Understanding Quantum Matter Data
Scientific advances on quantum mechanical properties of societies of electrons can result in new technological paradigms that can revolutionize human societies. In pursuit of the new technological paradigms that quantum systems will enable, modern quantum physics has focused attention on strongly correlated quantum matter (SCQM). In SCQM, strong interactions between individual electrons cause strong correlations in the motion of electrons, creating surprising physical qualities.
Intense research over the past decade has generated vast quantities of data on SCQM. Analyzing massive data sets in terms of theoretical models, however, presents a daunting challenge. Because quantum mechanical imaging is probabilistic and because inference from data must be subject to the laws of physics, researchers need new data science approaches to meet the data-driven challenges of SCQM.
To bridge the gap, Eun-Ah Kim, Physics, in collaboration with Kilian Q. Weinberger, Computer Science, and Andrew Gordon Wilson, New York University, is developing machine learning tools that will connect the experimental data on SCQM with theoretical understandings of quantum physics. Initially, the interdisciplinary team is building interpretable machine learning tools for analyzing atomic scale image data. They’re also constructing unsupervised machine learning tools for diffraction data obtained by x-ray scattering off of quantum material. The new tools will result in new understanding as well as insights derived from machine learning, which will guide development of new experiments. The team plans to integrate the resulting machine learning tools into the Cornell High Energy Synchrotron Source (CHESS) beamline.
This project—converging experimental data, quantum theory, and data science—represents the first step toward a future institute that would connect academic institutions and scattering experiment facilities nationwide.