In recent years, syntheses of nonplanar polycyclic aromatic hydrocarbons (PAHs) have attracted increasing attention due to their unique structures and novel physical properties1,2,3,4,5. Nonplanar carbon-based π-conjugated systems can be constructed with different topologies of bowls, hoops, and saddles, which can be considered as conjugated segments of larger 0D to 3D systems such as fullerenes6,7,8,9, carbon nanotubes10,11, helical nanocarbons12,13,14,15, and porous graphene sheets16,17. Riemann surfaces are curved helical structures in chemistry, described as deformed versions of the complex plane in mathematics. Carbon nanosolenoids can be considered to follow a Riemann surface18, in which one atomic carbon layer continuously spirals around the line perpendicular to the basal plane. Interestingly, Riemann surfaces were proposed as well-known objects in complex analysis and inspired scientists to design multidimensional carbon architectures (Fig. 1a)19. Some nonplanar structures were predicted to have many interesting mechanical, electronic, and magnetic properties20,21.

Fig. 1: Schematic illustrations of Riemann surface and nitrogen-doped carbon nanosolenoids (N-CNS).
figure 1

a An example of Riemann’s surface. b Design of fully π-extended N-CNS.

Narrow strips of graphene with abundant edges and high aspect ratios, called graphene nanoribbons (GNRs), have great potential applications in photonics, electronics, and energy science22,23,24. Using a top-down method, GNRs can be synthesized by breaking apart graphene or carbon nanotubes. However, it is difficult to atomically precise control over the GNR structure and further investigate their applications. Therefore, bottom-up synthesis seems to be a feasible approach to solve this problem. Well-defined GNRs can be constructed by surface-assisted or solution-based polymerization of precursors followed by cyclodehydrogenation reactions25,26,27. Among them, donor–acceptor (D–A) heterostructures can be fabricated through heteroatom doping in graphene strips with nitrogen, sulfur, and boron atoms at edges28,29,30,31. Owing to its electron-donating character and tendency to induce robust ferromagnetism in graphene at a high doping concentration, nitrogen has become the major heteroatom for the synthesis of heterostructures32,33. The electronic properties of GNRs could also be tuned by doping with nitrogen, which has been studied both theoretically and experimentally34,35. Fasel, Müllen, and coworkers used a stepwise cyclization/dehydrogenation sequence on a metal Au(111) surface to obtain nitrogen-doped graphene nanoribbons (N-GNRs), demonstrating a p–n junction character with an electric field of 2 × 108 V m–1 and a band shift of 0.5 eV at the heterojunction36. Moreover, the synthesis of porous N-GNRs was reported by the Meyer group via Ullmann coupling on an Ag(111) surface37. However, surface-assisted synthesis strategies are difficult to scale up and make nonplanar 3D strips. Very recently, we developed a facile solution-based strategy to synthesize all-carbon nanosolenoid material with Riemann surface18. Nanostructured graphitic carbon materials resembling a Riemann surface with helicoid topology are predicted to have interesting magnetic, electrical, and photonic properties19,20,21. For example, in 2016, Yakobson theoretically predicted a carbon solenoid with Riemann surfaces. This material could emerge a large magnetic field when a voltage is applied and become a brand new magnetic material19. The following experimental results proved that such carbon nanosolenoid material has magnetic properties18. Based on these previous results, we speculate that nitrogen-doped carbon nanosolenoid (N-CNS) heterojunction material with Riemann surfaces will possess special optoelectronic and photocatalytic properties, which is unexplored in this field. However, there is no report on the large-scale fabrication of nitrogen-doped carbon nanosolenoid (N-CNS) heterojunction material with Riemann surfaces.

Herein, we report the first synthesis and characterization of such a nitrogen-doped carbon material with interesting physical properties (Fig. 1b). The photophysical properties were investigated using UV–Vis absorption and fluorescence spectroscopy. Furthermore, the broad absorption of N-CNS in the visible region enables it as an excellent candidate for photocatalysis. The photocatalysis pathway for solar energy conversion helps to meet the growing demand for energy issues and environmental restoration38,39. Currently, efficient and environmentally friendly photocatalysts are in high demand for many important energy conversion applications, such as solar water splitting, photocatalytic organic reactions, and so on40,41,42. In addition, metal-free catalysts are also of great significance in green chemistry and attract much attention in research43,44. Metal-free catalysts have many advantages that include low cost and ready availability, low toxicity, higher stability in air and water, and increased synthetic efficiency due to the avoidance of the time-consuming removal of toxic metal traces. N-CNS, which does not contain any metal component and has photocatalytic properties under visible light, can be used as a potential metal-free catalyst for both visible light-driven H2 production and efficient photocatalytic organic transformations.


