Masayoshi Adachi1 Benjamin Pierce2 Ahmad Karimi2 Laura Wilson2 Roger French2 Jennifer Carter2 Hiroyuki Fukuyama1

1, Tohoku University, Sendai, , Japan
2, Case Western Reserve University, Cleveland, Ohio, United States

Aluminum nitride (AlN) is a promising substrate material for AlGaN-based ultra-violet light emitting diodes. In the Fukuyama group at Tohoku University, AlN crystal growth methods have been developed [1] with a recent focus on solution growth using a Ni-Al alloy. In order to design this technique, a fundamental study for the AlN formation on a Ni-Al droplet was undertaken. To understand the growth behavior and design an optimum crystal growth technique, an in-situ observation system for solution growth of AlN crystal using electromagnetic levitation [2] has been developed.
As part of the Tohoku University and Case Western Reserve University collaboration in Data Science for Life Sciences and Materials Science, an international and interdisciplinary research project started, focused on statistical significant quantification of AlN formation behavior, spanning nucleation, growth and coalescence so to design and define an optimum crystal growth technique.
The materials science goal was to profile the nucleation and growth rates of AlN crystals on the spherical liquid Ni-Al droplet The study protocol, representing the details of how the experiments are to be run, including varying the temperature, the Ni-Al composition ratio, the N2 gas pressure, and the static magnetic field (which controls the solution flow in the molten droplet). This design space of the crystal growth predictors allowed us to encompass from high to low nucleation rates, and from hundreds of AlN crystalites on the sphere, to controlled single crystal growth. The high speed cameras recorded the droplet from the top and the side simultaneously, and the video images analyzed in this project encompass over 530,000 single frame images. With this large dataset, analysis was done in our distributed and high performance computing environment at CWRU [3]. Image analysis was performed using Python (v2.7) libraries including Matplotlib, Numpy, Scipy, Pandas, Seaborn, Trackpy with Skimage and OpenCV [4-7].
In addition to the technical details, the code development involved graduate and undergraduate students, distributed and high performance computing, data exchanges of large datasets and producing robust codes with a number of students that can be validated and are sufficiently modular so that the pipeline is multi-functional. For code development over a two year period involved 3 undergraduate (UG) and 3 graduate (GS) students who participated sequentially, so proper code styling, commenting, documentation and Git code versioning were essential. In addition, communication between the image analysis students from Materials Science, Mechanical Engineering and Computer Science departments, was enabled by the students all having taken Applied Data Science courses at CWRU, so that the basics of an Open Data Science tool chain, gave them a common framework for both tools and data analysis project structure.
The project goal of the synergistic process may be advanced by French’s teaching Applied Data Science this summer at Tohoku University, with the goal of establishing a “nuclei” of Materials Data Scientists here, which can “grow” into a robust local community of students applying these new and complementary tools to Materials Science problems.

[1] M. Adachi et al., to be submitted.
[2] M. Adachi et al., SN Appl. Sci., 1(2019) 18.
[3] Yang Hu, et al., A Nonrelational Data Warehouse for the Analysis of Field and Laboratory Data From Multiple Heterogeneous Photovoltaic Test Sites, IEEE JPV. 7 (2017) 230–236.
[4] Python Software Foundation: Python 2.7.16.
[5] E. Jones, T. Oliphant, P. Peterson, others, SciPy: Open source scientific tools for Python, 2001.
[6] Stéfan van der Walt, et al., scikit-image: Image Processing in Python, PeerJ. 2 (2014) e453. doi:10.7717/peerj.453.
[7]G. Bradski, The OpenCV Library, Dr. Dobb’s Journal of Software Tools. (2000).