Synthesis and characterization
The bulky Co-MOF, [CoL(H2O)2]·0.5H2O (H2L = 5-(1H-1,2,4-triazol-1-yl)isophthalic acid), was synthesized via a solvothermal reaction of CoCl2·6H2O and H2L in DMF/H2O at 130 °C for 72 h (see the Methods). Single-crystal X-ray diffraction analyses reveal that Co-MOF crystallizes in a monoclinic crystal system with a space group of C2/c (Supplementary Table 1). As shown in Figure 1a, the asymmetric unit of Co-MOF contains one Co2+ cation coordinated by two aqua ligands, one N donor from the triazine moiety as well as three O atoms from two independent carboxylate groups in two L ligands. Through this coordination mode, one cobalt center connects with three organic ligands into a plane parallel to b axis, forming a 2D layer-like structure. The 2D layers are stacked together via H-bonding interactions (O-H···O = 2.760(4) Å, Supplementary Figure 1) between aqua molecules and carboxylate oxygen atoms, showing negligible voids in the framework (Supplementary Figure 2).
Afterwards, the bulk phase purity of Co-MOF was confirmed by powder X-ray diffraction (PXRD; Figure 1b) by comparison with that simulated from single crystal data. Importantly, the sharp peak at 26.7° corresponding to (404(—)) face was found to represent the stacking direction of MOLs with a lattice spacing of 0.33 nm, consistent with the distance between two 2D planes. Tightly stacking among these 2D layers in Co-MOF results in the bulky crystals with sizes over 50 μm, which can only be transformed into ca. 2 μm crystals after ultrasonic treatment (Supplementary Figure 3). The close stacking will inevitably decrease the exposed active sites and impede the mass transport / electron transfer during the photocatalysis.
To overcome these problems, we attempt to use GO as 2D template to graft and stabilize metal coordination layers of Co-MOF to construct ultrathin MOL nanosheets. The synthetic procedure includes the incorporation of Co2+ ions into GO and the subsequent in-situ growth of Co-MOLs with H2L ligand on the 2D GO template (see Methods and Figure 1c). First, both the PXRD (Figure 1b) and FT-IR (Figure 1d) measurements on Co-MOF, GO and Co-MOL@GO samples reveal the effective graft of Co-MOL layers on the GO support. It should be noted that the peak of (404(—)) face at 26.7° is absent in the PXRD pattern of Co-MOL@GO (Figure 1b), indicating an obvious reduction of the stacking effect in MOLs on the GO support and the effective separation between different layers. The Co-MOL@GO sample contains approximately 6.9% Co-MOL moiety (1.2% ± 0.02% Co) determined by ICP-MS via repeated measurements.
Subsequently, the morphology of Co-MOL@GO was studied by transmission electron microscopy (TEM; Figure 1e), where small nanoflakes (15-20 nm) were homogeneously distributed on the GO template. EDS mapping images indicate the even distribution of Co, N, C and O elements on the Co-MOL@GO sample (Supplementary Figure 4). Atomic force microscopic (AFM) results also show the distribution of nanosheets on GO substrates (Figure 1f), with an average diameter of ca. 20 nm and a thickness of ca. 1.5 nm, close to three metal coordination layers of 1.4 nm determined by the single-crystal X-ray diffraction analysis (Figure 1a). Furthermore, high-resolution TEM (HRTEM; Figure 1g) measurements clearly show the crystal spacing of 0.27 nm in the tiny nanocrystal, in good agreement with the appearance of (040) facets, which is also consistent with the PXRD results with the characteristic peak located at 33.4° (0.27 nm) (Figure 1b). Accordingly, all the above experimental results can prove the successful immobilization of ultrathin Co-MOLs on the GO template, resulting in the Co-MOL@GO composite. Upon increasing the loading amount of Co2+, larger size of 2D nanosheets can be observed on the GO substrate, as indicated by TEM and PXRD measurements (Supplementary Figure 5). These results further confirm that the GO template synthesis represents a facile strategy for the synthesis of ultrathin MOL nanosheets.
