Downsizing Porphyrin Covalent Organic Framework Particles Using Protected Precursors for Electrocatalytic CO2 Reduction

Covalent organic frameworks (COFs) are promising electrocatalyst platforms owing to their designability, porosity, and stability. Recently, COFs with various chemical structures are developed as efficient electrochemical CO2 reduction catalysts. However, controlling the morphology of COF catalysts remains a challenge, which can limit their electrocatalytic performance. Especially, while porphyrin COFs show promising catalytic properties, their particle size is mostly large and uncontrolled because of the severe aggregation of crystallites. In this work, a new synthetic methodology for rationally downsized COF catalyst particles is reported, where a tritylated amine is employed as a novel protected precursor for COF synthesis. Trityl protection provides high solubility to a porphyrin precursor, while its deprotection proceeds in situ under typical COF synthesis conditions. Subsequent homogeneous nucleation and colloidal growth yield smaller COF particles than a conventional synthesis, owing to suppressed crystallite aggregation. The downsized COF particles exhibit superior catalytic performance in electrochemical CO2 reduction, with higher CO production rate and faradaic efficiency compared to conventional COF particles. The improved performance is attributed to the higher contact area with a conductive agent. This study reveals particle size as an important factor for the evaluation of COF electrocatalysts and provides a strategy to control it.


Introduction
Covalent organic frameworks (COFs), organic crystalline 2D or 3D polymers formed by reversible linkages, [1] are promising platforms for heterogeneous catalysis owing to their high chemical and topological designability, tunable porosity, and stability. [2]he exploration of COFs for electrocatalysis has recently gained interest because of the impetus for a transition to renewable energy sources. [3]COF electrocatalysts show promise in important reactions such as hydrogen evolution, oxygen evolution, oxygen reduction, and electrochemical CO 2 reduction reactions (HER, OER, ORR, and eCO 2 RR, respectively).Among them, eCO 2 RR has gained interest as a technology for closing the carbon cycle. [4]Since the first report of COF-catalyzed eCO 2 RR by Chang, Yaghi, and co-workers in 2015, [5] a variety of COFs with different chemical structures have been developed affording remarkable eCO 2 RR activity and selectivity for eCO 2 RR over hydrogen evolution. [6]Among the reported COF eCO 2 RR catalysts, metalated porphyrin-and phthalocyanine-based COFs have demonstrated the superiority of combining high activity and >90% selectivity even in aqueous electrolytes, where hydrogen evolution from water reduction can severely compete with eCO 2 RR. [7]However, despite great success in the chemical structure design, morphology control at the COF micro-and macrostructure level remains largely unexplored.Most COF catalysts are obtained and used as large secondary particles resulting from uncontrolled aggregation and intergrowth of small crystallites, and the effects of COF particle size on eCO 2 RR performance have not been studied.Meanwhile, studies on other organic heterogeneous eCO 2 RR catalysts have revealed that smaller particle sizes improve catalytic performance because charge transport to the active sites becomes more efficient. [8]Therefore, downsizing COF particles is also expected to improve catalytic activity, and such studies on particle size effects are necessary to reveal the intrinsic catalytic properties of COFs.
Controlling the aggregation of COF crystallites represents one of the major challenges in the COF field, and several approaches have been developed over the past decade. [9]13b] However, it is worth mentioning that these strategies for suppressing aggregation usually have as a common prerequisite that all the COF pre-cursors need to be dissolved to form a homogeneous solution.Meanwhile, metalated porphyrins and phthalocyanines, which are typically used for constructing eCO 2 RR-active COFs, have a strong tendency toward intermolecular stacking and coordination.Although the weak intermolecular stacking of some COF precursors in a solution can be utilized for morphology control, [10d] the strong intermolecular interactions of porphyrins and phthalocyanines usually preclude their complete dissolution and thus hamper morphology control.Therefore, the morphology control of eCO 2 RR-active COF particles remains challenging.
Herein, we introduce a novel synthetic approach for controlling the morphology of porphyrin COF catalysts, which utilizes the trityl protection of amino groups (Figure 1).Trityl protection provides excellent solubility to a cobalt porphyrin precursor.Although trityl protection has never been employed in COF chemistry, we show that its deprotection proceeds in situ smoothly under typical COF synthesis conditions.This colloidal deprotection-polycondensation process offers downsized COF particles with less crystallite aggregation compared to the conventional unprotected precursor, without compromising the COF crystallinity and porosity.The downsized COF particles exhibit superior performance in eCO 2 RR compared to conventional large COF particles.Thus, this study provides a novel methodology to downsize COF electrocatalysts and shows its effectiveness in enhancing electrocatalytic performance.

