An artificial intelligence (AI) program has identified a material not found in nature that could reduce the amount of lithium used in batteries by up to 70%.
The new material, a mixture of sodium, lithium, yttrium and chloride ions, is a type of mixed metal chloride, and was found to be the best choice out of 32 million candidates.
Lithium is the main component of secondary batteries, and demand for this metal has skyrocketed in recent years. However, the mining process to obtain that element is particularly important. energy intensive and often causes permanent water and land pollution. This means that many companies are looking for alternative materials to make batteries.
Pacific Northwest National Laboratory (PNNL) teamed up with Microsoft to do just that. Researchers used Microsoft’s Azure Quantum Elements tool to screen new materials for potential use in low-lithium batteries. The scientists published their findings on January 8th. preprint server arXiv.
battery It works by moving charged particles back and forth between positive and negative terminals, known as electrodes. When you connect a wire, lithium ions move from the negative electrode to the positive electrode through a conductive material called an electrolyte. Meanwhile, electrons can move through the wire in the same direction and extract energy from the battery.
Related: New battery technology could reduce charging times and increase EV range before the decade ends
In this study, the researchers focused on solid electrolyte materials that scientists hope to develop as safer and more efficient alternatives to current liquid electrolytes. Importantly, the electrolyte material must be compatible with the electrodes, completely blocking the movement of electrons within the battery while allowing lithium ions to easily pass through.
They started with more than 32 million candidates generated by substituting different elements into existing electrolyte structures and combined AI techniques to filter materials based on their properties.
“Many of these theoretical computer-generated candidates are actually not stable enough to be made in the lab, so the first step was to filter by stability.” candler smithmechanical engineers at the National Renewable Energy Laboratory told Live Science. In this first screening, within hours his 32 million material increased to 500,000 material.
The team then selected nine other criteria and used AI to apply them sequentially, classifying candidates by electronic properties, cost, and strength, narrowing the field to 18 finalists. “I was very impressed that they accomplished all this in just 80 hours of his computer time. It would have taken him 20 years to experimentally examine all this material.” said Smith. “Combining their machine learning pipeline with physics-based models of molecular dynamics will yield significant results and really speed up research.”
The researchers synthesized a series of final materials containing varying proportions of lithium, sodium, the rare earth element yttrium, and chloride ions. Interestingly, this mixture of lithium and sodium allows the material to conduct both types of ions (something previously thought not possible) and may also work in sodium-ion batteries. In particular, one sodium-rich version contains 70% less lithium than traditional batteries, potentially significantly reducing the price and environmental impact of these batteries in the future.
The team then tested the candidates’ electronic properties. “Ionic conductivity, or how fast lithium ions can travel, is a key property of the electrolyte and determines how quickly the battery can be charged, which is critical for electric vehicles,” Smith said. he explained.
Traditional lithium-ion batteries use a liquid organic solvent electrolyte, which allows for faster ion movement and faster charging times. However, the solvent is flammable and side reactions with the electrodes will degrade the battery over time. “Solid electrolytes have the advantage of being chemically stable and much less flammable. The disadvantage is that they don’t move the lithium ions as quickly, so charging times are slower,” Smith said. Ta.
The best-performing candidates identified by AI are an order of magnitude less conductive than current liquid electrolytes, which equates to a difference in charging time of 30 minutes versus 5 hours. Therefore, it is necessary to improve the electronic performance of materials to make them suitable for practical use. application. That said, the researchers built a working prototype from the final material and used it to power a light bulb, a Microsoft representative told his Live Science in an email.
Smith believes this is a great starting point. But the use of AI to streamline materials discovery is the most impactful outcome of this research, with the potential for the same machine learning pipeline to support research in hundreds of other related fields. He explained that there is a gender.
This is something that both Microsoft and PNNL are keenly considering in the future. “The new battery results are just one example, a testament to that,” said Brian Abrahamson, PNNL’s chief digital officer. statement. “We realized early on that the appeal of this field was the speed of AI to help identify promising materials and the ability to quickly implement those ideas in the lab. We have cutting-edge technology and scientific expertise.”
