Argonne National Laboratory

Materials Science

Theme I: Method Development

We are developing novel neutron and X-ray scattering methods and data analysis tools that advance our scientific programs and enable new science.

Our group plays an active and important role in expanding the capabilities of existing instrumentation and developing novel instrumentation and methods. These enable our investigations of the relation of local disorder and short-range correlations embedded in crystalline materials to their properties of importance for future energy technologies

Single-Crystal Diffuse Scattering

Our group led the development of novel single-crystal diffuse neutron and synchrotron X-ray scattering techniques, which have the potential for broad impact in determining the relations between local correlations on the nano- and mesoscale and emergent physical phenomena. These efforts led to the development and funding of the CORELLI single-crystal diffuse scattering instrument at the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory (https://​neu​trons​.ornl​.gov/​c​o​relli).

Single-crystal diffuse neutron scattering (left) measured with inorganic halide perovskite (CsPbBr3). A Monte Carlo simulation of the local structure reveals that quasi-2D regions of correlated Br6 octahedra (right) give calculated diffuse scattering intensity (middle) in good agreement with measurement.

3D-ΔPDF Analysis

The availability of scattering intensities from single crystals over large volumes of momentum transfer enables transformation into three-dimensional pair-distribution functions (3D-PDF) via Fourier transforms. By subtracting the Bragg peak intensities from the observed data before performing the transforms, we obtain the 3D-ΔPDF, which is the probability of a two-particle pair vector with respect to the average structure.

The 3D-ΔPDF transform of measured diffuse scattering (left) from crystalline sample with complex disorder contains negative (blue) and positive (red) intensities that denote interatomic vectors that are less (blue) or more (red) likely to occur in the disordered than average structure.

Machine Learning and Spectral Analysis of Large Sets of Diffuse Scattering Data

We develop and utilize advanced computational tools for efficient analysis of large sets of single-crystal X-ray and neutron scattering data. These developments allow us to ensure that data reduction and analysis keep pace with the speed of data collection possible at modern, high-brightness light sources, such as the Advanced Photon Source at Argonne. They also provide completely novel modes of probing subtle structural correlations, providing insight on how such ​“hidden order” affects material properties.

Unsupervised machine learning of single-crystal X-ray scattering data finds two clusters of Bragg peaks (green and yellow) with different temperature dependences as well as diffuse scattering clusters (red and blue) from Goldstone mode fluctuations in Cd2Re2O7 (Proc. Natl. Acad. Sci.).

 

Further information can be found in the home page for the Analysis of X-rays with Machine Learning and Statistics project.