Synthesis and Characterization of a Trityl-Protected Precursor Co(ttpp) and a Conventional Precursor Co(tapp)
Co(tapp) (tapp = 5,10,15,20-tetrakis(4-aminophenyl)porphinato (2−)) is a precursor commonly used to synthesize COFs that catalyze electrochemical CO 2 reduction (Figure 1). [7]We noticed that Co(tapp) only partially dissolves under common COF synthesis conditions, which uses solvent mixtures such as o-dichlorobenzene/n-butanol 1:1(v/v) [5] or odichlorobenzene/benzyl alcohol 1:3(v/v) [14] at a nominal concentration around 10 mm and a temperature of 120 °C.The low solubility of Co(tapp) can be a general problem in the morphology control of its COF derivatives.Therefore, we aimed at improving the solubility of Co(tapp) using a protecting group.
We chose the trityl (Tr) group as a protecting group for Co(tapp).Bulky Tr groups are expected to effectively block the stacking of porphyrins, while they can be introduced and removed under mild conditions.Although tritylamines have never been employed in COF synthesis, they can be deprotected by acetic acid and water, [15] both of which are present in typical COF synthesis conditions.In addition, the deprotection is thermodynamically irreversible in the presence of 1 equiv. of water, yielding TrOH as a by-product. [16]This behavior would promote burst nucleation after an initial deprotection stage, in contrast to the suppression of nucleation by competitive modulators typically used for obtaining larger crystallites. [17]he fourfold Tr-protected Co(tapp), named Co(ttpp) (ttpp = 5,10,15,20-tetrakis(4-(tritylamino)phenyl)porphinato (2−)), is synthesized from free-base TAPP, which is the same precursor to Co(tapp) (Figure 1).TAPP was tritylated with TrBr first, and then metalated with a Co 2+ salt in one pot, requiring no additional purification step.The product identity and purity are confirmed by 1 H nuclear magnetic resonance (NMR), electrospray ionization mass spectrometry (ESI-MS), infrared (IR) absorbance, ultraviolet-visible (UV-vis) absorbance, inductively coupled plasma optical emission spectroscopy (ICP-OES), and scanning electron microscopy with energy dispersive X-ray spectroscopy (SEM-EDX) (Figures S3-S7, Supporting Information page 7), showing no residual TAPP, intermediate TTPP, or Br species (Figures S4 and S5, Supporting Information).
The solubility of Co(ttpp) was tested in various solvents.Co(ttpp) dissolves in a wide range of aprotic organic solvents (benzene, dichlorobenzene, nitrobenzene, dichloromethane, and chloroform) at a concentration of more than 10 mm at room temperature.In nitrobenzene/1,4-dioxane 1:3(v/v), which is the optimal solvent for our COF synthesis (vide infra), Co(ttpp) is completely dissolved at a concentration of 20 mm while Co(tapp) is almost insoluble (Figure 2a).To investigate the origins of the improved solubility of Co(ttpp), we compare solubility in three cases.First, free-base TTPP dissolves in chloroform at a concentration of 10 mm while the solubility of TAPP is lower, which assures that the bulky Tr groups inhibit the stacking of porphyrin rings and thus improve the solubility.Second, while TAPP dissolves in chloroform at a concentration of 0.5 mm, Co(tapp) is almost insoluble, indicating that the cobalt ion also plays a role in solubility.It is reported that cobalt porphyrins bearing amino or pyridyl groups form polymers via axial coordination at the cobalt centers, [18] suggesting similar coordination polymerization for Co(tapp) (Figure 2b).The bulky Tr groups in Co(ttpp) would inhibit such coordination and thus improve solubility.Third, Co(tapp) samples stored under air for a long time dissolve less than fresh samples, suggesting possible effects from the aerobic oxidation of Co(tapp).
