On-Surface Synthesis and Collective Spin Excitations of a Triangulene-Based Nanostar.

Triangulene nanographenes are open-shell molecules with predicted high spin state due to the frustration of their conjugated network. Their long-sought synthesis became recently possible over a metal surface. Here, we present a macrocycle formed by six [3]triangulenes, which was obtained by combining the solution synthesis of a dimethylphenyl-anthracene cyclic hexamer and the on-surface cyclodehydrogenation of this precursor over a gold substrate. The resulting triangulene nanostar exhibits a collective spin state generated by the interaction of its 12 unpaired π-electrons along the conjugated lattice, corresponding to the antiferromagnetic ordering of six S=1 sites (one per triangulene unit). Inelastic electron tunneling spectroscopy resolved three spin excitations connecting the singlet ground state with triplet states. The nanostar behaves close to predictions from the Heisenberg model of an S=1 spin ring, representing a unique system to test collective spin modes in cyclic systems.

Open‐shell nanographenes have emerged as privileged testing ground to study carbon‐based magnetism and to explore potential applications of organic molecules in spintronics and quantum technologies.[ , ] However, the atomically precise preparation of such nanostructures is challenging due to the high reactivity of these graphene derivatives with radical character under ambient conditions. In this sense, on‐surface synthesis under ultra‐high vacuum (UHV) has become an interesting alternative to fabricate and study the potential functionality of open‐shell carbon nanostructures.[ , ] A paramount example is the preparation of [3]triangulene (Figure 1 a), a non‐Kekulé triangular‐shaped polycyclic aromatic hydrocarbon with two unpaired π electrons, which was finally generated and visualized on a surface by means of scanning probe microscopy (SPM), after previous unsuccessful attempts by solution chemistry methods.[ , ] The particular diradical character of [3]triangulene with S=1, makes it a perfect building block for the design of open‐shell carbon‐based nanostructures with different ground state spin quantum numbers. In fact, during the last three years, the development of triangulene‐based nanographenes has been impressive; not only larger triangulenes, such as [4]‐, [5]‐ and [7]‐triangulene,[ , , ] but other appealing structures such as a [7]triangulene‐ring, an extended version of [3]triangulene, or [3]triangulene dimers were prepared by on‐surface synthesis. Even the iconic Clar's goblet, which is actually formed by two fused [3]triangulenes sharing one benzene ring, was generated and characterized on‐surface. Interestingly, many of these triangulene‐based nanostructures exhibit a characteristic low‐energy spectral fingerprint associated with spin excitations induced by inelastic tunneling electrons, which was not observed in isolated triangulene units. Such inelastic features provide unique access to quantify spin interactions in nanographenes and to test current theoretical models for describing collective spin states.

Figure 1

a) Previously reported triangulene derivatives. b) Synthesis of triangulene‐based nanostar (TNS) developed in this work, by combining in‐solution and on‐surface chemistry.

In this work, we present a new triangulene derivative consisting in six [3]triangulene moieties arranged in a macrocycle, forming a triangulene‐based nanostar (TNS) with a pore size of 1.2 nm (Figure 1 b). Besides the captivating structure of this carbon‐based nanosized molecular star, we were particularly intrigued by the spin configuration adopted by the 12 interacting unpaired electrons in the cyclic hexamer over a metal substrate.

We envisioned the preparation of TNS by the on‐surface cyclodehydrogenation (CDH) of macrocycle 1 (Figure 1 b), in particular by the efficient formation of 12 C−C bonds (in blue) and 12 π radicals in one annealing step promoted by Au(111) under UHV conditions. We selected 2,7‐dibromo‐10‐(2,6‐dimethylphenyl)‐anthracene (2) as triangulene building block with the appropriate bromine disposition for the generation of the cyclic hexamer 1 by surface‐assisted Ullmann‐coupling. Compound 2 was obtained in one step by reaction of dibromoanthrone 3 with (2,6‐dimethylphenyl)lithium, followed by acid‐promoted dehydration (see supporting information (SI) for details). However, in our hands, the on‐surface polymerization of compound 2 resulted in a wide variety of triangulene‐based oligomers, as it has been independently reported very recently. Therefore, inspired by the synthesis of “nano‐saturns” by Toyota and co‐workers, we decided to attempt the preparation of macrocycle 1 by means of in‐solution coupling of dibromoanthracene 2. In fact, Ni‐catalyzed Yamamoto coupling of compound 2 resulted in a mixture of soluble and insoluble oligomers, which, after purification by filtration and washing procedures, led to the isolation of TNS precursor 1 in 10 % yield.

