General purpose potential for glassy and crystalline phases of Cu-Zr alloys based on the ACE formalism

Niklas Leimeroth,Jochen Rohrer,and Karsten Albe

DOI: https://doi.org/10.1103/physrevmaterials.8.043602

IF: 3.98

2024-04-17

Physical Review Materials

Abstract:A general purpose machine-learning interatomic potential (MLIP) for the Cu-Zr system is presented based on the atomic cluster expansion formalism [R. Drautz, Phys. Rev. B 99 , 014104 (2019)]. By using an extensive set of Cu-Zr training data generated withdensity functional theory, this potential describes a wide range of properties of crystalline as well as amorphous phases within the whole compositional range. Therefore, the machine learning interatomic potential (MLIP) can reproduce the experimental phase diagram and amorphous structure with considerably improved accuracy. A massively different short-range order compared to classica interatomic potentials is found in glassy Cu-Zr samples, shedding light on the role of the full icosahedral motif in the material. Tensile tests of B2-CuZr inclusions in an Cu50Zr50 amorphous matrix reveal the occurrence of martensitic phase transformations in this crystal-glass nanocomposite. https://doi.org/10.1103/PhysRevMaterials.8.043602 Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI. Published by the American Physical Society

materials science, multidisciplinary

What problem does this paper attempt to address?

This paper mainly discusses the development of a universal machine learning potential (MLIP) for copper-zirconium (Cu-Zr) alloys, based on the atom cluster expansion (ACE) formalism, to describe their glassy and crystalline states. In this study, the authors generated a large amount of Cu-Zr training data using density functional theory (DFT) to construct an MLIP that accurately describes the properties of both crystalline and amorphous phases across the entire composition range. The potential successfully reproduced experimental phase diagrams and amorphous structures, highlighting the significant differences in short-range ordering in amorphous Cu-Zr samples and revealing the role of complete pentagonal atomic clusters in the material. Additionally, the paper simulated the tensile testing of B2-CuZr in a Cu50Zr50 amorphous matrix and observed the martensitic phase transition from the B2 phase to the B19' phase, which agreed with experimental results. These findings demonstrate that this MLIP model provides a new approach with DFT-level accuracy for studying the thermodynamic and mechanical properties of Cu-Zr binary alloys, especially in dealing with the complex deformation mechanisms of amorphous and crystalline-glass composites. In terms of methodology, the paper employed the ACE formalism with GPU acceleration to improve computational performance and implemented various types of MLIP in the LAMMPS simulation code to facilitate large-scale simulations. The training dataset included different structures and chemical information, covering elements, various intermetallic phases, and amorphous structures. The ability to reproduce the entire concentration-temperature phase diagram was demonstrated through thermodynamic integration. In summary, the objective of the paper is to address how to create an efficient machine learning potential model that comprehensively describes the Cu-Zr alloy system, including its crystalline and amorphous phases, as well as the interactions across the entire composition range.

General purpose potential for glassy and crystalline phases of Cu-Zr alloys based on the ACE formalism

The characterisation of atomic structure and glass-forming ability of the Zr–Cu–Co metallic glasses studied by molecular dynamics simulations

Different phase transformation behaviors of B2-CuZr crystalline phase and their associated mechanical properties by molecular dynamics using different potentials

Development of suitable interatomic potentials for simulation of liquid and amorphous Cu–Zr alloys

Thermodynamic Calculation And Interatomic Potential To Predict The Favored Composition Region For The Cu-Zr-Al Metallic Glass Formation

Optimization and Validation of a Deep Learning CuZr Atomistic Potential: Robust Applications for Crystalline and Amorphous Phases with near-DFT Accuracy

Formation and structure of Al-Zr metallic glasses studied by Monte Carlo simulations

Prediction of Glass-Forming Ability and Characterization of Atomic Structure of the Co-Ni-Zr Metallic Glasses by a Proposed Long Range Empirical Potential

Prediction of Favored and Optimized Compositions for Cu-Zr-Ni Metallic Glasses by Interatomic Potential

Developing an interatomic potential for martensitic phase transformations in zirconium by machine learning

Phase diagram, thermodynamic modeling and predictions of glass-forming range and ability in the Co–Cu–Zr system

Atomic structure and energetics of amorphous–crystalline CuZr interfaces: a molecular dynamics study

Construction of Al-Mg-Zn Interatomic Potential and the Prediction of Favored Glass Formation Compositions and Associated Driving Forces

Glass Forming Abilities of Binary Cu100−xZrx (34, 35.5, and 38.2 At. %) Metallic Glasses: A LAMMPS Study

Structure and glass-forming ability of simulated Ni-Zr alloys

Machine Learning Interatomic Potentials for Amorphous Zeolitic Imidazolate Frameworks

Prediction of Glass-Forming Ability and Atomic-Level Structure of the Al-Zr-Pd Metallic Glasses by Molecular Dynamics Simulations

First-principles prediction and experimental verification of glass-forming ability in Zr-Cu binary metallic glasses

Development of machine learning interatomic potential for zinc

Correlative study between elastic modulus and glass formation in ZrCuAl(X) amorphous system using a machine learning approach

Atomic-scale simulation to study the dynamical properties and local structure of Cu-Zr and Ni-Zr metallic glass-forming alloys