Nano-COF synthesis and characterizationTFP-BpyD and TFP-BD nano-COF were selected for this study because of their high stability, irreversible enol-to-keto tautomerization, and good photocatalytic activity29. The synthesis procedures for TFP-BpyD and TFP-BD nano-COF were based on a previous report (Fig. 1a)28. The amine and aldehyde monomers were first loaded into CTAB/SDS mixed micelles to form separate homogeneous micellar solutions. By mixing these amine and aldehyde solutions with an acetic acid catalyst, the reaction mixture turned orange, indicating an imine condensation reaction. Unlike solvothermal syntheses and sonochemical syntheses30, the reaction mixture remained clear and homogeneous with no apparent precipitation. A pronounced Willis-Tyndall scattering behavior from the two reaction mixtures confirms the presence of COF colloidal particles (Supplementary Fig. 1). To further verify the formation of nano-COF colloids, UV-visible absorption spectra and solution nuclear magnetic resonance (NMR) spectra were obtained directly for the colloidal solutions after 3 days. The 1H NMR spectra of TFP-BpyD and TFP-BD nano-COF revealed that the peak (δ ≈ 9.8 ppm) belonging to CHO hydrogens and the peaks (δ ≈ 5.0–8.0 ppm) attributed to NH2 and aromatic hydrogens disappeared after the reaction (Supplementary Figs. 2, 3). UV-visible spectra showed that the characteristic peaks corresponding to the TFP, BpyD and BD monomers disappeared too, and new peaks were formed at 415 and 428 nm for TFP-BpyD and TFP-BD nano-COF, respectively, further proving that the starting materials were consumed, and that condensed, polymeric imine species were formed (Supplementary Fig. 4).Fig. 1: Synthesis of nano-COFs and their morphologies.The synthetic routes for TFP-BpyD and TFP-BD nano-COF (a). Cryo-TEM and AFM images of TFP-BpyD nano-COF (b, c) and TFP-BD nano-COF (d, e). Inserts of c and e: AFM height images with the corresponding cross-section analysis along the indicated dashed blue lines. Source data for (c) and (e) are provided as a Source Data file.To further characterize the nano-COFs, ammonia and ethanol were added to the reaction mixtures, which precipitated the materials as insoluble yellow powders. Elemental analysis (EA) for the solid nano-COFs corroborates their structure (Supplementary Table 1). Fourier-transform infrared (FT-IR) spectra revealed the appearance of stretching bands at around 1570 cm−1 and 1286 cm−1, which are assigned to the C = C and CH–NH bonds (Supplementary Fig. 5). 13C CP-MAS solid-state NMR spectra showed clear signals near 184 and 148 ppm, corresponding to the carbonyl carbons and –CH-NH– (Supplementary Fig. 6), respectively. The full X-ray photoelectron spectroscopy (XPS) survey spectra confirmed the presence of C, N, and O elements of both nano-COFs (Supplementary Fig. 7a). High-resolution XPS spectra of C 1s for both nano-COFs show the presence of ketoenamine C = O (Supplementary Fig. 7b). All above results indicate the formation of β-ketoenamine linkage in the two nano-COFs. High-resolution XPS spectra of N 1 s for TFP-BpyD nano-COF shows two individual peaks at 399.1 and 397.9 eV, attributable to the keto-enamine N (−C − HN − C−) and pyridinic N (−C = N − C−), respectively (Supplementary Fig. 7c), where the more electronegative pyridinic N would be favorable for photocatalysis. Powder X-ray diffraction (PXRD) patterns of isolated nano-COFs showed no obvious peak from the monomers, further supporting the complete consumption of the amine and aldehyde monomers (Supplementary Fig. 8). PXRD suggested a slightly higher degree of structural order in TFP-BpyD nano-COF than TFP-BD nano-COF, possibly due to higher reversibility in the BpyD linker, although neither COF showed much evidence of long-range order and crystallinity in this precipitated, bulk form. A degree of permanent porosity was confirmed by nitrogen adsorption-desorption experiments at 77 K for these bulk precipitated powders (Supplementary Figs. 9, 10). Both nano-COFs exhibited a rapid uptake at low pressure P/Po < 0.1 with a sorption profile best described as a type IV isotherm, characteristic of microporous / mesoporous material. The Brunauer-Emmett-Teller (BET) surface areas of TFP-BpyD and TFP-BD nano-COF were calculated to be 113 m2 g−1 and 598 m2 g−1, respectively. Here, both nano-COFs showed lower crystallinity and porosity compared with bulk synthesized COFs. This stems from the lower reversibility of the covalent bond formation that is associated with enol-to-keto