Solution Synthesis and Characterization of a Long and Curved Graphene Nanoribbon with Hybrid Cove–Armchair–Gulf Edge Structures

Abstract

Curved graphene nanoribbons (GNRs) with hybrid edge structures have recently attracted increasing attention due to their unique band structures and electronic properties as a result of their nonplanar conformation. This work reports the solution synthesis of a long and curved multi‐edged GNR (cMGNR) with unprecedented cove–armchair–gulf edge structures. The synthesis involves an efficient A₂B₂‐type Diels–Alder polymerization between a diethynyl‐substituted prefused bichrysene monomer (3b) and a dicyclopenta[e,l]pyrene‐5,11‐dione derivative (6) followed by FeCl₃‐mediated Scholl oxidative cyclodehydrogenation of the obtained polyarylenes (P1). Model compounds 1a and 1b are first synthesized to examine the suitability and efficiency of the corresponding polymers for the Scholl reaction. The successful formation of cMGNR from polymer P1 bearing prefused bichrysene units is confirmed by FTIR, Raman, and solid‐state NMR analyses. The cove‐edge structure of the cMGNR imparts the ribbon with a unique nonplanar conformation as revealed by density functional theory (DFT) simulation, which effectively enhances its dispersibility in solution. The cMGNR has a narrow optical bandgap of 1.61 eV, as estimated from the UV–vis absorption spectrum, which is among the family of low‐bandgap solution‐synthesized GNRs. Moreover, the cMGNR exhibits a carrier mobility of ≈2 cm² V⁻¹ s⁻¹ inferred from contact‐free terahertz spectroscopy.

A curved multi‐edged graphene nanoribbon (cMGNR) with an unprecedented cove–armchair–gulf edge topology is synthesized through the efficient A₂B₂‐type Diels–Alder polymerization followed by the Scholl reaction. The resultant cMGNR possesses an average length of sub‐hundred nanometer and a narrow optical bandgap, making it a promising candidate for the fabrication of single GNR‐based nanoelectronic devices.

1. Introduction

Graphene nanoribbons (GNRs) have attracted enormous interest in recent decades due to their potential applications in transistors, photovoltaics and quantum electronic devices.^{[ 1 ]} The current synthetic methods for GNRs consist of top‐down and bottom‐up approaches.^{[ 2 ]} In contrast to top‐down methods by cutting graphene or unzipping carbon nanotubes, bottom‐up organic synthesis represents an efficient approach to constructing GNRs with atomic precision.^{[ 3 ]} To date, the representative bottom‐up synthetic route of GNRs includes the solution‐phase or surface‐assisted polymerization of tailor‐made molecular precursors to form polyarylene polymers followed by the subsequent intramolecular cyclodehydrogenation (or Scholl reaction).^{[ 3 ,‐ 4 ]} In particular, the solution‐synthesis protocol has enabled the preparation of large‐scale GNRs with control over the width, length, and edge structure. In addition, the rational geometry control of GNRs by edge topology engineering has recently received considerable attention owing to its high potential for optoelectronic property tuning.^{[ 5 ]} For example, the introduction of cove‐ or fjord‐type edges on the periphery of GNRs significantly alters their topological conformations and electronic structures. Very recently, we demonstrated that GNRs bearing cove edges exhibit nonplanar geometry with a low bandgap and high charge carrier mobility.^{[ 5d,e ]} Moreover, highly twisted GNRs containing fjord edges were recently synthesized, which represent potential candidates as chiral optical materials and chiral catalysts.^{[ 5c,f ]} Despite substantial progress toward the synthesis of nonplanar GNRs, strategies for the preparation of long and curved GNRs remain elusive, which hamper their further integration into single GNR‐based devices.^{[ 6 ]}

Early in 2015, we attempted to synthesize long and fully cove‐edged GNRs in solution from dibromo‐substituted bichrysene monomers through an Ullmann coupling reaction (Figure  1a),^{[ 7 ]} but only short oligomers were obtained because the steric hindrance at the bay positions of bichrysene impedes efficient polymerization in solution. In contrast, the Diels–Alder polymerization protocol has enabled the preparation of polyphenylene precursor polymers with high molecular weight, thereby producing longer GNRs,^{[ 8 ]} with lengths surpassing these of synthesized by other polymerization methods, such as Suzuki–Miyaura, Ullmann, and Yamamoto polymerization.^{[ 3a,d ]} For instance, we have realized the synthesis of exceptionally long GNRs (>200 nm) via AB‐type Diels–Alder polymerization from a bifunctional ethynyl‐substituted cyclopentadienone monomer (Figure 1b).^{[ 9 ]} However, AB‐type Diels–Alder polymerization for the synthesis of long GNRs is thus far limited to the above ethynyl‐substituted cyclopentadienone monomer, endowing GNRs with only a gulf‐edge topology and a rather planar conformation.^{[ 9} , ^{10 ]} Under this scenario, it is desirable to design new precursors for the synthesis of long GNRs with unique edge topologies and geometries based on the Diels–Alder polymerization approach.