Despite the large molecular mass of macrocycle 1 (C132H96, 1680 Da), its vacuum deposition was possible by flash‐annealing a silicon wafer loaded with grains of this compound. Figure 2 a shows an overview STM image of the Au(111) substrate after deposition. Molecular rings surviving thermal sublimation appear in closed‐packed molecular islands, surrounded by polymeric fragments, probably from rings broken during the thermal sublimation (for the coverage of Figure 2 a, about 8 % of the material deposited corresponds to intact precursor 1). Individual molecules 1, extracted from the domains by lateral manipulation with the STM tip, appear composed of six protrusions corresponding to the six dimethylphenyl groups attached to the anthracenes moieties (Figure 2 b).

Figure 2

On‐surface synthesis of triangulene‐based nanostar (TNS). a,b) STM images of precursor 1 as deposited on the Au(111) substrate. Molecular rings appear assembled in packed domains. The isolated molecule 1 in (b) was extracted using the STM tip, to prove the covalent nature of the macrocycle. c) STM image of the surface after annealing. TNS and fragments appear mostly planarized. d) STM image of a TNS formed after thermal annealing. e) BR‐STM image of a TNS. (a: V=0.5 V, I=60 pA; b: V=−0.2 V, I=40 pA; c: V=1 V, I=40 pA; d: V=1 V, I=100 pA; e: V=5 mV). Data analysis using WSxM.

To induce the full planarization of the macrocycle 1 into the targeted TNS, we annealed the substrate to 330 °C. At this temperature, the CDH reaction leading to the formation of π radicals takes place with the removal of 3 H atoms per methyl side group and ring closing through C−C bond formation. However, due to the high temperatures required to fully planarize the whole rings and to their radical character, the CDH competes with additional C−C coupling and bond‐breaking reactions, resulting in a few surviving isolated TNS (about 20 %) and many triangulene oligomers around (Figure 2 c and d). Bond‐resolved scanning tunneling microscopy (BR‐STM) images of molecular rings like in Figure 2 d, obtained using a CO functionalized STM tip, resolved the triangulene units connected in a closed ring structure (Figure 2 e), in agreement with the expected structure of TNS (Figure 1 b). The planarization reaction involves the formation of twelve six‐membered rings, each one at the position of the methyl groups of precursor 1. The resulting structure accommodates two unpaired π electrons in each triangulene unit, which are expected to interact throughout the conjugated π‐lattice and stabilize a collective spin state.

A first inspection of the magnetic interactions in the ring obtained by mean‐field Hubbard (MFH) simulations (Figure 3 a, see methods in the SI for details) reveals that intra‐triangulene parallel magnetic coupling remains strong in the TNS, building a total spin S=1 in each triangulene unit, and dominating over a weaker anti‐parallel coupling between triangulene sub‐units.[ , ] Such interaction pattern is expected to lead to a global spin singlet ground state in the TNS. Interestingly, antiferromagnetic (AFM) rings of integer spins have been widely studied for demonstrating the Haldane conjecture,[ , ] stating that the spin excitation spectrum of an infinite Heisenberg S=1 AFM chain remains gapped, with an energy value close to 0.4 J (J being the Heisenberg exchange constant). Such a gapped spectrum is an anomalous behavior of integer AFM spin chains, which contrasts with the gapless spin excitation of similar chains made of half‐integer spins.

Figure 3

Collective spin excitations in a triangulene‐based nanostar. a) Spin density map of a TNS obtained from mean‐field Hubbard simulations (on‐site Coulomb potential U=3.5 eV). b) BR‐STM image indicating the TNS site where spectra in (c) and line spectra in (d) were taken. c) dI/dV spectrum on a triangulene unit of a TNS, showing inelastic steps at E 1=±14 mV, E 2=±42 mV and E 3=±80 mV. The red line shows the simulated spectral function from the Heisenberg model with exchange constant J=18 meV. d) dI/dV spectral line measured across the single triangulene unit indicated in (b). The excitation steps appear around the borders of the TNS. Set‐point: V=100 mV, I=0.1 nA; individual spectra were offsetted to appear with the same zero‐bias conductance. e) Model of a six‐membered ring of spin‐1, simulated using the Spin Hamiltonian shown in the panel. f) Excitations of the collective spin states obtained from a Heisenberg model with J=18 meV (only states up to ST=3 are shown). Thin/thick lines indicate singly/doubly degenerate total spin states. Arrows mark the three dominant transitions induced by tunneling electrons (see SI), which coincide in energy and weight with the three steps in the experimental plot in (c).

