Finding new, effective electrolytes is a key challenge in developing next-generation batteries for electric vehicles, phones, laptops, and large-scale energy storage systems.
The ideal electrolyte must strike a balance between several factors—stability, conductivity, and efficiency—which often conflict with one another. According to Ritesh Kumar, a Postdoctoral Fellow at the University of Chicago, "The electrodes have to satisfy very different properties at the same time. They always conflict with each other."
Kumar is the first author of a new paper in Chemistry of Materials, where researchers have applied artificial intelligence (AI) and machine learning to tackle this issue. The paper introduces a new method for identifying molecules that can maximize the three key properties needed for a battery electrolyte: ionic conductivity, oxidative stability, and Coulombic efficiency.
Using a dataset derived from over 250 research papers, the team applied AI to calculate what they call the "eScore" for various molecules. This eScore weighs the three critical factors and identifies molecules that meet all three requirements. The team has already tested this approach and successfully identified a molecule that performs as well as the best electrolytes available today, marking a significant advancement in a field traditionally driven by trial-and-error.
"Electrolyte optimization is a slow and challenging process," said Jeffrey Lopez, an assistant professor at Northwestern University. "These data-driven frameworks are critical for accelerating the development of new battery materials and leveraging AI's potential to speed up scientific research."
AI technology helps researchers by highlighting promising molecules, saving time, energy, and resources by preventing wasted efforts on unproductive candidates. At the University of Chicago, AI is already being applied to research in areas such as cancer treatments, water purification, and quantum materials.
https://github.com/BrianTMN/Rainbow-Six ... ited-money
https://github.com/JasonBTD/Temple-Run- ... d-gems-MOD
https://github.com/KevinBMT/50-Tiny-Roo ... -coins-MOD
https://github.com/DavidFNT/Grand-Gangs ... -money-MOD
https://github.com/AndrewGNT/Happy-Dine ... d-gems-MOD
https://github.com/StephenGNA/Realms-of ... everything
https://github.com/EricBNT/Run-Goddess- ... y-and-gems
https://github.com/AlexanderPGN/State-C ... d-gems-MOD
https://github.com/KyleNBO/Spider-Fight ... -money-MOD
https://github.com/ScottBND/Flex-City-u ... d-gold-MOD
https://github.com/JustinDGT/Massive-Wa ... d-gold-MOD
https://github.com/TravisMBR/Agent-Hunt ... -money-MOD
https://github.com/ShaneGNC/War-Machine ... amonds-MOD
https://github.com/TimothyKBT/Standoff- ... d-gold-MOD
https://github.com/ChrisBNC/Zombie-Apoc ... d-gold-MOD
https://github.com/GregTNB/Critical-Str ... -money-MOD
https://github.com/DustinGVT/Bowmasters ... d-gems-MOD
https://github.com/NathanDTB/Kick-the-B ... d-gold-MOD
https://github.com/FrankJBD/Temple-Run- ... d-gems-MOD
https://github.com/CraigHTB/Shadow-Figh ... d-gems-MOD
https://github.com/JoelKNV/Space-Surviv ... d-gems-MOD
https://github.com/TroyMBC/World-War-He ... d-gold-MOD
https://github.com/AdamBNC/Bed-Wars-unlimited-money-MOD
https://github.com/BrentGMM/War-Thunder ... -money-MOD
https://github.com/BruceFNV/Critical-Op ... -money-MOD
https://github.com/CaseyBTJ/ODIN-VALHAL ... amonds-MOD
https://github.com/DarrenBMT/Chainsaw-J ... -money-MOD
https://github.com/DennisKMB/Once-Human ... d-Crystgin
https://github.com/DevonMNB/Suzy-s-Rest ... -money-MOD
https://github.com/DustinBRM/FPS-Strike ... -money-MOD
https://github.com/KevinHSN/Grunt-Rush- ... y-and-gems
https://github.com/DaveALG/Top-Heroes-K ... everything
https://github.com/jasonRBL/Archers-Onl ... d-gems-MOD
Given the vast number of possible molecules—1060, or a one with 60 zeros after it—AI's ability to filter through billions of possibilities and flag promising candidates is an invaluable tool. "It would have been impossible for us to go through hundreds of millions of compounds manually," said Chibueze Amanchukwu, an assistant professor at the University of Chicago.
Amanchukwu likened the research process to listening to music online. Imagine an AI trained on a person’s unique music preferences. This AI would be able to go through existing playlists and predict which songs the person would like. The next step would be an AI that can create new playlists based on these preferences. In the case of electrolyte research, the goal is for AI to design entirely new molecules that fit the necessary criteria.
The team has been curating the data for the AI since 2020, manually compiling thousands of potential electrolytes from over 50 years of research literature. One challenge in this process is the need to extract data from images in research papers, such as charts and diagrams, which are not readable by most large language models.
Despite the extensive dataset, there is still work to be done. "I don't want to find a molecule that was already in my training data," said Amanchukwu. "I want to look for molecules in very different chemical spaces." The team found that the AI performed well when identifying molecules similar to those in its training data, but it struggled with unfamiliar compounds, signaling that there is still much to improve as they continue developing AI for designing next-generation battery electrolytes.