Figure 1.

a) Attempts to synthesize fully cove‐edged graphene nanoribbons (GNRs) in solution. b) AB‐type Diels–Alder polymerization to realize exceptionally long gulf‐edged GNRs in solution. c) Synthetic route of long and curved GNRs based on A₂B₂‐type Diels–Alder polymerization. The functional groups in GNRs are omitted for clarity.

In this work, we demonstrated the successful synthesis of a novel, curved multi‐edged GNR (cMGNR) with unprecedented hybrid cove–armchair–gulf edges via the A₂B₂‐type Diels–Alder polymerization from the newly designed bisdienophile monomer based on a diethynyl‐substituted prefused bichrysene (3b) and functionalized dicyclopenta[e,l]pyrene‐5,11‐dione monomer (6), followed by the Scholl reaction (Figure 1c). As the “cutouts” from the cMGNR, compounds 1a and 1b were synthesized as model compounds to prove the feasibility of the Scholl reaction of bichrysene‐based precursor. Density functional theory (DFT) simulations reveal that the cMGNR adapts a unique curved conformation mostly due to the presence of abundant cove‐edge structures, and its lateral width lies within the range of 0.71‐1.07 nm (Figure 1c). The resultant cMGNR has an average length of ≈71 nm, as estimated by GPC analysis of the corresponding precursor polymer (P1), which is obviously longer than other reported curved GNRs.^{[ 5} , ⁷ , ^{11 ]} The chemical structure and optical properties of cMGNR were characterized by means of solid‐state NMR, FTIR, Raman, and UV–vis spectroscopic analyses. An optical bandgap of 1.61 eV is estimated for cMGNR, which is in line with the calculated value (1.56 eV). Moreover, time‐resolved terahertz spectroscopy reveals a charge‐carrier mobility of ≈2 cm² V^–1 s^–1 in the cMGNR. Our study presents an efficient synthesis strategy to achieve a low‐bandgap, long, and curved GNR with unique hybrid edge structures, providing a promising candidate for the fabrication of single GNR‐based nanoelectronic devices.

2. Results and Discussion

2.1. Synthesis and Characterization of the Model Compounds

To estimate the efficiency of the Scholl reaction for such particular geometry of precursor polymer for the targeted cMGNR, model compound 1a was first designed and synthesized from bichrysene‐based oligophenylene 5a. As depicted in Scheme  1 , the synthesis of 11,11“‐dibromo‐5,5”‐bichrysene (2a) was first performed according to our previously reported procedure.^{[ 7 ]} Then, the Sonogashira coupling of 2a with the commercially available triisopropylsilylacetylene followed by treatment with tetrabutylammonium fluoride provided 11,11“‐diethynyl‐5,5”‐bichrysene (3a) with a yield of 76% in two steps. Moreover, 2,7‐di‐tert‐butyl‐9,11‐bis(4‐(tert‐butyl)phenyl)‐10H‐cyclopenta[e]pyren‐10‐one (4) was prepared. Afterward, the key precursor 5a containing the bichrysene unit was achieved by a Diels–Alder cycloaddition reaction between 3a and 4 in o‐xylene at 170 °C for 10 h in 80% yield. With precursor 5a in hand, the Scholl reaction toward 1a was examined. As monitored by matrix‐assisted laser desorption/ionization‐time‐of‐flight mass spectrometry (MALDI‐TOF MS), the reaction always gave a mixture of the desired product (1a) and the partially cyclized byproduct (1a’), even with an extended reaction time (30 h) and a large amount of FeCl₃ (7 equiv./H) (Figures S1 and S2, Supporting Information). Interestingly, byproduct 1a' could be efficiently separated by silica‐based chromatography with a yield of 39%. The structure of 1a' was confirmed by high‐resolution MALDI‐TOF MS (Figure S4, Supporting Information) and NMR spectroscopy (Figures S6–S9, Supporting Information), in which the two C−C bond formations in the central bichrysene unit did not occur (pink dots in the structure of 1a' in Scheme 1). This result could be rationalized by the highly distorted structure of the partially cyclized byproduct 1a'. The carbon atoms (1/1' and 2/2') on the bichrysene units of 1a' (pink dots in the structure of 1a', Scheme 1) are distantly separated (3.28 Å) between the two blades (Figure S1, Supporting Information), which makes further dehydrogenations difficult. Therefore, target model compound 1a was obtained in 51% yield, which was also confirmed by high‐resolution MALDI‐TOF MS analysis (Figure S3, Supporting Information). Due to its strong aggregation in common organic solvents (i.e., C₂D₂Cl₄, toluene‐d₈), it is not possible to record its ¹H NMR spectrum, even at high temperature (120 °C). Based on the above results, we adopted a new strategy that employs precursor 5b by prefusing the bichrysene moiety at the early stage to improve the efficiency of the Scholl reaction.

