HEA 2023: Fundamental Theory and Modeling I
Program Organizers: Andrew Detor, DARPA/DSO; Amy Clarke, Los Alamos National Laboratory
Monday 1:40 PM
November 13, 2023
Room: William Penn Ballroom
Location: Omni William Penn
Session Chair: Shalini Roy Koneru, The Ohio State University
1:40 PM Introductory Comments
1:45 PM Invited
Quantifying Short-range Order and its Domain Size in High-entropy Alloys: Rodrigo Freitas1; 1Massachusetts Institute of Technology
Complete characterization of short-range order is challenging to realize because of the sheer number of local chemical configurations that must be accounted for. Moreover, the chemical complexity of concentrated solid-solution phases is often described as the tendency for some of these configurations to be more common than others (i.e., "slightly less random than completely random"), which does not translate easily into a physically intuitive and quantifiable picture. In this talk I will introduce an approach that combines statistical mechanics and information theory with machine learning techniques to quantify the space of local chemical configurations available for high-entropy alloys. I will show that this approach leads to a predictive framework for evaluation of short-range order domain sizes.
2:15 PM
Bond-stiffness Based ML Approach to Predict Atomic Level Properties in MPEAs: Nathan Linton1; Dharmendra Pant1; Dilpuneet Aidhy1; 1Clemson University
On the one hand, the presence of multiple elements in large proportions in multi-principal element alloys (MPEAs) present opportunities to unravel novel properties, whereas on the other, they pose a large computational challenge due to the large phase space, especially for density functional theory (DFT) calculations, that are inherently very expensive. We present PREDICT (PRedict properties from Existing Database In Complex alloys Territory), a machine learning framework coupled with DFT whereby properties in MPEAs could be predicted simply by learning from the binary alloys database. The physics is included via bond stiffness derived from DFT to predict elastic constants, vibrational entropy and other related properties in FCC based MPEAs leading to accurate predictions. This approach enables probing any MPEA composition by just including DFT information of constituting binary alloys, thereby altogether bypassing DFT calculations in MPEAs.
2:35 PM
Local Order Average-atom Approach for Simulating Refractory High Entropy Alloys: Chloe Zeller1; Ellad Tadmor1; 1University of Minnesota - Twin Cities
Refractory high-entropy alloys (RHEAs) are characterized by their high service temperature and superior strength and are thus promising materials for harsh-environment aerospace applications, such as exhaust nozzles and thermal protection materials on hypersonic vehicles. High-fidelity molecular dynamics (MD) simulations are a promising approach for predicting the thermomechanical response of RHEAs, however large simulations are required to obtain good statistics to account for local chemical fluctuations that can have a large effect on properties. To reduce system size, Average-Atom (AA) interatomic potentials, where interactions between species are weighted according to their occurrence probability in a random alloy, have been used to predict phase averages of properties of interest. We extend this approach by accounting for local ordering in multicomponent alloys, i.e. preferential short-range arrangements of atomic species. Using this new Local Order Average-Atom (LOAA) approach, we study the effect of local ordering on the mechanical properties of prototypical multicomponent systems.
2:55 PM
Quantitative Assessment of Local Chemical Ordering in Atomistic Simulations of High-entropy Alloys: Killian Sheriff1; Yifan Cao1; Rodrigo Freitas1; 1Massachusetts Institute of Technology
High-entropy alloys (HEAs) exhibit exceptionally good combinations of properties recently reported to correlate with chemical short-range ordering (cSRO). However, in atomistic simulations, their state of cSRO has only been so far characterized using the Warren-Cowley parameters. Yet, this approach is incomplete as distinct local atomic configurations sharing the same chemical concentration are indistinguishable. Here, we propose a generalized framework, based on graph-convolution neural networks equivariant to E(3) symmetry operations, statistical mechanics, and information theory, capable of completely identifying the set of distinct local atomic bonding environments and their associated population densities in HEAs. This approach leads to a quantitative characterization of the cSRO state and provides a predictive framework for evaluation of cSRO domain sizes, thus offering novel avenues to explore the relationships between processing, structure, and properties in HEAs.
