Reconstructed covalent organic frameworks

Covalent organic frameworks (COFs) are distinguished from other organic polymers by their crystallinity1–3, but it remains challenging to obtain robust, highly crystalline COFs because the framework-forming reactions are poorly reversible4,5. More reversible chemistry can improve crystallinity6–9, but this typically yields COFs with poor physicochemical stability and limited application scope5. Here we report a general and scalable protocol to prepare robust, highly crystalline imine COFs, based on an unexpected framework reconstruction. In contrast to standard approaches in which monomers are initially randomly aligned, our method involves the pre-organization of monomers using a reversible and removable covalent tether, followed by confined polymerization. This reconstruction route produces reconstructed COFs with greatly enhanced crystallinity and much higher porosity by means of a simple vacuum-free synthetic procedure. The increased crystallinity in the reconstructed COFs improves charge carrier transport, leading to sacrificial photocatalytic hydrogen evolution rates of up to 27.98 mmol h−1 g−1. This nanoconfinement-assisted reconstruction strategy is a step towards programming function in organic materials through atomistic structural control.


Chemical reconstruction (synthesis of RC-COF-1)
A Pyrex tube was charged with 1,3,5-triformylphloroglucinol (21.0 mg, 0.10 mmol), 1,1'-(1,4-5 phenylene)diurea (29.1 mg, 0.15 mmol), NMP (0.8 mL), o-DCB (0.2 mL) and aqueous acetic acid (6 M, 0.1 mL). The mixture was briefly sonicated for 10 seconds, and the tube was then flash-frozen at 77.3 K (liquid nitrogen bath) and degassed by three freeze-pump-thaw cycles before being evacuated to an internal pressure of 100 mtorr. The tube was sealed and heated at 90 °C for 72 h, and afterwards the temperatures was in situ elevated to 110, 120, 130, 150, 160 and 170 °C, respectively, and kept for 10 a further 72 h. The obtained precipitates were isolated by filtration, briefly washed with DMF and acetone to afford solvated COFs. To activate the COFs, solvent exchange was performed with DMF (20 mL × 12), methanol (20 mL × 6), THF (20 mL × 6) and hexane (20 mL × 6), respectively. The material was then dried under vacuum at 60 °C for 12 h to yield the orange, red and dark red solids, respectively. 15 Based on the above results, we further explored the solvent effect on the transformation. In detail, Urea-COF-1 was isolated from reaction system (90 °C, 72 h) then briefly washed with DMF and acetone. Without vacuum dry, the powder (~ 37.2 mg) was then transfer into a Pyrex tube which was charged with solvent of o-DCB (1.0 mL), or NMP (1.0 mL), or glacial acetic acid (1.0 mL), or NMP/H2O (9/1 v/v, 1.0 mL) or H2O (1.0 mL). The mixture was sonicated for 2 min, and the tube was 20 then flash-frozen at 77.3 K (liquid nitrogen bath) and degassed by three freeze-pump-thaw cycles before being evacuated to an internal pressure of 100 mtorr. The tube was sealed and heated at 160 °C for 72 h. Caution: gas could be released from reaction system. The precipitates were isolated by filtration, briefly washed with DMF and acetone to afford the solvated samples. To activate the samples, solvent exchange was performed with DMF (20 mL × 12), methanol (20 mL × 6), THF (20 25 mL × 6) and hexane (20 mL × 6), respectively. The materials were then dried under vacuum at 60 °C for 12 h to yield red and dark red solids, respectively. RC-COF-1 was synthesized by treating Urea-COF-1 with water at 160 °C for 72 h, which generated the product as a dark red solid (23.8 mg, yield 90%). The ~ 10% weight loss for isolation was mostly due to sample handling. Elemental analysis of

Synthesis of Urea-COF-2 (also known as COF-118)
Urea-COF-2 was synthesized according to previously reported procedures with some

Synthesis of RC-COF-2
In a manner similar to the preparation of RC-COF-1, treatment of Urea-COF-2 with water at 160 5 °C for 72 h generated the product as a dark red solid (yield 89%). Elemental analysis of activated sample: Calcd. for C51H48N6O6: C: 72.84; H: 5.75; N: 9.99%; Found: C: 71.17; H: 5.43; N: 8.36%. 10 In a manner similar to the preparation of DP-COF-1 using previously reported procedures 3