tautomerization, as well as the reduced crystallite sizes, and reduced long-range order29. Thermogravimetric analysis (TGA) indicated that these nano-COFs were thermally stable up under nitrogen to around 350 °C (Supplementary Fig. 11).The morphology of the two nano-COFs was characterized using SEM and TEM. These images suggested that TFP-BpyD and TFP-BD nano-COF had nanofiber and nanosphere morphologies, respectively (Supplementary Figs. 12–15). However, aggregation in the precipitated powders made it difficult to discern the real size of the primary particles for the two nano-COFs. We therefore used cryo-TEM to observe the morphology of the two nano-COFs closer to their native state in solution. TFP-BpyD nano-COF showed a nanofiber morphology with fiber lengths around 300 nm and fiber diameters of 10–20 nm (Fig. 1b and Supplementary Fig. 16a). Cryo-TEM suggested that TFP-BD nano-COF had a ‘spherical’ morphology with a particle diameter of around 30 nm (Fig. 1d and Supplementary Fig. 16b). AFM images confirmed this particle size, but the average heights of the two nano-COFs as measured by AFM were 8 nm and 4 nm, respectively, suggesting that nanoribbons (i.e., flat fibers) and nanosheets (i.e., disks) might be a more accurate description of the two particle morphologies (Fig. 1c, e and Supplementary Fig. 17). This morphology might stem from the strong tendency of these COFs to formed two-dimensional layered packings, and can explain the low crystallinity due to the lack of longer-range π–π stacking. The reason for the difference in morphologies between the two nano-COFs, which have similar chemical structures (Fig. 1a), is unclear.Photocatalytic hydrogen evolution reactionTo study the effect of particle size on photocatalytic activity, the bulk COF analogues of these two nano-COFs, TFP-BpyD COF and TFP-BD COF, were synthesized using our sonochemical method30 as a control. These bulk COFs were characterized by EA (Supplementary Table 1), FT-IR spectroscopy, SEM, nitrogen adsorption-desorption measurements, UV-visible absorption spectroscopy, PXRD and TGA measurements (see Supplementary Information). The bulk materials showed FT-IR spectra that were similar to the precipitated TFP-BpyD and TFP-BD nano-COFs as well as similar thermal stabilities (Supplementary Figs. 11 and 18), but the bulk materials showed much higher long-range order (c.f., Supplementary Figs. 19, 20) and, as a result, significantly higher surface areas (1149 m2 g−1 and 630 m2 g−1, respectively, Supplementary Figs. 21, 22). SEM images showed micron-scale aggregates for the bulk COFs (Supplementary Figs. 23, 24). Notably, when dispersing nano-COFs and bulk COF materials at the same concentrations in aqueous solution, the nano-COFs showed far better dispersion and more effective light-harvesting than the comparable bulk COFs. As shown in Fig. 2a and Supplementary Fig. 25, both nano-COFs show low light transmittances (10–25 %) in the wavelength range 350–550 nm, while the corresponding bulk COF dispersions absorb much less light in this range (70–90% transmission); that is, the two nano-COFs are far more effective at light harvesting at these concentrations. Compared to bulk COFs, the maximum absorption band of the nano-COFs showed a blue shift of around 100 nm. This can be attributed to the decreased π-conjugation in the ab plane, as well as reduced stacking in c-axis.Fig. 2: Photocatalytic hydrogen evolution for bulk COFs and nano-COFs, and inverse concentration dependence for TFP-BpyD nano-COF.a Transmittance spectra of TFP-BpyD nano-COF (blue) with bulk TFP-BpyD COF (pink) at the same concentration (11.4 μg/mL); see also Supplementary Fig. 25. b Comparison of photocatalytic H2 evolution of TFP-BpyD nano-COF (blue squares) with bulk TFP-BpyD COF (pink circles) over 5 h (conditions for nano-COF: 0.25 mL TFP-BpyD nano-COF in 24.75 water, diluted H2PtCl6 solution as a platinum precursor (15 wt.% loading), 880 mg AA (overall concentration is 0.2 M), λ  >  420 nm; Conditions for bulk COFs: 5 mg of bulk COF catalyst in water, diluted H2PtCl6 solution as a platinum precursor (4 wt.% loading), 0.2 M AA (25 mL), λ  >  420 nm). c Sacrificial photocatalytic H2 evolution using different concentrations of TFP-BpyD nano-COF. Conditions were as follows: TFP-BpyD nano-COF (1.0, 0.5, 0.1 and 0.05 mL) in 0.1 M AA solution (5 mL in final), 5 wt.% Pt loading, 1 h illumination (Oriel Solar Simulator, 1.0 sun). d Effect of varying Pt and AA concentration for TFP-BpyD nano-COF. For Pt concentration series, TFP-BpyD nano-COF (0.05 mL) in 0.1 M AA solution (5 mL total volume), 5, 10, or 15 wt.