Designed synthesis and characterizations. CN-COF was synthesized via a Schiff-base condensation reaction of 1,3,5-triformylphloroglucinol (Tp) with 4,4'-diamino-[1,1'-biphenyl]-3,3'-dicarbonitrile (BDCN) in the presence of 6 M aqueous acetic acid (Fig. 1a). The formation of β-ketoenamine linkage via a keto-enol tautomerization could increase the chemical stability of CN-COF and the ketene can also serve as an electron donor. Thus, a ketene-CN D-A pair was successfully incorporated in CN-COF, which was further confirmed by the computational study of charge distribution (Fig. 1b). A control sample, BD-COF with similar linkage and topology structure to CN-COF but without cyano as acceptor, was synthesized according to literature method using Tp and benzidine (BD) as monomers.
The FT-IR spectrum of CN-COF displayed vibration peaks at 1618 cm-1 and 1579 cm-1 respectively assigned to the C=O and C=C stretching vibrations together with aromatic ring skeleton vibrations at 1494 cm-1 and 1443 cm-1, showing the formation of β-ketoenamine linkage (Fig. 1c).[46, 47] Notably, the typical C≡N stretching vibration appeared at 2200 cm-1, indicating that cyano groups can endure the synthesis condition without decomposition. Further, 13C CP-TOSS NMR spectrum of CN-COF provided strong supportive structure information with apparent chemical shift for -C=O at 184 ppm, -NH-C=C at 149 ppm and 104 ppm, C≡N at 114 ppm and aromatic rings in the range of 150 to 95 ppm (Fig. 1d). All data above provided adequate chemical composition evidence for the successful preparation of CN-COF with cyano groups. The FT-IR spectrum of BD-COF was in accordance with a previous report, showing the presence of vibrations associated with β-ketoenamine structure (Supplementary Fig. 1). CN-COF and BD-COF with decomposition temperature beyond 350 oC in flowing air had high thermal stability evaluated by thermogravimetric analysis (TGA) (Supplementary Fig. 2).
The crystalline nature of CN-COF was characterized by powder X-ray diffraction (PXRD) technique. The PXRD pattern of CN-COF exhibited a predominant peak at 3.60o, corresponding to the reflection of (100) plane, with other weak peaks at 6.25o, 7.18o and a broad peak at 26.5o, which can be assigned to the (110), (200) and (001) plane, respectively (Fig. 1e). Furthermore, Pawley refinement confirmed that the diffraction patterns of CN-COF were consistent with a hexagonal lattice with P6/M space group (a = b = 28.78 Å, c = 3.60 Å; α = β = 90o, γ = 120o; Rp = 3.99%, Rwp = 4.91%) similar to an eclipsed model (Supplementary Table 1, Fig. 1e). A poor correlation with crystallographic structures of CN-COF was obtained with staggered AB model (Supplementary Table 2, Supplementary Fig. 3), further confirming the eclipsed AA stacking model of CN-COF. Analogously, the PXRD patterns of BD-COF were consistent with previous report, showing the eclipsed AA stacking model (Supplementary Fig. 4). In comparison with CN-COF, the (001) diffraction of BD-COF shifted to higher 2-theta angle, indicating that the interlayer distance of BD-COF was smaller than that of CN-COF, likely due to the increased charge repulsion between interlayer with the existence of strong polar cyano group. All organic semiconductors displayed quite different diffraction peaks with monomers, demonstrating the successful formation of corresponding polymer without residual monomers (Supplementary Fig. 4).
The Brunauer–Emmett–Teller (BET) surface area of CN-COF measured by nitrogen sorption isotherms at 77 K was 559 m2 g-1 with the pore size ranging from 1.0 to 2.5 nm calculated by nonlocal density functional theory (NLDFT) method (Supplementary Table 3, Supplementary Fig. 5). The BD-COF has a BET surface area of 519 m2 g-1 with pore size distribution from 1.0 to 2.5 nm (Supplementary Table 3, Supplementary Fig. 5). The scanning electron microscopy (SEM) images of both CN-COF and BD-COF depicted the rod-like morphology (Supplementary Fig. 6). After ball milling, CN-CON and BD-CON were obtained. The TEM images of CN-CON and BD-CON showed almost identical nanosheet morphology with lateral sizes close to 500 nm (Supplementary Fig. 7). The periodic framework structure of CN-CON was visualized by high-resolution transmission electron microscopy (HRTEM). The hexagonal straight pore feature of CN-CON can be clearly observed (Fig. 2a). Fourier-filtered image of enlarged red square area showed that the interplanar spacing of (100) lattice plane was calculated to be 2.3 nm, consistent with the pore size by N2 sorption isotherm and simulated eclipsed model (Fig. 2b). The atomic force microscopy (AFM) images of both CONs drop-coated onto mica from ethanol suspensions also displayed nanosheet morphology with thickness ranging from 2 to 3 nm, corresponding to the existence of only ~6-9 COF layers (Fig. 2c). The PXRD patterns and pore size distributions of CN-CON and BD-CON after ball milling were identical to pristine COFs, indicating that COFs can endure the high-energy ball milling process because of the high thermostability (Supplementary Fig. 4). The BET surface area of CONs decreased as compared with the pristine COFs due to the exfoliated ultra-thin nanosheet effect. Both CONs can be well dispersed in water to form colloid solution as verified by the conspicuous Tyndall effect, and colloid solution remains stable even over 4 months (Fig. 2e, Supplementary Fig. 8). The dynamic light scattering (DLS) measurement showed the dominant colloid size distribution at ~ 530 nm for both CONs coincided with corresponding TEM results (Fig. 2e, Supplementary Fig. 7 and 8). All evidences above proved that the 2D CONs were successfully obtained by mechanical exfoliation of bulk COFs.
