Microsoft, PNNL Use AI to Find Lithium Alternatives

Lithium batteries have long been used as the source of energy storage in electronic devices, ranging from mobile phones and laptops to energy storage systems. More recently, the emergence of electrical vehicles (EVs) has further increased dependence on lithium. However, lithium's high price and ever-increasing demand are beginning to pose a significant challenge to battery manufacturers. 

In a joint venture with Pacific Northwest National Laboratory, Microsoft has employed artificial intelligence (AI) to create a battery with less lithium. This technology may address the constraints scientists encounter in developing lithium batteries.

 

Lithium-ion batteries. Image used courtesy of Adobe Stock

 

Alternatives to Lithium Batteries?

Lithium's suitability for batteries lies in its unique combination of low atomic mass, high electrochemical potential, good conductivity, low self-discharge rate, and high energy density. These properties contribute to the efficiency, durability, and compactness of lithium batteries as compared to other competing battery chemistries. For example, in contrast to battery technologies such as nickel-cadmium or nickel-aluminum hybrids, lithium batteries exhibit rapid charging capabilities and can undergo numerous discharge cycles before reaching depletion.

However, lithium faces many geopolitical and logistical challenges, making it less than ideal for large-scale, long-term use. 

Lithium reserves only exist in a few countries, meaning any unrest in global politics could potentially initiate a global supply issue. Moreover, lithium is relatively expensive due to its limited production capacity, energy-intensive extraction processes, and high demand. These factors have sparked concerns among battery manufacturers around the globe and have led to researchers working on identifying alternatives to lithium. 

However, discovering, testing, and employing these alternatives could become time-consuming and expensive. Finding the right alternative could require brute-force testing of millions of materials in trial-and-error experiments. To make this process more efficient, researchers must find an automatic and cost-effective way.

 

The Role of Artificial Intelligence and High-Performance Computing 

Microsoft and Pacific Northwest National Laboratory recently published a paper highlighting using AI and cloud high-performance computing (HPC) to accelerate the discovery and experimental verification of new materials.

 

Block diagram of an AI and HPC-based environment. Image courtesy of ArXiv

 

AI is employed in this context because of its proficiency in handling large datasets and its ability to quickly process, analyze, and derive patterns from vast amounts of information. This allows for more accurate, fast, and insightful decision-making, especially in material discovery.

By employing AI integrated with cloud HPC, the Microsoft research team could create a shortlist of 18 promising candidate materials among a list of 32 million at the onset. Afterward, material scientists at Pacific Northwest National Laboratory recommended incorporating more precise screening criteria and increasing the number of elimination rounds. Following the guidance, the team identified a potential candidate and adopted an approach that entailed substituting about half of the lithium atoms in a battery with sodium atoms.

Upon developing a functional battery using the novel chemistry, the researchers discovered that, although the battery was stable, it exhibited lower conductivity than required. Scientists are working to improve the technology for future material discoveries.

 

Future Prospects of Lithium Batteries

The impact of Microsoft’s research lies beyond reducing reliance on lithium for battery production but towards accelerating the fields of material science with automated solutions. Hopefully, this method can lead to the discovery of a true alternative to lithium that is safer, cheaper, and less volatile.