At present, when interpreting the measurement results of X-ray absorption spectra, a calculation method called first-principles calculation is often used. First-principles calculations, however, are very complicated and take an enormous amount of time. If it were possible to directly estimate the properties of compound structures from the results of X-ray absorption spectra without using first-principles calculations, it would be possible to greatly reduce the analysis time. In recent years, materials informatics (MI) has attracted attention as an effort to realize this.
Therefore, in this blog, we will introduce an overview, research and success cases, and issues at the time of introduction under the theme of materials informatics .
- Overview of materials informatics
- History of materials informatics
- Research Cases of Materials Informatics
- Materials Informatics Success Stories
- Materials informatics challenges
- in conclusion
Materials informatics is a field and technology that uses information science technology such as machine learning and data mining to improve the efficiency of material development .
Figure 1 shows a comparison between conventional development and materials informatics development. In conventional material development, material design was mainly based on prior research and the experience of researchers. As a result, many simulations and experiments had to be performed for many candidate materials, and a lot of time was spent on development.
On the other hand, materials informatics-type materials development uses information science technology to efficiently search for materials that meet the target performance. This can significantly reduce development time.
The importance of materials informatics has increased with changes in the social environment and development of information science. For example, the evolution of mobility and the penetration of 5G have increased the need for development of new materials.
Here, we will look at the history of the development of materials informatics.
History around the world
With the announcement of the ” Materials Genome Initiative (MGI) ” by former US President Barack Obama in 2011, efforts in materials informatics have accelerated around the world.
|2011||America||Former President Obama announces the Materials Genome Initiative (MGI). Invest $500 million over the five years from 2012 to 2016, focusing on “
transformation of materials research culture,” “unification of experiments, computer calculations, and theory,” “access to digital data,” and “development of materials-related human resources.”
|2014||Switzerland||Launched Materials Revolution Computational Design and Discovery of Novel Materials (MARVEL).|
|2015||EU||Started Novel Materials Discovery (NOMAD).|
|2015||Korea||Launched the Creative Materials Discovery Project as a 10-year plan.|
|2015||China||Announced Made in China 2025. Develop new materials in 10 priority areas.|
History in Japan
In Japan, SIP ” Materials Integration ” started in 2014 .
“Innovative Structural Materials” Strategic Innovation Promotion Program (SIP) | Japan Science and Technology Agency defines as follows.
Materials integration refers to utilizing the results of materials science and integrating all science and technology such as theory, experiments, analyses, simulations, and databases to support materials research and development from an engineering perspective. Aiming for comprehensive materials technology tools
In 2015, the JST innovation hub construction support project “Information Integration Type Substances and Materials Development Initiative” was launched, followed by the NEDO “Ultra-advanced material ultra-rapid development basic technology project (super-super project)” in 2016.
Furthermore, in 2018, SIP Phase 2 “Materials Revolution by Integrated Materials Development System” was launched, and efforts are being made to further strengthen global competitiveness in materials informatics.
As a recent case study using materials informatics, here we introduce the development of spin thermoelectric materials using decision trees by the Japan Science and Technology Agency (JST).
A spin thermoelectric material is a thermoelectric material that uses the force of a spin (magnet). This spin thermoelectric material has a very complicated crystal structure, and its physicochemical properties have not been elucidated. Therefore, machine learning was used to solve this problem. The machine learning technique used is decision trees. We constructed a decision tree model to predict the thermoelectromotive force from crystal lattice mismatch, molecular weight of rare earth elements, spin magnetic moment, and orbital magnetic moment.
Fig. 2 shows the prediction results from the decision tree. The following three things became clear from this prediction result.
- The smaller the SR ( spin magnetic moment), the larger the thermoelectromotive force.
- The larger the nR (molecular weight of the rare earth element ), the larger the thermoelectromotive force.
- The smaller the ∆ 𝑎 (crystal lattice mismatch), the larger the thermoelectromotive force.
As in this research case, materials informatics can provide knowledge about things that have not yet been elucidated by chemistry and physics.
Here, we briefly introduce some successful cases of materials informatics.
Succeeded in developing high-performance heat conversion materials in cooperation with the Advanced Institute for Materials Science, Tohoku University
All simulations were performed on a computer, and Toyota succeeded in developing all-solid-state battery materials in just one year, which took five years.
Asahi Kasei Corporation
Focusing on human resource development, succeeded in implementing development that would take several years in half a year
Toray Industries, Inc.
Significantly shortens the development of carbon fiber reinforced plastics, which takes two to three years just for design
ENEOS Co., Ltd.
By using it together with our unique prediction model, it is possible to create accurate designs even for high-performance polymers with complex molecular structures.
As described above, manufacturers are making progress in introducing and working on materials informatics, and there are various success stories .
Materials informatics continues to attract attention from material manufacturers around the world, but there are some challenges when introducing it. In particular, I would like to introduce two issues that are currently problematic, so please refer to them.
Underdeveloped data infrastructure
- Data is paper-based and not digitized
- Data is not standardized in a parsable format
- No failure data saved
and other problems related to the data infrastructure.
In addition, in some cases, the data owned by the company may not be sufficient as analysis data, and the need for opening an open platform and open innovation is increasing.
Absence of data personnel
One of the challenges is the lack of human resources who can handle data appropriately, such as data scientists. In the field of materials informatics, we need human resources who have both a deep understanding of materials and knowledge of data science .
As a solution, expansion of university lecture programs and private study courses can be considered.
As the technology environment surrounding us changes dramatically, such as the evolution of mobility and the development of 5G, there is a growing need for new materials that can accelerate these changes. Fierce development competition is already taking place among material manufacturers around the world, and attention to materials informatics is likely to continue to increase in the future.
In Skill Up AI, we have opened a ” Materials Informatics Course “. In this course, you can learn the current trends in materials informatics and the major analysis methods often used in materials informatics. If you are interested in materials informatics, please consider taking this course.
In addition, we hold a practical AI study session ” Skill Up AI Camp ” every Wednesday. At the study sessions, we will cover various practical themes and provide hints that will lead to improved practical skills in data analysis and AI development. There is also a corner where the instructor answers questions and concerns from the participants.