1 Introduction

Converting CO2 into renewable energy via artificial photosynthesis is ideal pathway for addressing the escalating energy shortages and climate changes at the same time [1,2,3,4]. However, though countless effort has been paid till today, the solar-driven CO2 conversion is still suffered a low efficiency, which should be attributed to the high chemical stability of CO2 molecules and the fast recombination of photogenerated charge carriers [5,6,7]. These facts make the design and construction of highly efficient catalyst essential for further application of photocatalytic CO2 conversion [8,9,10]. Metal–organic frameworks (MOFs) are a class of crystalline porous materials consisting of tunable multitopic organic linkers and metal-based nodes [11,12,13]. With their typically high accessible surface area, MOFs are promising candidates for a wide range of applications including gas capture, separation, and storage [14,15,16,17]. It was fessible to modify MOFs with cocatalysts, and the modified MOFs exhibited great potential in photocatalysis such as pollutant degradation, hygogen evolution and CO2 reduction as a type of semiconductor-like materials [18,19,20,21,22,23]. For example, controllable manipulation of metal composites or single atoms have been used as MOF catalyst engineering strategies in recent years [24,25,26,27]. It is found that some MOFs such as UiO-66-NH2 with amine groups can have strong coordination with metal ions, which could be beneficial to achieving higher dispersion of metal atoms [28,29,30]. Such atomic modification with high atomic utilization efficiency often gives impressive activity in heterogeneous photocatalysis with MOF for introducing considerable reactive sites and it also offers an opportunity to promote the separation of exicted electrons and holes during the photocatalysis process.

Previous works have proven that cobalt is a promising catalytic center for CO2 photocatalytic reduction reaction [31,32,33]. Coordinatively unsaturated Co sites were proved to be able to accept excited electrons from substrate [34], which greatly enhanced the separation of photoinduced electrons and holes. It is also found that the Co sites have a great promotion effect on the adsorption and activation of CO2 by strong interaction between Co 3d electrons and C 2p electrons of adsorbed CO2, which could hence active CO2 effectively [35]. Notably, the application of cobalt cocatalyst was limited by the low dispersion on traditional semiconductors. On the other hand, as in the construction of Co modified MOFs materials, expansive linkers or metal–organic complexes are widely used. Regardless the good activity those catalysts have, it is not economical for further application. Herein, in order to increase the atomic utilization of Co, we realize the decoration of highly dispersed Co into UiO-66-NH2 through an economic NaBH4 treating method. With the promoted dispersion of cobalt, the obtained xCoNU photocatalyst exhibited outstanding catalytic performance. The CO generation rate was up to 1.93 mmol g−1 h−1, with TON of 38 within 3 h, which is 10.2 times higher than that of UiO-66-NH2. In addition, the improved photocatalytic performance caused by the introduction of cobalt has also been studied in detail. This work provides a promising approach to engineering the actives sites on MOF materials for solar-driven CO2 conversion.

2 Experiment Section

2.1 Preparation of UiO-66-NH2

UiO-66-NH2 (noted as NU66) was synthesized by a solvothermal method. 0.24 g ZrCl4, 3.75 g benzoic acid and 0.18 g 2-amino-1, 4-benzenedicarboxylic acid (NH2-BDC) were added into 60 mL dry dimethyl formamide (DMF). The mixture was sonicated and stirred for 10 min to achieve a homogenous solution. Then the solution was transferred into a 100 mL stainless autoclave with Teflon liner and heated at 120 °C for 24 h. After cooling down to room temperature, the resultant suspension was filtered, washed with DMF and methanol respectively. The final product was dried under vacuum for 12 h at 50 °C.

2.2 Preparation of xCoNU

Series of xCoNU (x presents the theoretical Co loading weight percentage) was obtained by NaBH4 treating method. Typically, for the synthesis of 1CoNU, 0.1 g NU66 was firstly dispersed in 35 mL acetonitrile (MeCN). Acetonitrile solution of CoCl2·6H2O dissolved (4 mg, 5 mL) was then added into the suspension, and the mixture was kept stirring for 30 min. Afterward acetonitrile solution of NaBH4 (6.8 mg, 10 mL) was slowly dropped into the mixture. To promote a complete reaction, the mixture was kept for another 10 min stirring. Then precipitate was collected by centrifugation and washed by acetonitrile several times. The final product was dried under vacuum without heat. 2CoNU, 5CoNU, 7CoNU and 10CoNU were synthesized by similar method which only scaled up the amount of CoCl2·6H2O and NaBH4 during reaction.

