Session: 03-03-01: Integrated Computational Materials Engineering (ICME)

Paper Number: 94934

94934 - Investigation of Mechanical Properties of Combinatorial Ti-Cu Film Using MD Simulation With Neural Network Potential 

(Abstract) To discover a novel material, combinatorial synthesis approach became a powerful tool, which enables synthesize samples with atomic elements deposited with sputtering, atomic composition is graded in-plane at one sample. Since these thin films play a role of materials library, combinatorial synthesis approach enables high-throughput materials screening to delineate composition-property relationship and identify atomic compositions with desired properties. Meanwhile, Molecular Dynamics (MD) simulation is an excellent tool to predict mechanical properties from the perspective of atomic scale, and it is compatible with combinatorial synthesis approach since the atomic composition ratios are controlled in one sample. However, MD simulations require interatomic potentials to depict the movements of atoms, therefore the existence of accurate interatomic potentials is critical to perform reliable MD simulations. The development of interatomic potentials generally requires numerous fittings and ab initio calculations, which is a difficult and tedious process. Therefore, in this study, a neural network (NN) based method to create interatomic potentials is developed, which are referred as neural network potentials (NNPs). With NNPs, MD simulations can be performed for combinatorial thin films to investigate the mechanism of mechanical properties from the perspective of atomic scale.

(Method) From the experiments, titanium copper (Ti-Cu) alloy was synthesized through combinatorial synthesis approach, and the Ti-Cu thin film with atomic composition graded in-plane in the sample was created. Ti-Cu alloy is widely used as electronic devices such as connectors and camera modules because of its excellent stress relaxation resistance, bond formality and workability. Subsequently, XRD analysis was conducted to elucidate the crystalline structures regards of atomic compositions. Finally, nanoindentation was carried out as well as reverse analysis, to determine the mechanical properties of Ti-Cu combinatorial thin films. For the atomistic simulations, VASP package was employed to create the training data of the NN, and NN was trained using DeePMD-kit package. First, ab initio molecular dynamics (AIMD) simulations were carried out to create training data. This data was constructed according to the experimental results such as XRD analysis. Subsequently the training process of NN was conducted by using DeePMD-kit. The optimized DNN was ready to use in LAMMPS package as NNP, and the mechanism of mechanical properties of Ti-Cu alloy was investigated.

(Results & discussions) From the experiments, the mapping of Young’s modulus and yield strength regards of atomic compositions were obtained. It is found that the higher Ti atomic compositions are associated with the higher yield strength. From the MD simulations, the Young’s modulus mappings was obtained, and it showed good agreements with experimental results. For the yield strength, the experimental and computational results were not comparable because of there was a huge gap in terms of size scale between experiments and MD simulations; however, the onset of plastic deformation, which corresponded to yield strength in experiments, was investigated. Since the onset of plastic deformation occurs due to dislocation motion, the dislocation motions were simulated in MD simulations. From the MD simulations, it turned out that with higher Ti atomic compositions, the critical stress, which is referred as Peierls stress, for dislocation motion became larger. Thus, the higher Ti atomic compositions contribute to higher Peierls stress, ultimately leading larger yield strength in the experimental results.

(Reference) Takeru Miyagawa, Kazuki Mori, Nobuhiko Kato, Akio Yonezu, Development of Neural Network Potential for MD Simulation and Its Application to TiN, Computational Materials Science, Volume 206, 15 April 2022, 111303.

Presenting Author: Takeru Miyagawa Chuo University

Presenting Author Biography: I am a graduate student at Chuo University.

Authors:

Takeru Miyagawa Chuo University
Yugo Sakai Chuo University
Akio Yonezu Chuo University
Kazuki Mori ITOCHU Techno-Solutions Corporation
Nobuhiko Kato ITOCHU Techno-Solutions Corporation
Keiji Ishibashi COMET