Synthesis of DP-COF-3 (also known as TpBD)
In a manner similar to the preparation of DP-COF-1 using previously reported procedures 4 ,  10 In a manner similar to the preparation of DP-COF-1 using previously reported procedures 5
For RC-COF-1, Urea-COF-1 was isolated by filtration, briefly washed with DMF and acetone. 15 The powder was then transfer into a Pyrex tube which was charged with deionized water (1.0 mL). The mixture was sonicated for 2 min, and the tube was then sealed in the air and heated at 160 °C for 72 h. The resulted powder was isolated by filtration, briefly washed with DMF and acetone. To activate the samples, solvent exchange was performed with DMF (20 mL × 12), methanol (20 mL × 6), THF (20 mL × 6) and hexane (20 mL × 6), respectively. The material was then dried under vacuum at 60 20 °C for 12 h to give a dark red solid (24.9 mg, yield 91%).

Transformation of the model compound
To further understand this unexpected crystal-to-crystal transformation, we tested a water-insoluble urea model compound (Fig. 1b) under the same conditions as for the Urea-COF (H2O, 160 °C). A βketoenamine product was obtained, but the isolated yield was much lower (~ 11%). By contrast, 5 decomposition occurred when the model compound was dissolved in an NMP/H2O mixture (9/1 v/v) at 160 °C, as determined by NMR (Supplementary Fig. 1, 2). These results suggest that the nanoconfinement effect and perhaps the porosity and 2D layered structure in the COFs are necessary for the reconstruction reaction, noting that the urea model compound is not porous.  The mixture was sonicated for 2 min, and the tube was then flash-frozen at 77.3 K (liquid nitrogen bath) and degassed by three freeze-pump-thaw cycles before being evacuated to an internal pressure of 100 mtorr. The tube was sealed and heated at 160 °C for 72 h. The precipitate was isolated by filtration and washed with water (50 mL) to afford 2,4,6-tris((phenylamino)methylene) cyclohexane-1,3,5-trione (17.1 mg, yield 11%). 1

Transformation in solution state (NMP/H2O):
A Pyrex tube was charged with (2,4,6-25 trioxocyclohexane-1,3,5-triylidene)tris(methanylylidene)tris(3-phenylurea) (100.0 mg, 0.18 mmol), NMP (1.8 mL) and water (0.2 mL). The mixture was sonicated for 2 min, and the tube was then flashfrozen at 77.3 K (liquid nitrogen bath) and degassed by three freeze-pump-thaw cycles before being evacuated to an internal pressure of 100 mtorr. The tube was sealed and heated at 160 °C for 12 h or 72 h before being cooled to room temperature. The reaction mixture was mixed with water (30 mL). 30 The precipitate was isolated by filtration and washed with water (50 mL). Supplementary Fig. 4 Ion chromatograms reports for standard ammonium ion solution (a) and 10 aqueous solution that was filtered from reaction system (hydrothermal treatment of Urea-COF-1 at 160 °C for 72 hours, b). The experiments were conducted with ICS-3000 system (Thermo Fisher Scientific, USA) equipped with a conductivity detector. 30 mM methanesulfonic acid (MSA) was used as eluent and the column temperature was maintained at 40 °C with a run time of 20 min. The aqueous solution filtered from reaction system was diluted by 500 times. A high concentration (1.574 mg L -1 ) of Supplementary Fig. 5 The evolution of the FTIR spectra with time for as-synthesized Urea-COF-1 by treatment with water at the elevated reaction temperatures of 160 °C (samples were measured after 5 activation).