Scheme 1.

Synthetic route toward model compounds 1a and 1b. Reagents and conditions: a) i) PdCl₂(PPh₃)₂, CuI, TIPS‐acetylene, tetrahydrofuran (THF), Et₃N, 70 °C, 12 h; ii) tetrabutylammonium fluoride , THF, 0 °C, 15 min. b) o‐xylene, 4, 170 °C, 10 h. c) FeCl₃ (7 equiv./H), CH₃NO₂, CH₂Cl₂, rt, 30 h for 5a; 2 h for 5b.

Based on a similar synthetic strategy as that for 5a, the key precursor 5b was synthesized starting from 9,18‐dibromo‐3,12‐di‐tert‐butylbenzo[a]dinaphtho[2,1,8‐cde:1’’2’’,3’’,4’’‐ghi]perylene (2b). Compared to precursor 5a, the central bichrysene unit in 5b was fully fused, leaving only the outer diphenylbenzo[e]pyrene units for the subsequent Scholl reaction. To our delight, the final cyclodehydrogenation of 5b using FeCl₃ (7 equiv./H) proceeded smoothly at room temperature, yielding the desired product 1b with a high yield of 95% within only 2 h of reaction time. The efficient formation of model compound 1b from 5b was first validated by MALDI‐TOF analysis (Figure  2a). In positive mode, the intense peak at m/z = 1815.0408 of 5b disappeared, and a new signal at 1794.8923 was clearly obtained, precisely matching the calculated molecular mass [M ⁺] = 1794.8915 for C₁₄₀H₁₁₄ and the simulated isotopic distribution pattern. Thanks to the additional two tert‐butyl groups (position 13 in 1b) on the armchair edges, compound 1b becomes more soluble than its analog 1a. The chemical structure of 1b was unambiguously confirmed by ¹H NMR spectroscopy (Figure 2b) with the help of 2D NMR measurements (Figures S10–S13, Supporting Information).

Figure 2.

a) High‐resolution matrix‐assisted laser desorption/ionization‐time‐of‐flight mass (MALDI‐TOF) mass spectra of 5b and 1b. b) ¹H NMR spectrum of 1b in toluene‐d₈ (500 MHz, 30 °C).

2.2. Synthesis and Characterization of the cMGNR

Encouraged by the successful synthesis of model compound 1b from 5b, we carried out synthesis of the cMGNR from the precursor polymer (P1) containing the prefused bichrysene unit. As shown in Scheme  2 , the bisdiene monomer, i.e., 2,8‐di‐tert‐butyl‐4,6,10,12‐tetrakis(4‐(tert‐butyl)phenyl)dicyclopenta[e,l]pyrene‐5,11‐dione (6), was first prepared in a similar manner to the synthesis of diene 4. Then, A₂B₂‐type Diels–Alder polymerization was carried out by refluxing a solution of monomer 3b and monomer 6 in diphenyl ether (120 × 10⁻³ m, 24 h) without the addition of any other reagent or catalyst, affording the polyphenylene precursor P1 in 96% yield. Three polymer fractions were obtained through recycling preparative gel permeation chromatography (rGPC). The number‐average molecular weight (M _n) of the main fraction of P1 (75 wt%) was estimated to be ≈57 000 g mol ⁻¹ with a polydispersity index (PDI) of 1.9 based on analytical GPC analysis against polystyrene (PS) standards (Figure S16, Supporting Information). Moreover, linear‐mode MALDI‐TOF MS analysis of P1 showed periodic pattern peaks up to m/z ≈25 000 with an interval of m/z = 1502, corresponding to the mass of one repeating unit (Figure S17, Supporting Information). The length characterization of cMGNR through the scanning tunneling microscopy /atomic force microscope techniques was not successful so far, due to its highly twisted geometry and weak affinity on highly oriented pyrolytic graphite . Finally, the targeted cMGNR with a unique curved conformation and hybrid cove–armchair–gulf edges (Scheme 2) was obtained from P1 through the optimized Scholl reaction with FeCl₃ (7 equiv. per H to be removed) in CH₂Cl₂ for 3 days.