With the verified morphology of Co-MOL@GO, a series of additional measurements were operated to verify the changes of Co-MOLs and GO in Co-MOL@GO sample. Initially, X-ray photoelectron spectroscopy (XPS; Figure 2a) reveals a negative shift in the binding energies of Co 2p in Co-MOL@GO compared to those of Co-MOF. This observation indicates the existence of interactions between Co-MOF and GO surface, which can facilitate the charge transfer between Co-MOF and GO to impede the recombination of charge carriers.35 Moreover, GO substrate was substantially reduced under the solvothermal condition, which was confirmed by the Raman spectra of Co-MOL@GO with an increased ratio between defective bands (ID) and graphitic band (IG) relative to that of GO (1.54 vs 1.33) (Figure 2b).36 The reduced GO can be more conductive and feasible for electron transfer during photocatalytic process. This conclusion was further supported by the results of electrochemical impedance spectroscopy (EIS), where Co-MOL@GO exhibits a smaller charge-transfer resistance compared to those of Co-MOF and GO (Figure 2c). Overall, all above results suggest that Co-MOL@GO can be a good candidate as a catalyst for photocatalytic CO2 reduction with its advantages in abundantly exposed active sites, excellent mass transport and charge-transfer ability.
Photocatalytic CO2 reduction
The catalytic performance of Co-MOL@GO for visible-light-driven (λ = 450 nm) CO2 reduction was investigated in a CO2-saturated CH3CN/H2O (v : v = 4 : 1) solution. Ru(phen)3(PF6)3 (phen = 1,10-phenanthroline; denoted as RuPS) was employed as the PS. Co-MOF, Co@GO and GO were used as the catalysts in the control experiments. The gaseous products, CO and H2, were analyzed by gas chromatography, and the liquid products, i.e. HCOOH, were checked by ion chromatography. As shown in Figure 3 and Table 1, the main products are CO and H2, and no liquid products can be detected. Remarkably, a CO yield of 216.2 mmol/g with a high selectivity of 95% can be achieved during 12 h irradiation of Co-MOL@GO-based photocatalytic system. In comparison, the photocatalytic experiment with bulky Co-MOF as the catalyst affords a CO yield of 91.5 mmol/gMOF, much smaller than that of Co-MOL@GO-containing system, as well as a lower selectivity of 82%. Control experiments reveal that GO shows no activity towards CO2 reduction under this photocatalytic condition, indicating that the MOLs should be the real active species in the Co-MOL@GO sample. As a result, the CO yield can be estimated as 3133 mmol/gMOL, ca. 34 times higher than that of bulky Co-MOF under similar condition, indicative of the remarkable intrinsic catalytic activity of ultrathin Co-MOF in Co-MOL@GO. Impressively, in terms of CO yield and selectivity, the catalytic performance of Co-MOL@GO (3133 mmol/gMOL, 95%) is superior to those of all the state-of-the-art MOF catalysts for visible-light-driven CO2 reduction (Supplementary Table 2).37 Additionally, the photocatalytic system with Co@GO as the catalyst can also produce CO and H2 under the same conditions, but with a much smaller amount than that of Co-MOL@GO (Supplementary Figure 6), showing that the formation of Co-MOL nanosheets on GO is the key to achieve high-performance CO2 reduction. Accordingly, GO template strategy is promising to fabricate high-performance catalysts for photocatalytic CO2 reduction. These comparative results clearly demonstrate much-enhanced intrinsic catalytic activity of 2D-nanosized Co-MOL in contrast to the bulky Co-MOF and other samples. This enhancement can be mainly attributed to its great exposure of catalytic active sites enabled by the ultrathin feature of the MOL, and the incorporation of graphene as charge-transfer mediator.
Table 1 | Photocatalytic results for CO2 reduction to CO.a
Entry
|
Catalyst
|
CO Yield (mmol/g)
|
H2 Yield (mmol/g)
|
CO (%)
|
1
|
Co-MOL@GO
|
216.2 (3133)[b]
|
11.2 (162)[b]
|
95
|
2
|
Co-MOF
|
91.5
|
19.8
|
82
|
3
|
Co@GO
|
66.7
|
3.16
|
95
|
4
|
GO
|
N.D.
|
< 0.10[c]
|
N.A.
|
5
|
No catalyst
|
N.D.
|
< 0.10[c]
|
N.A.
|
aConditions: CO2 saturated 5.0 mL CH3CN/H2O (v:v = 4:1), catalyst (10 mg/L), RuPS (0.4 mM), TEOA (0.3 M), light intensity of 100 mW/cm2 (λ = 450 nm), 12 h of irradiation. [b]Calculated based on Co-MOL moiety. [c]Near the detection limits.