To check the oxidation state of cobalt, we analyzed Co(tapp) and Co(ttpp) samples stored under air for 1 month or oxidized with I 2 by 1 H NMR and UV-vis absorbance spectroscopy.The 1 H NMR spectra of Co(tapp) samples show two sets of signals, which can be assigned to paramagnetic Co II and diamagnetic Co III species by comparison with the literature [19] (Figure 2d).An old sample shows a higher amount of Co III than a fresh sample, indicating gradual aerobic oxidation.In contrast, the 1 H NMR spectra of both fresh and old Co(ttpp) samples show only Co II signals [19a] (Figure 2e), and the absence of Co III species is confirmed by comparison to a sample oxidized by I 2 (Figure S8, Supporting Information).In the UV-vis absorbance spectra, old Co(tapp) shows a red-shifted Soret band compared to that of Co(ttpp) (Figure 2c).The redshift can be attributed to the oxidation of Co II to Co III , [20] which is supported by the spectrum of Co(ttpp) oxidized by I 2 (Figure S7, Supporting Information).These results indicate that Co(tapp) is susceptible to aerobic oxidation while Co(ttpp) is stable.The oxidation of Co(tapp) would be promoted by the coordination of amino groups to cobalt, while the stability of Co(ttpp) is similar to a typical tetracoordinate cobalt porphyrin. [21]We could not obtain Co(tapp) in a pure Co II state, even with an assynthesized sample (Figure 2d), a commercial source of Co(tapp), or after treatment with NaBH 4 . [22]Although the oxidation of Co(tapp) does not affect the resulting COF structure (vide infra), oxidation would strengthen the coordination of amino groups, lowering the solubility of Co(tapp).
To summarize, Tr-protected Co(ttpp) shows improved solubility by inhibiting porphyrin stacking, coordination polymerization, and aerobic oxidation observed in Co(tapp).The high solubility of Co(ttpp) is ideal for morphology-controlled COF syntheses.4a,23]

Comparison of the COF Formation from the Unprotected versus Protected Precursors
We chose COF-366-Co as a model porphyrin COF to synthesize because it is well-known as an eCO 2 RR catalyst. [5,24]COF-366-Co is a 2D square-lattice COF which forms from Co(tapp) and terephthalaldehyde (TPA) (Figure 1).
After optimization of the synthesis conditions, we found that COF-366-Co is formed as sub-μm particles when Co(ttpp) is solvothermally reacted with TPA in the presence of benzoic acid and water in nitrobenzene/1,4-dioxane 1:3(v/v).The solvent nitrobenzene/dioxane is advantageous to exploit the solubility of Co(ttpp), reaching more than 20 mm at room temperature, while two solvent systems typical in COF syntheses, mesitylene/1,4dioxane 1:3 (v/v) and o-dichlorobenzene/n-butanol 1:1 (v/v), show less solubility.Acid and water promote reversible imine condensation as well as in situ deprotection of tritylamines.The product is hereafter called COF-T.For comparison, Co(tapp) was reacted with TPA under the same conditions, whose product is named COF-A.
The reactions of the two precursors give visual differences (Figure 1).Co(tapp) does not dissolve in the solvent even after sonication for 15 min and sediments at the bottom before heating, and the product COF-A is also found as a solid at the bottom.In contrast, Co(ttpp) completely dissolves in the solvent to form a homogeneous solution.After heating, the product COF-T solids are homogeneously suspended in the solvent, indicating that the reaction proceeds through colloidal states.The formation of a colloid is presumably due to the following two aspects: First, the homogeneous initial state affords homogeneous nucleation.10a,c,d,25] These results mean that Tr protection changes the formation process of COF-366-Co from a solid-to-solid transformation to a solution-tocolloid process.