2.3 Characterizations

TEM images were recorded by a JEM-2100 equipped with fast Fourier transformation (FFT). The crystallinity of samples was characterized by Rigaku D/MAX 2550 diffraction meter with Cu K radiation (λ = 1.5406 Å) at 40 kV and 40 mA. The surface chemical states of samples were determined by high-resolution XPS (PerkinElmer PHI 5000C ESCA system: Al Kα radiation, operated at 250 W). Ultraviolet–visible (UV–vis) diffuse reflectance spectra were recorded by a UV–vis spectrophotometer (UV-2600, Shimadzu). The charge separation efficiency of all samples was evaluated on a luminescence spectrometer (Cary Eclipse) with the excitation wavelength at 315 nm. The charge transfer efficiency of samples was evaluated by Electrochemical Impedance Spectroscopy (EIS) on a Zennium electrochemical station with a three-electrode cell. Time-resolution photoluminescence (TRPL) spectra was performed on an Edinburgh Analytical Instruments FLS 980 coupled with a time-correlated single-photo-counting system at room temperature.

2.4 CO2 Photocatalytic Reduction (CO2PR) Evaluation

In the typical photocatalytic CO2 reduction reaction, 5 mg of catalyst, 5 mg of [Ru(bpy)3]Cl2·6H2O (abbreviated as Ru, bpy = 2′2-bipyridine), 1.0 mL of triethylamine (TEA), 1.0 mL of H2O and 8.0 mL of acetonitrile were added into a homemade quartz reactor with total volume of 200 mL. Then, high purity CO2 was introduced into the reactor with a partial pressure of 1.0 atm. A 300 W Xe lamp with a AM1.5 cut-off filter was used as the light source. During the reaction, the reaction system was viciously stirred by magnetic stirrer to ensure a homogeneous gaseous atmosphere. The temperature of the reaction system was kept at 25 °C by condensate water. The gaseous mixture was analyzed by a gas chromatograph (GC2014, Shimadazu, carrier gas: Ar) equipped with a thermal conductivity detector (TCD) and a flame ionization detector (FID) every half an hour. The GC was equipped with 3 columns: an FFAP column and a PN for the pre-separation of the gas, and a 13X column for the separation of H2, O2, CH4, CO and H2. H2 was measured by TCD. CH4 was measured by FID, while CO was converted to CH4 by a methanation reactor first and then analyzed by the FID.

3 Results and Discussion

The morphology of as-synthesized UiO-66-NH2 based catalysts was analyzed by TEM. As shown in Fig. 1A (UiO-66-NH2) and Fig. 1B (5CoNU), no significant change for the MOF was found after the introducing of Co, which suggested that UiO-66-NH2 was feasible for modification. The XRD patterns of blank UiO-66-NH2 and 5CoNU were shown in Fig. S1. Typical UiO-66-NH2 patterns were obtained and no significant crystallinity change was found after the Co modification [36]. On the other hand, no peaks belonging to cobalt-related species can be sighted, possibly due to the low loading amount of cobalt. Given that directly reducing CoCl2 in the same condition only resulted in an amorphous product, it was also possible that Co species in 5CoNU stayed amorphous.

Fig. 1
figure 1

TEM images of A NU66, B 5CoNU

Full size image

The loading amount of cobalt in xCoNU series samples was verified by ICP-AES. As listed in Table 1, the actual loading amount of Co in samples was all around 65% of calculated result, indicating a similar Co reduction level in xCoNU. Notably, when replacing UiO-66-NH2 with UiO-66 which lacks –NH2 to synthesize 5CoUiO-66, a slightly lower Co content was obtained (3.61% to 2.83%). It could be ascribed to the complexation between metal ions and –NH2 as reported [37]. With no –NH2, it could be harder for UiO-66 to adsorb Co ions before NaBH4 treating than UiO-66-NH2. The UV–vis absorption spectra of the samples were shown in Fig. 2A. After loading of Co, the absorption edge of xCoNU samples were barely changed, which suggest that the reduction process didn’t alter the structure of UiO-66-NH2. Due to the deep color of Co species, the absorbance of xCoNU increased with the increasing of Co amount [38]. The CO2 adsorption ability of our catalysts was also investigated. UiO-66-NH2 is reported to have better CO2 adsorption ability than UiO-66 [39, 40], which could be beneficial for CO2 reduction. It was found that after the loading of Co, the CO2 adsorption ability of UiO-66-NH2 was slightly decreased as exhibited in Fig. 2B, which could be ascribed to the partial blocking of micropore during the loading process. Therefore, the enhanced CO2 photoreduction efficiency of 5CoNU could not be attributed to the CO2 adsorption capacity.

