Speaker
Description
According to the development of new machine learning technologies and prediction models using it, the application of machine learning is already very active in all research areas, including material design even in life sciences. Since artificial intelligence type model using machine learning has the powerful potential to compensate for the shortcomings of material development, we are very interested in material design with machine learning, especially, smart material designs.
Many smart materials have been developed recently, such as perovskite, which is a solar material. Among them, there is a lot of interest in thermoelectric materials. Thermoelectric materials are materials that can generate electricity when heat is applied, and can be applied to various fields such as recycling waste heat from automobiles, weapons development, and electricity generation for spacecraft.
However, despite the advances in science, until recently, materials design has required very large computational resources, such as first-principles calculations, and long development times. For example, calculating the configuration of single atom or molecular structure can take weeks to a month or more on a cluster-class computing machine, and it can take several months to years to determine whether the results are synthesizable.
Shortening the development period is a very important factor in the development of these increasingly demanded smart materials. Machine learning is emerging as an effective technology that can shorten the development period of materials based on existing data. By predicting material properties that can predict the power generation efficiency of thermoelectric materials, the effect of shortening the material development period can be achieved.
We have developed a machine learning model to predict material properties that can predict power generation efficiency, which is the most important aspect of thermoelectric material development, and have been able to confirm its effectiveness. In this presentation, we will present experimental data on thermoelectric materials obtained through our collaborative research and introduce the machine learning methodologies developed to predict the properties.
References
J. Lee, J. Shin, T. Ko, S. Lee, H. Chang, and YunKyong Hyon, Descriptors of atoms and structure information for predicting properties of crystalline materials, Materials Research Express, 8(2) 026302, pp.1--8, 2021
Jino Im, Seongwon Lee, Tae-Wook Ko, Hyun Woo Kim, YunKyong Hyon and Hyunju Chang, Identifying Pb-free perovskites for solar cells by machine learning, npj Computational Materials 5, Article number:37, 2019.
| Keywords | machine learning, material design, thermoelectric, material properties |
|---|
