Abstract
Atomic-scale manufacturing of carbon-based quantum materials with single-bond precision holds immense potential in advancing tailor-made quantum materials with unconventional properties, which are crucial in developing next-generation spintronic devices and quantum information technologies. On-surface chemistry approaches, including surface-assisted synthesis and probe-assisted manipulation, are impeded by challenges in reaction selectivity control or restricted by scalability and production efficiency. Here we demonstrate the concept of the chemist-intuited atomic robotic probe by integrating probe chemistry knowledge and artificial intelligence, allowing for atomically precise single-molecule manipulation to fabricate single-molecule quantum π-magnets with single-bond precision. Our deep neural networks not only transform complex probe chemistry into machine-understandable tasks but also provide chemist intuition to elusive reaction mechanisms by extracting the critical chemical information within the data. A joint experimental and theoretical investigation demonstrates that a voltage-controlled two-electron-assisted electronic excitation enables synchronous six-bond transformations to extend the zigzag edge topology of single-molecule quantum π-magnets, triggered by phenyl C(sp2)–H bond activation, which aligns with initial conjectures given by the deep neural models. Our work represents a transition from autonomous fabrication to intelligent synthesis with levels of selectivity and precision beyond current synthetic tools for improved synthesis of organic quantum materials towards on-chip integration.
This is a preview of subscription content, access via your institution
Access options
Subscribe to this journal
Receive 12 digital issues and online access to articles
$119.00 per year
only $9.92 per issue
Buy this article
- Purchase on Springer Link
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Data availability
All experimental and theoretical calculation data are available in the main text or Supplementary Information.
Code availability
Codes for deep neural networks and STM-AI platform are available at GitHub (https://github.com/jiali1025/Intelligent-topological-engineering-of-quantum-p-magnets).
References
Yazyev, O. V. Emergence of magnetism in graphene materials and nanostructures. Rep. Prog. Phys. 73, 056501 (2010).
Wang, W. L., Yazyev, O. V., Meng, S. & Kaxiras, E. Topological frustration in graphene nanoflakes: magnetic order and spin logic devices. Phys. Rev. Lett. 102, 157201 (2009).
Morita, Y., Suzuki, S., Sato, K. & Takui, T. Synthetic organic spin chemistry for structurally well-defined open-shell graphene fragments. Nat. Chem. 3, 197–204 (2011).
Mishra, S. et al. Topological frustration induces unconventional magnetism in a nanographene. Nat. Nanotechnol. 15, 22–28 (2020).
Mishra, S. et al. Large magnetic exchange coupling in rhombus-shaped nanographenes with zigzag periphery. Nat. Chem. 13, 581–586 (2021).
Lieb, E. H. Two theorems on the Hubbard model. Phys. Rev. Lett. 62, 1201–1204 (1989).
Fernandez-Rossier, J. & Palacios, J. J. Magnetism in graphene nanoislands. Phys. Rev. Lett. 99, 177204 (2007).
Han, W., Kawakami, R. K., Gmitra, M. & Fabian, J. Graphene spintronics. Nat. Nanotechnol. 9, 794–807 (2014).
Song, S. et al. On-surface synthesis of graphene nanostructures with pi-magnetism. Chem. Soc. Rev. 50, 3238–3262 (2021).
Grill, L. et al. Nano-architectures by covalent assembly of molecular building blocks. Nat. Nanotechnol. 2, 687–691 (2007).
Su, J. et al. Atomically precise bottom-up synthesis of pi-extended [5]triangulene. Sci. Adv. 5, eaav7717 (2019).
Mishra, S. et al. Synthesis and characterization of pi-extended triangulene. J. Am. Chem. Soc. 141, 10621–10625 (2019).
Mishra, S. et al. Synthesis and characterization of [7]triangulene. Nanoscale 13, 1624–1628 (2021).
Stipe, B. C. et al. Single-molecule dissociation by tunneling electrons. Phys. Rev. Lett. 78, 4410–4413 (1997).
Hla, S. W., Bartels, L., Meyer, G. & Rieder, K. H. Inducing all steps of a chemical reaction with the scanning tunneling microscope tip: towards single molecule engineering. Phys. Rev. Lett. 85, 2777–2780 (2000).
Jiang, Y., Huan, Q., Fabris, L., Bazan, G. C. & Ho, W. Submolecular control, spectroscopy and imaging of bond-selective chemistry in single functionalized molecules. Nat. Chem. 5, 36–41 (2013).
Kaiser, K. et al. An sp-hybridized molecular carbon allotrope, cyclo[18]carbon. Science 365, 1299–1301 (2019).
Albrecht, F. et al. Selectivity in single-molecule reactions by tip-induced redox chemistry. Science 377, 298–301 (2022).