Thermal and chemical stability of Co-MOF and Co-MOL@GO were carefully investigated. For Co-MOF, thermogravimetric analysis (TGA) was conducted to investigate its thermal stability. As shown in Supplementary Figure 7, the TGA curve shows three continuous weight losses from 95 to 320 ºC, suggesting the loss of lattice and coordinated water molecules in the cavity of bulky Co-MOF. It could also be observed that a thermal decomposition of Co-MOF took place until heating up to 400 ºC, revealing its high thermal stabilty. Meanwhile, the bulk Co-MOF was soaked in a CO2-saturated CH3CN/H2O (v : v = 4 : 1) solution containing 0.3 M TEOA, a reaction medium for photocatalytic CO2 reduction. After one day, the solid samples were isolated for subsequent PXRD measurements. No obvious difference of the PXRD pattern can be observed compared to that of as-synthesized sample (Supplementary Figure 8). These results demonstrate remarkable thermal and chemical stability of this Co-MOF, assuring its robustness in photocatalytic CO2 reduction. For Co-MOL@GO, PXRD pattern of the solid sample isolated from the photocatalytic system shows similar signals with that of as-prepared Co-MOL@GO, indicating the intact crystalline composition of Co-MOL@GO catalyst after photocatalysis (Supplementary Figure 9). Moreover, recycle experiments show no substantial decrease in the activity after three runs of photocatalytic reactions, confirming the retained activity of the Co-MOL@GO catalyst (Figure 3b). On the other hand, isotope labeling experiment with 13CO2 shows that 13CO is the main product in this photocatalytic system (Figure 3c), manifesting that the CO product really derives from CO2 rather than the decomposition of TEOA, RuPS, GO or Co-MOLs. All these results demonstrate the excellent stability of Co-MOL@GO in the photocatalytic CO2-to-CO conversion.
Electron transfer pathway
To elucidate the electron transfer pathway, the emission quenching experiments of RuPS were conducted in detail with the quenchers including Co-MOL@GO, Co-MOF and TEOA (Figure 3d-3e and Supplementary Figure 10). In the fluorescence spectra of RuPS, an emission peak at 598 nm was detected with the excitation at 450 nm. Upon the gradual addition of Co-MOL@GO, the efficiency of fluorescence quenching was far higher than those obtained by addition of bulky Co-MOF and GO, indicating that the ultrathin feature of MOL and the incorporation of graphene mediator are beneficial to electron transfer. The Ksv quenching constants of RuPS quenched by Co-MOL@GO was calculated as 3250 L g-1 by Stern-Volmer plot,38,39 much higher than that obtained by addition of Co-MOF (813 L g-1), and no obvious quenching can be observed in the presence of isolated GO. These results suggest the fast electron transfer between homogeneous PS* and heterogeneous Co-MOL@GO catalyst mediated by GO, affording the high catalytic performance.38,39 Further, control experiments show that negligible fluorescent quenching can be detected by addition of various amounts of TEOA (Figure 3e), indicating that the exited RuPS* was directly quenched by Co-MOL@GO via an oxidation quenching mechanism.
The acceleration of charge transfer was confirmed by time-resolved absorption spectroscopy. As shown in Figure 3f, the kinetic traces for excited RuPS show that the lifetime of RuPS* (478.3 ns) was much longer than that obtained in the presence of Co-MOF (412.7 ns) catalyst, and this lifetime can be further shortened to be 357.0 ns when Co-MOL@GO composite was present. These results suggest the rapid separation and migration of photogenerated charge carriers between homogeneous PS and heterogeneous catalyst with graphene as the mediator. In the presence of TEOA, the excited lifetime of RuPS* was similar to that of isolated RuPS (481.9 vs. 478.3 ns), further confirming the oxidative quenching electron transfer pathway in this photocatalytic reaction for CO2 reduction.40
Further, the acceleration of electron transfer mediated by GO and the roles of different components in the photocatalytic system were evaluated by in situ transient photovoltage (TPV) measurements on GO, Co-MOF, RuPS, and Co-MOL@GO. As shown in Figure 4a, the photocurrent response of GO is the highest among the detected samples, and its curve in the CH3CN/H2O (v:v = 4:1) medium is similar to that in the air (Supplementary Figure 11). Figure 4b shows that the photocurrent intensity of Co-MOF/RuPS mixture is higher than that of isolated components of Co-MOF and RuPS. These results suggest a coupling effect between Co-MOF and RuPS that can enhance the signal, while the response of GO is still stronger than that of Co-MOF/RuPS (Figure 4c). Then, the photocurrent intensity decreased when GO was mixed with RuPS or Co-MOF (Co-MOL), respectively. Especially, the photocurrent intensity of Co-MOL@GO composite is much lower than that from the combination of GO and RuPS (Figure 4d). These observations infer the electron transfer pathway, in which the electrons can be transferred to both Co-MOL and RuPS from the GO surface, and Co-MOL can accept electrons more easily than RuPS.
