IISc researchers use machine learning and amorphous materials to build high energy density batteries

The Indian Institute of Science (IISc) researchers, in a new study using a machine learning model and amorphous materials, have built batteries with a higher energy density.

While Lithium-ion batteries power most electronics, they have limited energy density as they can store only a certain amount of energy per mass or volume of the battery.

“In order to store even more energy with the same mass or volume, you will have to explore alternative energy storage technologies,” said Sai Gautam Gopalakrishnan, assistant professor at the Department of Materials Engineering, IISc.

Mr. Gopalakrishnan and his team have studied how to boost the movement of ions in magnesium batteries, which can have a higher energy density.

In their study applying a machine learning model, the team has shown that using amorphous materials as positive electrodes to build these batteries can significantly increase their rate of energy transfer.

Lithium ion or magnesium batteries contain a positive (cathode) and a negative (anode) electrode, separated by a liquid electrolyte. Each time a lithium or magnesium ion goes from the cathode to the anode or vice versa, energy is exchanged with the device.

Twice the amount

“In magnesium batteries, each magnesium atom can actually exchange two electrons, whereas each lithium atom can only exchange one electron with the external circuit. So, you can get close to twice the amount of energy per atom moved,” he said.

He added that the main bottleneck in commercialising magnesium batteries is the lack of good materials that can act as cathodes.

According to IISc, so far, scientists have largely been looking at crystalline materials, which have a periodically ordered arrangement of atoms. However, because magnesium moves very slowly within these materials, they are unable to absorb and release magnesium ions at a fast enough rate.

The team built a computational model of an amorphous vanadium pentoxide material and calculated how fast magnesium ions can move within it.

To build such models, scientists typically use a method called density functional theory (DFT), which accurately models systems at an electronic level.

Speed and accuracy

To combine speed and accuracy, the team used a machine learning framework. They first used density functional theory (DFT) to generate data on how the amorphous cathode would function at a small scale.

After training their machine learning model on this data, they used the model to perform Molecular Dynamics (MD) simulations.

“With MD, they were able to model the material at a larger scale – to get a better picture of how far the magnesium moves within the amorphous material and how long it takes. Compared to state-of-the-art crystalline magnesium materials, the team observed about five orders of magnitude improvement in the rate of magnesium movement in the amorphous form,” IISc said.