Harnessing machine learning to analyze quantum matter

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

However, these research innovations are producing unprecedented – and so far indecipherable – volumes of data.

“The informational content of a document can quickly exceed the total informational content of the Library of Congress, which is about 20 terabytes,” said Eun-Ah Kim, professor of physics at the College of Arts and Sciences, who is at the forefront of quantum materials research and harnessing the power of machine learning to analyze data from quantum materials experiments.

An example of 3D X-ray diffraction data going through a phase transition on cooling. The magenta plot shows special spots associated with the formation of charge density waves as revealed by the X-TEC machine learning algorithm.

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

A group led by Kim has successfully used a machine learning technique developed with computer scientists at Cornell to analyze massive amounts of data from the quantum metal Cd2Re2O7, settling a debate about this particular material and setting the stage for future learning. automatic assisted by an overview of the new mater phases.

The paper, “Harnessing Interpretable and Unsupervised Machine Learning to Address Big Data from Modern X-ray Diffraction,” published June 9 in Nature.

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

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

“We used unsupervised machine learning algorithms, which are well suited for translating high-dimensional data into clusters that make sense to humans,” said Cornell Ann computer science professor Kilian Weinberger. S Bowers College of Computing and Information Science.

Through this analysis, the researchers uncovered important insights into the behavior of electrons in the material, detecting what is known as the pseudo-Goldstone mode. They were trying to understand how atoms and electrons position themselves in an orderly fashion to optimize interaction within the astronomical “community” of electrons and atoms.

“In complex crystalline materials, a specific structure of many atoms, the unit cell, repeats itself in a regular arrangement like in a high-rise apartment complex,” Kim said. “The repositioning we discovered is happening at the scale of each apartment, across the whole complex.”

Because the layout of the units remains the same, she said, it’s hard to detect this repositioning when 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 secret symmetries in the system that may otherwise be difficult to see,” Kim said. “Our discovery was made possible by X-TEC.”

This finding is significant for three reasons, Kim said. First, it shows that machine learning can be used to analyze large X-ray powder diffraction (XRD) data, serving as a prototype for applications of X-TEC as it evolves. X-TEC, available to researchers as a software package, will be integrated into the synchrotron as an analysis tool with the Advanced Photon Source and the Cornell High Energy Synchrotron Source.

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

“To the best of our knowledge, this is the first case of detection of a Goldstone mode using XRD,” Kim said. “This atomic-scale insight into the fluctuations of complex quantum material will be just the first example of answering the key scientific questions accompanying any discovery of new phases of matter…using large, information-rich diffraction data .”

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

“The mathematical 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 great fun because they’re so good at modeling the natural world. When it comes to data modeling, they’ve really gotten off to a good start.”

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

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.

Sherry J. Basler