Microsoft’s Quantum Computing Tools Expedite Battery Materials Research

January 26, 2024 by Shannon Cuthrell

Laboratory researchers are using Microsoft’s advanced computing tools to screen promising battery materials faster than traditional computation and testing methods.

Chemistry and materials science researchers often spend years narrowing down millions of potential structures and element combinations to find ideal candidates for new battery technologies. A co-innovation partnership between the Pacific Northwest National Laboratory (PNNL) and Microsoft aims to reduce that timescale to weeks or days, using the tech giant’s high-performance computing and cloud-based artificial intelligence (AI) software to develop battery materials with enhanced efficiency and energy storage. 


Samples of the solid electrolyte materials discovered with Microsoft’s high-performance computing and AI technology.

Samples of the solid electrolyte materials discovered with Microsoft’s high-performance computing and AI technology. Image used courtesy of Microsoft/by Dan DeLong


The Washington State-based partners will target computation-first processes, allowing researchers to focus the early stages of their projects on the most promising materials. 

The platform’s AI-informed calculations can quickly compare materials for energy, stress, force, band gap, and mechanical characteristics. The models can predict properties 1,500 times faster than conventional density functional theory calculations. 

PNNL and Microsoft’s Quantum team previously discovered a new battery electrolyte using 70% less lithium content than other industry innovations. Replacing part of the composition with sodium avoids the supply chain issues of standard lithium-ion batteries made with scarce and expensive materials. Conventional liquid electrolytes are also highly flammable, but alternatives like solid-state electrolytes offer increased safety and stability. 

Azure Quantum Elements accelerated the pair’s initial research and development phase, condensing years of trial-and-error into one week. With computational simulation and laboratory validation, researchers could prioritize screening high-potential materials. 

The partnership will initially focus on computational chemistry and materials science before expanding to other disciplines that would benefit from quantum computing resources, such as pharmaceuticals. PNNL recently revealed plans to bring its AI research projects together under one roof, launching a new Center for AI. 

The collaboration builds on their ongoing work with the Azure Quantum Elements platform, which combines Microsoft’s high-performance computing systems, Azure cloud service, and AI simulation models. 


Accelerating Energy Storage Research and Development

Usually, materials synthesis involves reading existing studies on competing materials and devising new approaches. Hypothesis testing comes next, with multiple changes along the way. Developing new materials requires rounds upon rounds of process of elimination and calculations. 

With AI, scientists can synthesize billions of data bits in far less time than incumbent automated or manual methods. Azure Quantum Elements uses advanced AI models to find materials capable of unlocking energy-on-demand. 

For example, PNNL and Microsoft researchers used AI simulations to synthesize and test millions of energy storage materials in under a week. Microsoft trained its AI to assess all workable battery elements and determine ideal combinations. Azure Quantum Elements evaluated the thermodynamic phase stabilities of 32.6 million potential inorganic materials, predicting 589,609 stable materials within days. 


Stages of Microsoft and PNNL’s battery materials research project.

Stages of Microsoft and PNNL’s battery materials research project. Image used courtesy of Microsoft


Then, another set of AI tools filtered out candidate molecules based on reactivity, redox potential, band gap, and conductivity. The models picked out 800 promising materials with sufficient stability. 

Using density functional theory, the high-performance computing verification process identified candidates based on each material’s energy relative to its other states. Molecular dynamics simulations studied the atom movements and molecules in each option, narrowing the list to 150. After weighing practical advantages and disadvantages, including cost and availability, the pool was reduced to 23, 18 of which featured previously unknown compositions. The whole process took only 80 hours.


PNNL’s battery materials scientists collaborate on a project.

PNNL’s battery materials scientists collaborate on a project. Image used courtesy of PNNL (screenshot)


In the following prototyping phase, the researchers manually ground solid precursors of the materials with a mortar and pestle and then compacted them into a small pellet with a hydraulic press. Each pellet was placed into a vacuum tube furnace, heated to 450-650°C, and then carefully transferred to an argon-filled glove box to minimize oxygen or water contamination. Then, researchers ground the samples into a fine powder for additional analysis. 

In the next phase, researchers will develop and test other material candidates recommended by Microsoft’s models.