To determine the active centers, the TPV measurements were performed in different atmospheres of N2 and CO2 to evaluate the real catalytic active centers (Figure 4e-4h). As shown in Figure 4e and 4f, similar photocurrent intensities were observed in the curves of either GO or RuPS under N2 or CO2. In contrast, a sharp decrease of photocurrent intensity was detected with Co-MOF under CO2 in comparison with that under N2 (Figure 4g). These results reveal that the photocatalytic CO2 reduction should occur on the surface of Co-MOF. That is, Co-MOL represents the active component in the system. To further confirm this proposal, the TPV of Co-MOL@GO/RuPS mixture was performed, and a similar trend to that of Co-MOF was observed, confirming the role of Co-MOL as the active center (Figure 4h). Overall, the above TPV analyses confirm that Co-MOL is the active component of photocatalytic CO2 reduction and that GO serves as the electron mediator to deliver electrons to Co-MOL.
Mechanistic studies
As the Co-MOL is the active component in photocatalysis, its molecular catalytic mechanism was further investigated by the electrochemical measurements. First, we studied the electrochemical behavior of GO and Co-MOL@GO loaded on the surface of the glassy carbon electrode (GCE), respectively. As shown in Supplementary Figure 12a, the irreversible waves at ca. -0.75 V versus normal hydrogen electrode (vs. NHE) were both observed in the cyclic voltammograms (CVs) of Co-MOL@GO and GO under N2, where the reduction currents are mainly attributed to the reduction events at GO. To avoid the interference of GO, Co-MOF was directly employed to investigate the redox behavior of the MOF catalyst. As shown in Supplementary Figure 12b, a quasi-reversible redox couple at E1/2 = -0.76 V vs. NHE (reduction wave at -0.85 V) appeared in the CV of Co-MOF under N2, corresponding to CoII/I reduction. Upon purging CO2 into the system, an irreversible reduction wave peaking at -0.94 V vs. NHE with a relatively large current was detected, indicating a chemical step driven by CoII/I reduction, most possibly for catalytic CO2 reduction,41 as the position of this reduction wave is more negative than the standard reduction potential of CO2/CO (-0.65 V vs. NHE).42 Moreover, the above results indicate that the catalysis driven by the photoexcited RuPS* should be thermodynamically accessible, as the driving force from the oxidative quenching pathway (E = -0.84 V vs. NHE)43 is more negative than the onset potential (ca. -0.75 V) of the catalytic wave.
According to above studies,41,44 the catalytic mechanism of Co-MOF in CO2-to-CO conversion can be tentatively proposed, which was further verified by DFT calculation. A molecular prototype presenting the Co-complex moiety was subjected in computational studies (Figure 5a). The overall free energy changes are downhill besides the rate-determining CO2-binding step, suggesting the calculated catalytic cycle is theoretically viable. As illustrated by Figure 5, the photocatalytic cycle begins with the photo-excitation of RuPS (Figure 5b). Then, the exited RuPS* species can be oxidatively quenched by Co-MOL@GO catalyst to drive the CoII/I reduction to form CoI species. The calculated potential for the reduction from CoII to CoI is -1.03 V vs. NHE, approaching to the measured value (ca. -0.94 V) in the presence of CO2, further confirming the accessibility of this proposed mechanism. The CoI intermediate can react with CO2 to generate a Co-CO2 adduct. Then a 1e-/1H+ proton-coupled electron transfer takes place to generate a Co-COOH species. The reduction and protonation from Co-CO2 to Co-COOH is also viable as the calculated reduction potential (+0.39 V) is much more positive than the CoII/I reduction potential. Finally, the C-OH bond in the Co-COOH intermediate will cleave and release CO to recover the CoII state. The remaining RuPS+ species from the oxidative quenching pathway can be reduced to original RuPS by the TEOA, completing the photocatalytic cycle. During this photocatalytic process, the 2D GO not only serves as the template to reduce the surface energy of ultrathin nanosheets for constructing ultrathin MOLs with more exposed active sites, but also to supply conductive channels that can facilitate photoexcited electron transfer, which both play key roles in promoting the photocatalysis.