Molecular design of N-CNS

Figure 2 shows the conceptual structure with helicoid topology, in which nanostructured N-doped carbon nanosolenoid material resembles a Riemann surface. The appropriate building units are rationally chosen to obtain the desired structural features. A hexaphenylbenzene (HPB) derivative was chosen as the PAH building unit to achieve the large π-extended feature of N-CNS. To obtain the heterostructure of N-CNS, 5-bromopyrimidine serves as the starting material for introducing nitrogen atoms and Sonogashira coupling can be performed to produce 1,2-di(pyrimidin-5-yl)ethyne. Next, the Diels–Alder reaction between 5,10-dibromo-1,3-diphenyl-2H-cyclopenta[l]phenanthren-2-one and 1,2-di(pyrimidin-5-yl)ethyne can afford the HPB derivative unit M1. Another building unit M2 was designed as the linker with two ortho-boryl groups to achieve the helicoid topology of N-CNS. Two ortho-tert-butyl groups in M2 were introduced to improve the solubility of precursor P1 and N-CNS.

Fig. 2: Synthetic approach to N-CNS.
figure 2

Reagents and conditions: (i) M1 (1.0 equiv.), M2 (1.0 equiv.), K2CO3 (10 equiv.), Aliquat 336 (5 mol%), Pd(PPh3)4 (10 mol%), Ar, toluene/H2O (v/v, 5:1), 110 °C, 60 h; (ii) P1 (1.0 equiv.), DDQ (16 equiv.), TfOH, Ar, anhydrous CH2Cl2, 0 °C, 24 h.

Solution synthesis of N-CNS with large lateral π-extension

The synthesis procedure of the N-CNS is summarized in Fig. 2. The essential precursor P1 was fabricated by combining the molecular building blocks 5,5’-(6,11-dibromo-1,4-diphenyltriphenylene-2,3-diyl)dipyrimidine (M1) and 2,2’-(4,4”-di-tert-butyl-[1,1’:2’,1”-terphenyl]-4’,5’-diyl)bis(4,4,5,5-tetramethyl-1,3,2-dioxaborolane) (M2). Then, P1 was further performed Scholl reaction to obtain large π-conjugated N-CNS. The detailed synthesis procedure of these molecular building blocks is described in Supplementary Information. The polymerization of M1 and M2 was performed by Suzuki–Miyaura coupling reaction using Pd(PPh3)4 as the catalyst, toluene and water as solvents, and reacted at 110 °C for 60 h to obtain the essential precursor (P1) with a yield of 83%. Finally, N-CNS was obtained by oxidative cyclodehydrogenation of P1 using 2,3-dichloro-5,6-dicyano-1,4-benzoquinone and trifluoromethanesulfonic acid at 0 °C for 24 h. During the reaction, we found that the solution color quickly turned black, and finally, a dark purple solid was obtained. Moreover, the fluorescence of the solution was changed from blue to red, further confirming the increasing degree of π-extended conjugation of the backbone.

Gel permeation chromatography (GPC) was used to measure the average molecular weight of N-CNS (Fig. 3c). The weight average molecular weight (MW), relative number-average molecular weight (Mn), and polydispersity index (PDI) of N-CNS were obtained. The molecular weight distribution of N-CNS shows a single broad peak with a PDI of 2.28. The Mn of N-CNS is 65,400 g mol−1. Considering that one helical pitch consists of three monomeric units, this Mn corresponds to ~76 monomer units and contains ~25 spiral pitches.

Fig. 3: Spectroscopic characterization of N-CNS.
figure 3

a and b Representative FTIR spectral regions of precursor P1 (blue lines) and N-CNS (red lines). c GPC trace of N-CNS. d Raman spectrum of N-CNS measured at 532 nm (2.33 eV) on a powder sample.