Scheme 2.

Synthetic route to the curved multi‐edged graphene nanoribbon (cMGNR) and density functional theory (DFT) optimized geometry of a short segment of cMGNR.

The chemical structure of cMGNR was then fully characterized by a combination of FTIR, Raman, and solid‐state NMR analyses. FTIR analysis of the cMGNR shows that, compared to polymer P1, the weak signals from the typical aromatic C−H stretching vibrations (broad peaks at 3037 and 3081 cm⁻¹ for P1) were diminished, indicating efficient “graphitization” (Figure  3a), which matched well with the simulation (Figure S28, Supporting Information).^{[ 12 ]} Furthermore, the most relevant fingerprint between P1 and cMGNR is the SOLO mode (wagging of an isolated aromatic C−H bond neighbored by two C−C bonds). The cMGNR is characterized by the SOLO mode of new peaks at 863 and 892 cm⁻¹ (blue in Figure 3a) of P1, while both peaks are absent in the spectrum of P1.^{[ 12 ]} We found that the intensity of the DUO mode (wagging of two adjacent aromatic C−H bonds) at 832 and 879 cm⁻¹ (sky blue in Figure 3a) significantly decreased, which indicates that the DUO mode arises only from the terminals of cMGNR (802 cm⁻¹). The peak at approximately 800 cm⁻¹, remains consistent in the spectra of both the cMGNR and P1, which represents the TRIO mode (wagging of triply adjacent C−H groups, pink in Figure 3a).^{[ 12 ]} Notably, the peaks at 758, 692, 674, and 637 cm⁻¹ are assigned to the out‐of‐plane and in‐plane bending modes of the carbon atoms of P1 and the cMGNR. In the Raman spectrum of cMGNR, two intense peaks at approximately 1322 and 1588 cm⁻¹ were observed (Figure 3b), which can be assigned to the D and G bands of graphitic materials, respectively. This result is consistent with the literature values for the reported bottom‐up solution‐synthesized GNRs.^{[ 13 ]} The relatively high intensity of the D band can be explained by the contribution of the edges as defects. Moreover, three well‐resolved double‐resonance signals were also observed at 2644, 2911, and 3179 cm⁻¹, which can be assigned to 2D, D + G, and 2G peaks, respectively. Solid‐state ¹H and ¹³C{¹H} MAS NMR experiments of P1 and cMGNR (Figures S18 and S19, Supporting Information) validate the successful formation of the cMGNR as obtained from quantitative analysis of the aromatic and aliphatic signals (Table S1, Supporting Information). 2D ¹H‐¹H DQ‐SQ NMR correlation experiments (Figure S18, Supporting Information) further reveal a reduced span of the aromatic ¹H‐¹H autocorrelation signal (up to ≈11 ppm) for cMGNR, which is associated with the reduced π–π stacking interactions in comparison to those of other GNR samples (up to ≈15 ppm).^{[ 3} , ⁹ , ¹⁰ , ^{14 ]} This observation further supports the nonplanar geometry of the cMGNR as depicted in Scheme 2.

Figure 3.

a) FTIR spectra of P1 and the cMGNR measured on powder samples. b) Raman spectrum of the cMGNR recorded at 514 nm. c) UV‐vis‐NIR absorption spectra of 5b, 1b, and P1 (10⁻⁵ m) as well as the cMGNR (0.1 mg mL⁻¹) in THF.

The UV–vis absorption spectra of model compound 1b and the cMGNR, as well as their precursors (5b and P1), were recorded in tetrahydrofuran solutions (Figure 3c). In comparison to those of precursors 5b and P1, the absorption peaks of 1b and the cMGNR display a significant redshift, due to their π‐extended conjugation after the cyclodehydrogenation reactions. The longest absorption peak appears at 605 nm for 1b, which is assignable to the HOMO to LUMO electronic transition according to time‐dependent DFT calculations (Table S3, Supporting Information). The optical bandgap of 1b is calculated to be 1.81 eV, based on the Tauc plot of UV–vis absorption, which is in good agreement with the calculated result (1.87 eV, Scheme S22, Supporting Information). Remarkably, the cMGNR exhibits a strong, broad absorption covering the UV, visible, and even near‐infrared (NIR) region, indicating the efficient “graphitization” of polymer precursor P1. From the onset of its longest absorption peak at 621 nm, the optical bandgap of cMGNR is calculated to be 1.61 eV, in excellent agreement with the calculated result (1.56 eV, Figure S23, Supporting Information).

