On‐Surface Synthesis of Chiral Graphene Nanoribbon Segments via the Quarter‐Anthryl on Au(111) Surface

Quarteranthene is predicted to manifest nontrivial edge states as the competition from the hybridization of localized frontier states and the Coulomb repulsion between valence electrons, which hosts much research potential in memory, spintronic devices fabrication, and quantum computation. The fabrication of ribbon‐like structures with up‐mentioned edge states, such as chiral graphene nanoribbon, possesses a high significance in the topological phase transition investigation. However, the synthesis of chiral graphene nanoribbon is limited in reactive substrate or with the edge‐brominated precursor. Here, the fabrication of quarteranthene and its chiral graphene nanoribbon segment on Au(111) substrate is reported. A combined bond‐resolved scanning tunneling microscopy, noncontact atomic force microscopy characterization, and corresponding density functional theory calculations confirm the chemical structure of fabricated products on the Au(111) substrate. The detailed analysis of the laterally extended products reveals that the lateral‐extended structures are acquired via the linkage of the hydrocarbon quarter‐anthryl. In addition, several strategies are used to modulate the yield of quarteranthene and its lateral‐extended products. These findings provide a new insight into the lateral extension strategy on metal substrates and provide another possible fabrication strategy of chiral graphene nanoribbons on the relative inert Au(111) substrate.

the H atom migration from the 2,2′-positions to the 10,10′positions, which could form the (3, 1)-chGNRs after the further heat treatment. [21] A similar behavior appears in 9,9′-bianthryl (BA) on a reactive Cu(111) substrate. [32] This precursor BA mainly involves two steps in the recent report: 1) the thermally induced cleavage of C-H bonds on the 10 and 10' positions produces the biradical products, which is similar to the debromination of the DBBA precursor; 2) hydrogen atom transfer [33] step from the 2, 2′-positions to the 10, 10′-positions occurs, leading to the formation of the 2, 2'-biradicals. The linkage between the radicals and the afterward annealing treatment forms the (3,1)-chGNR. Last, the fabrication of chGNRs could be directly realized by the lateral brominated precursors. [2][3][4][5]14] However, the precursors lacking halogenated atoms have seldom been reported to possess regular lateral extension behavior on an inert Au(111) surface.
Here, we report the on-surface synthesis and characterization of the quarteranthene along with its lateral extended products on an Au(111) substrate. From our observation of the lateral fusion behavior, we propose the extension strategy that is similar to the result of the chGNRs fabricated on Cu(111), and we thus engineer the productivity by our lateral extension analysis. We choose 10-bromo-9,9′-bianthracene (BBA) [34] as a precursor. As illustrated in Scheme 1, the BBA precursor undergoes Ullman coupling forming the quarter-anthryl (dimer). The heat treatment afterward induces the Scholl or cyclodehydrogenation reaction on the Au(111) surface forming the open-shell quarteranthene (1U).

