Science & Technology (Commonwealth Union) – In the wake of the recent events, energy security has become a key focus for governments across the world.
When it comes to large-scale energy storage and strengthening national energy security, the United States requires more advanced battery technologies. Scientists at Lawrence Livermore National Laboratory (LLNL) are exploring several solutions to this challenge, with one approach standing out: physics-informed machine learning.
In two recent studies, LLNL researchers investigated how combining molecular dynamics simulations with physics-informed machine learning can reveal deeper connections between the structure and performance of complex battery materials. Using this integrated approach, they studied carbon anodes used in sodium-ion batteries and liquid electrolytes found in lithium-ion batteries.
LLNL scientist and study author Liwen (Sabrina) Wan indicated that these studies demonstrate that the structural complexity of battery materials is not merely a challenge to overcome, but can actually provide new opportunities for innovation.
She further pointed out that by incorporating this complexity into physics-informed machine learning models, they have the ability to forecast material properties and uncover design strategies that conventional methods are unable to identify.
The first study, published in Energy Storage Materials, focuses on sodium-ion batteries. Since sodium is widely available and can be sourced domestically, this technology could play a key role in creating a more secure and reliable U.S. energy storage supply chain.
“Sodium ions can move into all of that disorder, slipping between layers, settling on surfaces and filling nanopores,” said LLNL scientist and author Nikhil Rampal. “That complexity is part of what makes hard carbon so promising, but it is also what makes it so challenging to design.”
Scientists have spent years trying to determine how the atomic structure of hard carbon influences the movement of sodium ions within the material. In this study, researchers used LLNL’s advanced high-performance computing systems to model the motion and interactions of every atom as the material evolved over time.
Rampal indicated that they effectively produced an atom-by-atom visualization showing sodium ions as they move through the carbon, form clusters, or become confined within the structure.
The researchers then used these detailed simulations to train a machine learning model capable of predicting atomic interactions. This approach enables larger, longer, and more precise simulations at a lower cost. The model was also used to categorize sodium ion movement into eight distinct patterns, each defined by how the ions interact with the hard carbon structure.
Rampal pointed out that higher carbon density and greater sodium concentration cause ions to aggregate or become trapped inside nanopores, which has important consequences for battery charging performance and thermal stability.
The outcome is a quantitative relationship linking the material’s microstructure with sodium-ion movement, along with practical strategies for improving hard carbon performance. The researchers say this approach offers a clear route toward safely optimizing sodium-ion mobility, helping accelerate the future adoption of sodium battery technology.
The second study, published in EES Batteries, uses a similar approach to tackle another major battery challenge: developing improved electrolytes for lithium-ion batteries. Creating an optimal electrolyte is extremely complex because the number of possible combinations of solvents, salts, additives and concentrations is enormous, making complete experimental screening nearly impossible.
Traditional electrolyte prediction methods often depend on text-based descriptions that overlook the three-dimensional arrangement of molecules. Instead, the LLNL researchers used molecular dynamics simulations to create realistic 3D molecular structures and then supplied those configurations to a machine learning model. The model was able to forecast the statistical stability of each molecular arrangement.
The researchers’ main finding was that electrochemical stability is determined by the behavior of the entire molecular system rather than simply the individual components added together.