Bulk spectroscopic characterizations of N-CNS

The highly efficient cyclodehydrogenation of precursor P1 into N-CNS was confirmed by Fourier transform infrared (FTIR), Raman spectroscopy, and solid-state 13C NMR measurements. FTIR analysis of the material N-CNS after the cyclodehydrogenation revealed that the signals at 3030, 3055, and 3079 cm−1 originating from aromatic C–H stretching vibrations in P1 were diminished26. Similar peaks (2906 and 2867 cm−1 for N-CNS, 2905 and 2867 cm−1 for P1) associated with the C–H stretching of alkyl chains were observed upon cyclodehydrogenation, demonstrating the integrity of the alkyl substituents. The vibration bands at 698, 813, and 838 cm−1 from mono- and disubstituted benzene rings were attenuated (Figs. 3a, b,  S2 and S3)45. Furthermore, the appearance of a new band maximized at 873 cm−1 confirmed the opla band which is a typical signal for aromatic C–H at the cove position (Fig. 3b)45,46,47. All these observations indicate the highly efficient conversion of precursor P1 into N-CNS.

As a useful tool for characterizing carbon nanomaterials, Raman spectroscopy was performed which is sensitive to the structural and edge defects to confirm the high structural quality of N-CNS. The result shows the typical structural feature for graphene nanoribbons with characteristic intense G and D bands25. The Raman spectrum of N-CNS (excited at 532 nm, powder sample) demonstrates the typical G-band peak and D-band peak for armchair-type GNRs observed at 1604 and 1329 cm–1, respectively (Fig. 3d), which is consistent with those of GNRs obtained by bottom-up synthesis48. Double formants at 2649, 2934, and 3206 cm–1 were initially assigned to 2D, D + D′, and 2D′ peaks, respectively. The D-band peak can be explained by collective modes of the confined hexagonal rings for the confinement of π-electronics into a finite-size domain in large π-extended PAHs49.

Solid-state 13C NMR spectra were performed to confirm the structural changes before and after the Scholl reaction (Fig. S4). The characteristic single peak (δ = 31.28 ppm for P1 and 31.92 ppm for N-CNS) can be assigned to t-butyl carbons. P1 shows multiple peaks (δ = 109.13–162.02 ppm), which can be assigned to the aromatic carbons. The main aromatic 13C NMR signals of N-CNS centered at ~124 ppm, probably caused by main π-conjugated carbon atoms in the N-CNS. Moreover, the peaks of N-CNS in the aromatic zone move slightly to the lower field because of the increasing π-conjugation.

Physical properties

The photophysical properties of N-CNS and P1 were characterized in solution by steady-state spectroscopy and time-resolved spectroscopy (Fig. 4). The absorption spectrum of P1 only shows a narrow absorption band at 250–380 nm maximized at 300 nm. In sharp contrast, the UV–Vis spectrum of N-CNS in DMF showed a broad redshifted absorption band in the range of 260–800 nm maximized at ~500 nm, indicating that a large π-conjugated structure was formed after cyclodehydrogenation. Furthermore, under an excitation at 350 nm, N-CNS exhibits a broad emission band at 440–820 nm maximized at 564 and 661 nm, demonstrating significant redshift compared to the reference precursor P1 (maximized at 413 nm). This emission redshift is consistent with the UV–Vis result, indicating a large and well-extended conjugation nature in N-CNS. As shown in Fig. 4b, the precursor P1 is a light-yellow solid, while N-CNS becomes a black solid. For comparison, the dilute DMF solution of N-CNS showed an intense red photoluminescence under irradiation by a hand-held UV lamp at λ = 365 nm, while the precursor P1 in solution presented an intense blue photoluminescence.

Fig. 4: Photophysical properties of N-CNS.
figure 4

a UV–Vis absorption (dots) and fluorescence spectra (lines) of N-CNS (wine) in DMF and precursor P1 (blue) in CH2Cl2. b Solid powder of precursor P1 (I) and N-CNS (II), precursor P1 in CH2Cl2 solution (III) and N-CNS in DMF solution (IV) under UV irradiation at 365 nm. c, d Emission lifetimes for precursor P1 in CH2Cl2 (black dots: data; blue line: fitting) and N-CNS in DMF (black dots: data; wine line: fitting).

Moreover, the luminescence lifetime (τs) of precursor N-CNS and P1 were measured by time-resolved fluorescence decay using the time-resolved photoluminescence (TRPL) technique (Figs. 4c, d). The luminescence lifetime of N-CNS follows second-order kinetics with lifetimes of 6.12 and 2.13 ns at 661 nm when excited at 420 nm. For comparison, P1 shows longer fluorescence lifetimes at 16.64 and 3.28 ns at 413 nm when excited at 320 nm. The different emission lifetimes probably result from the different excited states of P1 and N-CNS.