Pavlicek, N. et al. Synthesis and characterization of triangulene. Nat. Nanotechnol. 12, 308–311 (2017).
Wang, T. et al. Aza-triangulene: on-surface synthesis and electronic and magnetic properties. J. Am. Chem. Soc. 144, 4522–4529 (2022).
Mishra, S. et al. Nonbenzenoid high-spin polycyclic hydrocarbons generated by atom manipulation. ACS Nano 16, 3264–3271 (2022).
Butler, K. T., Davies, D. W., Cartwright, H., Isayev, O. & Walsh, A. Machine learning for molecular and materials science. Nature 559, 547–555 (2018).
Bures, J. & Larrosa, I. Organic reaction mechanism classification using machine learning. Nature 613, 689–695 (2023).
Davies, A. et al. Advancing mathematics by guiding human intuition with AI. Nature 600, 70–74 (2021).
Burger, B. et al. A mobile robotic chemist. Nature 583, 237–241 (2020).
Ruiz Euler, H. C. et al. A deep-learning approach to realizing functionality in nanoelectronic devices. Nat. Nanotechnol. 15, 992–998 (2020).
Li, J. et al. Machine vision automated chiral molecule detection and classification in molecular imaging. J. Am. Chem. Soc. 143, 10177–10188 (2021).
Lemm, D., von Rudorff, G. F. & von Lilienfeld, O. A. Machine learning based energy-free structure predictions of molecules, transition states, and solids. Nat. Commun. 12, 4468 (2021).
Oviedo, F., Ferres, J. L., Buonassisi, T. & Butler, K. T. Interpretable and explainable machine learning for materials science and chemistry. Acc. Mater. Res. 3, 597–607 (2022).
Rashidi, M. & Wolkow, R. A. Autonomous scanning probe microscopy in situ tip conditioning through machine learning. ACS Nano 12, 5185–5189 (2018).
Alldritt, B. et al. Automated structure discovery in atomic force microscopy. Sci. Adv. 6, eaay6913 (2020).
Leinen, P. et al. Autonomous robotic nanofabrication with reinforcement learning. Sci. Adv. 6, eabb6987 (2020).
Krull, A., Hirsch, P., Rother, C., Schiffrin, A. & Krull, C. Artificial-intelligence-driven scanning probe microscopy. Commun. Phys. 3, 54 (2020).
Carracedo-Cosme, J., Romero-Muniz, C. & Perez, R. A deep learning approach for molecular classification based on AFM images. Nanomaterials 11, 1658 (2021).
Chen, I. J. et al. Precise atom manipulation through deep reinforcement learning. Nat. Commun. 13, 7499 (2022).
Gordon, O. M. & Moriarty, P. J. Machine learning at the (sub)atomic scale: next generation scanning probe microscopy. Mach. Learn. Sci. Technol. 1, 023001 (2020).
Su, J., Lyu, P. & Lu, J. Atomically precise imprinting π-magnetism in nanographenes via probe chemistry. Precis. Chem. 1, 565–575 (2023).
Hapala, P. et al. Mechanism of high-resolution STM/AFM imaging with functionalized tips. Phys. Rev. B 90, 085421 (2014).
Su, J. et al. On-surface synthesis and characterization of [7]triangulene quantum ring. Nano Lett. 21, 861–867 (2021).
Li, J. et al. Uncovering the triplet ground state of triangular graphene nanoflakes engineered with atomic precision on a metal surface. Phys. Rev. Lett. 124, 177201 (2020).
Zhao, Y. et al. Quantum nanomagnets in on-surface metal-free porphyrin chains.Nat. Chem. 15, 53–60 (2022).
Su, X. et al. Atomically precise synthesis and characterization of heptauthrene with triplet ground state. Nano Lett. 20, 6859–6864 (2020).
Zheng, Y. et al. Designer spin order in diradical nanographenes. Nat. Commun. 11, 6076 (2020).
Ren, S., He, K., Girshick, R. & Sun, J. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Trans. Pattern Anal. Mach. Intell. 39, 1137−1149 (2017).
He, K., Gkioxari, G., Dollár, P. & Girshick, R. Mask R-CNN. 2017 IEEE International Conference on Computer Vision (ICCV) 2980–2988 (2017).
Toshev, A. & Szegedy, C. DeepPose: human pose estimation via deep neural networks. 2014 IEEE Conference on Computer Vision and Pattern Recognition 1653–1660 (2014).
Lundberg, S. M. & Lee, S.-I. A unified approach to interpreting model predictions. Adv. Neural Inf. Process. 30, 4768–4777 (2017).
