Prediction of material parameters of spin textures from LTEM images by machine learning models CNN and ViT

doi:10.1088/1674-1056/ae5db1

Abstract Machine learning (ML) models were trained to predict the Dzyaloshinskii-Moriya interaction constant $D$ and anisotropy constant $K$ of spin textures from Lorentz transmission electron microscopy (LTEM) images. Two ML models, convolutional neural network (CNN) and vision transformer (ViT), were trained, tested and employed to predict the values of $D$ and $K$. Firstly, training and testing datasets composed of 9300 topological spin textures were prepared using the micromagnetic simulation method with the values of $D$ and $K$ randomly chosen. Secondly, the performance of the CNN model for predicting $D$ and $K$ values was investigated by varying the number of training data, the pooling process, and the number of fully connected layers, which indicated high prediction accuracies. Thirdly, better performance of the ViT model for predicting $D$ and $K$ values was achieved with the coefficient of determination $R^{2}$ reaching as high as 0.9987 and 0.9991 for predicting $D$ and $K$ values, respectively. Particularly, the reliance on synthetic data and noise robustness was discussed. Finally, the prediction abilities of CNN and ViT models were evaluated and compared. The results of this research indicate that the ML models can achieve the material parameters directly and effectively from LTEM images, and this process may contribute to the design of advanced devices based on the topological spin textures.

Keywords: spin texture LTEM machine learning ViT model

Received: 23 December 2025
Revised: 24 March 2026
Accepted manuscript online: 10 April 2026

PACS:	75.25.-j	(Spin arrangements in magnetically ordered materials (including neutron And spin-polarized electron studies, synchrotron-source x-ray scattering, etc.))
	75.78.Cd	(Micromagnetic simulations ?)
	87.19.lv	(Learning and memory)
	07.05.Tp	(Computer modeling and simulation)

Corresponding Authors: Lin Chen
E-mail: cl@njupt.edu.cn

Cite this article:

Bohan Li(李博涵), Xinyuan Zhang(张歆媛), Guanhua Chen(陈冠桦), and Lin Chen(陈琳) Prediction of material parameters of spin textures from LTEM images by machine learning models CNN and ViT 2026 Chin. Phys. B 35 077507