Section 4 Breadth of the synthetic strategy
Supplementary Fig. 7 Synthetic schemes for RC-COF-2, RC-COF-3, RC-COF-4 reconstructed from 5 the respective Urea-COFs. Besides isocyanates (Urea-COF-2), COFs can also be constructed from arylamine starting materials (Urea-COF-3 and Urea-COF-4), significantly expanding the potential synthetic scope. All urea-COFs can reconstruct into RC-COFs by treatment with water at 160 °C for 72 hours, which represents a generalizable synthetic route for preparing high-crystallinity imine COFs using readily accessible monomers. Supplementary Fig. 8 The top view of eclipsed AA-stacking (a) and staggered AB-stacking (b) models for Urea-COF-1. 5 Supplementary Fig. 9 PXRD patterns for Urea-COF-1: experimental (dark), calculated with fully eclipsed AA-stacking (red) and staggered AB-stacking (blue) models. The eclipsed AA-stacking 10 model yielded a PXRD pattern that was consistent with the experimental profile. 21 Supplementary Fig. 10 The top view of eclipsed AA-stacking (a) and staggered AB-stacking (b) models for Urea-COF-2. 5 Supplementary Fig. 11 PXRD patterns for Urea-COF-2: experimental (dark), calculated with fully eclipsed AA-stacking (red) and staggered AB-stacking (blue) models. The eclipsed AA-stacking model yielded a PXRD pattern that was consistent with the experimental profile. Supplementary Fig. 12 The top view of eclipsed AA-stacking (a) and staggered AB-stacking (b) models for Urea-COF-3. 5 Supplementary Fig. 13 PXRD patterns for Urea-COF-3: experimental (dark), calculated with fully eclipsed AA-stacking (red) and staggered AB-stacking (blue) models. The eclipsed AA-stacking model yielded a PXRD pattern that was consistent with the experimental profile. Supplementary Fig. 14 The top view of eclipsed AA-stacking (a) and staggered AB-stacking (b) models for Urea-COF-4. 5 Supplementary Fig. 15 PXRD patterns for Urea-COF-4: experimental (dark), calculated with fully eclipsed AA-stacking (red) and staggered AB-stacking (blue) models. The eclipsed AA-stacking model yielded a PXRD pattern that was consistent with the experimental profile. Supplementary Fig. 16 The top view of eclipsed AA-stacking (a) and staggered AB-stacking (b) models for RC-COF-1. 5 Supplementary Fig. 17 PXRD patterns for RC-COF-1: experimental (dark), calculated with fully eclipsed AA-stacking (red) and staggered AB-stacking (blue) models. The eclipsed AA-stacking model yielded a PXRD pattern that was consistent with the experimental profile. Supplementary Fig. 18 The top view of eclipsed AA-stacking (a) and staggered AB-stacking (b) models for RC-COF-2. 5 Supplementary Fig. 19 PXRD patterns for RC-COF-2: experimental (dark), calculated with fully eclipsed AA-stacking (red) and staggered AB-stacking (blue) models. The eclipsed AA-stacking model yielded a PXRD pattern that was consistent with the experimental profile. Supplementary Fig. 20 The top view of eclipsed AA-stacking (a) and staggered AB-stacking (b) models for RC-COF-3. 5 Supplementary Fig. 21 PXRD patterns for RC-COF-3: experimental (dark), calculated with fully eclipsed AA-stacking (red) and staggered AB-stacking (blue) models. The eclipsed AA-stacking model yielded a PXRD pattern that was consistent with the experimental profile. Supplementary Fig. 22 The top view of eclipsed AA-stacking (a) and staggered AB-stacking (b) models for RC-COF-4.

Section 8 Control experiments: hydrothermal treatment
To assess the possibility that enhanced crystallinity was the result of hydrothermal cycling of the imine COFs, rather than preorganization, we performed a series of control experiments involving postsynthetic hydrothermal treatment of the four DP-COFs under precisely the same conditions as used for 5 reconstruction (H2O, 160 °C, 72 hours). PXRD was used to assess the crystallinity of the frameworks before and after hydrothermal treatment. As shown in Supplementary Fig. 38, below, PXRD suggested no change in crystallinity for the DP-COFs, apart from DP-COF-4, which showed a notable decrease in crystallinity. Likewise, no improvement in surface area was observed after hydrothermal treatment for DP-COF-2, DP-COF-3, or DP-COF-4. The surface area for DP-COF-1 was somewhat improved 10 by hydrothermal treatment, but it was still around half the value measured for RC-COF-1. These data show that the improvement in crystallinity in the RC-COFs is not a result of simple hydrothermal cycling or annealing of the resultant imine COFs. The greatly improved levels of crystallinity and porosity are only obtained via reconstruction of the highly crystalline urea COFs, and the associated preorganization and nanoconfined polymerization. Supplementary Fig. 38 Comparison of PXRD patterns (a, c, e, g) and nitrogen adsorption isotherms / BET surface areas (b, d, f, h) for RC-COFs along with DP-COFs, both as-synthesized and after hydrothermal treatment (H2O, 160 °C, 72 hours). The high levels of crystallinity and porosity observed 5 in the RC-COFs are not attained by hydrothermal treatment of DP-COFs; rather, it is a result of preorganization in the urea COF precursors, coupled with nanoconfined reconstruction.