3:15 PM Break
3:35 PM
Capturing Short-range Order in High-entropy Alloys with Machine-learning Potentials: Yifan Cao1; Killian Sheriff1; Rodrigo Freitas1; 1Massachusetts Institute of Technology
Chemical short-range order (cSRO) is recently reported to strongly influence the mechanical properties of various high-entropy alloys (HEA). However, the intricate nature of cSRO has made it challenging for current machine-learning potentials (MLP) to capture this feature, and many proposed approaches lack quantitative analysis of MLP performance on this task. In this work, we propose a generalized strategy to construct first-principles training databases and effectively train MLPs capable of characterizing cSRO in HEAs. We demonstrate this strategy by quantitatively analyzing the MLP performances in reproducing cSRO effects in various properties of CrCoNi HEA, including defect properties and phase stability.
3:55 PM
Accelerating the Discovery of Low-Energy Structure Configurations: A Computational Approach that Integrates First-principles Calculations, Monte Carlo Sampling, and Machine Learning: Md Rajib Khan Musa1; Yichen Qian1; Dr. David Cereceda1; 1Villanova University
In this work, we developed a novel and highly efficient computational approach that combines MC sampling, DFT calculations, and Machine Learning (ML) techniques to accelerate the discovery of low-energy structure configurations of alloys. Our method is inspired by the well-established cluster expansion technique, leveraging its strengths while addressing its limitations. Specifically, we enhanced the reliability of the cluster expansion by avoiding out-of-sample prediction using machine learning. We performed first-principle Density Functional Theory (DFT) calculations for those samples. We applied our novel approach to several tungsten-based alloys. Our results show a noteworthy reduction in root mean square error (RMSE) compared to cluster expansion, suggesting its superior accuracy and reliability.
4:15 PM
Computational Discovery of B2 Phases in the Refractory High Entropy Alloys: Junxin Wang1; Maryam Ghazisaeidi1; 1Ohio State University
The Multi-Cell Monte Carlo method for phase prediction for multicomponent alloys has demonstrated great potential in simulating coexisting phases in many-component crystalline systems. To find potential B2 structures in high entropy alloys, this method is applied to composition space, spanning through the refractory element range in the periodic table. First, we explore the refractory elements with BCC ground state structures (MoNbTaWV) and then the ones with HCP ground state structures (TiZrHfOsReRu). Ordered structures are found in both systems and their thermal stability are analyzed. We further look into the combination of all the refractory elements (both HCP and BCC elements) and try to identify the most possible groups to form a B2 structure.
4:35 PM
Deciphering the Strength-vs-ductility Trade-off for High-entropy Alloys with AI-driven Fully ab Initio-based Material Modeling: Max Hodapp1; Ivan Novikov2; Olga Kovalyova2; Alexander Shapeev2; Franco Moitzi1; Oleg Peil1; 1Materials Center Leoben; 2Skoltech
In this talk, we present a novel Bayesian multi-objective optimization framework that fully automatically predicts multicomponent refractory alloys with Pareto-optimal strength-vs-ductility ratios. Our framework involves predictive material models as objective functions that are fed with ab initio-based data exclusively, coming from efficient CPA calculations or atomistic simulations using machine-learned interatomic potentials, allowing for a screening over the whole alloy space at an acceptable cost. More broadly, our framework is neither limited to two objectives nor to specific mechanical properties, and, therefore, we anticipate that it also enables us to accelerate the discovery of new materials with exotic properties for various other applications. Further, we outline how magnetism can be brought into the game so that our methodology would allow for screening over a much larger space of (magnetic and non-magnetic) alloys than the space that can currently be approached with state-of-the-art methods.
4:55 PM
Atomic Representations of Local and Global Chemical Effects of Mechanical Strength: Mitchell Wood1; Megan McCarthy1; Mary Alice Cusentino1; 1Sandia National Laboratories
The exceptional properties observed in complex concentrated alloys (CCAs) arise from the interplay between crystalline order and chemical disorder at the atomic scale, complicating the determination of properties. The base metallurgical argument for CCAs’ observed strength is the maximization of solid solution strengthening effects, but is difficult to quantitatively address from experiments alone. Herein we present a quantum-accurate interatomic potential(IAP) for use in molecular dynamics simulations of MoNbTaTi that efficiently scales to systems that are converged with respect to size, time and chemical complexity. Furthermore, we use this IAP to quantify the relationship between inhomogeneous lattice strains and novel definitions of local chemical environment. We will highlight the improvement of this reduced order model over historical arguments of local atomic volume and element-wise attribution of strengthening in these complex alloys.