Showa Denko KK – Most stable structures of chemical compounds, key to prediction of material properties, identified twice as fast compared to expert researcher with 3 times higher probability of success

Showa Denko KK. (President and CEO Hidehito Takahashi) and Quantum Simulation Technologies, Inc. (CEO and co-founder Toru Shiozakihereinafter referred to as QSimulate), collaborated to develop a new capability for QSimulate’s quantum simulation platform (QSP Materials2) for materials development, which reduced the researcher’s workload by more than half experimented.

The new module automates the search for the most stable structure3 of chemical compounds, which is the key to discovering the properties of materials. Compounds with coordination bonds known to be difficult can be simulated using this module. Compared to a conventional manual search, this new method has a 3x higher probability of predicting the most stable structures while taking only 45% of the time used in the manual approach.

Issues that have been addressed

The key to materials development is to identify the most stable structures for chemical compounds in order to discover the properties of chemical compounds. However, it becomes too complex as the number of atoms in the molecule increases as the number of candidate structures increases exponentially.

For example, finding the most stable structure of activated vitamin B1, thiamine diphosphate (C12H19ClN4O7P2S), takes 1.5 years, assuming the time to simulate a structure is 12 hours. Previously, this task was performed using the intuition of an experienced researcher. Although effective, this process was time consuming and required significant human effort. Moreover, with much of their time spent on these manual tasks, expert researchers had little time to undertake innovative projects.

The solution

To solve these problems, Showa Denko used QSP materials, developed by QSimulate. QSP Materials is a platform that enables the automation of custom workflows for quantum mechanical simulations. We have collaborated with QSimulate to develop a new feature for this platform that automates the identification of the most stable structure for chemical compounds that have coordinate bonds. This form of automation was not possible before, and only manual techniques existed in its place. Now, using a single structure for the chemical compounds of interest as input, the platform automatically identifies the most stable structure as output.

What we have achieved

Our rigorous testing has demonstrated that the new QSP Materials feature more than halves the time required to complete this task compared to the previous approach. Additionally, our tests show that QSP Materials increases the likelihood of finding stable structures that could not be found manually.

Specifically, we compared the new QSP materials module to the manual approach by evaluating the structures of 25 chemical compounds, each comprising more than 100 atoms. For the manual approach, an experienced researcher used his intuition to generate 10 structures he thought were stable. The researcher then optimized each structure using quantum chemistry calculations4 to determine the most energetically stable. On the other hand, by using a random structure as input, QSP Materials automatically generated a most stable structure.

After comparing these approaches, we found that QSP Materials has a 3x higher probability of finding the correct structure than the manual approach. In addition, the new module on QSP Materials allowed us to reduce the resolution time to 45% of the manual approach.

Conclusion

As the performance of the new feature, which automatically finds the most stable structures of chemical compounds, has been successfully demonstrated, we plan to expand the use of this platform to other areas of materials development within of the company with the hope that it will lead to new innovations. We are committed to developing new capabilities like these that will drive us to our competitive advantage.

Contact:

Tel: +81 3-5470-3235