Abstract
Determining the number of constituent atoms in metallic nanoclusters (NCs) directly from imaging is key to understanding how atomicity governs their size-dependent properties. Scanning transmission electron microscopy (STEM), which captures real-space images of materials with tunable magnification down to the atomic scale, provides an invaluable means to probe such structures. However, despite these advantages, automated and accurate identification of NC atomicity remains challenging, requiring robust extraction of features such as projected shape and contrast distribution from imaging data. To address this challenge, we present a deep learning framework that classifies platinum NCs (Ptn; n = 19, 30, 41, 55, 70) using high-resolution aberration-corrected STEM images. A convolutional neural network extracts structural features that are separable in UMAP (Uniform Manifold Approximation and Projection) space, with class-specific focus visualized using Grad-CAM (Gradient-weighted Class Activation Mapping). The model achieves high accuracy, even for mixed-atomicity samples (n = 19, 41, 70) on a shared substrate. To address domain shift, we apply fine-tuning with high-confidence pseudo-labels, significantly recovering performance. A dual-channel model integrating Local Contrast Normalization (LCN) filtering achieves a coefficient of determination of R² = 0.94 ± 0.03, outperforming size-based classification. This framework automates atomic-scale classification from STEM images and advances autonomous workflows via real-time analysis and machine learning based decisions.
Similar content being viewed by others
Data availability
The datasets generated and analyzed during the current study, including STEM image patches and class labels, are available at Zenodo under the https://doi.org/10.5281/zenodo.18014239. All relevant documentation, including data format specifications and Python environment files, is provided to ensure full reproducibility of the results.
Code availability
The code used for training, inference, and visualization in this study is available at Zenodo under the https://doi.org/10.5281/zenodo.18014239. The repository includes scripts for data preprocessing, model training, evaluation, Grad-CAM visualization, and UMAP projection, along with documentation and environment configuration files to ensure reproducibility.
References
Bawendi, M. G., Steigerwald, M. L. & Brus, L. E. The quantum mechanics of larger semiconductor clusters (‘quantum dots’). Annu. Rev. Phys. Chem. 41, 477–496 (1990).
Brack, M. The physics of simple metal clusters: self-consistent jellium model and semiclassical approaches. Rev. Mod. Phys. 65, 677–732 (1993).
de Heer, W. A. The physics of simple metal clusters: experimental aspects and simple models. Rev. Mod. Phys. 65, 611–676 (1993).
Alivisatos, A. P. Semiconductor clusters, nanocrystals, and quantum dots. Science 271, 933–937 (1996).
Haruta, M. Size- and support-dependency in the catalysis of gold. Catal. Today 36, 153–166 (1997).
Heiz, U., Sanchez, A., Abbet, S. & Schneider, W.-D. Catalytic oxidation of carbon monoxide on monodispersed platinum clusters: each atom counts. J. Am. Chem. Soc. 121, 3214–3217 (1999).
Dulub, O., Hebenstreit, W. & Diebold, U. Imaging cluster surfaces with atomic resolution: the strong metal-support interaction state of Pt supported on TiO 2 (110). Phys. Rev. Lett. 84, 3646–3649 (2000).
Franz, D. et al. Atomic structure of Pt nanoclusters supported by graphene/Ir(111) and reversible transformation under CO exposure. Phys. Rev. B 93, 045426 (2016).
Tsunoyama, H., Shibuta, M., Nakaya, M., Eguchi, T. & Nakajima, A. Synthesis and characterization of metal-encapsulating Si16 cage superatoms. Acc. Chem. Res. 51, 1735–1745 (2018).
Ohshimo, K. et al. Correlation between electronic shell structure and inertness of Cun+ toward O2 adsorption at n = 15, 21, 41, and 49. J. Phys. Chem. A 122, 2927–2932 (2018).
Shibuta, M., Inoue, T., Kamoshida, T., Eguchi, T. & Nakajima, A. Al13− and B@Al12− superatoms on a molecularly decorated substrate. Nat. Commun. 13, 1336 (2022).
El-Sayed, M. A. Small is different: shape-, size-, and composition-dependent properties of some colloidal semiconductor nanocrystals. Acc. Chem. Res. 37, 326–333 (2004).
Landman, U. & Luedtke, W. D. Small is different: energetic, structural, thermal, and mechanical properties of passivated nanocluster assemblies. Faraday Discuss. 125, 1–22 (2004).
Häkkinen, H. & Landman, U. Gas-phase catalytic oxidation of CO by Au2-. J. Am. Chem. Soc. 123, 9704–9705 (2001).