The chemical stability of CN-COF under harsh condition, especially under prolonged light irradiation, is the prerequisite for photocatalysis application. To our delight, CN-COF could withstand different harsh conditions, such as upon 3-day immersion in THF, DMSO, DMF, 3 M aqueous HCl, and 3 M NaOH, as evidenced by the almost identical PXRD patterns and FT-IR spectra before and after treatments (Supplementary Fig. 9). Even after 3-day Xenon lamp irradiation in water, no obvious changes in PXRD pattern or FT-IR spectrum could be observed for CN-COF (Supplementary Fig. 9). The above results revealed the excellent chemical stability and photo-stability of CN-COF.
Photocatalytic hydrogen evolution reaction. The UV-vis diffusion reflectance spectroscopy (DRS) spectrum of CN-COF exhibited an absorption band with edges at 627 nm, implying the obvious visible light responsive nature (Fig. 3a). In comparison with BD-COF, CN-COF presented apparent red shifts. It is well established that the absorption edge of π conjugated systems will red shift with the incorporation of chromophore in frameworks. The optical band gaps of CN-CON and BD-COF were calculated to be 2.17 and 2.24 eV by Tauc plots, respectively (Supplementary Fig. 10). This result indicated that cyano groups could narrow the band gap, as a result, increasing light trapping ability. Furthermore, the conduction band (CB) and valence band (VB) positions of the two COFs estimated by electrochemical Mott-Schotty plots and their optical band gaps were enough for both proton reduction and water oxidation reaction (Supplementary Fig. 10). The positive slope of Mott-Schottky plots indicated typical n-type semiconductor feature for both COFs. CN-COF exhibited more negative CB position than that of BD-COF, implying stronger driving force for proton reduction.
We next evaluated the activity of CN-COF for HER under visible light (λ > 420 nm). The loading amount of Pt and different types of sacrificial reagent (sodium ascorbate, Na2SO3 or TEOA) were screened first (Supplementary Table 4, Supplementary Fig. 11). 1 wt% Pt with 0.1 M ascorbic acid as a sacrificial reagent was the optimized reaction conditions for CN-COF. Notably, no obvious change in Pt size was observed with the Pt loading used for the test (Supplementary Fig. 11). The volcano curve of HER rate and Pt loading implied the combined effect of the electron trapping for proton reduction and light absorbance by Pt. Under optimized conditions, the average H2 evolution rate of CN-COF and BD-COF was 1217 and 39.5 μmol h-1 (Fig. 3b), respectively. Thirty-folds increasing of H2 evolution rate of CN-COF in comparison with BD-COF demonstrated the promotion effect of cyano groups in photocatalytic HER.
Amazingly, the photocatalytic HER rate of CN-CON was as high as 2684 μmol h-1 (normalized by mass: 134200 μmol g-1 h-1), more than twice that of CN-COF. The BD-CON also showed an increased hydrogen production rate (159 µmol h-1), which was 4 times higher than that of pristine bulk COF. The enhanced hydrogen production rate of the several-layered nanosheet as compared with the bulk COFs was related with the short migration distance of photogenerated charge carriers and also more exposing reaction surface.[51,52] Deuterium isotope experiments were carried out using D2O, and the evaluated gases were detected by mass spectrometry (MS). Nearly all mass-to-charge contributions are D2, indicating that the produced H2 was indeed from water molecules (Supplementary Fig. 12). The AQE as high as 82.6% at 450 nm (Fig. 3c) was achieved for CN-CON. The AQE decreased as the wavelength of the irradiation light increasing, identical to the light adsorption properties of CN-CON. Notably, the AQE as high as 4.2% could still be obtained even irradiated with 650 nm red light, implying the efficiency of CN-CON in photocatalytic HER. In the literature, the photocatalytic activity was always compared using H2 evolution rate normalized by the mass of the photocatalyst. In fact, this is not a right way due to the fact that the H2 evolution amounts do not increase linearly with the content of photocatalyst and AQE is a suitable parameter for comparing the photocatalytic activity, as pointed out by Prof. C. Li and Prof. K. Domen recently. We summarized the reported AQE and photocatalytic activity of COFs and polymers in H2 evolution reaction (Fig. 3d, Supplementary Table 5 and 6).[32-39,54-66] As compared with the reported COF/polymer-based photocatalyst, CN-CON showed the highest AQE as far as we know, which was in a comparable level to that of previously reported star inorganic Pt-PdS/CdS photocatalyst. Even comparing with the H2 evolution rate normalized by the mass, CN-CON is suprior to most of the COFs/polymer-based photocatalyst.