The product morphologies are analyzed by scanning electron microscopy (SEM), transmission electron microscopy (TEM), and powder X-ray diffraction (PXRD) in a powder state after filtration and rinsing with solvents and supercritical CO 2 .In SEM images, COF-A shows rod-like particles with a size of around 10 μm, similar to that of Co(tapp) (Figures 3a and S9, Supporting Infor-mation).While the surfaces of Co(tapp) particles before the reaction are smooth, the surfaces of COF-A particles are covered by many small crystallites and their spherical aggregates, similar to the originally reported COF-366-Co (Figure S10, Supporting Information). [5]This result indicates that COF-A grows on undissolved Co(tapp) particles by heterogeneous nucleation.On the other hand, SEM images of COF-T show only small spherical aggregates of crystallites, indicating that COF-T forms mainly from homogeneous nucleation from a solution (Figure 3a and Figure S11, Supporting Information).TEM images of COF-T show crystallite aggregates with diameters of 162 ± 41 nm (Figure 3b and Figure S12, Supporting Information), which are two orders of magnitude smaller than those of COF-A.Meanwhile, the crystallite (primary particle) sizes of COF-A and COF-T are similar, as indicated by the similar peak widths in PXRD patterns (Figure 3c).This result supports that the Tr groups in Co(ttpp) are irreversibly removed from the amino groups in the initial stage and do not interfere with crystal growth as modulators. [17]he crystal structures of each sample are analyzed by PXRD and N 2 and CO 2 sorption.The PXRD patterns of COF-A and COF-T are similar to each other and to a previous report of COF-366-Co, [5] indicating the formation of the same crystal structure (Figure 3c).Although the structure previously proposed was an eclipsed stacking model, [5] the presence of stacking slippage [26] is supported by further analysis with Rietveld refinement (Figure 1, Supporting Information pages 31-34).The PXRD patterns are also in good agreement with the fast Fourier transform (FFT) of the TEM image of COF-T (Figure S12, Supporting Information).The N 2 sorption isotherms of COF-A and COF-T show uptake curves in the microporous region similar to each other and to a previous report, [5] indicating similar micropore structures (Figure 3d).Accordingly, pore size distribution (PSD) analysis shows similar profiles with maxima at 2.1 nm for COF-A and COF-T (Figures S20 and S21, Supporting Information), which agrees with the model crystal structure (Figure S53, Supporting Information).The Brunauer-Emmett-Teller (BET) surface areas of COF-A (1643 m 2 g −1 ) and COF-T (1998 m 2 g −1 ) are higher than that of the originally reported COF-366-Co (1360 m 2 g −1 ), [5] indicating higher purities.The CO 2 sorption isotherms of COF-A and COF-T are similar to each other (Figure S22, Supporting Information), with a slightly higher uptake for COF-A, which can be due to unreacted Co(tapp) (vide infra).These results indicate that COF-A and COF-T have similar crystal structures despite the morphological differences, and Tr-protected amines work as a surrogate COF precursor without disturbing the crystal structure of the resultant COF.
The complete removal of Tr groups is further checked by solidstate 13 C NMR, digestion 1 H NMR, IR absorbance, and reaction yield.The solid-state 13 C NMR spectrum of COF-T is similar to that of COF-A, showing no signals from Tr groups (Figure S25, Supporting Information).We also digested each COF into monomer solutions in DCl/D 2 O/DMSO-d 6 solvent and analyzed them with 1 H NMR. [27] The 1 H NMR spectrum of digested COF-T shows only Co(tapp) and TPA signals similar to COF-A, without Tr species (Figures S23 and S24, Supporting Information).The IR absorbance spectrum of COF-T is also similar to those of COF-A, showing no signals from Tr groups (Figure S26, Supporting Information).These results show that Tr groups are not incorporated into the COF, and the deprotection by-product TrOH is removed by rinsing with solvents and supercritical CO 2 .Meanwhile, the yield of COF-T is 87% from Co(ttpp) regarding the mass of cobalt determined by ICP-OES, with a small loss due to the particles trapped in the filter paper.The filtrate from the COF-T reaction mixture is pale orange in contrast to the dark red Co(ttpp) solution.These results indicate that the in situ deprotection of Co(ttpp) proceeds completely under these COF synthesis conditions.
The oxidation states of Co centers are analyzed by magnetometry and solid-state 13 C NMR. Magnetometry of COF-T shows temperature-dependent magnetism of  M T = 0.48 cm 3 K mol −1 in the high-temperature region (Figure S27, Supporting Information), which agrees with a square-planar low-spin Co II porphyrin (0.49 cm 3 K mol −1 ). [28]The solid-state 13 C NMR spectra of COF-T and COF-A are almost identical, indicating the majority of Co atoms take the same low-spin Co II states in the two substances (Figure S25, Supporting Information).
