AI Alchemy for Batteries: Atomic to Microstructure Insights from The Royal Institution

In this Royal Institution event, three researchers explore how artificial intelligence is transforming battery science from the atomic scale to the microstructure and system level. The talks trace a thread from Humphry Davy and Faraday to today’s AI powered energy storage research, highlighting generative models, data integration, and automated laboratories that speed up discovery and deployment of better batteries.

• AI as a digital alchemist turning data into material designs for next generation batteries
• Co scientist agents and automated labs accelerate experiments and discovery
• From atomic scale design to grid scale and microstructure understanding, AI weaves across battery scales
• Balancing data, computing power, and physics to deliver transferable, real world insights

Overview

The Royal Institution hosts a three speaker session on how artificial intelligence accelerates battery science from atomic design to microstructure, and finally to system level health and lifetime assessment. The talks connect a long historical thread from the discovery of electrochemistry to the modern use of AI to understand and accelerate the development of batteries that power devices, vehicles, and grids.

AI at the Atomic Scale

The first talk surveys how scale changes the battery problem. AI converts the vast body of chemistry and materials data into machine learning friendly representations, often vectors that encode elemental and structural information. The speaker explains the idea of chemical space and the enormous number of potential materials, far beyond what can be tested experimentally. Generative AI and co scientists emerge as key tools. The chameleon model demonstrates how text descriptions can be transformed into new material candidates, while tools like Crystallize show how AI can design and test hypotheses through experiments, potentially in collaboration with automated labs. The takeaway is that AI and materials theory are converging to accelerate discovery while respecting the fundamental electrochemistry that Davy and Faraday laid down. A startup, CUSP AI, provides an industrial scale engine for materials intelligence and co scientists that can operate across knowledge bases, simulations, and lab automation. The speaker notes that this is not a replacement for scientists but a force multiplier, enabling campaigns that would have taken months to run in hours or days when coupled with automated laboratories.

AI at the System Scale

The second talk moves up to the grid and application scale. Here the focus is battery state of health and lifetime in real world use. The speaker emphasizes the complexity of degradation, which arises from mechanical distress, particle cracking, lithium plating, interface film formation, and environmental factors. A battery management system must predict remaining useful life and adapt charging strategies. Data driven models, including time series approaches and attention based learning, are combined with physics informed or hybrid models to ensure physically plausible predictions and generalizability across chemistry and form factor. The talk also addresses challenges in privacy, data scarcity, and deployment in real world vehicles and infrastructure, as well as the need for scalable hardware suitable for on device inference.

AI for Microstructure

The final talk focuses on microstructure, the internal architecture of electrodes that governs performance. The speaker discusses how AI supports four activities: characterize the microstructure from imaging, generate three dimensional structures from two dimensional data, design manufacturing parameter spaces to optimize microstructure, and validate AI driven predictions through robotic labs. The role of vision transformers and generative models in image based segmentation and 3D structure inference is highlighted, along with a spin out company that translates AI microstructure models into industrial workflows. The talk concludes with the promise of agentic AI that can plan, reason, and operate tools to execute experiments autonomously while maintaining rigorous validation and reproducibility across scales and disciplines.

Conclusion

Across the atomic scale, the systems scale, and the microstructure layer, the speakers emphasize that AI accelerates discovery by integrating data with physics and enabling orchestration of campaigns that would be impractical with humans alone. The session links a historical lineage from Davy and Faraday to a modern energy ecosystem where AI helps design, test, and deploy batteries capable of powering a sustainable future while addressing practical issues of scale, safety, and governance.