Using machine learning to analyze quantum material

Electrons and their behavior pose fascinating questions to quantum physicists, and recent innovations in sources, instruments and facilities may allow researchers to access even more information encoded in quantum materials.

However, these research innovations produce unprecedented – and as yet undecipherable – amounts of data.

“The information content in a piece of material can quickly exceed the total information content in the Library of Congress, which is about 20 terabytes,” said Eun-Ah KimProfessor of Physics at the School of Arts and Sciences, who is at the forefront of both research on quantum materials and harnessing the power of machine learning to analyze data from experiments with quantum materials.

An example of 3D X-ray diffraction data undergoing a phase transition upon cooling. The magenta graph shows special points associated with charge density waveforming as revealed by the machine learning algorithm X-TEC.

“The limited capacity of the traditional analysis method – largely manual – is quickly becoming the critical bottleneck,” Kim said.

A group led by Kim has successfully used a machine learning technique developed in collaboration with computer scientists at Cornell to analyze massive amounts of data from the quantum metal Cd2Re2O7, settle a debate on this particular material and pave the way for future machine learning, aided insight into new phases of issue.

The newspaper, “Using interpretable and unsupervised machine learning to tackle big data from modern X-ray diffraction,” published June 9 in the Proceedings of the National Academy of Sciences.

Cornell physicists and computer scientists collaborated to build an unsupervised and interpretable machine learning algorithm called XRD Temperature Clustering (X-TEC). The researchers then applied X-TEC to investigate the key elements of the pyrochlore oxide metal, Cd2Re2O7.

X-TEC analyzed eight terabytes of X-ray data in minutes, spanning 15,000 Brillouin zones (uniquely defined cells).

“We used unsupervised machine learning algorithms, which are a perfect fit to translate high-dimensional data into clusters that make sense to humans,” says Kilian Weinberger, professor of computer science in the Cornell Ann. S Bowers College of Computing and Information Sciences.

Thanks to this analysis, the researchers uncovered important insights into the electron behavior in the material and discovered what is known as the pseudo-Goldstone mode. They tried to understand how atoms and electrons position themselves in an orderly manner to optimize the interaction within the astronomically large ‘community’ of electrons and atoms.

“In complex crystalline materials, a specific multi-atom structure called the unit cell repeats itself in a normal arrangement, such as in a high-rise apartment building,” Kim said. “The repositioning we discovered is happening on a scale of every apartment unit, across the whole complex.”

Because the arrangement of the units remains the same, she said, it’s difficult to detect this repositioning by looking from the outside. However, the repositioning almost spontaneously breaks a continuous symmetry, resulting in a pseudo-Goldstone mode.

“The existence of the pseudo-Goldstone mode can reveal the secret symmetries in the system that are otherwise hard to see,” Kim said. “Our discovery was made possible by X-TEC.”

This discovery is important for three reasons, Kim said. First, it shows that machine learning can be used to analyze massive X-ray powder diffraction (XRD) data, prototyping applications of X-TEC as it scales. Available to researchers as a software package, X-TEC will be integrated into the synchrotron as an analysis tool at the Advanced Photon Source and at the Cornell High Energy Synchrotron Source.

Second, the discovery settles a debate about the physics of Cd2Re2O7.

“To our knowledge, this is the first time a Goldstone mode has been detected using XRD,” Kim said. “This atomic-scale insight into fluctuations in a complex quantum material will be just the first example of answering important scientific questions associated with any discovery of new phases of matter … using information-rich voluminous diffraction data.”

Third, the discovery shows what collaboration between physicists and computer scientists can achieve.

“The mathematical inner workings of machine learning algorithms are often no different from models in physics, but applied to high-dimensional data,” Weinberger said. “Working with physicists is a lot of fun because they are so good at modeling the natural world. When it comes to data modelling, they are really taking off.”

Co-authors include Geoff Pleiss, MS ’18, Ph.D. ’20; Jordan Venderley, MS ’17, Ph.D. ’19; Krishnanand Mallayya, postdoctoral researcher in the Lab of Atomic and Solid State Physics; and Michael Matty, a PhD student in the field of physics. The research was done in collaboration with colleagues from Argonne National Lab.

This research was supported by a grant from the National Science Foundation and a grant from the Department of Energy.

Kate Blackwood is a writer for the College of Arts and Sciences.

Leave a Comment

Your email address will not be published.