Accelerated Discovery and Insertion of Next Generation Structural Materials: Process Driven Techniques for Materials Discovery; Investigation of Thin Film Materials
Sponsored by: TMS Materials Processing and Manufacturing Division, TMS: Phase Transformations Committee, TMS: Integrated Computational Materials Engineering Committee
Program Organizers: Soumya Nag, Oak Ridge National Laboratory; Andrew Bobel, General Motors Corporation; Bharat Gwalani, North Carolina State Universtiy; Jonah Klemm-Toole, Colorado School of Mines; Antonio Ramirez, Ohio State University; Matthew Steiner, University of Cincinnati
Wednesday 2:00 PM
March 22, 2023
Room: Sapphire M
Location: Hilton
Session Chair: Andrew Bobel, GM; Jonah Klemm-Toole, Colorado School of Mines; Matt Steiner, University of Cincinnati
2:00 PM
Accelerating Multimodal Data Collection: A Workflow for Metallic Films: Kimberly Bassett1; Brad Boyce1; 1Sandia National Laboratories
Accelerated materials discovery through statistical or machine learning methods necessitates the generation of large data sets. While automated workflows for materials characterization proliferate in industry, specific tools for smaller laboratories are scarce in the literature. Here, we detail a general workflow for accelerating the multimodal characterization of metallic electrodeposited films in pursuit of improving their tribological properties. Methods such as sample tracking, maximization of available instrument features, custom sample fixturing, and strategies to overcome instrument limitations facilitate process acceleration to achieve a high throughput workflow, minimize human bias during characterization, and expedite repetitive tasks. The workflow efficacy is evaluated by key factors such as data repeatability and human capital saved. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC., a wholly owned subsidiary of Honeywell International, Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA-0003525.
2:20 PM
AI and Machine Learning Tools for Development and Analysis of Image Driven 2D Materials: Hafiz Munsub Ali1; Venkata A. S. Kandadai2; Bharat Jasthi2; Venkataramana Gadhamshetty2; Etienne Gnimpieba1; 1University of South Dakota; 2South Dakota School of Mines and Techonology
Recent advances in data driven computing leverage artificial intelligence and machine learning techniques to solve diverse challenges involved 2D material. Microscopic and spectroscopic data are involved in complete life cycle of 2D material such as design, discovery. characterization, maintenance, etc. This review uses PRISMA as a systemic review process to ensure relevance and reproducibility. Articles are searched in three publication databases (WoS, PubMed, Dimensions) with two key research questions. PRISMA identifies 31 articles relevant to the topic. The most studied questions are 2D new material synthesis and engineering with reasonable appreciation for functional discovery/property, defect characterization and grain boundary, but work on corrosion application/detection is not done yet as per our analysis. From the relevant set of articles, the tasks prediction, classification, deep learning, clustering, and AI appeared 38.98%, 28.81%, 13.56%, 10.17%, and 8.47% respectively. Results show that AI (i.e., 8.47%) is still an underexplored computing technique.
2:40 PM
Efficient Conductivity and Hardness Optimization in Cu-Ag-Ni Alloys using Bayesian Active Learning: Terrance Life1; Shankarachary Ragi1; Bharat Jasthi1; Ananth Kandadai1; 1South Dakota School of Mines and Technology
The traditional method to develop and improve materials with desired properties is through trial and error, a time-consuming process requiring domain experts to propose, fabricate, and test numerous potential materials. Our proposed method to accelerate this is to utilize combinatorial deposition and Bayesian active learning to efficiently solve a multi-objective optimization problem for thin-film materials. We demonstrate our method through the fabrication of a copper-silver-nickel alloy in which the percentages of each element vary throughout the sample. This allows the sample to represent a range of potential alloys which our method explores to optimize the conductivity and surface hardness, guided by a Bayesian framework based on Gaussian Process models. The predictions from these models permit the system to selectively test only the alloys most likely to yield the optimal properties. This concludes with an analysis of the global Pareto front, which guides the synthesis of future test samples.
3:00 PM
High-throughput Synthesis and Mechanical Characterization of Sputtered Metallic Alloys: Adie Alwen1; Vignesh Manoharan1; Andrea Hodge1; 1University of Southern California
This work seeks to expedite alloy discovery by using combinatorial and high-throughput synthesis and characterization to develop material libraries that link alloy composition, microstructure, and mechanical properties. To correlate alloying with changes in material characteristics, binary and ternary alloys are investigated by co-sputtering thin films with compositional gradients, creating 169 distinct samples per sputtering run. Material libraries are generated by characterizing each sample’s composition, morphology, texturing, and mechanical properties. To identify compositions of interest and elucidate relationships between material properties, data from the material libraries is analyzed and used to train machine learning models. Using this information, alloys with novel material properties are selected and further investigated using TEM. In total, the material property space for a ternary alloy is explored and compared to binary alloys in order to understand the effects of alloying on material characteristics and microstructure.