Link, S. & El-Sayed, M. A. Optical properties and ultrafast dynamics of metallic nanocrystals. Annu. Rev. Phys. Chem. 54, 331–366 (2003).
Bergeron, D. E., Roach, P. J., Castleman, A. W., Jones, N. O. & Khanna, S. N. Al cluster superatoms as halogens in polyhalides and as alkaline earths in iodide salts. Science 307, 231–235 (2005).
Burda, C., Chen, X., Narayanan, R. & El-Sayed, M. A. Chemistry and properties of nanocrystals of different shapes. Chem. Rev. 105, 1025–1102 (2005).
Häkkinen, H. The gold–sulfur interface at the nanoscale. Nat. Chem. 4, 443–455 (2012).
Yamamoto, K. & Imaoka, T. Precision synthesis of subnanoparticles using dendrimers as a superatom synthesizer. Acc. Chem. Res. 47, 1127–1136 (2014).
Imaoka, T. et al. Platinum clusters with precise numbers of atoms for preparative-scale catalysis. Nat. Commun. 8, 688 (2017).
Ohnuma, A. et al. Enhanced oxygen reduction activity of size-selected platinum subnanocluster catalysts: Ptn (n = 3–9). Catal. Sci. Technol. 12, 1400–1407 (2022).
Tsunoyama, H. et al. Enhanced oxygen reduction activity of platinum subnanocluster catalysts through charge redistribution. Chem. Commun. 55, 12603–12606 (2019).
Ohnuma, A. et al. Size-selected platinum subnanocluster catalysts alloyed with nickel (PtnNim, n = 3–7, m = 1–3) or cobalt (PtnCom, n = 4–8, m = 1–2) for the oxygen reduction reaction. ACS Appl. Nano Mater. 8, 16748–16758 (2025).
Nair, L. V. et al. Pt17(CO)12(PPh3)8]n+ (n = 1, 2): synthesis and geometric and electronic structures. J. Phys. Chem. C 121, 11002–11009 (2017).
Negishi, Y. et al. -Alumina-supported Pt17 cluster: controlled loading, geometrical structure, and size-specific catalytic activity for carbon monoxide and propylene oxidation. Nanoscale Adv. 2, 669–678 (2020).
Schmitt, C. et al. Continuous flow synthesis of atom-precise platinum clusters. Nanoscale Adv. 6, 2459–2468 (2024).
Chakraborty, P. et al. Exploring the activation of atomically precise [Pt17(CO)12(PPh3)8]2+ clusters: mechanism and energetics in gas phase and on an inert surface. ACS Nano 19, 3624–3634 (2025).
Hirata, N. et al. Platinum nanocluster catalysts supported on Marimo carbon via scalable dry deposition synthesis. RSC Adv. 11, 39216–39222 (2021).
Tsukamoto, K. et al. Evaluating atomic counts in metal nanoclusters via scanning transmission electron microscopy. Ultramicroscopy 278, 114242 (2025).
Butler, K. T., Davies, D. W., Cartwright, H., Isayev, O. & Walsh, A. Machine learning for molecular and materials science. Nature 559, 547–555 (2018).
Toyao, T. et al. Machine learning for catalysis informatics: recent applications and prospects. ACS Catal. 10, 2260–2297 (2020).
Choudhary, K. et al. Recent advances and applications of deep learning methods in materials science. npj Comput. Mater. 8, 59 (2022).
Zhao, X.-G. et al. Machine learning for experimental reactivity of a set of metal clusters toward C–H activation. J. Am. Chem. Soc. 146, 12485–12495 (2024).
Mou, L.-H. et al. Unveiling universal reactivity descriptors of metal clusters toward dinitrogen activation: a machine learning protocol with three-level feature extraction. ACS Catal. 15, 6618–6627 (2025).
Helveg, S. et al. Atomic-scale imaging of carbon nanofibre growth. Nature 427, 426–429 (2004).
Midgley, P. A. & Dunin-Borkowski, R. E. Electron tomography and holography in materials science. Nat. Mater. 8, 271–280 (2009).
Bals, S. et al. Atomic scale dynamics of ultrasmall germanium clusters. Nat. Commun. 3, 897 (2012).
Lee, J., Zhou, W., Pennycook, S. J., Idrobo, J.-C. & Pantelides, S. T. Direct visualization of reversible dynamics in a Si6 cluster embedded in a graphene pore. Nat. Commun. 4, 1650 (2013).
Inazu, M. et al. Dynamic hetero-metallic bondings visualized by sequential atom imaging. Nat. Commun. 13, 2968 (2022).
Ziatdinov, M. et al. Deep learning of atomically resolved scanning transmission electron microscopy images: chemical identification and tracking local transformations. ACS Nano 11, 12742–12752 (2017).
