Harnessing machine learning to analyze quantum material
An example of 3D X-ray diffraction data going through a phase transition upon cooling. The magenta plot shows special points associated with charge density wave formation as they were revealed by the machine learning algorithm X-TEC. Credit: Krishna Mallayya/Provided

Recent innovations in sources, instruments and facilities allow researchers to potentially access more of the information contained in quantum materials.

Volumes of data are indecipherable thanks to these research innovations.

According to Eun-Ah Kim, professor of physics in the College of Arts and Sciences, the total information content in the Library of Congress is about 20 terabytes.

Kim said that the limited capacity of the traditional mode of analysis was becoming a critical problem.

A group led by Kim has successfully used a machine learning technique developed with Cornell computer scientists to analyze massive amounts of data from the quantum metal Cd 2 Re 2 O 7

The paper "Harnessing Interpretable and Unsupervised Machine Learning to Address Big Data from Modern X-ray Diffraction" was published in June.

Cornell physicists and computer scientists worked together to build a machine learning tool. The researchers used X-TEC to investigate the elements of the metal.

In a matter of minutes, X-TEC analyzed eight terabytes of X-ray data.

Kilian Weinberger is a professor of computer science at the Cornell Ann. The college of computing and information science is called the S Bowers College.

The researchers were able to detect the pseudo-Goldstone mode thanks to this analysis. They were trying to understand how atoms and electrons position themselves in a way that maximizes their interaction with one another.

Kim said that in a high-rise apartment complex the unit cell repeats itself in a regular arrangement. At a scale of each apartment unit, the repositioning takes place.

She said that it is difficult to detect the positioning of the units because they remain the same. The pseudo-goldstone mode is caused by the repositioning almost spontaneously.

Kim said that the secret symmetries in the system can be revealed by the existence of pseudo-Goldstone mode. X-TEC helped us find our way.

The discovery is significant for three reasons. It shows that machine learning can be used to analyze X-ray powder diffraction data and serve as a prototype for applications of X-TEC. X-TEC will be used at the Advanced Photon Source and the Cornell High Energy Synchrotron Source.

There was a debate about the physics of Cd 2 Re 2 O 7.

This is the first instance of the detection of a Goldstone mode using XRD. This atomic scale insight into fluctuations in a complex quantum material will be the first example of answering key scientific questions.

Physicists and computer scientists can work together.

Weinberger said that the mathematical inner workings of machine-learning programs are similar to models in physics. Physicists are great at modeling the natural world. They hit the ground running when it came to data modeling.

More information: Jordan Venderley et al, Harnessing interpretable and unsupervised machine learning to address big data from modern X-ray diffraction, Proceedings of the National Academy of Sciences (2022). DOI: 10.1073/pnas.2109665119 Journal information: Proceedings of the National Academy of Sciences