The morphology and microstructures of N-CNS were further investigated by scanning electron microscopy (SEM) and atomic force microscope (AFM). As shown in Fig. 5a, bulk N-CNS has layered structures consisting of thin sheets with micrometer-long wrinkles. AFM image showed that N-CNS can self-assemble into well-defined helix bundles on the surface (inset, Fig. 5a), which is consistent with our previous study.18 To further investigate the electronic properties of N-CNS, EPR measurements were carried out using a solid-state sample at room temperature (Fig. 5b). N-CNS displayed a typical single EPR peak, with a ge value of ~2.0006, which confirms unpaired electrons and magnetic properties exist in this present solenoid material18,19. Moreover, the chemical compositions and states of N-CNS were investigated by X-ray photoelectron spectroscopy (XPS). In Fig. 5c, C 1s spectrum of N-CNS shows two peaks at 284.8 and 286.3 eV, corresponding to sp2 bonded carbon–carbon, and sp2 bonded carbon-nitrogen (N = C–N), respectively. Figure 5d exhibits N 1s XPS spectra, the peak at 399.5 eV could correspond to sp2 N bonding of (C = N–C). Ni 2p, Co 2p, Fe 2p, and Pd 3d XPS spectra of N-CNS are shown in Fig. S8. These results indicate that all metal elements involved in the synthetic reactions are absent from the samples, further demonstrating that N-CNS is a metal-free carbon material.

Fig. 5: SEM, AFM, room temperature EPR spectrum and XPS spectra characterizations of N-CNS.
figure 5

a SEM image of bulk N-CNS, scale bar 4 μm; inset, AFM image of N-CNS, scale bar 6 Å. b Room temperature EPR spectrum of solid powder. c C 1s and d N 1s XPS spectra of N-CNS.

The explicit real-space structural elucidation of such organic molecules requires direct high-resolution structural imaging using electron microscopy. However, these strands of N-CNS organic molecules are extremely vulnerable to electron beam irradiation and thus prohibit traditional electron microscopy techniques from imaging the integral structures. The recent advances in low-dose electron microscopy allow the direct imaging of beam-sensitive materials50,51,52, including metal-organic frameworks3,52,53,54, covalent-organic framework52,55,56 and organic molecules55. With the development of low-dose electron microscopy methods, integrated differential phase contrast scanning transmission electron microscopy (iDPC-STEM) shows great advantages in the imaging of low-Z elements and electron beam-sensitive materials18,57,58. Accordingly, we tried to image the organic N-CNS helices using the low-dose iDPC-STEM technique to maintain the structural integrity of these helices. Although the strands of N-CNS helices tend to bundle together due to the strong intermolecular interactions, prolonged probe sonication allows the segregation of a considerable fraction of single-strand N-CNS helices that can be clearly identified and elucidated by low-dose iDPC-STEM imaging. As we can see in Fig. 6, the intrinsic single-strand molecular structures of N-CNS helices can be clearly resolved from both [100] and [201] projections of the proposed 1D periodic structural model (Fig. S5). Figure 6a shows the iDPC-STEM image of a single-strand N-CNS helix projected along [100] axis, which exhibits fringe contrast characteristics for the periodic helical structures. The measured fringe spacing (b′) of the N-CNS helix is 0.40 ± 0.03 nm and the helical width (d′) is 2.4 ± 0.2 nm, respectively, which matches well with the proposed molecular structural model of N-CNS helix. As shown in Figs. 6c and  S7, the calculated helical width of d is 25.12 Å (without the consideration of tert-butyl) and the fringe spacing of b is 4.110 Å; then the resulting angle γ is 2.974o (γ = argtan (b/(d*π))). It has been widely acknowledged that the iDPC-STEM contrast closely resembles the electrostatic potential of the chemical structure18,59. The present results demonstrate the iDPC-STEM image contrast closely resembles the simulated projected potential map of the proposed N-CNS helix model, as shown in Figs. 6b and c. Moreover, the helical structural model is further validated from another iDPC-STEM image taken along the [201] direction of the structural model (Fig. 6d), from which the N-CNS helix is tilted away from the image plane as shown in Fig. 6f. The STEM image exhibits arrays of bright dots that corresponding to the projected columns of benzene rings in the N-CNS helix, which match well with the simulated projected potential map as shown in Fig. 6e. Taken together, low-dose iDPC-STEM imaging along both flat-lying and tilted projections of single-strand N-CNS helices unambiguously confirms the proposed molecular structural mode of nitrogen-doped carbon nanosolenoid with Riemann Surfaces.