Table 1 Cobalt content in xCoNU (x = 1, 2, 5, 7, 10) and 5CoUiO-66
Full size table
Fig. 2
figure 2

A UV–Vis DRS spectra of NU66 and xCoNU with different Co content; B CO2 adsorption performance of NU66 and 5CoNU at 298 K

Full size image

The valence states of all elements in NU66 and 5CoNU were verified through X-ray photoelectron spectroscopy (Fig. 3). Compared to NU66, the binding energy of Zr 3d in 5CoNU was slightly decreased, which reflected the increased electron density of Zr by Co loading. The binding energy for O 1s was reduced and the peak width was broadened, which could be ascribed to the formation of Co–O bound. On the contrary, binding energy for N 1s was barely changed, which ruled out the formation of Co–N bound. In Co 2p spectra, the binding energy of Co 2p1/2 and Co 2p3/2 were respectively located at 780.5 eV and 796.5 eV, corresponding to Co3+ CoOOH species [41, 42], which was reported as a result of insufficient reaction between Co precursor and NaBH4 [43]. According to the TEM and XRD results, there are no Co species particles on xCoNU. Therefore, the Co exists as highly dispersed CoOOH in xCoNU.

Fig. 3
figure 3

XPS A Zr 3d spectra, B O 1s spectra, C N 1s spectra of NU66 and 5CoNU, and D Co 2p spectra of 5CoNU

Full size image

CO2 photocatalytic reduction performance of xCoNU was evaluated in a homemade quartz reactor with Ru as the absorber. As illustrated in Fig. 4A and Table S1, only CO and H2 were detected to be reductive products. It was found that blank UiO-66-NH2 showed a rather low activity for both CO and H2 generation (Fig. 4A). After loading Co, the CO2PR activity of catalysts was significantly improved compared to the bare UiO-66-NH2, which clearly demonstrated that Co sites could work as the main active centers for CO2 photocatalytic reduction reaction. Notably, the selectivity for CO in all xCoNU samples was not altered and fixed at around 65% (Table S1), which suggest the main active sites in the reaction were the same. The optimistic loading amounts of Co was found to be 3.61 wt% (5CoNU), which exhibited an average CO generation rate of 1.93 mmol g−1 h−1 (9.67 μmol h−1 in 3 h, Fig. 4A) and H2 generation rate of 1.18 mmol g−1 h−1 in 3 h. The TON for CO was 38.66 with 5CoNU in 3 h, which is 10.2 times higher than that of UiO-66-NH2. It was also observed that catalytic performance of this system was suppressed after 3 h, mainly due to the degradation of Ru (Fig. S2). Furthermore, durability test was conducted for 5CoNU in Fig. 4B. It was found that the CO2PR performance of the catalyst can still hold after 5 cycles, suggesting good stability of 5CoNU.

Fig. 4
figure 4

A CO2 reduction performance of xCoNU samples with different Co content in 3 h; B durability experiment of 5CoNU, every cycle lasted for 3 h

Full size image

To further explore the charge separation ability of xCoNU series catalysts, room-temperature photoluminescence (PL) and electrochemical impedance spectroscopy (EIS) analysis were performed. After the loading of Co, the impedance of UiO-66-NH2 was decreased, which reflected an improved conductivity (Fig. 5A). Furthermore, as illustrated in Fig. 5B, the main PL emission peak of UiO-66-NH2 at 470 nm was greatly reduced after the loading of Co, demonstrating that the suppressed recombination of MOF-origined electron–hole pair. Since the absorber Ru itself also had strong PL response, we believe it essential to verify if the addition of Co can reduce the total photocarrier recombination within the overall reaction system composed of photocatalyst, Ru, and the mixed solution. Figure 5C showed the PL emission of the mix suspension with the same mixture ratio in our CO2PR test. The fluorescence was firstly quenched as the content of Co raised from 0 to 3.61%, which is consisted with the CO2PR result. The lifetime of photoluminescence was also extended (Fig. 5D; Table S3), indicating an enhanced charge separation ability of xCoNU. Additionally, compared with mixed Ru and NU66, τ2 of mixed Ru and 5CoNU extended better, which reflected a faster electron transition from Ru to 5CoNU [44]. Still, the fluorescence intensity increased again after excessive Co loading, suggesting a lower photoinduced carriers separation efficiency. This result demonstrated that overloading of Co may cause the formation of new recombination centers and reduce the reaction efficiency, while proper loading amount of Co could help photoinduced electrons and holes separation within the reaction system.