Wang, S., Zhu, J., Blackwell, R. & Fischer, F. R. Automated tip conditioning for scanning tunneling spectroscopy. J. Phys. Chem. A 125, 1384–1390 (2021).
Alldritt, B. et al. Automated tip functionalization via machine learning in scanning probe microscopy. Comput. Phys. Commun. 273, 108258 (2022).
Arikawa, S., Shimizu, A., Shiomi, D., Sato, K. & Shintani, R. Synthesis and isolation of a kinetically stabilized crystalline triangulene. J. Am. Chem. Soc. 143, 19599–19605 (2021).
Cocker, T. L., Peller, D., Yu, P., Repp, J. & Huber, R. Tracking the ultrafast motion of a single molecule by femtosecond orbital imaging. Nature 539, 263–267 (2016).
Repp, J., Meyer, G., Stojkovic, S. M., Gourdon, A. & Joachim, C. Molecules on insulating films: scanning-tunneling microscopy imaging of individual molecular orbitals. Phys. Rev. Lett. 94, 026803 (2005).
Neaton, J. B., Hybertsen, M. S. & Louie, S. G. Renormalization of molecular electronic levels at metal-molecule interfaces. Phys. Rev. Lett. 97, 216405 (2006).
Yamazaki, S. et al. Interplay between switching driven by the tunneling current and atomic force of a bistable four-atom Si quantum dot. Nano Lett. 15, 4356–4363 (2015).
Erpenbeck, A., Ke, Y., Peskin, U. & Thoss, M. Current-induced dissociation in molecular junctions beyond the paradigm of vibrational heating: the role of antibonding electronic states. Phys. Rev. B 102, 195421 (2020).
Ke, Y., Erpenbeck, A., Peskin, U. & Thoss, M. Unraveling current-induced dissociation mechanisms in single-molecule junctions. J. Chem. Phys. 154, 234702 (2021).
Giessibl, F. J. Advances in atomic force microscopy. Rev. Mod. Phys. 75, 949–983 (2003).
Giessibl, F. J. The qPlus sensor, a powerful core for the atomic force microscope. Rev. Sci. Instrum. 90, 011101 (2019).
Gross, L., Mohn, F., Moll, N., Liljeroth, P. & Meyer, G. The chemical structure of a molecule resolved by atomic force microscopy. Science 325, 1110–1114 (2009).
Bartels, L., Meyer, G. & Rieder, K. H. Basic steps of lateral manipulation of single atoms and diatomic clusters with a scanning tunneling microscope tip. Phys. Rev. Lett. 79, 697–700 (1997).
Rizzo, D. J. et al. Topological band engineering of graphene nanoribbons. Nature 560, 204–208 (2018).
Groning, O. et al. Engineering of robust topological quantum phases in graphene nanoribbons. Nature 560, 209–213 (2018).
Acknowledgements
J.L. acknowledges the support from the NRF, Prime Minister’s Office, Singapore, under the Competitive Research Program Award (NRF-CRP29-2022-0004), MOE Tier 2 grants (MOE-T2EP10221-0005 and MOE-T2EP10123-0004), Agency for Science, Technology and Research (A*STAR) under MTC Individual Research Grants (Project ID M21K2c0113) and Science and Technology Project of Jiangsu Province (grant number BZ2022056). X.W. acknowledges the support from the National Key R&D Program of China (no. 2022ZD0117501), Agency for Science, Technology and Research (A*STAR) Singapore RIE2020 Advanced Manufacturing and Engineering Programmatic Grant A1898b0043, and Tsinghua University Initiative Scientific Research Program. J.S. acknowledges the support from Agency for Science, Technology and Research (A*STAR) Advanced Manufacturing & Engineering (AME) Young Individual Research Grant (YIRG) A2084c0171. C.Z. acknowledges the support from MOE Tier 2 grants (MOE2019-T2-2-030), and computational resources in NUS High Performance Computing (HPC) facilities and National Supercomputing Centre (NSCC) in Singapore.
Author information
Authors and Affiliations
Contributions
C.Z., X.W. and J. Lu supervised the project. J.S., J. Li, X.W. and J. Lu conceptualized this project. J. Wu synthesized and characterized molecular precursor. J.S. and X.P. performed the on-surface experiments and analysis. N.G. and C.Z. performed the DFT calculations. P.L. performed the illustration of conceptual figure. J. Li, J.Y., J. Wang, Z.L., J.S., K.M., J. Lu and X.W. designed and constructed the deep neural networks. All authors contributed towards writing the manuscript.
Corresponding authors
Ethics declarations
Competing interests
The authors declare that they have no competing interests.
Peer review
Peer review information
Nature Synthesis thanks Shiyong Wang and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Peter Seavill, in collaboration with the Nature Synthesis team.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 Additional nc-AFM characterization data for molecule 1 and 2.