Fig. 6: iDPC-STEM characterizations of N-CNS.
figure 6

Low-dose iDPC-STEM images (a, d), simulated projected potentials (b, e), and structural models (c, f) of single-strand N-CNS along [100] (upper figures) and [201] (lower figures) axes respectively. The false-colored images are rendered in aquatic color code. The structural model used for simulating the electrostatic potential is constructed by adhering the single-strand N-CNS along a specific projection with a properly situated amorphous carbon layer with a thickness of 1 nm. A specific point-spread-function (PSF) width of 1.6 Å was used for the simulation.

Photocatalytic properties

Since N-CNS has excellent photophysical properties, it could be a potential metal-free photocatalyst for energy conversion. The photocatalytic H2 evolution performance using N-CNS was further evaluated under the light irradiation of a 300 W Xe lamp equipped with a UV cut-off filter (λ > 420 nm). The typical photocatalytic H2 production of N-CNS is schematically illustrated in Fig. 7a. Impressively, N-CNS can be used as a metal-free photocatalyst for hydrogen production in water. Figure 7b shows the photocatalytic activity of N-CNS for H2 production in the presence of electron donors. With increasing concentrations of the electron donors (Na2S and Na2SO3), the photocatalytic H2 evolution rate rapidly increased and reached a maximum value of 190.2 ± 21.2 µmol g−1 h−1 in the presence of 0.75 M Na2S and 1.05 M Na2SO3. When the concentrations of Na2S and Na2SO3 increased to 1.0 and 1.4 M, respectively, the H2 evolution rate decreased to 182.8 ± 17.5 µmol g−1 h−1, which is probably due to the loss of visible light energy hindered by the undissolved Na2SO3. Compared to N-CNS, the H2 evolution rate of commercial photocatalyst TiO2 P25 was only 5.5 µmol g−1 h−1 under the same conditions upon visible light irradiation (Table S2), indicating that N-CNS has better photocatalytic activity than P25 under visible light using the same reaction conditions. We also compared N-CNS with other metal-free materials reported in the literature for hydrogen evolution and these results are shown in Table S3. Most of these materials exhibited low H2 evolution rates, and only a few materials demonstrated rates higher than 100 µmol g−1 h−1. In contrast, the hydrogen evolution rate of N-CNS is 190 µmol g−1 h−1. This result demonstrated that N-CNS has good photocatalytic properties for H2 production. Besides, the photocatalytic activity of N-CNS using different sacrificial electron donors was also measured for comparison (Fig. S9). As shown in Fig. 7d, the N-CNS photocatalyst exhibited excellent stability and maintained a similar photocatalytic activity for 36 h. An obvious transient photocurrent response was observed, suggesting the good photo-induced electron separation/transport property of N-CNS (Fig. S13). To confirm the electronic band positions of the as-prepared materials, Mott–Schottky (MS) measurements were carried out (Fig. S11). The conduction-band position is −0.46 eV (vs. NHE) for N-CNS. The absorption edge of N-CNS is located at 670 nm, corresponding to the band gap of 1.85 eV (Fig. S12). Combined with the band gap of N-CNS, the valence-band position can be determined as 1.39 eV (vs. NHE). All these results demonstrate that N-CNS can perform as a good metal-free photocatalyst for visible-light-driven H2 evolution.

Fig. 7: Photocatalytic H2 production by N-CNS.
figure 7

a Schematic illustration of photocatalytic H2 production by N-CNS. b H2 evolution rate of N-CNS photocatalyst at different concentrations of the hole scavenger under visible light (wine: 0.25 M Na2S and 0.35 M Na2SO3; green: 0.5 M Na2S and 0.7 M Na2SO3; orange: 0.75 M Na2S and 1.05 M Na2SO3; cyan: 1 M Na2S and 1.4 M Na2SO3;). c UV–Vis absorption spectrum of N-CNS in DMF with calculated optical bandgap. d Cycling runs of N-CNS for photocatalytic H2 evolution in the presence of 1.0 mg N-CNS photocatalyst in a 30 mL aqueous solution containing 0.75 M Na2S and 1.05 M Na2SO3 (black: cycle 1; red: cycle 2; blue: cycle 3). The error bars in b and d show standard deviation (SD).

