Computational Discovery and Design of Materials : Session I
Sponsored by: TMS Materials Processing and Manufacturing Division, TMS: Computational Materials Science and Engineering Committee, TMS: Integrated Computational Materials Engineering Committee
Program Organizers: Houlong Zhuang, Arizona State University; Duyu Chen, University of California, Santa Barbara; Ismaila Dabo, Pennsylvania State University; Yang Jiao, Arizona State University; Sara Kadkhodaei, University Of Illinois Chicago; Mahesh Neupane, Army Research Laboratory; Xiaofeng Qian, Texas A&M University; Arunima Singh, Arizona State University; Natasha Vermaak, Lehigh University
Monday 8:30 AM
March 20, 2023
Room: Cobalt 502A
Location: Hilton
Session Chair: Sara Kadkhodaei, University of Illinois Chicago; Mahesh Neupane, Army Research Laboratory
8:30 AM Invited
Ultra-fast Interpretable Machine-learning Potentials for Accelerated Structure Prediction of Materials: Richard Hennig1; Stephen Xie2; Pawan Prakash1; Ajinkya Hire1; Robert Schmid3; Hendrik Kraß3; Matthias Rupp3; 1University of Florida; 2KBR, NASA Ames Research Center; 3University of Konstanz
Crystal structure predictions and all-atom dynamics simulations are indispensable quantitative tools in chemistry, physics, and materials science. Still, large systems and long simulation times remain elusive due to the trade-off between computational efficiency and predictive accuracy. Machine-learning potentials (MLPs) can provide efficient surrogate models for accurate ab-initio electronic structure methods. However, current limitations of MLPs include data inefficiency, instabilities, and lack of interpretability. To address these challenges, we combine effective many-body potentials in a cubic B-spline basis with second-order regularized linear regression (https://arxiv.org/abs/2110.00624). We demonstrate that these ultra-fast potentials are data-efficient, physically interpretable, sufficiently accurate for applications, can be parametrized automatically, and are as fast as the fastest traditional empirical potentials. We illustrate that combining the ultra-fast force field (UF3) method with genetic algorithms can enable large-scale molecular dynamics simulations and accelerate crystal structure predictions of materials.
9:00 AM Cancelled
Computational Reconnoiter for the Design of Amorphous Transition Metal Oxides for Surface Transfer Doping of Diamond: Peter Greaney1; Cameron Chevalier1; Harsha Antony1; Pegah Mirabedini1; Sarah Allec2; Mahesh Neupane3; 1University of California, Riverside; 2Pacific Northwest National Laboratory; 3Army Research Laboratory
Diamond, with its wide band gap and high carrier mobility, is a promising material for the next generation of high frequency power electronics. Unfortunately, diamond cannot be easily doped by substitutional impurities due to the poor solubility of potential dopants. In this talk I present our recent work on elucidating the electronic-structure property relationships in a range of transition metal oxides that are candidates for surface transfer doping of diamond. With the use of amorphous oxides we seek to take advantage of both the inherent lack of the translational symmetry in glassy materials, and the continuum of structures that can be obtained in amorphous materials by tuning the processing conditions. While this tunability is advantageous for functionality it raises the challenge for computationally led design that one must sample a vast configurational space of structures. We have sought to overcome this using a multi-scale modeling approach.
9:30 AM Invited
Modeling of Local Lattice Distortion Effects on Vacancy Migrations in Multicomponent FCC Alloys: Zhucong Xi1; Louis Hector, Jr2; Amit Misra1; Liang Qi1; 1University of Michigan; 2GM Global Technical Center, General Motors Company
Accurate prediction of vacancy migration energy barriers in multi-component alloys is extremely challenging yet critical for the development of diffusional transformation kinetics needed to model alloy behavior in many technological applications. In these alloys, the local chemical environment along the migration pathway effectively generates local lattice distortion causing fluctuations in energy barriers during vacancy migration. In this paper, we studied vacancy migrations in face-centered cubic (FCC) Al-Mg-Zn-based ternary and quaternary alloys. A quartic function of the reaction coordinate is proposed to accurately describe the energy landscape of the minimum energy path (MEP) for each vacancy migration event. We also used the local lattice occupations as inputs to train surrogate models based on the cluster expansion method to predict the coefficients of the quartic function, which accurately and efficiently output the vacancy migration energy barriers for the kinetic Monte Carlo simulations of diffusional transformation in Al alloys.
10:00 AM Break
10:20 AM Invited
Computation Discovery of Materials for Solid-state Batteries: Yifei Mo1; 1University of Maryland, College Park
All-solid-state Li-ion battery based on solid electrolytes is a promising next-generation battery technology with high energy density, intrinsic safety, long cycle life, and wide operational temperatures. However, the lack of solid electrolyte materials that satisfy multiple requirements, such as high ionic conductivity, good stability, and interfacial compatibility with the electrode, are impeding the development of this new battery technology. To resolve these materials challenges, we develop and leverage an array of data-driven computation techniques to discover and design novel solid-state Li-ion conductors as solid electrolytes for all-solid-state batteries. The data-driven approach enables rapid searching over a large materials space of tens of thousands of materials with highly diverse structures and chemistries. Dozens of novel solid-state conductors are discovered through our data-driven materials search. Our study demonstrates a new paradigm of using machine learning techniques for materials discovery that overcome the data-scarcity challenges.
10:50 AM
Machine-learning-boosted Searching and Optimization of Emergent Quantum Materials: Mingda Li1; 1Massachusetts Institute of Technology
Machine learning (ML) has demonstrated great power in materials science. However, the complex interplay between the charge, spin, orbital and lattice degrees of freedom in quantum materials poses challenge in implementing ML. Here I introduce how ML can accelerate quantum materials research combining measurement techniques. First we will introduce a ML-based topology predictor, where we show that the band topology is indeed encoded in a simple spectral signal with 90%+ accuracy. Hundreds of new topological materials are found accordingly [1]. Second, we use ML to better resolve a fine magnetic effect, which cuts the experimental resolution by a two [2]. Finally, we show how ML can assist in ultrafast diffraction analysis that leads to a panoramic mapping of phonon thermal transport [3]. We conclude by envisioning a variety of quantum-materials-related problems which ML play a key role [4]. [1] arXiv:2003.00994 (Adv. Mater. 2022) [2] https://aip.scitation.org/doi/10.1063/5.0078814 [3] arXiv:2202.06199 [4] https://aip.scitation.org/doi/10.1063/5.0049111