Synthesis and Characterization of the Quarteranthene
After the deposition of the BBA precursor on the Au(111) surface held at 180 °C followed by the subsequent annealing at 280 °C (heating rate : 6 °C min −1 ), a relative low coverage of the 1U (about 14.8% shown in Figure S1a, Supporting Information) is fabricated. The overall STM image in Figure 1a depicts the individual covalently bonding oligomers. The highresolution STM image (HR-STM) in Figure 1b is acquired in the yellow rectangle area of Figure 1a, revealing a three-lobelike local density of states (LDOSs) appearance at both terminals of 1U. According to the previous reports, [17][18][19]35,36] these three-lobe-like LDOSs at the end are the fingerprint of the singe occupied electron. As illustrated in Figure 1c, the constant height bond-resolved scanning tunneling microscope (BR-STM) image of 1U shows a visible ring feature, which fits well with the chemical structure of the 1U. An experimental scanning tunneling spectroscopy (STS) measurement is conducted to reveal the electronic structure of the 1U. As shown in Figure 1d, dI/dV spectra of the 1U with the acquired positions marked in Figure 1b, shed light on the three obvious resonances (-1350 mV, 1700 mV, and near the Fermi level). Density functional theory (DFT) calculated energies for the free-standing 1U (closed-shell, antiferromagnetic, and ferromagnetic) are displayed in Figure S2, Supporting Information, and the experimental open-shell evidence by the additional H atom passivation is shown in Figure S3, Supporting Information. The results show that the AFM ground state is energy preferable, which is in conflict with the observed additional states distribution. [37] Further, we performed the low bias range dI/dV spectra measurement to figure out the resonances near the Fermi level, which is displayed in Figure S4, Supporting Information. The spectra reveal the presence of two resonances near Fermi level and the dependence of points acquire positions. Elaboration in detail is that the spectrum taken at the bottom position depicts an obvious resonance at the Fermi level while the upper one shows two resonances at the Fermi level and at about 30 mV, which corresponds to the previous reported end-state. [20] The presence of the resonance at the Fermi level reminds us to take the charge transfer into consideration. We performed the deformation charge density of the optimized 1U on the Au(111) ( Figure S5a,b, Supporting Information), and the result shows the existence of charge transfer behavior between the 1U and the underneath substrate (more than 1 e − from the 1U to substrate). Corresponding simulated energy level after the charging ( Figure S5c, Supporting Information) reveals that only one singly occupied electron is left leading to a net spin of S = 1/2, which indicates the observed resonances at the Fermi level resulting from the spin screen (Kondo resonances). Furthermore, the calculated spin density distribution of the charge transferred 1U locates around both termini which fits well with the additional states in the BR-STM image ( Figure 1c) and previous report. [37] Our simulated LDOS in Figure 1f shows the state distribution of HOMO-1 (HOMO represents the highest occupied molecular orbital) and LUMO+1 (LUMO represents the lowest Scheme 1. Reaction pathway toward the 1U structure on Au(111) surface. unoccupied molecular orbital), which are coincident with the two dI/dV maps at −1350 and 1700 mV, respectively.
Except for the resonance near the Fermi level, there are no visible resonances between the HOMO-1 orbital and the LUMO+1 orbital. We then speculate that the frontier orbitals of 1U on the Au(111) substrate are merged together. To ascertain the frontier orbital and the spatial electron distributions, we obtained several dI/dV maps with the bias voltage set around the Fermi level to fit with the calculated LDOS maps of the frontier orbitals (Figure 1f). [18] The results in Figure 1e show that two spatial electron distributions with the bias voltage located at 60 and −40 mV display the node-like appearance with slight differences, which may correspond to our calculated LDOS of the single occupied molecular orbitals (SOMOs) and single unoccupied molecular orbitals (SUMOs). This leads to a frontier gap of 100 mV for the NG 1U.
The statistics result of the productivity shows that the yield of 1U is relatively low (e.g., 13.8% in Figure 1a). Comparing with the fabricated 1U structure, by-product 1 along with its chiral by-product 1′ hosts a main proportion (e.g., 41% in Figure S1a, Supporting Information, and Figure 1a). In a deeper analysis of these products and roll-backing to their C-C coupling process, we speculate the existence of the radical/hydrogen migration progress. After the debromination, the radical in the precursor could migrate to other energy comparable sites, which leads to the formation of these by-products by the afterward linkage and the cyclodehydrogenation. [21,33] We thus performed the calculations toward the possible radical-located structures ( Figure S6, Supporting Information). and the results show that there are two additional situations (5-radical, 4-radical) whose energy is comparable to the 10-radical. The three situations induce the additional linkage pathways, which could fabricate the structure of 1U and by-products under further heat treatment.

