An artificial intelligence (AI) program has identified a groundbreaking material that could revolutionize battery technology by reducing the amount of lithium used by up to 70%. Developed through a collaboration between the Pacific Northwest National Laboratory (PNNL) and Microsoft, the new material, a mix of sodium, lithium, yttrium, and chloride ions, was selected from a pool of 32 million candidates.

The demand for lithium, a key component in rechargeable batteries, has surged in recent years. However, the extraction process for lithium is energy-intensive and often leads to environmental pollution. In light of these challenges, researchers have been exploring alternative materials for battery production.

Using Microsoft's Azure Quantum Elements tool, scientists at PNNL screened potential materials for low-lithium batteries. Their findings, published on January 8th in the pre-print server arXiv, revealed the selected material to be a solid electrolyte that enables the easy passage of lithium ions while blocking the movement of electrons in a battery.

The researchers began with over 32 million candidate materials, generated by substituting various elements into existing electrolyte structures. Through the implementation of AI techniques, they rapidly filtered out unstable options, reducing the pool to half a million materials in a matter of hours.

Applying a series of nine additional criteria, the team further narrowed down the options by considering electronic properties, cost, and strength. Sequentially applying AI algorithms, the final 18 materials were identified from the filtered candidates. This process, which traditionally would have taken two decades to accomplish experimentally, was completed in just 80 computer hours through the combination of machine learning and molecular dynamics models.

The final selection of materials synthesized by the researchers contained varying proportions of lithium, sodium, yttrium, and chloride ions. Remarkably, the mixture of lithium and sodium in the material enables conduction of both types of ions, which was thought to be impossible until now. One variant demonstrated a 70% reduction in lithium content compared to conventional batteries, potentially leading to significantly lower prices and decreased environmental impact in the future.

To test the electronic properties of the candidates, the researchers found that the top-performing material identified by the AI program had conductivity approximately one order of magnitude lower than current liquid electrolytes. This discrepancy translates to slower charge times, highlighting the need for further improvement before practical application. Nonetheless, a prototype built from the final material successfully powered a lightbulb, showcasing the potential of the discovery.

Kandler Smith, a mechanical engineer from the National Renewable Energy Laboratory, praised the AI program's efficiency in material discovery. Beyond its immediate impact on battery technology, Smith emphasized the potential for the machine learning pipeline to support research in numerous related fields.

Both Microsoft and PNNL expressed enthusiasm for the future exploration of AI in material discovery. Brian Abrahamson, PNNL's chief digital officer, acknowledged the significance of the speed and accuracy of AI in identifying promising materials, stating their intention to continue pushing the boundaries of technology and scientific expertise.

The recent breakthrough in battery material discovery marks a promising step forward in the search for alternative materials to reduce the reliance on lithium. Through the integration of AI and scientific expertise, researchers are poised to accelerate the development of more efficient and environmentally friendly battery technologies.