N-CNS can also be proposed as a novel photocatalyst to realize the C–H functionalization of 2-phenyl-1,2,3,4-tetrahydroisoquinoline (1a)60,61,62,63. Catalytic activity of N-CNS toward photocatalytic aza-Henry reaction was further evaluated with 1a and nitromethane as starting materials. The reaction was carried out in the presence of O2 with a 9 W blue LED as the light source. To our delight, the desired C–H functionalization product 1b was isolated in an excellent 95% yield by irradiation for 5 h with a catalytic amount of N-CNS (Table 1). And the use of visible light (300 W Xe lamp equipped with a UV cut-off filter, λ > 420 nm) resulted in a 90% yield. In sharp contrast, only a 9% yield of 1b was obtained when the reaction was performed using the precursor P1. Control experiments revealed that catalyst, light, and oxygen are crucial for the formation of 1b. Moreover, we also performed the absorption spectrum to investigate the stability of N-CNS before and after the photocatalytic aza-Henry reaction (Fig. S14). After the reaction, a three-step procedure including centrifugation, separation, and washing was done and N-CNS was performed in a new cycle, no significant loss of photocatalytic activity was observed after 10 cycles of 12 h (Fig. S22).

Table 1 The photocatalytic aza-Henry reactionsa

Besides the easy functionalization of tetrahydroisoquinoline, various other substituents were also subjected to photocatalytic organic transformations using N-CNS as the photocatalyst. As shown in Table 2, the catalytic results are also excellent for the benzylamine homocoupling reaction and oxygenation reaction of thioanisoles with all the yields >90%. Under light irradiation, the benzylamine homocoupling produced the product of N-benzyl-1-phenylmethanimine (3a) in a yield of 90.1%, as checked by 1H NMR. Oxygenation reaction of methyl(phenyl)sulfane was performed under similar photocatalytic reaction conditions and the product of (methylsulfinyl)benzene (3b) had a yield as high as 94.9% in only 5 h. In addition, more sulfinylbenzene compounds with the amendment of different groups at the para-position, such as –OMe (2c), –F (2d), were examined. The yields of 1-methoxy-4-(methylsulfinyl)benzene (3c) and 1-fluoro-4-(methylsulfinyl)benzene (3d) were calcualted to be 90.0% and 90.9%, respectively. All the high yields indicated that the N-CNS is an excellent metal-free photocatalyst in many photoredox reactions.

Table 2 The photocatalytic oxygenation of benzylamine and thioanisolesa


In summary, we reported a facile bottom-up synthesis of a large well-extended nitrogen-doped carbon nanosolenoid (N-CNS) with Riemann surface by a Pd-mediated Suzuki–Miyaura coupling followed by a Scholl reaction for cyclodehydrogenation. N-CNS was fully characterized by GPC, FTIR, Raman, solid-state 13C NMR spectra, and iDPC-STEM techniques. Its unique photophysical properties were investigated using UV–Vis, fluorescence, TRPL spectroscopy. Notably, N-CNS can be used as a simple polymeric metal-free photocatalyst for hydrogen production in water and a highly efficient photocatalyst such as aza-Henry reactions and oxygenation reactions of thioanisoles in the absence of any noble metals. The present results can shed light on the synthesis of novel carbon nanomaterials and explore their applications in solar energy conversion.


Compounds preparation

See Supplementary Information for details of the syntheses and characterization of M1 and M2.