The long-term cycling experiment with CN-CON as model catalyst showed no obvious decline in H2 production rate for more than 24 h (Fig. 3e) and the PXRD pattern and FT-IR spectrum of CN-CON remained almost the same before and after photocatalysis (Supplementary Fig. 13), signifying the excellent stability of CN-CON for photocatalysis. We also drop-casted Pt-CN-CON colloid solution onto a glass support (size of 1 cm × 6 cm, Supplementary Fig. 14). Hydrogen bubbles could be clearly observed over CN-CON film under visible light irraditaion for 10 h (Supplementary Fig. 14, Supplementory Video). The avergae HER rate of CN-CON film can reach to 292 mmol m-2, much higher than the reported COF film.
Incidentally, the photocatalytic oxygen evolution reaction (OER) was also performed here. To our delight, CN-CON could catalyze photocatalytic OER to afford OER rate of 1.933 μmol h-1 with CoxOy as co-catalyst and AgNO3 as electron sacrificial reagent (Supplementary Fig. 15). The 18O-labeled water experiment confirmed that the oxygen was sourced from water. The reason of relatively low OER rate may be due to the less positive valence band position of CN-CON (1.28 eV vs water oxidizing potentials 1.23 eV, pH = 0) and the sluggish four-electron transfer kinetic process of oxygen generation.
Mechanistic studies to uncover the role of CN. CN-CON afforded much higher activity than BD-CON though the two CONs have similar BET surface area, pore size and topology structure. The above electrochemical and optical characterization results revealed that the existence of CN would lead to an increase of light trapping and a negative shift of CB position. However, these thermodynamic properties did not allow for any insight into the exciton and charge carriers properties, which are important for understanding the photocatalysis process.
To understand the exciton properties of CONs, temperature-dependent photoluminescence (PL) measurement was carried out to measure the exciton binding energies. The integrated PL intensity of both CONs decreased with increasing temperature from 77 K to 253 K, which can be mainly attributed to the thermally activated nonradiative recombination process (Fig. 4a, 4c).[32,41,67] Further, based on a simple model, the temperature dependent PL intensity of two CONs can be expressed by the following equation:
where I0 is the intensity at 0 K, Eb is the binding energy, A is a proportional constant and kB is the Boltzmann constant.[67,68] By fitting the experimental data, the exciton binding energies of CN-CON and BD-CON were estimated to be 31.2 and 44.2 meV, respectively (Fig. 4b, 4d), demonstrating that the excitons of CN-CON were more prone to dissociation than that of BD-CON, and thus improved the ratios of free charge carriers for CN-CON and contributed to its high photocatalytic activity.
Further, femtosecond transient absorption (fs-TA) measurements were applied to investigate in photocatalysis difference between the two COFs. First, a typical colloidal CN-CON sample was excited using a 400 nm pump pulse, and the TA spectra was acquired with a broadband probe pulse (Fig. 5a). The spectra exhibited a broad negative bleaching signal in the range of 420-540 nm assigned to the ground state bleach (GSB), which indeed coincided with the steady-state absorption spectra (Supplementary Fig. 16). In addition, the spectra were also characterized by a weak and broad positive signal (from 550 to 750 nm) (Fig. 5a). This positive signal could be attribute to trapped carriers (so-called “polarons” in polymers), because its formation was complementary to the decay of the GSB signal within 0.8 ps (Fig. 5b). The trapped carriers could be further assigned specifically as trapped holes on the basis of their rapid decay in the presence of the hole scavenger AA. As shown in Supplementary Fig. 17, these trapped holes were transferred to AA in < 2 ps. Following the initial hole trapping process, the holes and electrons recombined slowly, leading to the simultaneous decay of the hole signal and a negative broad feature centered at ~560 nm. The latter was likely assignable to the stimulated emission of the trapped exciton, as its spectral feature coincided with the steady-state PL spectra (Supplementary Fig. 18).
The lifetime of charge carriers was studied by TA kinetics of CONs. As shown in Fig. 5c, the decay curves for trapped hole of two CONs revealed significantly different lifetimes. The lifetime of CN-CON (14.2 ± 2.3 ps) was three times longer than that of BD-CON (4.3 ± 0.6 ps). The longer charge carrier lifetime will decrease the probability of electron-hole recombination which is a competitive and detrimental process in real photocatalysis system. Notably, after quenching the holes with AA, a new broad negative signal ranged from 500 to 750 nm clearly emerged within ns time scale, attributed to the generation of long-lived free electrons (Supplementary Fig. 19). Fitting kinetics revealed that long-lived electrons of CN-CON had a much slower rate constant (ke = 1.3 ns-1) than that of BD-CON (ke = 5.3 ns-1) (Fig. 5d), which correlated well with H2 production activity. All the observations indicated that the existence of CN can effectively extend the life-time of charge carriers and eventually increase the photocatalytic hydrogen evolution activity.