SEMI-SUPERVISED CLASSIFICATION OF 2D MATERIALS USING SELF-TRAINING CONVOLUTIONAL NEURAL NETWORKS

Abstract

Deep learning algorithms require large amounts of data, and their accuracy rates are directly related to the amount and quality of the data. Moreover, supervised learning models require the data to be labeled. However, data labeling is always a time-consuming and laborious process. Labeling data obtained from microscope images can be more laborious. Molybdenum disulfide (MoS2) in monolayer form, which can be produced on large surfaces with the chemical vapor deposition method (CVD) and has advantages for potential electronic applications, is a frequently studied material in the field of nanotechnology. However, MoS2 produced on these large surfaces usually has defective surfaces and needs to be detected. This process is a difficult process to be performed with a microscope by an expert. Artificial intelligence-based supervised learning algorithms, which need labeled data, provide an effective solution for these detections. Furthermore, increasing the number of labeled data increases the accuracy of these algorithms. In this study, a teacher-student model is explored using self-training, a semi-supervised learning technique, to effectively train a deep convolutional neural network to detect defects on MoS2 samples. Initially, the teacher model is trained using a small amount of data labeled by an expert. This trained model is enriched by generating pseudo-labels for previously unlabeled data. Then, a student model is trained using these real and pseudo-labeled data. The trained model then replaces the teacher model, and the process repeats, gradually improving labeling accuracy. The results show that the self-training method increases accuracy from 77% to 82% compared to the CNN model trained only on the existing labeled data, and the defect regions in MoS2 are effectively classified with minimal manual labeling.

Keywords

• Self-training
• Deep learning
• CNN
• Two-dimensional materials
• MoS2

References

1. [1] Alpaydin E. Introduction to machine learning: MIT press, 2020.
2. [2] Ahmed SF, Alam MSB, Hassan M, Rozbu MR, Ishtiak T, Rafa N, et al. Deep learning modelling techniques: current progress, applications, advantages, and challenges. Artificial Intelligence Review, 2023;56:13521-617.
3. [3] Liu Y, Yang Z, Yu Z, Liu Z, Liu D, Lin H, et al. Generative artificial intelligence and its applications in materials science: Current situation and future perspectives. Journal of Materiomics, 2023;9:798-816.
4. [4] Vasoya N. Revolutionizing nano materials processing through IoT-AI integration: opportunities and challenges. Journal of Materials Science Research and Reviews, 2023;6:294-328.
5. [5] Sobral JA, Obernauer S, Turkel S, Pasupathy AN, Scheurer MS. Machine learning the microscopic form of nematic order in twisted double-bilayer graphene. Nature Communications, 2023;14:5012.
6. [6] Nandipati M, Fatoki O, Desai S. Bridging Nanomanufacturing and Artificial Intelligence—A Comprehensive Review. Materials, 2024;17:1621.
7. [7] Geim AK, Novoselov KS. The rise of graphene. Nature materials, 2007;6:183-91.
8. [8] Zhang Y, Yao Y, Sendeku MG, Yin L, Zhan X, Wang F, et al. Recent progress in CVD growth of 2D transition metal dichalcogenides and related heterostructures. Advanced materials, 2019;31:1901694.