The increased solubility of the precursor can also improve reaction progress.To investigate if this is the case, we analyze the digestion 1 H NMR, IR absorbance, and N 2 sorption data in detail.Integration of the TPA and Co(tapp) peaks in the digestion 1 H NMR spectra reveals their molar ratio, which should be 2 in pure COF-366-Co (Figures S23 and S24, Supporting Information).However, COF-A shows a TPA/Co(tapp) ratio of 1.66, while COF-T shows a TPA/Co(tapp) ratio of 1.99.This result indicates COF-A contains 17 mol% of unreacted Co(tapp).In the IR absorbance spectra, COF-A shows a weaker C = N stretch band than COF-T when normalized by the C = C stretch bands, which supports the incomplete Co(tapp) conversion in COF-A (Figure S28, Supporting Information).The N 2 sorption data show that COF-A adsorbs 15% less N 2 in micropores (Figure 3d and Figure S15, Supporting Information), has 18% less BET surface area (Figures S16-S19, Supporting Information), and has 15% less micropore volume than COF-T.These values agree with the digestion 1 H NMR results, regarding unreacted Co(tapp) as a non-porous impurity and assuming COF micropores have similar accessibility reflecting similar crystallite sizes.To investigate the reason for incomplete conversion in COF-A, we tested two alternative sets of synthetic conditions.The use of excess TPA (3 equiv.to Co(tapp)) as in a previous report [5] results in an IR spectrum similar to COF-A (Figure S28, Supporting Information), indicating similar incomplete conversion and ruling out the possibility that the consumption of TPA by side reactions is limiting the reaction progress.Increasing reaction time from 3 days to 9 days does not significantly change the product IR spectrum either (Figure S28, Supporting Information).Therefore, we speculate that during the formation of COF-A, COF crystallites growing on the surfaces of undissolved Co(tapp) particles become passivation layers, leaving unreacted Co(tapp) cores.On the other hand, the high solubility of Co(ttpp) facilitates full conversion to a COF.
To probe the generality of the Co(ttpp) method in downsizing COF particles, we have conducted the synthesis of isoreticular analogs of COF-366-Co: COF-367-Co [5] and Co(tapp)-TPDA-COF (Figure S29, Supporting Information).The formation of isoreticular crystalline frameworks is confirmed by PXRD, while the secondary particle sizes were measured as 0.41 ± 0.13 and 0.52 ± 0.16 μm, respectively, by SEM.This result showcases that Co(ttpp) is useful in preparing various Co porphyrin COFs as subμm particles.
These results demonstrate that Co(ttpp) serves as a precursor to COF-366-Co via in situ deprotection, and the resulting COF-T has an identical crystal structure and a similar crystallite size compared to COF-A from Co(tapp).Meanwhile, the higher solubility of Co(ttpp) than Co(tapp) enables the formation of smaller secondary particles.The smaller particle size of COF-T with a similar crystal structure and crystallite size to COF-A motivated us to study the effects of COF particle size on electrocatalytic performance.

Electrochemical CO 2 Reduction Catalyzed by COFs from the Unprotected and Protected Precursors
The catalytic properties of COF-A and COF-T for electrochemical CO 2 reduction are compared in 0.5 m aqueous KHCO 3 solution under CO 2 flow.The working electrodes were prepared with the same loading of a COF sample as a catalyst, carbon black as a conducting agent, [29] and Nafion as a binder on pieces of carbon paper as current collectors.Cyclic voltammetry (CV) of the COF-A and COF-T electrodes shows catalytic currents at reductive potentials (Figure 4a).To quantify the catalysis products, potentiostatic electrolysis was conducted at certain potentials while the evolution of gaseous products was quantified by gas chromatography, and that of liquid products was checked by 1 H NMR. Both COFs give CO gas as the major product and H 2 gas as the only side product, while the formation of liquid products such as formate is negligible (Figures S30 and S31, Supporting Information).The combined initial faradaic efficiency for CO and H 2 was 96% for COF-A and 99% for COF-T at −0.78 V versus the reversible hydrogen electrode (RHE), supporting that there is no other major product.The product CO is confirmed to originate from the CO 2 gas by 13 C-labeling experiments with 13 CO 2 (Figure S32, Supporting Information).