[1] Fert A, Reyren N and Cros V 2017 Nat. Rev. Mater. 2 17031
[2] Göbel B, Mertig I and Tretiakov O A 2021 Phys. Rep. 895 1
[3] He Q L, Hughes T L, Armitage N P, Tokura Y andWang K L 2022 Nat. Mater. 21 15
[4] Jiang L X, Li Q C, Zhang X, Li J F, Zhang J, Chen Z X, Zeng M and Wu H 2024 Acta Phys. Sin. 73 017505 (in Chinese)
[5] Yang H X, Liang J H and Cui Q R 2023 Nat. Rev. Phys. 5 43
[6] Zhao X P and Zhao J H 2025 npj Spintronics 3 25
[7] Mühlbauer S, Binz B, Jonietz F, Pfleiderer C, Rosch A, Neubauer A, Georgii R and Boni P 2009 Science 323 915
[8] Kahmann T, Laureen Rösch E, Enpuku K, Yoshida T and Ludwig F 2021 J. Magn. Magn. Mater. 519 167402
[9] Masood K I, Raich R, Jander A and Dhagat P 2024 J. Magn. Magn. Mater. 610 172562
[10] Kuepferling M, Casiraghi A, Soares G, Durin G, Garcia-Sanchez F, Chen L, Back C H, Marrows C H, Tacchi S and Carlotti G 2023 Rev. Mod. Phys. 95 015003
[11] Phatak C, Petford-Long A K and De Graef M 2016 Curr. Opin. Solid St. M. 20 107
[12] Kang S, Töllner M, Wang D, Minnert C, Durst K, Caron A, Dunin- Borkowski R E, McCord J, Kübel C and Mu X 2025 Nat. Commun. 16 1305
[13] Tang J, Kong L Y,WangWW, Du H F and Tian M L 2019 Chin. Phys. B 28 087503
[14] Marchiori E, Ceccarelli L, Rossi N, Lorenzelli L, Degen C L and Poggio M 2022 Nat. Rev. Phys. 4 49
[15] Habenschaden C, Sievers S, Klasen A, Cerreta A and Schumacher H W 2024 Rev. Sci. Instrum. 95 113704
[16] Kazakova O, Puttock R, Barton C, Corte-León H, Jaafar M, Neu V and Asenjo A 2019 J. Appl. Phys. 125 060901
[17] Allen E A and Tkatchenko A 2022 Sci. Adv. 8 eabm7185
[18] Li Q, Fu N, Omee S S and Hu J 2024 npj Comput. Mater. 10 245
[19] Mobarak M H, Mimona M A, Islam M A, Hossain N, Zohura F T, Imtiaz I and Rimon M I H 2023 Appl. Surf. Sci. Adv. 18 100523
[20] Ma W, Liu Z C, Kudyshev Z A, Boltasseva A, Cai W S and Liu Y M 2021 Nat. Photon. 15 77
[21] Liu D Y, Xu L M, Lin X M, Wei X, Yu W J, Wang Y and Wei Z M 2022 Chip 1 100033
[22] He W W, Li J Z, Kong X and Deng L 2024 Commun. Eng. 3 151
[23] Li J, Fang W Q, Jin S J, Suo C Y, Zhang T D, Wu Y L, Xu X D, Liu Y and Yao D X 2025 AI Sci. 1 015001
[24] Karniadakis G E, Kevrekidis I G, Lu L, Perdikaris P, Wang S F and Yang L 2021 Nat. Rev. Phys. 3 422
[25] Choubisa H, Todorović P, Pina J M, Parmar D H, Li Z L, Voznyy O, Tamblyn I and Sargent E H 2023 npj Comput. Mater. 9 117
[26] Shi S W, Chu S B, Xie Y E and Chen Y P 2025 Phys. Scr. 100 015957
[27] Chen G H, Yao J C, Zhu H F, Zhi T, Wang J, Xue J J, Chen L, Tao T and Tao Z K 2025 Acta Phys. Sin. 74 190203 (in Chinese)
[28] Mehmood N, Wang J B and Liu Q F 2022 J. Appl. Phys. 132 043904
[29] McCray A R C, Zhou T, Kandel S, Petford-Long A, CherukaraMJ and Phatak C 2024 npj Comput. Mater. 10 111
[30] McCray A R C, Bender A, Petford-Long A and Phatak C 2024 APL Mach. Learn. 2 026120
[31] Beg M, Lang M and Fangohr H 2022 IEEE Trans. Magn. 58 7300205
[32] Beleggia M and Zhu Y 2010 Philos. Mag. 83 1045
[33] Chen X W, Yang L Y, Savchenko A, Yang D H, Shi W, Denneulin T, Kiselev N S, Jin L, Song D S, Dunin-Borkowski R E and Zheng F S 2025 Phys. Rev. B 112 214455
[34] Haug T, Otto S, Schneider M and Zweck J 2003 Ultramicroscopy 96 201
[35] Walton S K, Zeissler K, Branford W R and Felton S 2013 IEEE Trans. Magn. 49 4795
[36] Holt S J R, Lang M, Loudon J C, Hichken T J, Suess D, Cortes-Ortuno D, Pathak S A, Beg M, Zulfiqar K and Fangohr H 2025 npj Comput. Mater. 11 20
[37] Yu X Z, Onose Y, Kanazawa N, Park J H, Han J H, Matsui Y, Nagaosa N and Tokura Y 2010 Nature 465 901
[38] Yu X Z, Kanazawa N, Onose Y, Kimoto K, Zhang W Z, Ishiwata S, Matsui Y and Tokura Y 2011 Nat. Mater. 10 106
[39] Yu X Z, Koshibae W, Tokunaga Y, Shibata K, Taguchi Y, Nagaosa N and Tokura Y 2018 Nature 564 95
[40] Zhao X, Wang L M, Zhang Y F, Han X M, Deveci M and Parmar M 2024 Artif. Intell. Rev. 57 99
[41] Alzubaidi L, Zhang J, Humaidi A J, Al-Dujaili A, Duan Y, Al-Shamma O, Santamaría J, Fadhel M A, Al-Amidie M and Farhan L 2021 J. Big Data 8 53
[42] Dosovitskiy A, Beyer L, Kolesnikov A,Weissenborn D, Zhai X H, Unterthiner T, Dehghani M, Minderer M, Heigold G, Gelly S, Uszkoreit J and Houlsby N 2010 ArXiv abs/2010.11929
[43] Maurício J, Domingues I and Bernardino J 2023 Appl. Sci. 13 5521
[44] Chicco D, Warrens M J and Jurman G 2021 PeerJ Comput. Sci. 7 e623
[45] Schmidhuber J 2015 Neural Networks 61 85
[46] Akhtar N and Ragavendran U 2020 Neural Comput. Appl. 32 879
[47] Lyu B Y, Zhao S H, Zhang Y B, Wang W W, Zheng F S, Dunin- Borkowski R E, Zang J D and Du H F 2024 Sci. China-Phys. Mech. Astron. 67 117511
[48] García-Palacios J L and Lázaro F J 1998 Phys. Rev. B 58 14937

1. 2026-077507-SI.pdf(1128KB)

