Inference-based Approaches for Material Discovery and Property Optimisation: Structure-Property Inference from Experiments
Sponsored by: TMS Advanced Characterization, Testing, and Simulation Committee, TMS Chemistry and Physics of Materials Committee
Program Organizers: Felix Hofmann, University of Oxford; Michael Short, Massachusetts Institute of Technology; Cody Dennett, Commonwealth Fusion Systems; Mohamed Abdallah Reza, University Of Oxford; Daniel Mason, UK Atomic Energy Authority
Wednesday 8:00 AM
October 12, 2022
Room: 326
Location: David L. Lawrence Convention Center
Session Chair: Felix Hofmann, University of Oxford
8:00 AM Introductory Comments
8:10 AM Invited
Uncover Hidden Materials Properties with the Lens of Machine Learning: Mingda Li1; 1MIT
Despite the significant progress in experimental characterization techniques, understanding the microscopic interaction mechanisms in complex material families remains a grand challenge. Machine learning (ML) brings new hope and can even serve as a new probe to study the complex interplay between the charge, orbital, spin, and lattice degrees of freedom. In this talk, I will introduce how ML can be used to reveal the hidden information in experimental data and elucidate the microscopic interactions. I will provide a few examples from our research, that 1)how ML can help identify a nuanced effect that can lead to electronics without energy dissipation, 2)how ML can be used to rapidly screen materials with superior thermal properties, and lastly, 3)how ML can result in interfacial defects identification and hidden phonon transport with unprecedented knowledge. We highlight the importance of the representations and envision a variety of measurement problems that can benefit from machine learning.
8:50 AM
Comparing High-dose Simulated Irradiation in Tungsten to Experiments: Daniel Mason1; Max Boleininger1; Fredric Granberg2; Guanze He3; Felix Hofmann3; Sergei Dudarev1; 1UKAEA; 2University of Helsinki; 3Oxford University
Massively overlapping molecular dynamics cascade simulations can be used to efficiently generate microstructures corresponding to high irradiation doses. Recent work has shown that the properties of the microstructures match experimental measurements well, provided we restrict our interest to temperatures below stage III recovery- migration of vacancies. In this talk we will compare properties of nanoscale defects generated by simulated irradiation of tungsten with comparable experimental measurements, to test the quality of the microstructures produced. We show that not only does simulated irradiation reproduce physical mechanisms responsible for the observed saturation effects, we can also predict hydrogen retention capacity, thermal diffusivity changes, and simulate TEM images which are in good quantitative agreement with the respective experiments. These predictions are essentially parameter-free, requiring only an empirical potential and boundary conditions of stress, temperature and dose, and therefore suggest a physically-informed way to infer properties of the defect microstructure from experimental measurements.
9:10 AM
Using Local Thermal Transport Property to Characterize Microstructure of Materials from Additive and Advanced Manufacturing Technologies: Zilong Hua1; Jorgen Rufner1; Arin Preston1; Amey Khanolkar1; Cody Dennett2; Robert Schley1; Caleb Picklesimer1; Asa Monson1; Michael McMurtrey1; David Hurley1; 1Idaho National Laboratory; 2Commonwealth Fusion Systems
Additive and Advanced Manufacturing technologies have been widely used to develop advanced materials and produce components with complicated geometry. By raising the temperature of the raw materials in a short time period, a significant temperature gradient is created across the specimen, causing microstructure inhomogeneity. In this talk, we use photothermal radiometry to locally measure thermal diffusivity of Additive and Advanced Manufacturing produced materials, the results of which can be correlated to microstructure variation. Measurements were conducted across a series of samples with different porosity, and on individual samples along the radial direction. Based on the similar technique, an instrument is under development to perform in situ thermal property measurements, with the ultimate objective to provide the real-time feedback control of Spark Plasma Sintering.
9:30 AM Invited
Exploring the Evolution of Irradiation-induced Defects Through Their Energetic Signatures: Charles Hirst1; Fredric Granberg2; Boopathy Kombaiah3; Penghui Cao4; Scott Middlemas3; R. Scott Kemp1; Ju Li1; Kai Nordlund2; Michael Short1; 1Massachusetts Institute of Technology; 2University of Helsinki; 3Idaho National Laboratory; 4University of California, Irvine
Accurate prediction of a material’s properties relies on the comprehensive characterization of all defects within. Yet for irradiated metals the smallest and most numerous defects are unable to be imaged in TEM. Instead of spatial characterization, defects can be detected and quantified through their excess energy. Differential scanning calorimetry (DSC) experiments of neutron-irradiated Ti measure defect densities that are 3 times greater than those determined using TEM. Two energetically-distinct annealing processes are also observed where the established recovery model predicts only one. Comparing calorimetry experiments to molecular dynamics (MD) simulations reveals the mechanism of recovery with dislocation loops gliding through the material, annihilating defects in their path. Positron annihilation spectroscopy (PAS) and XRD experiments will be used to validate the proposed mechanism. Through a combination of multiple techniques the evolution of irradiation-defects can be deduced, demonstrating the power of inference-based approaches to probe radiation damage at the atomic scale.
10:10 AM Break
10:30 AM
Multi-technique Characterisation of Ion-irradiation Effects on High-pressure-Torsion (HPT) Processed EUROFER-97: Kay Song1; Gregory Strangward-Pryce1; Abdallah Reza1; Guanze He1; David Yang1; Kenichiro Mizohata2; Felix Hofmann1; 1University of Oxford; 2University of Helsinki
High-pressure torsion (HPT) is a technique that involves subjecting materials to a compressive force (order of GPa) at the same time as torsional straining. This process modifies the microstructure and dislocation content of materials and has the advantage of producing samples with a large range of strain for investigation. The refinement of grains increases grain boundary area which could be favourable for enhanced irradiation resistance of materials for fusion reactors. This study investigates the effect of HPT on EUROFER-97, a leading candidate for reactor structural components. Different characterisation techniques have been combined to study the changes to the material properties (hardness and thermal diffusivity) of EUROFER-97 as a function of mechanical strain and irradiation dose. Their respective contributions to material changes have also been studied through dislocation content measured with X-ray diffraction. Results provide valuable insights into the potential of grain refinement to alter irradiation resistance of nuclear structural materials.
10:50 AM Panel Discussion