% Pt loading, 2 h illumination; For AA concentration series, TFP-BpyD nano-COF (0.05 mL) in 0.1, 0.2 or 0.5 M AA solution (5 mL total volume), 15 wt.% Pt loading, 2 h illumination. Blue bars, H2 evolution rate; pink bars, H2 evolution amount. e Plot showing sacrificial photocatalytic H2 evolution as a function of time for TFP-BpyD nano-COF over 42 h (0.25 mL TFP-BpyD nano-COF in 24.75 water, diluted H2PtCl6 solution as a platinum precursor (15 wt.% loading), 880 mg AA (overall concentration is 0.2 M), λ  >  420 nm). f Overlay of the UV/vis absorption spectrum (blue curves) of TFP-BpyD nano-COF with external quantum efficiency (EQE, pink dots) at three different incident light wavelengths. Source data for Fig. 2 are provided as a Source Data file.Photocatalytic hydrogen production experiments were carried out with both nano-COFs and bulk COFs as the photocatalysts, using ascorbic acid (AA) as the sacrificial agent and platinum as a cocatalyst. Here, we selected AA as a sacrificial electron donor because of its strong reducibility for ketoenamine COFs, as evidenced by previous work18. As shown in Fig. 2b, TFP-BpyD bulk COF powder (5 mg) produced 23.2 μmol H2 in 5 h under visible light, with a mass-normalized hydrogen evolution rate (HER) of 0.89 mmol g−1 h−1 (4.5 μmol h−1). For the corresponding nano-COF, the same photocatalytic experiment was carried out by diluting a stock solution of the nano-COF colloid with water. Remarkably, by using 0.25 mL of the TFP-BpyD nano-COF stock colloid solution, which contains just 0.085 mg of the COF, we generated 162.3 μmol H2 in 5 h. The amount of H2 evolved for the TFP-BpyD nano-COF was 7-fold higher than for the bulk TFP-BpyD material in absolute terms, while the enhancement in the mass-normalized HER was a factor of 440, achieving an average HER of 392.0 mmol g−1 h−1 (33.3 μmol h−1). Similarly, the TFP-BD nano-COF showed enhanced photocatalytic H2 production compared with bulk TFP-BD COF (Supplementary Fig. 26) with an average HER of 183.0 mmol h−1 g−1 (15.6 μmol h−1). Even though the bulk COFs showed improved crystallinity, higher levels of porosity, and a larger thermodynamic driving force for proton reduction (Supplementary Fig. 35), all of which are important factors for photocatalytic efficiency, the nano-COFs exhibited greatly enhanced hydrogen evolution because of their much better light harvesting characteristics.We sought next to study the effect of the colloid concentration on the photocatalytic performance of the TFP-BpyD nano-COF. As shown in Fig. 2c, decreasing the volume of TFP-BpyD nano-COF added from 1 to 0.5, 0.1 and 0.05 mL (corresponding to nano-COF concentrations of 68.6, 34.3, 6.86 and 3.43 μg/ mL, respectively) led to a roughly 10-fold increase in the absolute amount of hydrogen produced, form 0.4 mmol to 3.5 mmol. This differs markedly from previous studies involving bulk COF and polymer catalysts31,32, where the amount hydrogen evolved barely changes when increasing the concentration of photocatalyst up to some saturation value. We discuss this counterintuitive ‘reverse-concentration’ phenomenon below. We also optimized the amount of Pt and the AA sacrificial electron donor (Fig. 2d); the best condition was found to be 0.00343 mg/ mL TFP-BpyD nano-COF (that is, the original colloidal solution diluted 100 times) with 0.2 M AA and 15 wt. % Pt loading with respect to the COF. The longer-term photocatalytic performance of this optimized TFP-BpyD nano-COF system was investigated. As shown in Fig. 2e and Supplementary Fig. 36, TFP-BpyD nano-COF showed sustained H2 productions up to 42 h in the presence of AA as a sacrificial electron donor and Pt as a cocatalyst, albeit with a decrease in rate after around 20 h, giving an average HER of 392.0 mmol h–1 g–1 over the first 5 h under visible light (λ > 420 nm, 300 W Xenon lamp). To our knowledge, this is one of the highest mass-normalized sacrificial photocatalytic H2 evolution rates reported for an organic photocatalyst so far (Supplementary Table 3). The external quantum efficiency (EQE) was measured at different wavelengths to evaluate the photocatalytic H2 production performance. The EQE was determined to be 10.0% and 5.5% at 420 nm for TFP-BpyD and TFP-BD nano-COF, respectively, and the EQE followed the absorption spectrum, supporting a photoinduced H2 evolution process (Fig. 2f and Supplementary Fig. 37). After photocatalysis, the TFP-BpyD and TFP-BD nano-COF materials