To experimentally determine the spin configuration of the TNS, we probed their excitation spectrum by Inelastic Electron Tunneling Spectroscopy (IETS).[ , ] Specifically, we measured differential conductance (dI/dV) vs. bias spectra over the sides of triangulene units of a TNS and found bias‐symmetric stepped features (Figure 3 c), which are a characteristic fingerprint of spin excitation induced by inelastic electron tunneling.[ , , ] All triangulene units show spectra with the same shape (see additional data in SI), composed of three steps of different height at E 1=±14(1) mV, E 2= ±42(2) mV and E 3=±80(2) mV. The first two dI/dV steps have similar height (each amounting to a relative increase of conductance of ≈35 %), while the third one appears with small spectral weight (<7 %). As we show in the stacked dI/dV plot of Figure 3 d, the inelastic signal appears localized over the edges of the triangulene units. Such localization pattern is clearly pictured by mapping the inelastic signal d2 I/dV 2 over the whole TNS (shown in SI), confirming that the excitation probability of collective spin states coincides with the larger spin density over edge sites expected for π‐radical states.

The presence of three spin excitation steps in the dI/dV spectra is a remarkable feature that contrasts with the single‐stepped spectra measured in triangulene dimers and other graphene nanostructures.[ , , , ] Due to the ΔS=0, ±1 selection rule imposed by conservation of angular momentum, single IETS steps were attributed to singlet‐triplet transitions induced by the tunneling electrons. The multi‐step spectrum found for the TNS unveils a complex excitation pattern of collective spin states in the triangulene ring, accessible by inelastic tunneling electrons.

To unravel the origin of the inelastic steps, we modeled the coupled spin system with an isotropic Heisenberg Hamiltonian describing the AFM exchange interaction of six S=1 spins in a closed ring, as in Figure 3 e (theoretical details in SI). The many‐body ground state of the TNS, obtained by numerically exact diagonalization of the spin Hamiltonian (see methods in SI), has a total spin ST=0 (singlet), and consists of the superposition of the classical solutions of antiferromagnetically coupled triangulene segments (i.e. ground state from the MFH model, in Figure 3 a), and other combinations of the triangulene spins with global spin ST=0. Similarly, the set of excited states of the TNS are collective spin modes with total spin amounting from ST=0 to ST=6, which can be described as a superposition of triangulene S=1 multiplet states.

To explain the spin excitation spectrum, we assume that steps correspond to transitions with ΔS=±1 and adjust the Heisenberg exchange constant J to fit the lowest allowed excitation with the first dI/dV step (centered at ±14 mV). The best spectral match occurs using J=18 meV. This value is larger than the exchange constant derived from the Heisenberg model in triangulene dimers. Figure 3 f shows the computed excitations with energy up to 100 meV, classified according to their global spin (up to ST=3). Several singlet and triplet excited states appear within the energy window of the IETS spectrum, including triplet states grouped around excitation energy values (13 meV, 45 meV and 88 meV) that agree with the three dI/dV steps in our experimental spectra. However, not all transitions are equally accessible by tunneling electrons. To account for the cross‐section of each spin transition and explain their different dI/dV step height, we calculated the correlation of collective TNS spin states with tunneling electrons following the expression by Ternes. This quantity reflects the scattering probability of a tunnel electron with every collective spin eigenstate and, thus, it is related to the inelastic signal and, hence, to the step height. The first excitation at 13 meV has the larger spectral weight (≈60 %, as shown by arrows in Figure 3 f), while several states bundled at around E2=45 meV contribute to most of the remaining weight (39 %). Higher states are seemingly weaker in terms of spectral weight and produce a faint step in the spectrum. A simulated spin excitation spectrum shown in Figure 3 c (red plot) reproduces with good agreement the position and relative intensity of the three steps observed in the experimental plot. There are, however, small differences in step position and height that can be accounted for by including a more precise description of spin interactions. For example, including a finite intra‐triangulene coupling can correctly reproduce the equal height of first and second steps (see analysis for twelve 1/2 spins in the SI). Other effects such as a renormalization of spin interactions in the TNS induced by the metal substrate, or higher order correction to the Heisenberg coupling between spins,[ , ] can be additionally incorporated into the model to quantitatively reproduce the experimental spectra.

In summary, we have demonstrated the successful fabrication of a macrocycle formed by the covalent attachment of six [3]triangulenes, through a combination of solution synthesis coupling and on‐surface planarization. The design‐driven strategy of precursor molecules is thus a possible method to produce monodispersed ensembles of complex open‐shell nanostructures. For the presented case, the triangulene cyclic hexamer accumulated 12 unpaired electrons. Tunnel spectroscopy revealed that the unique spin states expected for such interacting systems survive on a metal surface. We detected three collective spin excitations from the many‐body singlet ground state to triplet states of the spin ring. These findings proved that the spin states of the nanostar are compatible with an antiferromagnetic S=1 spin ring.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Click here for additional data file.