As the demand for renewable energy grows, scientists are searching for better batteries that can store large amounts of electricity safely, cheaply and for long periods.
These batteries are especially important for storing power from wind and solar farms and strengthening the electricity grid.
Researchers at Lawrence Livermore National Laboratory (LLNL) have now developed a powerful new approach that combines artificial intelligence (AI) with physics-based computer simulations to better understand how batteries work at the atomic level.
Their findings, published in Energy Storage Materials and EES Batteries, could help speed up the development of the next generation of high-performance batteries.
Designing better batteries has always been difficult because the materials inside them are incredibly complex. Tiny changes in the way atoms are arranged can have a major effect on how well a battery stores and delivers energy. Traditional experiments and computer models often struggle to capture this level of detail.
The LLNL team tackled this challenge by combining molecular dynamics simulations with physics-informed machine learning. Molecular dynamics simulations create detailed computer models that show how individual atoms move and interact over time.
The AI system then learns from these simulations, allowing it to predict material behavior much faster than conventional methods while still following the laws of physics.
One study focused on sodium-ion batteries, which are attracting growing interest because sodium is abundant, inexpensive and widely available. Unlike lithium, sodium is found in many parts of the world, making it an attractive option for large-scale energy storage.
The researchers examined hard carbon, one of the most common materials used in sodium-ion battery anodes. Hard carbon has a highly irregular structure made up of tiny layers, pores and empty spaces. This complex structure gives sodium ions many possible pathways to move through the material, but it also makes the material extremely difficult to study.
Using LLNL’s powerful supercomputers, the researchers created detailed simulations showing how sodium ions move inside the hard carbon. These simulations revealed that the ions can travel between carbon layers, attach to surfaces or become trapped inside tiny pores. The AI model learned from these atomic-scale “movies” and identified eight different types of sodium-ion movement.
The study showed that as more sodium enters the material and the carbon becomes denser, the ions are more likely to cluster together or become trapped. These behaviors directly affect how quickly a battery can charge and discharge, as well as its overall safety. By understanding these patterns, scientists now have a clearer guide for designing hard carbon materials that allow sodium ions to move more freely.
The second study focused on lithium-ion batteries, which power smartphones, laptops and electric vehicles. Instead of studying the battery’s solid materials, the researchers investigated the liquid electrolyte that carries lithium ions between the battery’s electrodes.
Choosing the best electrolyte has traditionally relied on years of trial and error because countless combinations of solvents, salts and additives are possible. The LLNL team used molecular dynamics to build realistic three-dimensional models of these liquid mixtures and then trained AI to predict how stable they would be during battery operation.
The researchers discovered that the way molecules arrange themselves around lithium ions can dramatically change battery performance. In one example, simply replacing one lithium salt with another increased the predicted stability window by 57%, even though this important difference could not be detected using simpler computer models.
The team believes this new AI-guided approach could greatly reduce the time needed to discover better battery materials. Instead of testing thousands of possibilities in the laboratory, researchers can now use AI to identify the most promising candidates first.
Although the current work focuses on lithium-ion and sodium-ion batteries, the scientists say the same method could also help develop many other advanced energy storage technologies, accelerating the search for safer, longer-lasting and more efficient batteries.
Source: KSR.