Colliard-Granero, A. et al. Deep learning for the automation of particle analysis in catalyst layers for polymer electrolyte fuel cells. Nanoscale 14, 10–18 (2022).
Kim, J. et al. Self-supervised machine learning framework for high-throughput electron microscopy. Sci. Adv. 11, eads5552 (2025).
Zhong, X. et al. Explainable machine learning in materials science. npj Comput. Mater. 8, 204 (2022).
Ziletti, A., Kumar, D., Scheffler, M. & Ghiringhelli, L. M. Insightful classification of crystal structures using deep learning. Nat. Commun. 9, 2775 (2018).
Lu, Y. & Chen, W. Sub-nanometre sized metal clusters: from synthetic challenges to the unique property discoveries. Chem. Soc. Rev. 41, 3594 (2012).
Tsunoyama, H. et al. Development of high-flux ion source for size-selected nanocluster ions based on high-power impulse magnetron sputtering. Chem. Lett. 42, 857–859 (2013).
Zhang, C. et al. Advanced nanocluster ion source based on high-power impulse magnetron sputtering and time-resolved measurements of nanocluster formation. J. Phys. Chem. A 117, 10211–10217 (2013).
Duffe, S. et al. Penetration of thin C60 films by metal nanoparticles. Nat. Nanotechnol. 5, 335–339 (2010).
He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. In Proc. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 770–778 (2016). https://doi.org/10.1109/CVPR.2016.90.
Pinto, N., Cox, D. D. & DiCarlo, J. J. Why is real-world visual object recognition hard? PLoS Comput. Biol. 4, e27 (2008).
Jarrett, K., Kavukcuoglu, K., Ranzato, M. & LeCun, Y. What is the best multi-stage architecture for object recognition? In Proc. 2009 IEEE 12th International Conference on Computer Vision 2146–2153 (2009). https://doi.org/10.1109/ICCV.2009.5459469.
McInnes, L., Healy, J. & Melville, J. UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. J. Open Source Softw. 3, 861 (2018).
Selvaraju, R. R. et al. Grad-CAM: visual explanations from deep networks via gradient-based localization. In Proc. 2017 IEEE International Conference on Computer Vision (ICCV) 618–626 (2017). https://doi.org/10.1109/ICCV.2017.74.
Wang, M. & Deng, W. Deep visual domain adaptation: a survey. Neurocomputing 312, 135–153 (2018).
Wilson, G. & Cook, D. J. A survey of unsupervised deep domain adaptation. ACM Trans. Intell. Syst. Technol. 11, 46 (2020). 51:1-51.
Harris et al. Array programming with NumPy. Nature 585, 357–362 (2020).
OpenCV: Image filtering. https://docs.opencv.org/4.11.0/d4/d86/group__imgproc__filter.html#gaa0c3e5d762b3d7a8d4bde781ba6dc522.
Event — PyTorch 2.7 documentation. https://docs.pytorch.org/docs/stable/generated/torch.cuda.Event.html.
Liu, Z. et al. A ConvNet for the 2020s. In Proc. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 11966–11976 (2022). https://doi.org/10.1109/CVPR52688.2022.01167.
Acknowledgements
This work was partly supported by the Aichi Prefectural Government under the “Knowledge Hub Aichi”, Priority Research Project (4th term), by the Advanced Research Infrastructure for Materials and Nanotechnology in Japan (ARIM) (Project Issue Number: 24NM0110), and by JSPS KAKENHI Grant-in-Aid for Scientific Research (B) (No. 24K01442). The authors are grateful to Prof. Y. Adachi (Nagoya University) and Prof. M. Hatanaka (Keio University / IMS) for valuable discussion on machine learning procedures.
Author information
Authors and Affiliations
Contributions
Conceptualization: K.T., N.H., M.T., A.N. Methodology: K.T., N.H., M.T., A.N. Investigation: K.T., N.H., M.T., A.N. Visualization: K.T., N.H., M.T., A.N. Supervision: K.T., A.N. Writing—original draft: K.T., N.H., M.T., A.N. Writing—review & editing: K.T., N.H., M.T., A.N.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no Competing Non-Financial Interests but the following Competing Financial Interests: They all are inventors on JAPAN patent JP 2025-142865, submitted by Ayabo Corporation and Keio University, which cover the application related to the classification system using electron microscopy images of NCs described in this manuscript.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Tsukamoto, K., Hirata, N., Tona, M. et al. Interpretable deep learning for atomicity classification of platinum nanoclusters in STEM images. npj Comput Mater (2026). https://doi.org/10.1038/s41524-026-02014-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41524-026-02014-z
