dimension emerged when innovation types were considered separately. Product-oriented innovations tend to show different knowledge flow patterns compared to process innovations, reflecting how material discoveries versus production techniques traverse chemical domains. Mixed-innovation patents highlighted the increasingly hybrid nature of battery development, where advancements simultaneously target device performance and manufacturing scalability, underscoring the complex innovation ecosystem driving battery technology.
Temporal analyses illuminate dynamic trends in knowledge dependencies. The assembled datasets, spanning multiple decades and jurisdictions, reveal that while LIBs maintain robust internal knowledge flows, SIBs largely depend on LIB advancements initially but show gradual growth in internal citation strength over time. This evolutionary trajectory suggests maturing SIB technologies gaining confidence and independence, a vital insight for guiding future research investments and policy considerations.
Visualization tools play a crucial role in distilling these complex relationships into accessible, interpretable forms. Heat maps articulate relative knowledge flow intensities powerfully, while Sankey diagrams elegantly depict innovation-type interplays. Scatter plots tracing citation trajectories over years capture the emergence of new knowledge paradigms and potential technological shifts, creating rich, nuanced depictions of the battery innovation landscape.
Despite its robustness, the study acknowledges limitations rooted in the inherent nature of patent data. Patents capture explicit codified knowledge but miss tacit or informal knowledge exchanges intrinsic to scientific collaboration and personnel movement. Citation biases and examiner-added references can muddle interpretations, and temporal lags between filing, publication, and citation further complicate recent data analysis. Nevertheless, patents remain one of the most systematic, legally mandated sources for mapping technological knowledge flows.
This comprehensive framework not only elucidates the deep interconnectedness of lithium- and sodium-ion battery innovations but also offers a powerful methodology adaptable to other emergent technologies confronting heterogeneous maturity levels and innovation pathways. The fusion of AI-powered classification, rigorous normalization techniques, and multi-dimensional network analysis represents a significant leap forward in understanding how cutting-edge battery technologies co-evolve and influence each other within a complex, competitive innovation ecosystem.
As global energy storage demands escalate, understanding these knowledge flows is more than academic curiosity—it is a strategic imperative. Insights derived from such analysis can inform targeted innovation support, intellectual property strategies, and collaborative research agendas essential for accelerating the development of safer, more efficient, and scalable battery solutions that underpin the transition to a sustainable energy future.
The implications extend beyond batteries. This approach exemplifies how big data, AI, and network science can converge to unravel intricate innovation dynamics in fast-evolving technological fields. By revealing hidden dependencies and pathways of knowledge transfer, policymakers, industry leaders, and researchers can better anticipate technological trajectories, identify strategic leverage points, and cultivate ecosystems that maximize collective innovation potential without redundancy or fragmentation.
In sum, this pioneering research dissects the technological kinship and knowledge dependencies shaping the future of energy storage, advancing not only the battery field but also setting a new standard for how innovation crossing chemical and technological boundaries can be studied and leveraged through sophisticated patent analytics coupled with AI-driven classification methods.
Subject of Research:
Investigation of technological knowledge flows and interdependencies between lithium-ion and sodium-ion battery chemistries using patent citation analysis.
Article Title:
Knowledge interdependencies between lithium- and sodium-ion battery chemistries.
Article References:
Hemmelder, A., Panda, A., Peiseler, L. et al. Knowledge interdependencies between lithium- and sodium-ion battery chemistries. Nat Energy 11, 313–323 (2026). https://doi.org/10.1038/s41560-026-01985-z
Image Credits: AI Generated
DOI: February 2026