Section 11 Optical and electronic properties
Highly ordered COF lattices should allow for efficient photo-generated charge migration with fewer trapping sites. This is desirable in optoelectronic applications and in photocatalysis, particularly when coupled with high physical surface areas. First, the UV-visible reflectance spectra of the COFs were 5 measured in the solid state to determine their optical gaps (Extended Data Fig. 6e). RC-COF-1 and DP-COF-1 showed a broadened visible-light absorption response compared to Urea-COF-1. Kubelka-Munk analysis gave an optical band gap of ∼ 2.12 eV for Urea-COF-1, and narrower optical gaps of ∼ 1.87 eV and ∼ 1.90 eV for RC-COF-1 and DP-COF-1, respectively (Supplementary Fig. 44). This is attributed to the enhanced conjugation backbone and intramolecular charge transfer from the 10 phenylenediamine donor to the triketone acceptor. Room temperature steady-state photoluminescence (PL) spectroscopy was performed on these COFs with an excitation wavelength of 400 nm. It showed a decreased intensity at ∼ 620 nm for RC-COF-1 compared to DP-COF-1 ( Supplementary Fig. 45), suggesting a larger barrier for charge recombination in the highly crystalline reconstructed framework 6 . Time-correlated single-photon counting (TCSPC) was used to estimate the excited-state lifetimes 15 ( Supplementary Fig. 46, Table 2), and RC-COF-1 exhibited a longer average weighted lifetime (τavg = 3.93 ns) with respect to DP-COF-1 (τavg = 1.55 ns) in aqueous suspensions. Electron paramagnetic resonance (EPR) studies were also used to explore the electronic band structures (Extended Data Fig.  6a). A single Lorentzian line centered at a g value of 2.006 was observed for RC-COF-1 which intensified dramatically upon light excitation, suggesting an effective light-induced charge carrier 20 generation 7 , whereas DP-COF-1 displayed much lower intensity under same test conditions. The charge transfer in these COFs was also investigated by photo-electrochemistry. Electrochemical impedance spectroscopies (EIS) were measured with COF films under both dark and light conditions. The Nyquist plots for RC-COF-1 showed a semicircle with a smaller diameter than that was observed for DP-COF-1, suggesting an improved interfacial charge transport ( Supplementary Fig. 47). 25 Photocurrent measurements showed that RC-COF-1 produced a significantly enhanced photocurrent compared to its semi-crystalline counterpart, DP-COF-1, indicating more efficient separation of photogenerated charge carriers (Extended Data Fig. 6b). Hence, while RC-COF-1 is significantly more porous than DP-COF-1, and has a much lower bulk density, it nonetheless exhibits markedly better charge carrier transport. We attribute this to its greatly improved crystallinity.   Supplementary Fig. 47 Nyquist plots from electrochemical impedance spectroscopy for RC-COF-1, DP-COF-1 and Urea-COF-1 in the dark (a) and under 300 W Xe lamp irradiation (λ > 420 nm) (b) 5 with a bias potential of -0.35 V vs Ag/AgCl reference electrode and Pt as a counter electrode.