The catalytic activities of the COF-T and COF-A electrodes are compared in terms of CO production turnover frequency (TOF) per all cobalt.The COF-T electrode achieves a higher initial TOF than that of the COF-A electrode at a potential of −0.78 V versus RHE (Figure 4b,c).A similar enhancement in activity is observed over the wide range of potentials from −0.88 to −0.58 V versus RHE (Figure S33, Supporting Information).Meanwhile, the catalytic stability of COF-T and COF-A electrodes are similar (Figure 4c), and the higher activity of the COF-T electrode continues over 10 h (Figure S34, Supporting Information).To check the possible interference by the unreacted Co(tapp) in COF-A, we also tested an electrode prepared from pristine Co(tapp).In our setup, pristine Co(tapp) shows higher activity than COF-A (Figure S35, Supporting Information), indicating that the low activity of the COF-A electrode is not caused by the unreacted Co(tapp) in COF-A.Therefore, it is demonstrated that particle downsizing can enhance the eCO 2 RR activity of COFs significantly.
The faradaic efficiency (FE) of COF-T and COF-A electrodes are compared in Figure 4d,e.The COF-T electrode shows higher initial FE(CO) (86-95%) over a wide potential range from −0.58 to −0.88 V versus RHE, while the COF-A electrode shows lower initial FE(CO) (38-67%).Moreover, the COF-T electrode exhibits better FE(CO) stability, maintaining a high FE(CO) of 89-91% for 45 min at −0.78 V versus RHE.In comparison, the FE(CO) of the COF-A electrode reduces from 58% to 38% under the same conditions, while FE(H 2 ) increases from 26% to 51%.The lower selectivity of COF-A is not due to the unreacted Co(tapp), because the pristine Co(tapp) electrode shows higher FE(CO) (Figure S36, Supporting Information).To investigate the origin of hydrogen evolution, we measured hydrogen evolution from a blank electrode without cobalt catalysts.Blank electrodes show H 2 evolution rates similar to COF-A and higher than COF-T electrodes, with a gradual increase in the H 2 evolution rate attributed to the electrowetting of the hydrophobic carbon materials (Figure 4f). [30]This result indicates that the small particles of COF-T suppress hydrogen evolution from electrode materials by efficient coverage, while the large particles of COF-A do not.Therefore, the higher FE(CO) of COF-T compared to COF-A is likely due to the suppression of hydrogen evolution from the electrode materials and the higher eCO 2 RR activity from the COF-T particles.
We next carried out a series of characterizations to elucidate the origin of the superior electrocatalytic performance of COF-T.The electrode macrostructures are investigated by SEM as shown in Figure 4g, Figures S39-S42 (Supporting Information).The COF-A electrode shows ≈10 μm particles similar to the original COF-A particles sitting on the fibers of carbon paper both before and after electrolysis.In contrast, the COF-T electrode shows no large particles (>1 μm) before and after electrolysis.Although the original particles of COF-T (≈160 nm) are not clearly visible among carbon black particles (≈50 nm), the PXRD and IR absorbance measurements of the COF-T/carbon black/Nafion catalyst ink show that the crystal and chemical structure of COF-T is maintained in the mixture (Figures S43 and S44, Supporting Information).Hence, it can be assumed that COF-T is well dispersed as small particles among carbon black on the electrode.The water contact angles of COF-T, COF-A, and blank electrodes show hydrophilic surfaces ( < 90°) in all cases, ruling out the formation of trapped gas layers on the electrodes in aqueous electrolyte (Figure S45, Supporting Information).The number of electrochemically active sites could not be quantified from CV experiments, because of too broad redox waves and parasitic hydrogen evolution currents from carbon materials (Figure S37, Supporting Information).To gain insights into the electrochemical accessibility of catalytic sites, we conducted control experiments without carbon black, which had been included in the abovementioned electrodes as conductive agent. [29]The resulting CO evolution rates show two orders of magnitude lower activity for electrodes without carbon black (Figure S38, Supporting Information).This result indicates that efficient eCO 2 RR occurs at the interface between COF particles and carbon black particles, as the conductivity of COFs is insufficient.Similar behavior is reported for other organic heterogeneous catalysts for eCO 2 RR. [8,31]In our system, the smaller particle size of COF-T compared to COF-A provides a larger interfacial area with carbon black (Figure 4h).In addition, the large particles of COF-A can disrupt or protrude from the carbon black layer, further diminishing the interface.Therefore, downsized COF-T particles show higher electrocatalytic activity.