Synthesis of P1

To a degassed suspension of 5,5’-(6,11-dibromo-1,4-diphenyltriphenylene-2,3-diyl)dipyrimidine (M1) (234 mg, 0.336 mmol), 2,2’-(4,4”-di-tert-butyl-[1,1’:2’,1”-terphenyl]-4’,5’-diyl)bis(4,4,5,5-tetramethyl-1,3,2-dioxaborolane) (M2) (200 mg. 0.336 mmol), potassium carbonate (465 mg, 3.36 mmol), Aliquat 336 (7.0 mg, 0.017 mmol, 5 mol%) in toluene (5 mL) and H2O (1 mL) was added Pd(PPh3)4 (39.3 mg, 0.034 mmol), then the mixture was degassed for 15 min. The mixture was then heated at 110 °C for 60 h under a nitrogen atmosphere. Upon the reaction cooling to room temperature, the mixture was diluted with 1 M aqueous hydrochloric acid, extracted with CH2Cl2, and concentrated under vacuum. The resulting precipitate was washed intensively with 0.5 M HCl solution in water, filtered, and washed with water, methanol, acetone, methanol, and hexane. The precipitate was collected, giving 304 mg of polymer P1 as a pale yellow solid (yield: 83%).

Synthesis of N-CNS

After a solution of polymer P1 (500 mg, 0.55 mmol), 2,3-dichloro-5,6-dicyanobenzoquinone (DDQ:2.01 g, 8.9 mmol) in anhydrous DCM (50 ml) was degassed with argon for 20 min, trifluoromethanesulfonic acid (3.0 mL) was added to the mixture. Then the reaction mixture was stirred at 0 °C for another 24 h under an argon atmosphere and quenched with saturated NaHCO3 solution, filtered, and washed with methanol, acetone, CH2Cl2, and methanol. The precipitate was collected to obtain the title polymer N-CNS as a dark black solid (yield: 76%): Mn = 65,391 g mol−1 and Mw = 149,204 g mol−1 by GPC.

Aza-Henry reaction

To a flame-dried 10 mL vial equipped with a magnetic stir bar with 1a (104.6 mg, 0.5 mmol) and N-CNS (1.5 mg). Nitromethane solvent (3 mL) was transferred to the vial via syringe under oxygen and stirred at a blue LED lamp (9 W) for 5 h. After the reaction, the N-CNS was separated by centrifugation and washed with DCM which could be used for the catalyzer again. The crude product 1b was in the solvent and purified by silica gel chromatography to afford the desired 1b as a yellow oil in a yield of 95% (127.5 mg). 1H NMR (CDCl3, 400 MHz): δ 7.33–7.16 (m, 5H), 7.13 (d, J = 7.2 Hz, 1H), 6.98 (d, J = 8.2 Hz, 2H), 6.85 (t, J = 7.3 Hz, 1H), 5.55 (t, J = 7.2 Hz, 1H), 4.87 (dd, J = 11.8, 7.8 Hz, 1H), 4.56 (dd, J = 11.8, 6.6 Hz, 1H), 3.71–3.56 (m, 2H), 3.14–3.04 (m, 1H), 2.84–2.75 (m, 1H).

The recycling process of N-CNS photocatalytic for aza-Henry reaction: To a flame-dried 10 mL vial equipped with a magnetic stir bar with 1a (104.6 mg, 0.5 mmol) and recycling N-CNS (1.5 mg). Nitromethane as the solvent (3 mL) was transferred into the vial via syringe under oxygen and stirred under a blue LED lamp (9 W) for 5 h. Then, the N-CNS catalyst was recovered for another cycle. After a few new cycles, the product 1b can be still obtained with a similar yield.

Benzylamine homocoupling reaction

To a flame-dried 10 mL vial equipped with a magnetic stir bar with 2a (1 mmol) and N-CNS (1.5 mg). MeCN (3 mL) was transferred to the vial via syringe under oxygen and stirred at a blue LED lamp (9 W) for 5 h. The mixture was filtered to remove the catalyst to get main product 3a. The yields were determined by 1H NMR analysis (Fig. S18).

Photocatalytic oxidation of thioethers

To a flame-dried 10 mL vial equipped with a magnetic stir bar with 2b–2d (1 mmol) and N-CNS (1.5 mg). MeOH (3 mL) was transferred to the vial via syringe under oxygen and stirred at a blue LED lamp (9 W) for 5 h. The mixture was filtered to remove the catalyst to get the main product 3b–3d. The yields were determined by 1H NMR analysis (Figs. S19–S21).