Characterization and Analysis of the Laterally Fused Structure
Apart from the 1U and its by-products, we observed plenty of regular (3,2,8)-chiral graphene nanoribbon (chGNRs) segments simultaneously, such as the 2U, 3U, and 4U structures lateral fused by corresponding 1U (Figures 2 and 4; Figure S7, Supporting Information). Figure 2a shows occasionally acquired STM image (the same sample with Figure 1a) that includes the 2U structure. The inset figure in Figure 2a depicts an HR-STM image obtained with a CO functionalized-tip revealing the symmetric node-like LDOS feature. To make sure of the chemical structure of 2U, we performed the BR-STM (Figure 2b). The result shows that a bright feature appears at two termini of the 2U structure. In order to exclude that this bright feature is caused by distorted adsorption configuration, we further performed the non-contact atomic force microscope (NC-AFM) characterization as shown in Figure 2c. The force signal depicts the relatively flat adoption configuration which means that the bright states at the termini attribute to the non-trivial states distribution instead of the structural distortion. The STS characterization in Figure 2d shows abundant resonance signals including an obvious resonance near the Fermi level, which leads to the unusual state-distribution in the BR-STM. Except for this, there are several resonances in the STS curves, such as −1100, −300, 510, and 1100 mV. The experimental dI/dV maps of these four energies fit well with the calculated LDOS maps of the HOMO-2, HOMO-1, LUMO+1, and LUMO+2 displayed in Figure 2e,f. Detailed experimental dI/dV maps near the Fermi level are performed as shown in Figure S8, Supporting Information, which displays the coincident state distribution with the calculated results from the SOMOs and SUMOs ( Figure  S9, Supporting Information, and Figure 2f). The results reveal that the frontier gap is about 150 mV of the lateral fused 2U structure. The appearance of additional states at two termini ( Figure 2b) and the positive shifted peak with high amplitude (Figure 2d) may result from the previous reported topological boundary state. [14] The configurations of this uniform lateral fused structures and a relatively low fabrication temperature (280 °C) confused us a lot. To the best of our knowledge, the lateral fusion in AGNRs on Au(111) requires a high annealing temperature and the obtained structures are uncontrollable, such as the formation of 14-, 21-AGNR, [28] and 6-AGNR. [29] The incomplete cyclization product in Figure 2a (left-up panel) confirms that the lateral fused structures can be fabricated at a low temperature (verified by height-dependent STM images in Figure S10, Supporting Information). In combination with the observed chGNR-like structures in Figure S7, Supporting Information (4U), we recognized that the lateral linkage strategy of 2U structure on the Au(111) may be similar to that of chGNRs on the Cu(111) substrate. Systematically, we propose two possible reaction routes: 1) Similar to the previous reports about the wider GNRs on Au(111), [24] the lateral fusion reaction takes place after the complete cyclodehydrogenation into the 1U (Figure 3, route 2); 2) similar to previous reports of fabricating chGNRs on Cu(111) by the precursor BA, [38] lateral C-C coupling happens before the cyclodehydrogenation (Figure 3, route 3). Considering the uniform lateral fused 2U products, we launched the consideration of two segments linked with the same positions. For a further description, the possibility of lateral fusion of two 1U into 2U structure is 3″-2′, 2″-7, 6′-6, 7″′-3′, and 6″′-7″ ( Figure S11, Supporting Information). Obviously, except for the linkage between 3″ and 2′, other linkage pairs will induce additional lateral structures, which are in conflict with our observations. To verify the lateral fusion strategy by 3″-2′ linkage, we calculated energy differences of several  Table 1 and Figure S12, Supporting Information. The linkage product between 3″ and 2′ in route 2 is proved to be energy nonpreferable, as the energy is higher than the linkage with zigzag terminal carbons (870 meV). This is comprehensible as the existence of radicals at two termini which could direct linkage into a longer one [39] (upper panel of route 1, Figure 3). In another way, the linkage at the lateral positions might involve the radical migration, which requires additional energy. On the contrary, the 3″-2′ linkage products are the most energy preferable in route 3. Furthermore, the lateral linkage in route 3 involves the C-H activation, and our calculated energy difference results show that the internal and lateral C-H cleavage are energy comparable ( Figure S14, Supporting Information), which favors the lateral linkage on Au(111). To testify this, we performed the annealing progress under a lower temperature (265 °C), as shown in Figure S13, Supporting Information. The distinguishable lateral fused structure with a partial cyclodehydrogenation demonstrated a supporting evidence toward route 3, even though we cannot exclude the possibility of the lateral linkage starting with the partial Scholl reaction ( Figure S14, Supporting Information).

The Productivity Control toward the Quarteranthene and the Fused Structure
As discussed in Figure S1, Supporting Information, the productivity of 1U is relatively low, which hinders the lateral extended products (occasionally found, such as in Figure 2a). Basing on our analysis, we further try to control the yield of 1U and 2U by adjusting the surface temperature during the molecular deposition and annealing rate in the second step. We first try to change the deposit strategy. We deposited the precursor BBA Figure 3. Two possible extension pathways from the dimer. In route 1, the quarter-anthryl first undergoes the cyclodehydrogenation into the structure 1U. After that, the 1U links by the terminal radicals forming the terminal extended 2U′ under a higher annealing temperature. In route 3, the quarteranthryl undergoes the C-H cleavage, the lateral linkage, and the cyclodehydrogenation forming the observed lateral extended 2U. onto the Au(111) held at room temperature (RT). The obtained self-assembly structures and the overlaid chemical structures are shown in Figure S15, Supporting Information. After that, we adopt the two-step annealing method [26] with the same annealing rate and similar sample coverage with the upper result ( Figure 1a). The statistics results in Figure 4b show that this gradual heat treatment experiment displays two times the productivity (28.2%), higher than in Figure 1a. The higher productivity might involve undergoing the self-assembly step and molecular diffusion. Previous results [40,41] reported that the debromination reaction start at about 100 °C on Au(111) and the heat treatment afterward promotes diffusion on the surface and C-C coupling. Considering several energy comparable products, the heat treatment enables the migration of the radicals, which promotes the formation of by-products. The assembly structure coordinated by Br-H and Br-Br bonds ( Figure S15, Supporting Information) facilitates direct C-C coupling after debromination. Based on this sample deposition strategy, we tried another controlling parameter toward the 1U and 2U structure with different annealing rates (6 °C min −1 for Figure 4c; 10 C min −1 for Figure 4d) in the second heat treatment. The total productivity including 1U and 2U of these two samples are almost the same (28% for Figure 4c; 29.6% for Figure 4d), while the higher annealing rate in Figure 4d leads to a higher productively of the 1U structure and the lower productively of the 2U structure. We attribute these phenomena to the competitive relationship of the two structures. The faster annealing rates will inhibit the diffusion step, which decreases the possibility of the lateral fusion. Additionally, by combining the results in Figure 4b,c ( Figure S16, Supporting Information), we find the sample coverage (12.5% for Figure 4b, 19.5% for Figure 4c) do have an influence on the productivity. The higher coverage in Figure 4c leads to a higher productivity of the 2U and a lower productivity of the 1U, which may result from the more contact possibility of the 1U structure, which leads to the existence of larger lateral and longer terminal extended structures. The observed 3U and longer segments host the same reaction pathway with the 2U structure indicating a proportional relationship to 2U and highly related sample preparation parameters (Figure 4b-d).

