The 7th International Congress on 3D Materials Science (3DMS 2025): 3D Data Processing, Software, and Reconstruction Algorithms III
Program Organizers: Henry Proudhon, Mines Paris Centre Des Materiaux; Can Yildirim, European Synchrotron Radiation Facility
Thursday 12:10 PM
June 19, 2025
Room: Platinum Ballroom 7&8
Location: Anaheim Marriott
Session Chair: Henry Proudhon, Mines Paris Centre Des Materiaux
12:10 PM
Uncertainty in High-Angular-Precision 3D-EBSD Orientation Measurements Using Spherical Harmonic Indexing and Global Pattern Center Optimization on Deformed and Undeformed Samples: Gregory Sparks1; Michael Uchic2; Paul Shade2; Mark Obstalecki2; 1University of Dayton Research Institute; 2Air Force Research Laboratory
In previous work, we demonstrated the use of high-precision EBSD orientation measurements from spherical harmonic indexing (SHI) combined with global pattern center (PC) optimization in 3D-EBSD reconstruction of an undeformed Ni superalloy sample. In this talk, we discuss the layer-to-layer consistency of the orientations measured using this technique both from the previous undeformed sample and a separate, deformed Ni superalloy sample, and compare this with the consistency of repeated measurements on a single layer for both samples. We also investigate the difference between orientation measurements from a Si wafer at 70° tilt (using the same technique of high-precision SHI with PC optimization) and orientation measurements from the same Si wafer at 0° tilt (obtained from matching experimental electron channeling patterns (ECP) with simulated ECP) as a potential explanation for residual rigid-body orientation errors in 3D-EBSD relative to other orientation measurement modalities.
12:30 PM
On the Inverse Problem of Recovering Admissible Intragranular Strain Fields From High-Energy X-Ray Diffraction Data: Carter Cocke1; Andrew Akerson1; Sara Gorske1; Katherine Faber1; Kaushik Bhattacharya1; 1California Institute of Technology
Recovering strain fields from high-energy X-ray diffraction data is a challenging inverse problem that is commonly simplified to only obtain grain-averaged strains. Advanced techniques that obtain sub-granular strain fields often relax the requirement of either strain compatibility or stress equilibrium. Here, we present two methods that strongly enforce these requirements. The first approach is a post-processing step using grain-averaged strains and microstructural information from preexisting reconstruction algorithms. Using a finite-element formulation, the problem is posed as a linearly constrained least-squares problem that may be efficiently solved. We demonstrate the method through synthetic and experimental examples and recover intragranular strains with low error. Eschewing the need for a pre-reconstruction step, the second method directly inverts raw diffraction measurements. We discuss preliminary results and the associated challenges. The methods presented highlight that mechanical admissibility significantly constrains the solution space, which may be exploited to recover intragranular strains with high accuracy.
12:50 PM
Enhancing Polycrystalline-Microstructure Reconstruction From Diffraction Microscopy With Phase-Field Post-Processing: Marcel Chlupsa1; Zach Croft1; Katsuyo Thornton1; Ashwin Shahani1; 1University of Michigan
A new protocol using a phase-field model processes 3D reconstructions of polycrystalline microstructures from synchrotron-based high-energy X-ray diffraction microscopy (HEDM) data. These reconstructions face challenges such as heterogeneous grain sizes, detector limitations, and overlapping Bragg peaks, among other errors. The noise often results in non-physical roughness at grain boundaries (GBs), complicating property measurements like tortuosity, curvature, and GB character. Such uncertainties hinder estimates of diffusivity, corrosion resistance, electrical resistance, and fracture strength. To address these issues, we employ phase-field equations to create a space-filling grain map that adheres to the physics of the microstructure. It penalizes high-energy grain shapes and configurations while promoting GB smoothing. High-confidence regions are preserved using a completeness-based mobility parameter in the phase-field model. This protocol offers an alternative to conventional image processing-based approaches. It can be applied to any diffraction-based, 2D or 3D reconstruction with grain and confidence data, including polyphase materials.