To estimate the effectiveness of this downsizing strategy for COF electrocatalysts, the eCO 2 RR performance of COF-T is compared with previously reported Co porphyrin COFs in terms of CO evolution rate per catalyst mass and faradaic efficiency.First, when compared with previous reports on particulate COF-366-Co, the performance of COF-T is significantly higher (Figure S46 and Table S2, Supporting Information).Second, when compared with other Co porphyrin COFs, the performance of COF-T is superior to most of them (Figure S47 and Table S3, Supporting Information).Although the activity of COF-T is not as high as that of the thin film catalyst COF-366-Co/HOPG, [24a] COF-T has an advantage as nanoparticles in terms of processability and applicability to non-flat electrodes.Besides, despite being based on a simple COF-366-Co structure, COF-T outperforms most of the previously reported Co porphyrin COFs including those with elaborate structure designs such as an electron transporter, [14] a donor-acceptor structure, [32] and a postsynthetic modification. [13]These comparisons demonstrate that downsizing COF particles is a powerful strategy for developing high-performance COF electrocatalysts.
The long-term catalytic stability of a COF-T electrode is examined for 10 h at −0.78 V versus RHE (Figure S34, Supporting Information).The activity gradually decreased to 48% of the initial value, which is similar to the original report of COF-366-Co (67% after 12 h at −0.68 V vs RHE) [5] and a structurally similar Co porphyrin imine COF (39% after 10 h at −0.7 V vs RHE). [33]he selectivity gradually decreased to 72% faradaic efficiency for CO evolution, which can be explained by the gradually increasing parasitic hydrogen evolution from the carbon materials discussed above.

Conclusion
In this work, we demonstrated a rational approach for suppressing the crystallite aggregation of COF eCO 2 RR catalysts, which enables the particle downsizing of COF-366-Co and its isoreticular analogs to hundreds of nanometers.The conventional synthesis of COF-366-Co is based on poorly soluble Co(tapp) and results in large crystallite aggregates with a size of around 10 μm because of heterogeneous nucleation, along with the incomplete conversion of Co(tapp) into the COF.By introducing trityl protection groups, the solubility of the porphyrin COF precursor, i.e., Co(ttpp), is significantly improved owing to the suppression of intermolecular stacking and amine-cobalt coordination interac-tions.The use of Co(ttpp) as a precursor affords homogeneous nucleation and downsizes COF-366-Co particles to 162 ± 41 nm, suppressing crystallite aggregation.It is worth mentioning that trityl groups can be introduced easily to amino groups and can be deprotected in situ under standard COF synthesis conditions, thus rendering the method introduced here generic and universal for many types of COFs.13b] This work showcases for the first time the effects of crystallite aggregation and the resulting particle size effects in COF electrocatalysts.The downsizing of COF-366-Co particles enhances eCO 2 RR activity owing to the higher contact area with a conductive agent, achieving a higher CO production TOF compared to the larger particles obtained via the conventional approach.Therefore, particle downsizing has the potential to systematically and significantly improve the electrocatalytic performance of COFs.Finally, this insight implies that the effects of particle size should be always considered when different COF electrocatalysts are compared to reveal intrinsic structure-performance relationships.

Figure 1 .
Figure 1.Synthetic schemes of downsized COF catalyst particles COF-T from a protected precursor Co(ttpp) in comparison with conventional COF particles COF-A from a conventional precursor Co(tapp), schematic representation of the dispersion states throughout the COF syntheses, and model crystal structure in a space-filling representation.

Figure 4 .
Figure 4. Electrochemical CO 2 reduction with COF-A and COF-T as catalysts (0.5 m KHCO 3 aq, CO 2 flow).a) Cyclic voltammograms normalized by the mass of COF catalysts.b) CO production turnover number (TON) per all cobalt atoms over time (−0.78V vs RHE).Three measurements with different synthetic batches are shown for each material.c) CO production turnover frequency (TOF) per all cobalt atoms over time (−0.78V vs RHE, logarithmic scale).Three measurements with different synthetic batches are shown for each material.d) Initial CO production faradaic efficiencies at different potentials (after headspace composition reaches a steady state in 10 min).e) Faradaic efficiencies over time (−0.78V vs RHE).f) H 2 evolution rates over time in comparison with a blank electrode without COFs (−0.78 V vs RHE).Three measurements with different synthetic batches of COFs or different blank electrodes are shown for each material.g) SEM images after electrolysis.h) Schematic representation of the electrode structures and electrochemical reactions.