were characterized by TEM, STEM, SEM, NMR and FT-IR. TEM image showed the retention of nanoscale morphology and Pt nanoparticle were observed to be distributed uniformly on the nano-COFs (Supplementary Figs. 38, 39). HAADF-STEM images and elemental mappings (Supplementary Figs. 40, 41) also indicated good Pt cocatalyst dispersion on the nano-COFs. FTIR and NMR characterizations indicated no significant changes after photocatalysis (Supplementary Figs. 42, 43), while TEM and SEM (Supplementary Fig. 44) analysis suggested some aggregation of the COF crystallites. We therefore consider the aggregation of nano-COFs to be the main reason for nonlinearity of the photocatalytic activity after 20 h, rather than the destruction of the COF skeleton. The depletion of AA and/or the accumulation of AA degradation products may also hinder the HER, as discussed in previous studies33,34.Mechanistic studyTo better understand the reverse concentration dependence for H2 evolution for the TFP-BpyD nano-COF (Fig. 2c), we measured UV-vis absorption, photoluminescence spectra (PL) and transient absorption spectra (TAS) for nano-COF aqueous solutions at different colloid concentrations. The UV-vis absorption spectra showed a concentration-dependent absorption for nano-COFs, and the position and shape of the absorption band were unchanged at these different concentrations (Supplementary Figs. 45–48), suggesting that the nano-COF was in a dispersed state without any apparent particle aggregation over the concentration ranges studied here. Transmittance spectra of TFP-BpyD nano-COF indicated increased light-harvesting when increasing the amount of nano-COF solution added from 0.01 to 1 mL (in a 5 mL total volume). When 0.5 mL and 1.0 mL of these stock TFP-BpyD nano-COF solution was added, the light absorption was saturated in the range of 350 to 550 nm, indicating total light absorbance (Fig. 3a).Fig. 3: Photophysical properties of TFP-BpyD nano-COF solutions at various concentrations.a Transmittance spectra of TFP-BpyD nano-COF at different concentrations (conditions: 0.01, 0.05, 0.10, 0.50 and 1.00 mL of the stock TFP-BpyD nano-COF solutions were diluted with water; the total solution volume is 5 mL). b Photoluminescence (PL) spectra of TFP-BpyD nano-COF with different concentrations (conditions: 0.01, 0.05, 0.10, 0.25, 0.50 and 1.00 mL of the stock TFP-BpyD nano-COF solutions were diluted with water; the total solution volume is 5 mL). c PL spectra recorded in the presence of AA (0.10 and 1.00 mL of the TFP-BpyD nano-COF stock solutions were diluted by 0.2 M AA, open symbols) and Pt (0.10 and 1.00 mL of the TFP-BpyD nano-COF stock solutions were diluted by water with Pt co-catalyst at 15 wt. % loading, half-closed symbols). The total solution volume is 5 mL. d TCSPC experiment for TFP-BpyD nano-COF at different concentrations in water (5 mL); the volumes refer to the amount of stock colloid solution added. 0.01, 0.05, 0.10, 0.25, 0.50 and 1.00 mL are represented by black stars, green diamonds, blue lower triangles, purple upper triangles, pink circles and yellow squares, respectively. Source data for Fig. 3 are provided as a Source Data file.When we used less than 0.1 mL of TFP-BpyD nano-COF of the stock solution (5 mL total volume), an intense emission peak was observed at 421 nm upon excitation at 340 nm (Fig. 3b). The intensity of this emission peak showed a maximum at 0.1 mL. Increasing the concentration of the nano-COF beyond 0.1 mL to 0.25 mL or higher (Fig. 3b) caused significant PL quenching and a dramatic reduction in the intensity of this emission peak. Also, at higher TFP-BpyD nano-COF concentrations (0.25 mL), a dual-emission was observed with an additional emission band emerging at 525 nm (E1 @ 421 nm, E2 @ 525 nm; Fig. 3b). The different lifetimes of E1 and E2 (Supplementary Fig. 49) and the redshift of the excitation spectra (Supplementary Fig. 50) indicated two different dynamic processes. We assign E1 to singlet emission from an individual nano-COF particle based on the small Stokes-shift, while E2 with a larger Stokes-shift can be assigned to a radiative decay process due to the charge-transfer (CT) that occurs between the nano-COF particles at higher concentrations. We believe that this PL change as a function of nano-COF concentration is analogous to the small molecule pyrene, which forms an excimer with a bathochromic shift in its emission peak and, likewise, shows emission quenching at higher concentrations (Supplementary Fig. 51)35,36,37. Typically, the