9. [9] Zhang H. Introduction: 2D materials chemistry. ACS Publications, 2018. p. 6089-90.
10. [10] Hua Q, Gao G, Jiang C, Yu J, Sun J, Zhang T, et al. Atomic threshold-switching enabled MoS2 transistors towards ultralow-power electronics. Nature Communications, 2020;11:6207.
11. [11] Zhang Y, Wan Q, Yang N. Recent advances of porous graphene: synthesis, functionalization, and electrochemical applications. Small, 2019;15:1903780.
12. [12] Fortin E, Sears W. Photovoltaic effect and optical absorption in MoS2. Journal of Physics and Chemistry of Solids, 1982;43:881-4.
13. [13] Yi M, Shen Z. A review on mechanical exfoliation for the scalable production of graphene. Journal of Materials Chemistry A, 2015;3:11700-15.
14. [14] Bonaccorso F, Lombardo A, Hasan T, Sun Z, Colombo L, Ferrari AC. Production and processing of graphene and 2d crystals. Materials today, 2012;15:564-89.
15. [15] Liu H, Wong SL, Chi D. CVD growth of MoS2‐based two‐dimensional materials. Chemical Vapor Deposition, 2015;21:241-59.
16. [16] Liu D, Chen X, Yan Y, Zhang Z, Jin Z, Yi K, et al. Conformal hexagonal-boron nitride dielectric interface for tungsten diselenide devices with improved mobility and thermal dissipation. Nature Communications, 2019;10:1188.
17. [17] Özden A, Şar H, Yeltik A, Madenoğlu B, Sevik C, Ay F, et al. CVD grown 2D MoS2 layers: A photoluminescence and fluorescence lifetime imaging study. physica status solidi (RRL)–Rapid Research Letters, 2016;10:792-6.
18. [18] Zhang J, Yu Y, Wang P, Luo C, Wu X, Sun Z, et al. Characterization of atomic defects on the photoluminescence in two‐dimensional materials using transmission electron microscope. InfoMat, 2019;1:85-97.
19. [19] Yorulmaz B, Özden A, Şar H, Ay F, Sevik C, Perkgöz NK. CVD growth of monolayer WS2 through controlled seed formation and vapor density. Materials Science in Semiconductor Processing, 2019;93:158-63.
20. [20] Lin X, Si Z, Fu W, Yang J, Guo S, Cao Y, et al. Intelligent identification of two-dimensional nanostructures by machine-learning optical microscopy. Nano Research, 2018;11:6316-24.
21. [21] Ngome Okello OF, Yang D-H, Chu Y-S, Yang S, Choi S-Y. Atomic-level defect modulation and characterization methods in 2D materials. APL Materials, 2021;9.
22. [22] Al‐Waisy AS, Ibrahim DA, Zebari DA, Hammadi S, Mohammed H, Mohammed MA, et al. Identifying defective solar cells in electroluminescence images using deep feature representations. PeerJ Computer Science, 2022;8:e992.
23. [23] Yao G, Lei T, Zhong J. A review of convolutional-neural-network-based action recognition. Pattern Recognition Letters, 2019;118:14-22.
24. [24] Alzubaidi L, Zhang J, Humaidi AJ, Al-Dujaili A, Duan Y, Al-Shamma O, et al. Review of deep learning: concepts, CNN architectures, challenges, applications, future directions. Journal of big Data, 2021;8:1-74.
25. [25] Bhuvaneswari V, Priyadharshini M, Deepa C, Balaji D, Rajeshkumar L, Ramesh M. Deep learning for material synthesis and manufacturing systems: A review. Materials Today: Proceedings, 2021;46:3263-9.
26. [26] Perkgoz C. Identifying optical microscope images of CVD-grown two-dimensional MoS2 by convolutional neural networks and transfer learning. PeerJ Computer Science, 2024;10:e1885.
27. [27] Xie Q, Luong M-T, Hovy E, Le QV. Self-training with noisy student improves imagenet classification. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020. p. 10687-98.
28. [28] Zou Y, Yu Z, Liu X, Kumar B, Wang J. Confidence regularized self-training. Proceedings of the IEEE/CVF international conference on computer vision, 2019. p. 5982-91.
29. [29] Yu L, Liu X, Van de Weijer J. Self-training for class-incremental semantic segmentation. IEEE Transactions on Neural Networks and Learning Systems, 2022;34:9116-27.
30. [30] Ke R, Aviles-Rivero AI, Pandey S, Reddy S, Schönlieb C-B. A three-stage self-training framework for semi-supervised semantic segmentation. IEEE Transactions on Image Processing, 2022;31:1805-15.
31. [31] Yamashita R, Nishio M, Do RKG, Togashi K. Convolutional neural networks: an overview and application in radiology. Insights into imaging, 2018;9:611-29.
32. [32] Soffer S, Ben-Cohen A, Shimon O, Amitai MM, Greenspan H, Klang E. Convolutional neural networks for radiologic images: a radiologist’s guide. Radiology, 2019;290:590-606.
33. [33] He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition. Proceedings of the IEEE conference on computer vision and pattern recognition, 2016. p. 770-8.