Section 12 Photocatalytic hydrogen evolution experiments
We tested both DP-COF-1 and RC-COF-1 for sacrificial photocatalytic hydrogen evolution using platinum (Pt) as the co-catalyst. The optimized catalytic condition was 3 wt.% Pt loading with ascorbic acid as the sacrificial agent ( Supplementary Fig. 48, 49). Urea-COF-1 showed an average hydrogen 5 evolution rate (HER) of 2.08 mmol h -1 g -1 during 5 h visible-light photolysis (λ > 420 nm), and DP-COF-1 showed an average HER of 7.04 mmol h -1 g -1 (Extended Data Fig. 6c). RC-COF-1 exhibited an average HER of 27.98 mmol h -1 g -1 ; that is, four times higher than DP-COF-1, despite having the same chemical formula. This activity was also twice as high as a state-of-art COF for this reaction, FS-COF 8 , as measured under identical conditions (12.38 mmol h -1 g -1 ). The mass-normalized HER for RC-COF-1 is among the highest reported for a COF photocatalyst (Supplementary Table 4). Control experiments were performed by removing either the light irradiation, the COF, or the sacrificial agent, and no hydrogen evolution was observed, supporting a sacrificial photocatalytic process mediated by the COF (Supplementary Fig. 50). The RC-COF-1 material also showed good reproducibility across different synthetic batches (Extended Data Fig. 6d), reinforcing the reliability of this reconstruction synthesis 15 route. The external quantum efficiencies (EQEs) of RC-COF-1 were estimated to be 6.39% at 420 nm, 5.92% at 490 nm, 5.20% at 515 nm, and 1.62% at 595 nm, respectively (Extended Data Fig. 6e). By comparison, DP-COF-1 exhibited a much-lower EQEs of 1.97%, 1.61%, 1.37%, and 0.54% at the same wavelengths. The quantum efficiencies for the reconstructed COF were three to four times higher 20 than its directly polymerized analog, despite having the same nominal chemical composition. Again, we attribute this to the greatly enhanced crystallinity in RC-COF-1. The highly ordered donor and acceptor molecular columns could enable independent pathways for exciton migration and electron/hole transport, resulting in long-lived charge-separation states 9 . The long-term stability of RC-COF-1 under photocatalytic conditions was monitored for 60 hours, and no obvious decrease in 25 activity was observed during this period (Extended Data Fig. 6f). Fig. 48 Time course of photocatalytic hydrogen evolution for RC-COF-1 under visible light irradiation (λ > 420 nm) with different Pt content as co-catalyst (2.5 mg COF and 0.1 M 5 ascorbic acid as sacrificial agent) (a). Corresponding hydrogen evolution rates (b). Fig. 49 Time course of photocatalytic hydrogen evolution for RC-COF-1 under 10 visible light irradiation (λ > 420 nm) with different sacrificial agents (2.5 mg COF and 3 wt.% Pt as co-catalyst) (a). Corresponding hydrogen evolution rates (b). sacrificial agent) (a). Corresponding hydrogen evolution rates (b). RC-COF-1 exhibited the highest photocatalytic activity in this series of COFs.

Section 13 Post-photocatalysis characterization
The isolated material after photocatalytic experiments was characterized by PXRD, FTIR spectrum, UV-vis reflectance spectrum, and SEM ( Supplementary Fig. 52-55, Extended Data Fig. 7), which showed retention of crystalline structure, demonstrating the high stability of RC-COF-1 as a 5 photocatalyst. Notably, TEM images also showed RC-COF-1 crystallites that were decorated with photo-deposited Pt co-catalyst nanoparticles (Extended Data Fig. 7). The Pt nanoparticles had an average size of 2.5 nm and were uniformly dispersed on the rod-like crystallite of RC-COF-1. The uniform morphology of the reconstructed COF and the good Pt cocatalyst dispersion might also contribute to its enhanced activity. 10 Supplementary Fig. 52 Experimental PXRD patterns for RC-COF-1 before (orange) and after 60 h photocatalysis (blue).
Supplementary Fig. 53 FTIR spectra for RC-COF-1 before (orange) and after 60 h photocatalysis (blue). 5 Supplementary Fig. 54 Solid-state UV-vis spectra for RC-COF-1 before (orange) and after 60 h photocatalysis (blue).  be used as a guide only. The FS-COF data, however, were collected in our laboratories on the same equipment and are therefore directly comparable. RC-COF-1 is much more active than the molecularly engineered FS-COF (Ref. 8), suggesting that the level of crystallinity in COFs might be at least as important for photochemical activity as their chemical composition.