Conclusions
In conclusion, we have demonstrated the on-surface synthesis of atomically precise quarteranthene and the uniform chirallike lateral-extended products on Au(111) surface. The lateral extended products observed at a relatively low temperature involve a similar extension strategy to form the chGNRs on the Cu(111) substrate. The lateral fusion strategy to fabricate the chGNRs of the anthryls, such as bianthryl, has been reported on a relative reactive Cu(111) surface while no reports on the inert Au(111) surface. The successful fabrication of chiral like GNRs on Au(111) surface via the longer precursor extends the chiral GNRs' fabrication toolkits. Based on the guidance of reaction pathway, we succeed in engineering the yield of the structures 1U and 2U by sample deposition strategy, annealing rate, and sample coverage.

Experimental Section
Sample Preparation and STM/STS Measurements: All the STM measurements were performed with a commercial low-temperature LT-STM from Scienta Omicron operating at a temperature of 3.9 K and base pressure below 2.1 × 10 −11 mbar. Au(111) single crystal surface (MaTeck) was prepared via standard cycles of sputtering with Ar + ions at a pressure of p = 1.6 × 10 −6 mbar, followed by annealing to 750 K for 20 min. Precursor BBA was contained in a quartz crucible and deposited at 358 K from a commercial four-cell evaporator on the Au(111) surface held at 453 K. After deposition, the sample was held at 453 K for 10 min. To form the NG, the sample was then annealed to 553 K for 20 min. For the gradual heat treatment experiment, all the samples were deposited on the Au(111) substrate held at room temperature (10 min) and fabricated by a two-step heat treatment (453 K, 10 min; 553 K, 20 min). Additional parameters including the coverage and the annealing rate are marked in the main-text. STM images and dI/dV maps were acquired in constant-current mode with the bias applies to the sample. Gold-coated tungsten tip was used for STM imaging and spectroscopy measurements were fabricated by indenting tungsten tip to Au(111) substrate. dI/dV spectra and maps were obtained with a lock-in amplifier (HF2Li PLL by Zurich Instruments) operating at a frequency of 601 Hz. Modulation voltages (root mean square amplitude, V rms ) for individual STS and maps measurements are provided in the respective figure captions. NC-AFM images were recorded with a CO-functionalized tip attached to a quartz tuning fork sensor [42] (the resonance frequency was 28235 Hz). CO decorated tip [43] was fabricated as in the reference. The data were processed with WSxM software. [44] DFT Calculations: Spin-polarized DFT calculations were performed using the Vienna ab-initio simulation package (VASP). [45,46] A generalized gradient approximation in the form of Perdew-Burke-Ernzerhof [47] was used for the exchange and correlation functional. H atoms were included to saturate dangling bonds. The electronic wavefunctions were expanded in a plane wave basis with an energy cutoff of 600 eV. For the unit cell, Brillouin zone sampling was done using a 1 × 1 × 1 grid Gamma point only for all calculations. In all the calculations, 15 Å vacuum layers in all three directions were used, and all atoms were fully relaxed until the residual forces on each atom were smaller than 0.01 eV Å −1 .
To consider the charge transfer, Au(111) surface was modeled by a two-layer slab separated by vacuum in the surface normal direction, and the topmost one layer and molecular were allowed to relax during geometry optimizations. In order to visualize the charge transfer, the charge density difference of the system was plotted. The electron density difference (Δρ) between the combined system (ρ 1U/Au ) and the sum of its separated constituents (calculated as freestanding species in the frozen geometry) is defined as Δρ = ρ 1u/Au − (ρ 1u + ρ Au ). The number of transferred charges was calculated by integrating Δρ from the bottom layer to vacuum. LDOS maps were simulated with an s-wave tip model with the spin open. The models and spin densities and the orbital were plotted with the Visualization for Electronic and Structural Analysis (VESTA) program [48] and the VASPMO.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.