Fig. 5
figure 5

A EIS spectra of xCoNU; Room-temperature photoluminescence spectra of B xCoNU alone and C xCoNU with Ru in MeCN solution; D TRPL of xCoNU with Ru in MeCN solution

Full size image

The influence of dissolvent used for synthesize of 5CoNU was investigated. Results show that there was no difference between samples obtained in nonaqueous solvent, while the sample obtained in aqueous solution showed significant decrease in both H2 and CO generation (Fig. 6A). It could be attributed to the inadequate reaction between Co precursor and NaBH4 due to NaBH4’s self-decomposition in water. As shown in Table 1, 5CoNU obtained from aqueous solution had a lower Co content compared to that from MeCN solution, which confirmed our suspicion.

Fig. 6
figure 6

A Photocatalytic performance of 5CoNUs synthesized in different solutions in 3 h; B control experiments, where all samples were synthesized in MeCN

Full size image

Series control experiments were employed to explore CO2PR with 5CoNU (obtained in MeCN) under different conditions. As illustrated in Fig. 6B, with no catalyst, only a small amount of gaseous product was detected. In the absence of Ru absorber, CO2PR with 5CoNU was also greatly suppressed, and only a small number of products were detected. Only H2 was produced when CO2 was replaced by Ar in the system, which demonstrated that CO came from the reduction of CO2 rather than the decomposition of other reactants. Also, without light, the reaction was terminated with no product detected, which confirmed the reaction was driven by photoirradiation. These results suggested that photo-electrons from excited Ru to active sites on 5CoNU triggered CO2PR in this system. When using Co powder of the same weight as catalyst, only about half the product generation rates were expressed, and the TON for CO was 0.95, which was about fortieth of 5CoNU. As shown in Fig. 1, introduction of UiO-66-NH2 could promote the dispersion of Co sites, which would increase the active Co atoms utilization and promote the CO2PR reaction. A decreased activity was also observed when using UiO-66 as the supporter, which was structurally similar to UiO-66-NH2. Notably, it was confirmed that 5CoUiO-66 had a lower Co loading capacity than 5CoNU, as analyzed by ICP-AES in Table 1. These results suggested that the linker with –NH2 group could help the loading of Co sites, which resulted in a higher photocatalytic reactivity of 5CoNU.

The process for the CO2PR on xCoNU can be illustrated in Fig. 7 based on our experimental results. Upon photoexcitation, the light absorber Ru is excited to give electrons and holes. The photoexcited electrons are then transferred to the cobalt sites on xCoNU where CO2 molecule is activated and reduced to CO. The oxidized light absorber is reduced back to Ru by the sacrificial reductant TEA to form an entire catalytic cycle. The cobalt sites on xCoNU can work as CO2 active sites and promoted the charge separation within UiO-66-NH2 supporter and between the supporter and absorber, which results in a considerable CO2 conversion efficiency.

Fig. 7
figure 7

Schematic process for the photocatalytic reduction of CO2 to CO using [Ru(bpy)3]Cl2 as a light absorber and xCoNU as a catalyst

Full size image

4 Conclusions

In summary, UiO-66-NH2 decorated with highly dispersed cobalt has been developed as an efficient catalyst with NaBH4 treating method for CO2 photoreduction. It was proved that the decorated Co atoms could enhance the charge separation and perform as the active centers for CO2 reduction. Our catalyst showed good performance for the photoreduction reaction under simulated solar irradiation. 5CoNU achieved a high CO generation rate of 1.93 mmol g−1 h−1 and H2 generation rate of 1.18 mmol g−1 h−1 in 3 h. Compared to Co powder, the TON of 5CoNU was lifted approximately 40 times due to the highly dispersed cobalt. This work provides a simple and convenient method to modify MOFs with metal cocatalyst for photocatalytic application.