Plenary Talk - 2

 

AI-assisted Optimization of End Group in Amorphous Fluorinated Polymer Electret for Vibration Energy Harvesting

Yuji Suzuki1

ysuzuki@mesl.t.u-tokyo.ac.jp

1Dept. of Mechanical Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan

1.  Introduction

Electret-based vibration energy harvesting (EH) is advantageous in terms of higher output power at low frequencies and small volumes [1]. Figure 1 shows our wrist-worn rotational electret energy harvester [2], which can provide 0.4 mW during human walking. We previously found that CYTOP (AGC Chemicals), which is amorphous fluorinated polymer, offers high surface potential. Although the number of repeat units is over 1000, the charging performance is very sensitive to the functional group at both ends [3]. We recently found through the quantum chemical analysis that the trapped charge is localized at the amide bond, which is formed by the dehydration reaction between the amine added and the carboxyl end group of CYTOP CTX-A [4,5]. In this talk, our recent studies on deep-learning-assisted discovery of new CYTOP- based polymer electret are discussed.

 

2.  AI-driven Discovery of CYTOP Electret [6]

We performed optimization on the functional groups of CYTOP by seeking in a large chemical database for promising amines which can react with the carboxylic groups in CTX-A. We employ the solid-state vertical ionization potential (IP) as a descriptor for the charging performance estimation. The electret discovery workflow is presented in Figure 2. After training MatErial Graph Network (MEGNet) models [7] with 5,522 IP data via quantum chemical computation, three molecules out of 1,176,591 candidates from the database are successfully identified and synthesized to validate their charging performance experimentally. The computational speed is 107-108 times faster than the DFT/PCM-based rigorous computations. One of the best candidates is 1,4-Bis(3-amino-propyl)piperazine (BAPP) with the solid-state IP of 5.32 eV, which exhibits -3.49 kV of surface potential with a 15 mm-thick film. Even in the high temperature of 115 ºC, which is above the glass transition temperature, the decay of surface potential is slow. Its estimated decay time at 80 ºC, which is the maximum operation temperature for consumer products, is over 140 years, and is much longer than our previous electret material [3].

3.  Conclusions

Machine learning based on MEGNet is used to evaluate the charging performance in the amorphous polymer electret. The potential of the deep-learning-based model is clearly shown for polymer electret discovery.

Figure   1:   Wrist-worn   rotational   electret   energy harvester [2]

Figure 2: Work procedure for fast screening promising electret candidates from PubChem and experimentally validate the charging performance [6].

 

This work was supported by JST CREST JPMJCR15Q3 and JPMJCR19Q1, Japan as well as "Advanced Research Infrastructure for Materials and Nanotechnology in Japan (ARIM)" of MEXT Proposal Number JPMXP1222UT1009.

References:

[1] Y. Suzuki, “Recent Progress in MEMS Electret Generator for Energy Harvesting,” IEEJ Trans. Electr. Electron. Eng., Vol. 6, pp. 101-111, 2011.

[2] T. Miyoshi, et al., “Output Energy Evaluation of Wrist-worn Electret Energy Harvester for Day-long Activities of Daily Living,” PowerMEMS 2022, Salt Lake City, T1A-02 (2022).

[3] K. Kashiwagi, et al., “Nano-Cluster-Enhanced High- Performance Perfluoro-Polymer Electrets for Energy Harvesting,” J. Micromech. Microeng., Vol. 21, 125016, 2011.

[4] S. Kim, et al., “Solid-State Electron Affinity Analysis of Amorphous Fluorinated Polymer Electret,” J. Phys. Chem. B, Vol. 124, pp. 10507-10513, 2020.

[5] Y. Zhang, et al., “Discovery of Polymer Electret Material via de Novo Molecule Generation and Functional Group Enrichment Analysis,” Appl. Phys. Lett. Vol. 118, 223904, 2021.

[6] Z. Mao, et al., “AI-driven Discovery of Amorphous Fluorinated Polymer Electret with Improved Charge Stability for Energy Harvesting,” Adv. Mater., in press.

[7] C. Chen, et al., “Graph Networks as a Universal Machine Learning Framework for Molecules and Crystals,” Chem. Mater. Vol. 31, 3564–3572, 2019

 

About speaker:

Professor Yuji Suzuki received the B.S., M.S., and Dr.Eng. degrees in mechanical engineering from the University of Tokyo, Tokyo, Japan, in 1987, 1989, and 1993, respectively. He is currently with the Department of Mechanical Engineering, the University of Tokyo, as a Professor. He received IEEE Fellow, Fellow of Combustion Institute, and Fellow of JSME. He serves as Steering Committee Member of PowerMEMS Conference, Organizing Committee member of IEEE International Symposium on Electret (ISE). He also served as the general co-chair of IEEE MEMS2010 (Hong Kong) and the general chair of PowerMEMS 2017 (Kanazawa). His research interests include MEMS-based energy harvesting using electrets, micro energy conversion such as microscale combustion, and optimal design/control of micro heat and fluid flow.