[1]	Curation and featurization of multiple topological materials databases Yuqing He(贺雨晴), Matteo Giantomassi, Gian-Marco Rignanese, and Hongming Weng(翁红明). Chin. Phys. B, 2026, 35(5): 050701.
[2]	Unveiling stable and efficient antiperovskite semiconductors via high-throughput computation and interpretable machine learning Hao Qu(瞿浩), Tao Hu(胡涛), Mingjun Li(李明军), Jiangyu Yang(杨江渝), Yunyi Zhou(周云逸), Shichang Li(李世长), Dengfeng Li(李登峰), Gang Tang(唐刚), and Chunbao Feng(冯春宝). Chin. Phys. B, 2026, 35(4): 046102.
[3]	Machine learning prediction of HSE06-level band gaps in two-dimensional semiconductors with reference-guided graph neural networks Zhen Wan(万振), Shun-Bo Jiang(姜顺波), Yuan Li(李圆), Hui Wang(王辉), Zong-Liang Li(李宗良), and Guang-Ping Zhang(张广平). Chin. Phys. B, 2026, 35(3): 037102.
[4]	Machine learning of chaotic characteristics in classical nonlinear dynamics using variational quantum circuit Sheng-Chen Bai(白生辰) and Shi-Ju Ran(冉仕举). Chin. Phys. B, 2026, 35(2): 020303.
[5]	EDIS: A simulation software for dynamic ion intercalation/deintercalation processes in electrode materials Liqi Wang(王力奇), Ruijuan Xiao(肖睿娟), and Hong Li(李泓). Chin. Phys. B, 2026, 35(1): 018201.
[6]	Revealing the dynamic responses of Pb under shock loading based on DFT-accuracy machine learning potential Enze Hou(侯恩则), Xiaoyang Wang(王啸洋), and Han Wang(王涵). Chin. Phys. B, 2026, 35(1): 018701.
[7]	Machine learning-assisted optimization of MTO basis sets Zhiqiang Li(李志强) and Lei Wang(王蕾). Chin. Phys. B, 2026, 35(1): 016301.
[8]	Review of machine learning tight-binding models: Route to accurate and scalable electronic simulations Jijie Zou(邹暨捷), Zhanghao Zhouyin(周寅张皓), Shishir Kumar Pandey, and Qiangqiang Gu(顾强强). Chin. Phys. B, 2026, 35(1): 017101.
[9]	Machine learning approach to reconstruct dephasing time from solid HHG spectra Jiahao Liu(刘佳豪), Xi Zhao(赵曦), Jun Wang(王俊), and Songbin Zhang(张松斌). Chin. Phys. B, 2025, 34(9): 097804.
[10]	Hyperparameter optimization and force error correction of neuroevolution potential for predicting thermal conductivity of wurtzite GaN Zhuo Chen(陈卓), Yuejin Yuan(袁越锦), Wenyang Ding(丁文扬), Shouhang Li(李寿航), Meng An(安盟), and Gang Zhang(张刚). Chin. Phys. B, 2025, 34(8): 086110.
[11]	Significant increase in thermal conductivity of cathode material LiFePO₄ by Na substitution: A machine learning interatomic potential-assisted investigation Shi-Yi Li(李诗怡), Qian Liu(刘骞), Yu-Jia Zeng(曾育佳), Guofeng Xie(谢国锋), and Wu-Xing Zhou(周五星). Chin. Phys. B, 2025, 34(2): 028201.
[12]	Quantum algorithm for marginal Fisher analysis Jing Li(李静), Yanqi Song(宋燕琪), Sujuan Qin(秦素娟), Wenmin Li(李文敏), and Fei Gao(高飞). Chin. Phys. B, 2025, 34(12): 120302.
[13]	Coupling of magnon modes in nanodisk with spin texture Zijie Zhou(周子杰), Junning Zhao(赵俊宁), Xinhui Ma(马心慧), Rong Wang(王熔), and Fusheng Ma(马付胜). Chin. Phys. B, 2025, 34(10): 107506.
[14]	Database of ternary amorphous alloys based on machine learning Xuhe Gong(巩旭菏), Ran Li(李然), Ruijuan Xiao(肖睿娟), Tao Zhang(张涛), and Hong Li(李泓). Chin. Phys. B, 2025, 34(1): 016101.
[15]	Classifying extended, localized and critical states in quasiperiodic lattices via unsupervised learning Bohan Zheng(郑博涵), Siyu Zhu(朱思宇), Xingping Zhou(周兴平), and Tong Liu(刘通). Chin. Phys. B, 2025, 34(1): 017103.

No Suggested Reading articles found!

Viewed

Full text

Abstract

Cited

Metrics
Related Articles

Prediction of material parameters of spin textures from LTEM images by machine learning models CNN and ViT

Cite this article:

Online attention