Researchers at the University of Technology Sydney have used artificial intelligence to discover several new lithium-ion battery cathode materials with superior energy density and stability compared to commercially used compounds.

The AI system screens millions of potential material compositions computationally, identifying candidates with promising characteristics. Those candidates are then synthesised and tested in the lab, a process that takes weeks rather than the years typically required for materials discovery.

Dr. Neeraj Sharma, who leads UTS’s battery materials research, said the approach fundamentally changes how materials science can operate. “Traditionally, researchers test maybe a few dozen compositions over several years. The AI lets us explore orders of magnitude more possibilities and focus our lab work on the most promising candidates.”

The system uses a combination of physics-based models that predict material properties and machine learning algorithms trained on databases of known materials. By combining these approaches, the AI can make predictions about materials that have never been synthesised.

One challenge is that materials databases contain mostly information about stable compounds that occur naturally or have been studied extensively. Novel battery materials often involve unusual compositions or crystal structures not well represented in existing data.

The UTS team addressed this by generating synthetic training data using computational chemistry methods. They calculated properties of hundreds of thousands of hypothetical compounds, then used that data to train machine learning models that generalise beyond the training set.

This hybrid approach outperforms pure machine learning or pure computational chemistry. The machine learning provides speed, while computational chemistry provides physical constraints that prevent the AI from suggesting impossible materials.

The newly discovered cathode materials include lithium-manganese-nickel-oxygen compounds with modified crystal structures that allow more lithium insertion and extraction without degrading. That translates to higher energy density and longer cycle life.

One material showed 320 milliamp-hours per gram capacity in initial testing, about 25% higher than NMC811, the best commercial cathode material. Cycle life testing is ongoing, but early results suggest the material maintains capacity better than conventional materials.

If those results hold up through extensive testing, this material could enable electric vehicles with 20-25% longer range using the same battery pack size, or equivalently smaller, lighter, cheaper battery packs for the same range.

But transitioning new battery materials from laboratory discovery to commercial production typically takes 5-10 years. Materials need to be tested thoroughly for safety, cycle life, manufacturing scalability, and cost before companies will risk deploying them in products.

Several factors can derail promising materials during this validation process. Manufacturing might prove difficult at scale. Subtle degradation mechanisms might only appear after thousands of charge-discharge cycles. Cost might be prohibitive even if performance is superior.

The AI-discovered materials still face all these challenges. But AI-guided discovery at least accelerates the initial identification phase, potentially compressing overall development timelines.

The research received funding from the Australian Research Council and several international battery companies interested in next-generation materials. One participating company has expressed interest in licensing the most promising compounds for further development.

AI for materials discovery is a rapidly growing field globally. Google DeepMind, Microsoft, and several startups are applying similar approaches to discover materials for batteries, solar cells, catalysts, and other applications.

Australian researchers bring particular expertise in battery materials characterisation and testing to this field. Facilities like the Australian Synchrotron enable detailed structural analysis of battery materials during operation, providing insights that improve AI models and validate predictions.

The UTS project is also applying AI to other battery problems beyond cathode materials. They’re exploring electrolyte formulations, anode materials, and interface coatings that improve battery safety and performance.

Electrolyte discovery is particularly interesting because liquid electrolytes involve complex mixtures of solvents, salts, and additives where composition space is enormous. AI can help identify optimal formulations that traditional trial-and-error approaches might never find.

One technical challenge is establishing validation pipelines that can quickly test AI predictions. The team has automated much of their materials synthesis and testing using robotic systems that operate overnight and weekends. This dramatically increases throughput compared to manual processes.

Building these automated pipelines required substantial upfront investment but pays dividends in enabling rapid iteration between AI predictions and experimental validation. Other research groups are adopting similar approaches, recognising that AI-guided discovery only works if you can validate predictions quickly.

The battery materials research is part of UTS’s broader efforts to establish Sydney as a hub for battery technology development. The university is building a new Battery Manufacturing Hub that will enable prototype battery cell production, bringing together researchers and industry partners.

Whether Australia can build significant battery manufacturing capability remains debated. The supply chain is currently dominated by Asian manufacturers with established scale and cost advantages. Australian initiatives focus on niches like high-performance batteries or next-generation chemistries where innovation might offset manufacturing cost disadvantages.

AI-enabled materials discovery could support this strategy by enabling Australian researchers to develop superior materials that justify higher manufacturing costs. Whether that works depends on achieving sufficient performance advantages and protecting intellectual property.

The UTS research demonstrates how AI is changing scientific research processes. Rather than replacing human researchers, AI augments their capabilities, allowing them to explore vastly larger possibility spaces and focus effort on the most promising opportunities.

Whether this leads to breakthroughs in battery technology that enable better electric vehicles and grid storage remains to be seen. But the approach is showing early promise and attracting significant research investment globally.