1:10 PM
Identifying SOFC Triple Phase Boundaries From 3D FIBSEM: Alexander Hall1; Chengge Jiao1; Bartłomiej Winiarski1; Patrick Barthelemy1; V. V. Rohit Bukka1; 1Thermo Fisher Scientific
Electrochemical reactions at Triple Phase Boudaries (TPB) largely determine the efficiency of Solid Oxide Fuel Cells (SOFC). SOFC microstructure can be viewed in 3D using FIB-SEM. Specifically, the combination of Backscattered Electron (BSE) and Through-the-Lens Detector (TLD) Secondary Electron (SE) images allows easy identification of the three phases in a SOFC electrode sample. By using a plasma ion source, we reconstructed a sample volume roughly ten times larger than prior efforts. Additionally, we used deep learning segmentation of the Pore, YSZ, and Ni phases of the SOFC electrode. With this information, we characterized the volume fraction, two-phase interface densities, and TPBs from a volume large enough to be representative of the sample. Our work represents many technical improvements to SOFC characterization, but also a standardized workflow on which future SOFC characterization efforts can be based.
1:30 PM
Unlocking 3D Nanoparticle Shapes From 2D HRTEM images: Deep Learning for Classification and Denoising at Atomic Resolution: Romain Moreau1; Hakim Amara1; Maxime Moreaud2; Jaysen Nelayah3; Adrien Moncomble3; Christian Ricolleau3; Damien Alloyeau3; Riccardo Gatti1; 1Université Paris-Saclay, ONERA, CNRS, Laboratoire d’Étude des Microstructures, 92320; 2IFP Énergies Nouvelles, 69360; 3Université Paris Cité, CNRS, Laboratoire Matériaux et Phénomènes Quantiques (MPQ), 75013
The study focuses on leveraging Deep Learning (DL) to enhance the analysis of nanoparticles (NPs) using High Resolution Transmission Electron Microscopy (HRTEM), down to atomic resolution. A Convolutional Neural Network (CNN) model was developed to automate the identification of 3D NP shapes from 2D images, overcoming challenges related to manual post-processing and noise. The model was trained on a carefully curated dataset of simulated HRTEM images, capturing various orientations, defocuses and NP sizes. Additionally, a UNet-type model was created for denoising and enhancing contrast between NPs and substrates, addressing limitations in image clarity due to microscope aberrations and environment. Even though this denoising model gives accurate predictions while processing simulated images, we have noticed that the inference on experimental images could sometimes fail. Therefore, we have added a segmentation model to improve the robustness of predictions on experimental micrographs.
1:50 PM
Classification of Abnormal Grain Growth Using 3D Convolutional Neural Networks: Woohyun Eum1; Yi Wang2; Amanda Krause2; Michael Tonks1; Joel Harley1; 1University of Florida; 2Carnegie Mellon University
Understanding what triggers abnormal grain growth (AGG) remains a challenge, as conventional descriptors do not fully explain it. This paper hypothesizes that more complex descriptors play a role in driving AGG. To test this, we train a 3D convolutional neural networks (CNN) to identify abnormal grains (identified manually after heat treatment) from the microstructure boundaries as-sintered Nickel samples before heat treatment, where human experts cannot visually distinguish abnormal grains. Our results demonstrate that the model achieves a 81.6% accuracy, as compared with a 66% accuracy for classifier based only on grain size. Note our input data does not contain orientation information, implying that abnormal grain growth may be observed from the grain boundary network alone. This work marks the first use of 3D CNNs with 3D measurements to capture multi-dimensional information about microstructures, highlighting their critical role in predicting AGG at early stages, without relying on traditional metrics.
2:10 PM
Multimodal Microstructural Image Segmentation of Low-Temperature Sn3Ag0.5Cu7Bi Solder Using Multi-Channel Deep Learning: Eshan Ganju1; John Wu1; Nikhilesh Chawla1; 1Purdue University
High-throughput and accurate semantic segmentation of electron microscopy data is crucial for efficient microstructural analysis, particularly in complex low-temperature solders. To obtain a complete picture of a materials system, various complementary streams of data are often required. By combining backscattered electron images, secondary electron images, and energy dispersive spectroscopy data as input for a deep learning-based approach, we can effectively exploit the complementary information provided by each imaging modality. When applied to a 3D dataset, this approach capitalizes on the unique material contrast from backscatter, surface topography from secondary electrons, and precise elemental distribution from compositional maps, leveraging the power of a multi-channel Unet++ inspired architecture to improve segmentation accuracy. Evaluation on an Sn3Ag0.5Cu7Bi dataset demonstrated that the multi-channel approach performs well and enables accurate identification and quantification of microstructural features. This study highlights the potential of multi-channel deep learning for advanced characterization of complex microstructures.