AI was utilized by Microsoft and Pacific Northwest National Laboratory to reduce a list of 32 million potential materials to just 18 in a matter of hours rather than years.
Although there has been some progress made, the world is going in the direction of renewable energy. Battery technology is not yet sophisticated enough to smooth out the inconsistency that is caused by power sources such as wind and solar. This is one of the factors that contributes to the problem. The development of novel battery materials is a laborious process; nevertheless, researchers at the Pacific Northwest National Laboratory (PNNL) of the Department of Energy have collaborated with Microsoft to accelerate the progression of this process through the application of artificial intelligence. By utilizing the cloud-based Azure Quantum Elements service offered by Microsoft, researchers were able to narrow down a list of 32 million potential battery materials to only a few dozen that are capable of being tested instantly.
In most cases, the process of searching for new battery materials begins with an examination of the literature that is currently available. In spite of this, scientists have a tendency to disclose their achievements rather than their failures, which results in gaps in the data that is available to the public. The computation was the initial step in the relationship between Microsoft and PNNL. The artificial intelligence system compiled a list of 32 million potential materials based on the makeup of those elements. Next, the algorithm eliminated substances that lacked the ability to maintain their chemical stability. This resulted in the list being reduced to 500,000, and then finally to 800.
In the present moment, Azure Quantum Elements does not make use of quantum computing; nonetheless, this is the direction that Microsoft intends to take the service. According to the company, it is already working on establishing the workflows and technologies that are necessary for that leap. Despite this, the cloud service is now operating on conventional computer gear, which allows it to handle both high-performance computing (HPC) and artificial intelligence (AI) workloads.
The past year has shown us all that artificial intelligence is quick, but it is not reliable. Following the process of narrowing down the list of materials, Microsoft claims that it utilized an HPC layer in order to carry out a more precise examination of the potential energy states of each contender. After that, artificial intelligence and high-performance computing were integrated to do molecular dynamics simulations in order to forecast the movement of atoms within each material, which is an essential component of batteries.
Due to the aforementioned factors, the list was narrowed down to a total of 150 applicants, each of whom was evaluated by HPC for their practicability, taking into account factors such as availability and cost. The list was reduced to just 23 elements as a result of this, and five of those entries turned out to be known information. When compared to the years of trial and error that were spent in the laboratory, this approach only took eighty hours.
The availability of these tools on the cloud could be a boon for scientists, including those working in institutions that have their very own supercomputers already in operation. Because these resources are shared, it is possible that teams will have to wait until there is time available for computing before they can continue their research. Due to the involvement of Microsoft, the project was able to move forward more quickly, which enabled researchers at PNNL to immediately test the 18 unique materials.
These compounds have been produced by scientists, one of which is a solid electrolyte that has the potential to be more stable and cost-effective than the materials that are now available (see top). In the past, scientists believed that the presence of sodium and lithium in this substance was damaging to batteries due to the fact that these atoms had the same charge but different sizes. However, salts are emerging as an intriguing possibility for the development of future battery technology. The PNNL researchers discovered that the two appear to be beneficial to one another, and the battery that was produced as a result uses seventy percent less lithium, which is far more expensive than sodium.
At this point, the team at PNNL is unsure as to whether or not the AI-assisted battery materials will result in significant advancements; however, the rate at which we will learn this is a crucial factor. Microsoft is of the opinion that artificial intelligence systems such as Azure Quantum Elements will revolutionize all discovery-based fields, such as chemistry and material science, which are just the beginning.