Data Science and Analytics for Materials Imaging and Quantification: Session II: Data-led Approaches for 3D Characterization & X-Ray Imaging
Sponsored by: TMS Structural Materials Division, TMS Materials Processing and Manufacturing Division, TMS: Advanced Characterization, Testing, and Simulation Committee, TMS: Integrated Computational Materials Engineering Committee
Program Organizers: Emine Gulsoy, Northwestern University; Charudatta Phatak, Argonne National Laboratory; Stephan Wagner-Conrad, Carl Zeiss Microscopy; Marcus Hanwell, Brookhaven National Laboratory; David Rowenhorst, Naval Research Laboratory; Tiberiu Stan, Asml
Monday 2:00 PM
March 15, 2021
Room: RM 16
Location: TMS2021 Virtual
Session Chair: Charudatta Phatak, Argonne National Laboratory
2:00 PM
Convolutional neural network-assisted recognition of nanoscale L12 ordered structures in face-centred cubic alloys: Yue Li1; Leigh Stephenson1; Raabe Dierk1; Baptiste Gault1; 1Max-Planck-Institut für Eisenforschung GmbH
Nano-size L12-type ordered structures are commonly used in FCC-based alloys to improve mechanical properties. They are often coherent with matrix, which makes them challenging to be characterized. Spatial distribution maps (SDMs) are used to probe local order by interrogating three-dimensional (3D) distribution of atoms within reconstructed atom probe tomography (APT) data. However, it is inefficient to manually analyse a large APT data and make crystal structure recognitions. Here, we introduce an intelligent L12-ordered structure recognition method based on convolutional neural network (CNN). The SDMs of simulated L12-ordered structure and FCC matrix were generated as training and validation datasets. These simulated images were used to train a CNN-based L12-ordered structure recognition model. Finally, the approach was tested on of L12–type δ'–Al3(LiMg) particles with an average radius of 3nm in an FCC-based Al-Li-Mg system. The minimum radius of detectable nanodomain is even down to 0.5nm. The proposed method is promising to be extended to other ordered structures in future.
2:20 PM
Deep Neural Network Facilitated Complex Imaging of Phase Domains: Longlong Wu1; Pavol Juhas1; Shinjae Yoo1; Ian Robinsion1; 1Brookhaven National Lab
Single-particle imaging by using inversion of coherent x-ray diffraction was put forward more than decades ago. Phase retrieval methods for the reconstruction of a single particle image from the modulus of its Fourier transform have been extensively applied in X-ray Structural Science. Here, we will present a deep neural work model, which gives a rapid and accurate estimate of the complex single-particle image. We demonstrate a way to combine the model with conventional iterative methods to refine the accuracy of the reconstructed results starting from the proposed deep neural work model. This developed deep neural network model opens up opportunities for fundamental research on using Machine Learning to do phase retrieval at high speed and accuracy. This is important for real-time inversion of coherent x-ray diffraction patterns for ultrafast time-resolved studies at XFELs as well as strong-phase objects where the phase domains found inside crystals by Bragg Coherent Diffraction Imaging.
2:40 PM
Quantitative X-ray Fluorescence Nanotomography: Mingyuan Ge1; Xiaojing Huang1; Hanfei Yan1; Wilson Chiu2; Kyle Brinkman3; Yong Chu1; 1Brookhaven National Laboratory; 2University of Connecticut; 3Clemson University
The Hard X-ray Nanoprobe at NSLS-II provides nanoscale 3D multi-modality imaging capability which is extremely useful in studying materials with complex internal structures. An important scientific problem is to investigate phase/grain boundaries of multi-component materials during or after material processing such as sintering, which generally has profound impacts on material’s property and functionality. X-ray fluorescent (XRF) nanotomgraphy is well-suited for the task. However, accurate quantification of XRF tomography is hampered by a well-known self-absorption problem. In this work, we present an approach to tackle the problem and apply to investigate the phase separation in a mixed ionic and electronic conducting ceramic material (Ce1-xGdxO2), a modal system generates broad interests in high temperature membranes and solid fuel cell applications. We will show how accurate absorption correction leads to a discovery of new phase in this system and providing key information in elucidating the structure evolution during the material synthesis process.
3:00 PM
Materials Characterization in 3D Using High Energy X-ray Diffraction Microscopy: Irradiated and Deformed Materials: Hemant Sharma1; Peter Kenesei1; Jun-Sang Park1; Zhengchun Liu1; Jon Almer1; 1Argonne National Laboratory
This talk will focus on recent developments in the High Energy Diffraction Microscopy (HEDM) technique for studying irradiated and deformed materials. Highlighting the various challenges in both experimentation and data analysis, we will discuss development of novel experimental techniques at the Advanced Photon Source and advanced data analysis techniques leveraging machine learning capabilities. Deep-Learning reinforced data analysis enables real-time reconstructions of highly complex microstructures. Furthermore, we will demonstrate a new data archival format which enables users to directly correlate diffraction information at the sub-grain level with grain-resolved properties.
3:20 PM
Understanding the Keyhole Dynamics in Laser Processing Using Time-resolved X-ray Imaging Coupled With Computer Vision and Data Analytics: Jongchan Pyeon1; Joseph Aroh1; Runbo Jiang1; Andy Ramlatchan2; Benjamin Gould3; Anthony Rollett1; 1Carnegie Mellon University; 2NASA Langley Research Center; 3Argonne National Laboratory
During laser processing of metals, localized metal evaporation resulting in the formation of a keyhole shaped cavity can occur if high enough energy densities are used. An unstable keyhole can have deleterious effects in certain applications (e.g. laser powder bed fusion) as it increases the likelihood of producing defects such as spatter or porosity. In this work, the dynamics of keyhole fluctuations were probed using in-situ synchrotron x-ray imaging at the Advanced Photon Source across a range of materials and laser parameters. The high temporal and spatial resolution of these experiments results in large datasets which were processed using computer vision techniques in order to extract time-resolved quantitative geometric features. These features were analyzed and a correlation was made between local keyhole geometry variation and the presence/absence of processing defects. Likewise, multivariate statistical tools were employed to understand the relationship between processing parameters, material properties, and keyhole geometry.
3:40 PM Question and Answer Period