2.3. THz Study of the cMGNR

To study the photoconductivity dynamics and transport properties of charge carriers in cMGNR, we employed contact‐free optical pump‐terahertz (THz) probe (OPTP) spectroscopy. Figure  4a presents the complex photoconductivity dynamics of the cMGNR dispersed in 1,2,4‐trichlorobenzene. An initial rise in photoconductivity is observed due to the transient population of free carriers following interband excitations (by a laser pulse with a photon energy of 3.10 eV). Subsequently, free carriers are transformed into exciton species, resulting in a rapid ≈ps decay in the real conductivity and a long‐lived imaginary contribution to the photoconductivity.^{[ 5} , ⁹ , ^{15 ]} To further study the free charge carrier transport properties at an early timescale, we measured the frequency‐resolved complex conductivity spectrum (at ≈1.3 ps after photoexcitation), and described the data using the Drude–Smith (DS) model (Figure 4b; see Supporting Information for more details).^{[ 16 ]} In the model, the charge carriers are assumed to be subject to anisotropic momentum scattering processes, particularly with an increasing preferential backscattering probability, due to the presence of, e.g., grain boundaries and torsional defects in the materials.^{[ 5b ]} A parameter c is introduced to characterize the backscattering probability, which ranges between 0 (isotropic scattering) and −1 (fully backscattering). DS model analysis (solid lines in Figure 4b) yields the charge scattering time τ = 36 ± 2 fs and c = −0.97 ± 0.01. Furthermore, knowing the effective mass of the charge carriers (m*_h = 2.16 m ₀ and m*_e = 1.27 m ₀, see details in Supporting Information), we infer the mobility of the charge carriers in the dc limit to be 2.4 cm² V⁻¹ s⁻¹, following μ =$\frac{e\tau }{m^{\ast }}(1+c)$. The obtained mobility of the cMGNR is relatively lower than other reported curved GNRs,^{[ 5d,e ]} mostly due to its large effective mass and flat band dispersion (Figure S27, Supporting Information).

Figure 4.

a) Time‐resolved complex terahertz photoconductivity of the cMGNR dispersed in 1,2,4‐trichlorobenzene, normalized to the absorbed photon density. b) Frequency‐resolved terahertz conductivity measured at ≈1.3 ps after photoexcitation (denoted by the dashed vertical line in (a)). The solid lines are fits using the Drude–Smith model.

3. Conclusion

In summary, we have demonstrated an efficient strategy to synthesize a novel long and curved GNR with combined cove, armchair, and gulf edge structures, namely, cMGNR. The synthetic route of the cMGNR involves efficient A₂B₂‐type Diels–Alder polymerization by using a diethynyl‐substituted fused bichrysene derivative (3b) as a bisdienophile monomer and dicyclopenta[e,l]pyrene‐5,11‐dione (6) as a bisdiene monomer, followed by intramolecular cyclodehydrogenation. Model compound 1b was efficiently synthesized from prefused bichrysene‐based oligomer 5b, which demonstrates the suitable structural design of precursor polymer P1. Remarkably, the obtained cMGNR had an average length of ≈71 nm and a low optical bandgap of 1.61 eV. In addition, THz photoconductivity analysis of the cMGNR revealed a charge carrier mobility exceeding 2 cm² V⁻¹ s⁻¹, which makes it a promising candidate for single‐GNR based nanoelectronic devices. This work opens a door for the synthesis of novel GNRs with high longitudinal extension and nonplanar conformation via A₂B₂‐type Diels–Alder polymerization by using other bisdiene and bisdienophile monomers.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Yang L., Ma J., Zheng W., Osella S., Droste J., Komber H., Liu K., Böckmann S., Beljonne D., Hansen M. R., Bonn M., Wang H. I., Liu J., Feng X., Solution Synthesis and Characterization of a Long and Curved Graphene Nanoribbon with Hybrid Cove–Armchair–Gulf Edge Structures. Adv. Sci. 2022, 9, 2200708. 10.1002/advs.202200708

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.