Supercomputer Models Speed New Materials Discovery Share your comment!

Stanene is softer and consequently much more rippled than its cousins graphene and silicene. Image: Mathew Cherukara, Badri Narayanan and Subramanian Sankaranarayanan, Argonne National Laboratory.

Courtesy Argonne National Laboratory

Machine Learning helps confirm properties of new “nanomaterials” 10x+ faster

Researchers using supercomputers are taking advantage of machine-learning algorithms to accurately predict the physical, chemical and mechanical properties of nanomaterials, which could accelerate the discovery and development of new materials.

The supercomputer simulations reduce the time it takes to produce such predictions, from years to months, according to a recent press release by Lawrence Berkeley National Laboratory (LBNL).

The researchers at Argonne National Laboratory used the Intel Xeon-based Edison supercomputer at LBNL’s National Energy Research Scientific Computing Center to study the structure and thermal conductivity of stanene, a 2D material made up of a one-atom-thick sheet of tin, the release said.

Their efforts, which involved parameters known as a “force field,” produced the first atomic-level computer model that accurately predicted stanene’s structural, elastic and thermal properties, the release said.

By using machine-learning algorithms, they were able to produce an accurate model within a few months. In the past, the researchers would have had to rely on their own intuition to build the model, which traditionally has taken years, the release said.

“Characterizing a force field like this takes a really long time to do by hand – a matter of years – so machine learning is a way of automating this process, speeding it up, so that we can explore more of the parameter space than a human can do and get a faster, more accurate fit,” said Mathew Cherukara, an Argonne postdoctoral researcher and lead author of a study that was recently published in The Journal of Physical Chemistry Letters.

The process the researchers developed is not material dependent, so in the future, they can use the machine-learning algorithms to analyze different classes of materials, the release said.

“What surprised us is that we can get more out of these models than we thought possible,” Cherukara said. “We can use, for example, the same model framework that we used for stanene to describe water – two very different materials. The underlying analytical model is the same. The difference is the numbers that go in and the parameterization of the model.”

Other Argonne researchers are already using the modeling framework that the study’s authors built. In the future, the study’s authors said they envision a future when non-experts can even use their process to explore the properties of materials.

Posted on March 17, 2017 by Wylie Wong, Slashdot Media Contributing Editor