formation of excimer in small molecules is promoted by monomer density (here the concentration of nano-COF), because of the dependence on a bimolecular interaction. Also, excimers usually show longer emission wavelengths than the excited monomer’s emission (here the emission of a single nano-COF particle)35. As such, it seems that this nano-COF exhibits a molecule-like excitonic nature, showing CT character at high particle concentrations.Both emission bands of E1 and E2 were decreased when further increasing the concentration of TFP-BpyD nano-COF to 0.5 and 1 mL (Fig. 3b). We attribute this to the increased collision rate of the nano-COF particles at higher concentrations, which suppresses the radiative recombination of free charge carriers. The magnitude of the PL quench for E1 was more pronounced than E2, which bears charge-transfer character, indicating that particle collision has a greater impact on the singlet emission (E1). Adding a sacrificial donor (AA) to the TFP-BpyD nano-COF solution gave rise to a similar quenching phenomenon (Fig. 3c and Supplementary Fig. 52). Again, the magnitude for the PL quench with AA for E1 greater than for E2, suggesting that singlet emission is more easily be extracted by AA. TSCPC was also carried out with TFP-BpyD nano-COF for emission at 371 nm. The lifetime for TFP-BpyD nano-COF with low-concentration (0.1 mL, τavg  =  3.04 ns) was estimated to be significantly longer than at higher nano-COF concentrations (0.5 mL, τavg  =  1.39 ns and 1.0 mL, τavg  =  1.18 ns, Fig. 3d and Supplementary Figs. 53–55), supporting an interparticle charge transfer process at high nano-COF concentrations. We note that the corresponding bulk materials did not show an equivalent PL quench phenomenon at higher COF concentrations (Supplementary Figs. 57, 58).Femtosecond (fs) transient absorption (TA) measurements were performed to study the dynamic processes for the TFP-BpyD nano-COF (Fig. 4a and Supplementary Fig. 64). The TFP-BpyD nano-COF exhibited a broadband photoinduced absorption (PIA) signal around 510 nm, independent of the concentration, which could be attributed to polaron pairs originating from individual nano-COF paricles38,39. At lower nano-COF concentration (0.35 mL), this PIA signal was found to decay very rapidly. By comparison, at higher nano-COF concentrations, a long-lifetime process was also observed for the PIA signals (Fig. 4b). This long lifetime may result from increased particle collision rates at high concentrations that induce singlet-singlet annihilation, note that the polaron pairs and separated charges can give rise to similar spectral signatures40,41.Fig. 4: Transient absorption spectra (TAS) for TFP-BpyD nano-COF at different concentrations and mechanistic interpretation.a TAS spectra under the following conditions: 0.35 mL (blue) or 1.0 mL (yellow) TFP-BpyD nano-COF, diluted by water to a total volume is 5 mL. b Life time test of TFP-BpyD nano-COF by TAS measurement. 0.35, 0.50 and 1.00 mL are represented by the color of blue, pink and yellow, respectively. The experimental and fitted data correspond to the symbols and curves. Scheme showing free charge carriers produced for nano-COFs at (c) low concentration and (d) high concentration (ascorbic acid: AA; dehydroascorbic acid: DHA). Source data for Fig. 4a and b are provided as a Source Data file.The reverse concentration-dependent photocatalytic performance can be rationalized by these various observations. At low nano-COF concentration, a greater number of electron-hole pairs (the initial excitation, E1) can be generated and extracted by the sacrificial agent, AA, to promote photocatalytic proton reduction. However, increasing the COF concentration causes increased collision rates between the nano-COF particles, together with a charge-transfer process and singlet-singlet annihilation, which finally produces free charges that are spatially isolated, along with longer lifetimes than at low concentrations (Fig. 4c, d)38. Given that this charge transfer state is barely extracted by AA, we ascribe this loss of the initial excitation state and resulting anabatic charge recombination to be the primary cause for the decreased sacrificial photocatalytic activity. In addition to these effects, at even higher concentrations (>1 mL nano-COF stock solution), the solutions become totally absorbing (Fig. 3a); as such, one would expect the mass-normalized hydrogen evolution rate to decrease in this regime, irrespective of interparticle charge transfer effects.