Experimental (a) and simulated AFM (b) image of a single molecule 1 on Au(111) surface. Scanning parameters of (a): Setpoint before switching off the feedback: Vs = 0.5 V, It = 400 pA. Scale bar: 4 Å. (c-d) Force spectroscopic measurements taken at different sites over a single molecule 2. The height difference between the cyclized area (lower panel c) and non-cyclized area (upper panel c) is ~150 pm, indicating that the non-cyclized benzylic groups remain intact after cyclizing the adjacent corner. Numbers indicate the positions where the spectra are recorded. Scanning parameters: Vs = 1 mV. Setpoint before switching off the feedback: Vs = 0.5 V, It = 400 pA. Scale bar: 4 Å.
Extended Data Fig. 2 Electronic structure of the molecular products obtained via sequential probe-assisted cyclodehydrogenation.
(a) Spin-polarized DFT calculations of molecular orbital diagrams of freestanding molecule 1 and molecule 3. (b) dI/dV spectra acquired at the edge of molecule 1 and 3 with the reference spectra collected over the bare Au(111) surface. dI/dV spectrum collected at the zigzag edge of molecule 1 shows two pronounced peaks located at −1.25 V (±0.05 V) and +1.46 V (±0.05 V), respectively (Upper curve). Similarly, dI/dV spectrum collected at the edges of molecule 3 after a second cyclization shows two small peaks (Middle curve) located at −0.54 V (±0.05 V) and +1.12 V (±0.05 V). As illustrated in Extended Data Fig. 2a, the frontier orbitals of a neutral molecule 1 contain two degenerate singly occupied molecular orbitals (ψ5↑ and ψ6↑, denoted as SOMOs) at −0.52 eV below EF and two singly unoccupied molecular orbitals (ψ5↓ and ψ6↓), denoted as SUMOs) at +1.47 eV above EF, in line with the experimental energetic position of occupied (−1.25 V) and unoccupied (+1.46 V) electronic states. Similarly, a neutral molecule 3 is predicted to have two pairs of frontier orbitals assigned to the SOMOs (ψ5↑ and ψ6↑, −0.47 eV) and SUMOs (ψ5↓ and ψ6↓, +1.11 eV), consistent with the experimentally recorded electronic states at −0.54 V and +1.12 V, respectively (Extended Data Fig. 2a, right). (c) Point dI/dV spectra acquired over different sites (indicated by coloured circles in the STM images shown on the right) of molecule 1–4 on Au(111) surface. Scale bar: 5 Å.
Extended Data Fig. 3 Additional dI/dV imaging and spectra measurements of molecule 1 and molecule 3.
Constant-current dI/dV maps of the SOMOs and SUMOs of molecule 1 (a-b) and the corresponding STM images (d-e) taken simultaneously. Constant-current dI/dV maps of the SOMOs and SUMOs of molecule 3 (c-d) and the corresponding STM images (f-g) taken simultaneously. (h) dI/dV spectra recorded near the EF. The curve corresponds to the red circle shown in (f). Setpoint: Vs = 0.06 V, It = 1.2 nA. Modulation voltage: 2 mV. Scale bar: 5 Å.
Extended Data Fig. 4 DFT-calculated cyclization of molecule 2 to form molecule 3, and the cyclization of molecule 3 to form molecule 4.
DFT-calculated molecular structures of the cyclization process from molecule 2 to 3 (a–e) and 3 to 4 (g–k). (f) The DFT-calculated energy diagram for the cyclization process of molecule 2 to 3 (f), and molecule 3 to 4 (i).
Extended Data Fig. 5 Schematic representation of the workflow and underlying mechanisms for interpreting the trained model.
(a) Application of a game-theoretic approach to decode the trained models. (b) Construction of feature blocks from STM images employed in model explanation analysis. Details can be referred to Supplementary Information Section 5.
Supplementary information
Supplementary Information
Experimental details and Supplementary Notes Sections 1–5, Figs. 1–36 and Tables 1 and 2.
Supplementary Video 1
Video demonstration of the CARP operation.
Source data
Source Data Fig. 5
Statistical source data.
Source Data Extended Data Fig./Table 1
Statistical source data.
Source Data Extended Data Fig./Table 2
Statistical source data.
Source Data Extended Data Fig./Table 3
Statistical source data.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Su, J., Li, J., Guo, N. et al. Intelligent synthesis of magnetic nanographenes via chemist-intuited atomic robotic probe. Nat. Synth 3, 466–476 (2024). https://doi.org/10.1038/s44160-024-00488-7
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s44160-024-00488-7
This article is cited by
-
Single-molecule chemistry with a smart robot
Nature Synthesis (2024)