NJIT
The open and spongy network within a porous transition metal oxide allows larger, double or triple charged ions to travel during the battery charging and discharge cycles.
Artificial intelligence is helping scientists decipher the new generation battery code that can replace lithium -ion technology. US researchers have discovered new porous materials capable of revolutionizing multivalent ion batteries.
New Jersey Institute of Technology (NJIT) researchers have resorted to artificial intelligence to face a critical problem for the future of energy storage: finding affordable alternatives and sustainable to lithium -ion batteries.
Upon discovering new porous materials, researchers may have paved the way for more powerful and sustainable energy storage, using abundant elements such as magnesium.
In published in Cell Reports Physical Sciencethe NJIT team, led by the teacher Burned DATTAsuccessfully applied general AI techniques to Quickly discover new porous materials capable of revolutionizing the batteries of multivalent ion.
These batteries, which Use abundant elementss such as magnesium, calcium, aluminum and zinc, represent a promising and economic alternative Lithium -ion batteries, which face supply and sustainability problems at global level.
Unlike traditional lithium -ion batteries, which resort to ions with only one positive charge, multivalent -ion batteries use elements whose ions have two or even three positive charges.
This means that they can potentially Store much more energybecoming highly appealing to future energy storage solutions.
However, the largest size and upper electrical charge of multivalent ions make their efficient accommodation difficult In batteries materials – an obstacle that NJIT’s AI -based investigation seeks to resolve directly.
“One of the biggest obstacles was not the lack of promising chemicals for batteries, but the Practical impossibility of testing millions of combinations of materials, ”explains Datta, in Da NJIT.
“We resorted to the Generative AI as a rapid and systematic way to go through this vast territory and identify the few structures that could really make multivalent batteries viable,” he adds.
“This approach allows us quickly explore thousands of candidates Potentials, dramatically accelerating the demand for more efficient and sustainable alternatives to lithium ion technology. ”
To overcome these obstacles, the team has developed an innovative approach based on two AI systems: A CRYSTAL DIFFUSION VARATIONAL AUTOENCODER (CDVAE) and a carefully adjusted large -scale (LLM) language model (LLM).
Together, these tools quickly explored thousands of new crystalline structures – Something so far impossible with traditional laboratory methods.
The model CDVAE was trained with extensive data sets of known crystal clear structures, allowing it to propose totally new materials with great structural diversity. In turn, the LLM was tuned to identify The closest materials to thermodynamic stability, a crucial factor for their practical synthesis.
“Our AI tools dramatically accelerated the process Of course, allowing us to identify five new porous transition metal oxide structures with remarkable potential, ”says Datta.“ These materials have large and open channels, ideal for the fast and safe transport of these bulky multivalent ions-a critical advance for next generation batteries. ”
The team validated the structures generated by IA Through mechanical-quantic simulations and stability tests, confirming that materials can, in fact, be synthesized experimentally and have great potential for real applications.
Datta also highlighted the broader implications of this approach: “This is more than discovering new battery materials-it’s about creating a quick and scalable method for explore any kind of advanced materialsfrom electronics to clean energy solutions, without resorting to long trial and error processes. ”
