Distribution of the mechanical properties of Ti–Cu combinatorial thin film evaluated using nanoindentation experiments and molecular dynamics with a neural network potential. (December 2022)

• Record Type: Journal Article
• Authors: Miyagawa, Takeru; Sakai, Yugo; Mori, Kazuki; Kato, Nobuhiko; Yonezu, Akio; Ishibashi, Keiji
• Is Part Of: Materials today communications. Volume 33(2022)
• Publication Date: 2022-12
• Subjects: Molecular dynamics simulation -- Neural network potential -- Mechanical property -- Combinatorial thin film -- Titanium–copper alloy; Materials science -- Periodicals; 620.11
• DOI: 10.1016/j.mtcomm.2022.104750
• Languages: English
• ISSNs: 2352-4928

Abstract

Abstract: The combinatorial approach is a prominent method for synthesizing samples with composition gradients and that enables the high-throughput discovery of new materials. It enables high-throughput property screening to delineate the composition–structure–property relationship and identify atomic compositions with desired properties. Titanium–copper (Ti–Cu) alloy is widely used in electronic devices because of its excellent mechanical properties, such as stress relaxation resistance, bond formality, and workability. Synthesizing Ti–Cu thin film using the combinatorial approach could reveal a new application of the alloy. Aiming to discover a new application of the alloy, Ti-Cu thin film was synthesized using combinatorial approach. After synthesizing Ti-Cu thin film, we conducted XRD analysis of Ti-Cu thin film and then nanoindentation test. In addition to these experimental procedures, we have utilized reverse analysis method to obtain the mechanical properties of the thin film. We finally obtained the mapping of crystal structures and mechanical properties with respect to the atomic composition. To reveal the mechanical properties and their mechanisms, we employed molecular dynamics (MD) simulation as a tool to predict mechanical properties from the perspective of atomic simulation; however, it requires interatomic potentials to depict the movements of atoms. Because of the complex structure of Ti–Cu thin film synthesized using the combinatorial approach, the creation of interatomic potentials is a difficult and time-consuming process. Therefore, interatomic potentials were developed in this study using a neural network (NN), which enabled MD simulations of the Ti–Cu thin film, which is our particular interest. These were referred to as neural network potentials (NNPs). MD simulations were conducted with NNPs to study the mechanical properties of Ti–Cu thin film. First, ab initio molecular dynamics (AIMD) simulations were performed to obtain the relationships between the atomic coordinates and the corresponding total potential energies (as well as interatomic forces) of each atom. Subsequently, the relationship between the atomic coordinates and atomic energies were used as a training dataset for the NN, and a desirable NNP was developed by optimizing the parameters of the NN. The developed NNP was then tested to check its reliability and accuracy. It was found that the NNP could accurately predict the energies and forces calculated by the AIMD simulations. The mechanical properties of the Ti–Cu thin film were then computed using the NNP and MD simulations. Some material cases were verified with many-body interatomic potentials (e.g., modified embedded-atom method (MEAM) potential) and results obtained by experiments and ab initio calculations. Finally, using MD simulations with the developed NNP, we investigated the mechanism of the mechanical properties from the perspective of atomic scales. The mechanical properties and their distribution were examined via a comprehensive experiment and MD simulation approach. Graphical Abstract: ga1 Highlights: This study reveals mechanical properties and their mechanisms of Ti-Cu thin film using nanoindentation experiments and Molecular Dynamics (MD) simulation. Ti–Cu thin films with different composition ratios were created using a combinatorial synthesis approach. The mechanical properties of the film and its distribution were measured using nanoindentation experiments. Ab initio MD simulations were performed to create Neural Network potential (NNP). MD simulation using a NNP was conducted to clarify the mechanical properties of the film.