| Project/Area Number |
22KJ0780
|
|---|
| Project/Area Number (Other) |
22J10567 (2022)
|
|---|
| Research Category |
Grant-in-Aid for JSPS Fellows
|
| Allocation Type | Multi-year Fund (2023)
Single-year Grants (2022) |
|---|
| Section | 国内 |
|---|
| Review Section |
Basic Section 19020:Thermal engineering-related
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
ZHANG YUCHENG 東京大学, 工学系研究科, 特別研究員(DC2)
|
|---|
| Project Period (FY) |
2023-03-08 – 2024-03-31
|
| Project Status |
Completed (Fiscal Year 2023)
|
| Budget Amount *help |
¥1,700,000 (Direct Cost: ¥1,700,000)
Fiscal Year 2023: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 2022: ¥900,000 (Direct Cost: ¥900,000)
|
|---|
| Keywords | electret material / CYTOP / deep learning / molecule optimization / PCM / DFT / energy material / polymer electret / ionization potential / molecule generation / machine learning |
|---|
| Outline of Research at the Start |
Previous material design is often by trial and error. Recently, DFT and machine learning is used to accelerate the material design. However, the proposed molecules by simulation are usually difficult for practical synthesis and application, where collaborative intelligence is developed and coupled.
|
| Outline of Annual Research Achievements |
Density functional theory (DFT) with polarizable continuum model (PCM) is introduced to analyze the charge trapping mechanism of CYTOP-based electrets. It is found that the computational cost of PCM-DFT is 16 times smaller than that of MD-DFT and requires no manual intervention. Thereafter, a quantum chemical dataset consisted of 10k molecules is built via high-throughput PCM-DFT computations.
ChemTS-based de novo molecule generation algorithm has been employed to design new molecules. To avoid the huge computational cost and the difficulty in the chemical synthesis, graph neural networks such as MEGNET have been used to screen amine end groups and to predict the charging performance of CYTOP electrets. Functional group enrichment analysis is made to extract interpretable knowledge from abundant data where hydroxyl group and piperazine substructure are found effective. Thereafter, quantum chemical formula, deep reinforcement learning and expert knowledge are coupled for successfully building an automatic collaborative intelligence system.
Brand-new superior electrets such as CTX-A/APDEA, DHPEDA, BAPP, APPCA are proposed and synthesized based on the proposed simulation methods and AI-based algorithms. Taking the developed CTX-A/BAPP as an example. It can retain the surface potential of over +/- 3kV after 2135 hours under room temperature. Its TSD peak temperature is around 236 °C, while the previously developed CTX-A/APDEA is around 180 °C. Its lifetime is estimated as 146 years at 80 ℃, which is much better than previously commercialized CTYOP-EGG (12.4 years).
|