3:20 PM
A High-throughput Setup for Materials Exposure to Simultaneous Irradiation-corrosion Conditions: Franziska Schmidt1; Hyosim Kim2; Yongqiang Wang2; Peter Hosemann1; 1University of California Berkeley; 2Los Alamos National Laboratory
The development of new materials suitable for irradiation-corrosion environments, such as nuclear reactors, requires extensive testing of proposed alloy compositions. Simultaneous irradiation-corrosion experiments are notoriously complicated, especially if repeatable quantitative results are desired to inform machine learning approaches. We propose a high-throughput approach for such experiments for thin (<10 µm) bulk samples or deposited thin-films. This is achieved by reducing the corrosive-medium volume, which allows us to produce tens of samples per irradiation, compared to one sample per experiment in similar recent work. A major disadvantage is that the achievable total dpa is inevitably low (<<0.1 dpa). However, the goal is to eliminate those alloys that show unsuitable performance, even at low dpa levels. In this talk, we will show initial results of Fe and stainless steel corrosion by lead-bismuth eutectic under simultaneous proton irradiation and discuss the applicability of this method for other corrosive media.
3:40 PM Break
4:00 PM
A Design Space for Tunable Ceramic-polymer Composites: Yan Li1; 1Dartmouth College
The route of devising polymer-derived ceramics (PDCs), which relies on heat treatment to convert preceramic polymers to ceramics, presents a flexible and energy-efficient approach to fabricate a broad spectrum of ceramics and in-situ ceramic-polymer composites with binary or multinary phases. Understanding the relationship among processing parameters, phase composition and material response holds an important key for property tailoring of PDC composites in different engineering applications. An integrated computational materials engineering (ICME) approach is developed to fundamentally understand how phase transition and microstructure design combine to affect the key mechanical properties of polymer derived ceramic composites. A few case studies will be provided to illustrate how to tailor the mechanical response by redistributing the energy dissipation in a controllable path.
4:20 PM
Combinatorial Mechanical Microscopy via Correlated Nanoindentation and EDX Mapping: Jeff Wheeler1; 1Femto Tools Ag
Mechanical microscopy is an emerging technique using high-speed nanoindentation to map the mechanical behavior and extract phase-level properties from complex microstructures with micron-scale lateral resolution. As such, it is a powerful technique for phase identification in combinatorial materials science investigations in a high-throughput manner on samples with compositional gradients, such as diffusion couples. A significant challenge for nanoindentation mapping is the statistical separation of phases with adjacent compositions and mechanical properties. In this work, we address this by using correlative mapping with analytical electron microscopy, particularly EDX, to accurately determine the relationships between mechanical properties and composition.
4:40 PM
High-throughput Electric-Field-assisted Sintering and Characterization Techniques for Materials Discovery: Michael Moorehead1; Arin Preston1; Zilong Hua1; Jorgen Rufner1; 1Idaho National Laboratory
Despite improvements in computing and modeling capabilities, the performance of new materials, particularly those which deviate greatly in composition from well-studied materials (e.g., high-entropy alloys), can be difficult to simulate given the lack of available experimental property data. While some modeling techniques may attempt to predict the properties of these exotic materials, most are forced to make extrapolations from more traditional materials. To fulfill the need for accelerated material synthesis and property measurement, a high-throughput methodology has been developed. Utilizing electric-field-assisted sintering (EFAS), also known as spark plasma sintering (SPS), equipped with custom tooling, samples of differing alloy compositions can be produced simultaneously as a single alloy array. Several arrays have been produced with compositions spanning the Co-Cr-Fe-Mn-Ni alloy family, including many high-entropy alloys, while the novel array geometry has enabled the samples to be polished and characterized in parallel, using X-ray diffraction, scanning-electron microscopy, and laser-based thermal diffusivity measurements.
5:00 PM
How Should You Select an Algorithm for a Materials Discovery Campaign with Multiple Objectives, Complex and High-dimensional Structure-processing-property Relationships, and a Small Adaptive Design Budget?: Sterling Baird1; Jeet Parikh2; Trupti Mohanti1; Taylor Sparks1; 1University of Utah; 2
Industry-relevant materials discovery tasks are often hierarchical, noisy, multi-fidelity, multi-objective, non-linearly correlated, and exhibit mixed numerical and categorical variables subject to linear and non-linear constraints. Examples include formulation optimization, compositional design of high entropy alloys, and multi-step synthesis. Choosing an algorithm that can expertly navigate such complex design spaces is a non-trivial task, and no single algorithm is supreme. So, how do you pair an algorithm to a design task? Here, we introduce PseudoCrab: a high-dimensional property predictor framed as a pseudo-materials discovery benchmark with fake compositional (linear) and "no-more-than-X-components" (non-linear) constraints. We apply a state-of-the-art high-dimensional Bayesian optimization algorithm (SAASBO) in conjunction with a multi-objective parallel Noisy Expected Hypervolume Improvement (qNEHVI) acquisition function and compare it against other high-performing models. Because PseudoCrab is customizable, researchers can adjust the PseudoCrab benchmark to more closely match their applications of interest during the algorithm downselection process prior to expensive materials discovery campaigns.