Spectroscopic analysis

High-resolution mass spectrometry (HR-MS) analyses were carried out using MALDI-TOF-MS techniques. NMR spectra were recorded on Bruker BioSpin (1H 400 MHz, 13C 100 MHz) spectrometer, and chemical shifts were reported as the delta scale in ppm relative to CDCl3 (δ = 7.26 ppm) for 1H NMR and CDCl3 (δ = 77.0 ppm) for 13C NMR. Data are reported as follows: chemical shift, multiplicity (s = singlet, d = doublet, t = triplet, m = multiplet, br = broad signal), coupling constant (Hz), and integration. UV–vis absorption spectra were performed on a UNIC-3802 spectrophotometer. Atomic force microscope (AFM) measurements were performed using a Dimension ICON microscope (Bruker) in the tapping mode in a clean room environment. All anhydrous solvents and starting chemical reagents for syntheses were purchased from commercial suppliers (Aldrich or Acros) and used without further purification unless otherwise noted. All moisture- or air-sensitive reactions were carried out in a dry reaction vessel under an inert atmosphere (argon or nitrogen) using standard Schlenk techniques. Preparative column chromatography was performed on silica gel (size 200–300 mesh). Analytical thin-layer chromatographies (TLC) were performed with silica gel HSGF 254. Flash chromatography was performed on silica gel (300–400 mesh). Gel permeation chromatography (GPC) was carried out using a G1316A PL gel column with a rate of 1.0 min/mL in DMF assured by G1310B Iso.pump and detected by a G1362A differential refractive index detector. The solid-state 13C NMR experiments were conducted on a Bruker AVANCE NEO600WB spectrometer (1H 600.16 MHz, 13C 150.93 MHz) using a Bruker 3.2 mm H/X/Y probe. The samples were packed in 3.2 mm o.d. ZrO2 rotos. Gly was used as an external reference to calibrate the radiofrequency (rf) field strength and chemical shift scale (δ (13C) = 176.03 ppm).

The iDPC-STEM imaging

The low-dose high-resolution iDPC-STEM images were obtained under a Cs-corrected electron microscope operated at 300 kV. The beam current was reduced to 1 pA, the convergence angle was 25 mrad and the collection angle of iDPC-STEM imaging was set to 7–29 mrad. The projected electrostatic potential was simulated using the QSTEM code with a point-spread-function (PSF) width of 1.5 Å over a single-strand CNS structural model embedded in a 1 nm-thick amorphous carbon layer.

Computational details

Geometrical structure and electronic structures of N-CNS

Geometrical optimization of the whole cell was carried out utilizing the Vienna an initio simulation package (version: VASP 6.3.1)64. Perdew–Burke–Ernzerhof (PBE)65 in terms of the gradient of electronic density described the exchange and correlation (XC) interaction in the Kohn–Sham equation. The interaction between ions and electrons was treated by the projector-augmented wave (PAW)66-based pseudopotential, featuring greater computational efficiency as well as high accuracy. Specifically, outer electrons of C, H, and N atoms were explicitly treated as valence electrons. Plane-wave function with kinetic energy less than the energy of 450 eV is included in the basic set. The 1 × 1 × 16 k-point grids based on the strategy of Monkhorst–Pack67, featuring enough accuracy in the calculation of total energy via convergence test, were used to sample in the Brillouin zone. During the calculations, the convergence value was set to 1.0 × 10−5 eV for self-consistent field calculations, and the geometrical optimization will keep running until <0.02 eV/Å of Hellmann–Feynman force per atom.

Photocatalytic experiments

The photocatalytic H2 evolution experiments were carried out in a 50 mL flask with stirring at room temperature using a 300 W Xe lamp equipped with a UV cut-off filter (λ > 420 nm). 2.0 mg of the photocatalyst was dispersed in 30 mL of aqueous solution containing 6 mL DMF for dissolution of N-CNS and Na2S and Na2SO3 as sacrificial reagents. Before irradiation, the solution was bubbled with high-purity nitrogen for 20 min to remove the air.

Photoelectrochemical measurements

Photoelectrochemical measurements were carried out on a CHI 660e electrochemical workstation in a standard three-electrode electrochemical cell with the photocatalyst-coated FTO as the working electrode, a platinum wire as a counter electrode, and a saturated Ag/AgCl electrode as a reference electrode. A sodium sulfate solution (0.2 M Na2SO4) was used as the electrolyte. 1 mg N-CNS powder was mixed with 1.0 mL acetonitrile and 50 μL Nafion and sonicated for 10 min. The working electrodes were prepared by dropping the suspension (50 μL) evenly onto the surface of the FTO plates and dried at room temperature. The Mott–Schottky plots were measured at the frequencies of 1500, 2000, and 2500 Hz.