AI-driven closed-loop control of lithium plating in batteries
The paper introduces AI‑LOMAC, an AI‑driven closed‑loop framework that integrates dynamic impedance spectroscopy and vision‑language models to detect lithium plating onset over five minutes early and autonomously mitigate it in real time, thereby enhancing battery safety, cycle life, and scalability across various cell formats.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Batteries power our modern world, from the smartphones in our pockets to the electric cars on our roads, but they carry a hidden vulnerability that threatens their safety and lifespan. Inside a battery, electricity is generated by lithium ions shuttling back and forth between two electrodes. Under normal conditions, these ions slide neatly into the layers of the electrode material, like books sliding onto a shelf. However, when a battery is charged too quickly, in cold weather, or pushed beyond its limits, the ions can no longer fit inside. Instead, they pile up on the surface, forming a metallic layer of lithium. This phenomenon, known as lithium plating, is dangerous. The metal can grow into sharp, needle-like structures called dendrites that pierce the internal separator, causing a short circuit that can lead to fire or explosion. Even if a short circuit does not occur, this plated lithium often becomes stuck and inactive, permanently reducing the battery's capacity. For years, scientists have struggled to detect this plating early enough to stop it, usually only noticing the problem after the damage is already done.
A team of researchers has now introduced a new system that changes this dynamic, shifting from waiting for failure to preventing it in real time. They developed an artificial intelligence framework called AI-LOMAC, which acts as a continuous, closed-loop guardian for the battery. Unlike traditional battery management systems that rely on fixed rules, such as stopping charging when a specific voltage is reached, this new system watches the battery's internal state as it happens. It uses a technique called dynamic impedance spectroscopy, which sends tiny, harmless electrical signals through the battery to measure how the surface of the electrode is behaving. The system focuses on a specific property called surface capacitance, which changes subtly as lithium begins to settle on the anode. By tracking these changes, the AI can identify the exact moment lithium plating is about to start, detecting it more than five minutes before any visible metal dendrites would appear.
Once the system spots the early warning signs, it does not simply shut down the battery. Instead, it takes immediate, autonomous action. The AI instantly reduces the charging current to a much lower, safer rate. This pause allows the newly formed lithium to dissolve and slide back into the electrode layers, a process called re-intercalation, effectively undoing the damage before it becomes permanent. The researchers tested this approach across a wide range of battery sizes, from small coin-sized cells used in research to large, 314-amp-hour prismatic batteries used in energy storage systems. In every case, the AI-driven system successfully suppressed the growth of lithium dendrites. In tests with commercial coin cells, the system reduced the amount of inactive, residual metallic lithium by 73 percent compared to unregulated batteries. Furthermore, batteries managed by this AI retained 84 percent of their capacity after 100 cycles of harsh, fast charging, whereas unregulated batteries under the same conditions lost their capacity much faster.
The power of this approach lies in its ability to see the invisible. Using custom-made cells that allowed the researchers to look directly at the electrode surface with optical microscopes, they confirmed that the AI's warning signal arrived long before any lithium could be seen with the naked eye. In one experiment, the system detected the onset of plating and reduced the current while the electrode surface remained perfectly clean. In contrast, batteries without this intervention developed visible clusters of lithium deposits that eventually turned into a messy, inactive layer. The system proved robust enough to handle the complexities of real-world batteries, which vary in size, chemistry, and age. It successfully managed everything from small laboratory cells to massive commercial pouch cells, adapting its detection and control strategies to each specific format without needing human intervention.
This work represents a significant step forward in battery safety and longevity, moving the industry away from reactive measures toward proactive protection. By bridging the gap between transient electrical signals and immediate physical control, the AI-LOMAC framework demonstrates that it is possible to manage the internal chemistry of a battery with precision. The researchers validated that this method works not just in theory, but in practice, across diverse battery types and under demanding conditions like fast charging and low temperatures. While the system currently focuses on lithium plating, the underlying architecture suggests a future where artificial intelligence acts as a continuous partner in battery management, capable of learning from operational data to refine strategies and perhaps even discover new ways to store energy. For now, the result is a tangible improvement: batteries that are safer, last longer, and can be charged faster without the fear of hidden damage accumulating inside.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.