A CO2 adsorption-enhanced semiconductor/metal-complex hybrid photoelectrocatalytic interface for efficient formate production

In photoelectrochemical CO2 conversion, the concentration of fixed CO2 on the photocathode surface is of primary concern. Herein, a CO2 adsorption-enhanced semiconductor/metal-complex hybrid photoelectrocatalytic interface was established by utilizing a carbon aerogel as the CO2 fixation substrate. In CO2 reduction photoelectrocatalysis, Co3O4 was employed as the light harvester, and Ru(bpy)2dppz was utilized as the electron transfer mediator and CO2 activator. The CO2 surface concentration exhibited a 380-fold increase on this hybrid interface than that on Co3O4/FTO. The CO2 conversion to formate occurred at an onset potential of 0.45 V (vs. normal hydrogen electrode, NHE) under photoelectrochemical conditions, 160 mV more positive than its thermodynamic redox potential. At an applied potential of 0.60 V (vs. NHE), the selectivity of the formate yield reached 99.95%, with a production rate of approximately 110 mmol cm 2 h 1 and a Faradaic efficiency of 86%. Such a conversion has an electron transfer rate of 2.94 10 3 cm s . The CO2 conversion to formate was confirmed to be an instantaneous proton-coupled electron transfer process, originating from the rapid photoelectrochemical activation of bpy and dppz in Ru(bpy)2dppz as well as the synergic effect of the promoted CO2 adsorption and the applied molecular catalysis.

A CO 2 adsorption-enhanced semiconductor/ metal-complex hybrid photoelectrocatalytic interface for efficient formate production † Xiaofeng Huang, a Qi Shen, a Jibo Liu, a Nianjun Yang* b and Guohua Zhao* a In photoelectrochemical CO 2 conversion, the concentration of fixed CO 2 on the photocathode surface is of primary concern. Herein, a CO 2 adsorption-enhanced semiconductor/metal-complex hybrid photoelectrocatalytic interface was established by utilizing a carbon aerogel as the CO 2 fixation substrate.
In CO 2 reduction photoelectrocatalysis, Co 3 O 4 was employed as the light harvester, and Ru(bpy) 2 dppz was utilized as the electron transfer mediator and CO 2 activator. The CO 2 surface concentration exhibited a 380-fold increase on this hybrid interface than that on Co 3 O 4 /FTO. The CO 2 conversion to formate occurred at an onset potential of À0.45 V (vs. normal hydrogen electrode, NHE) under photoelectrochemical conditions, 160 mV more positive than its thermodynamic redox potential. At an applied potential of À0.60 V (vs. NHE), the selectivity of the formate yield reached 99.95%, with a production rate of approximately 110 mmol cm À2 h À1 and a Faradaic efficiency of 86%. Such a conversion has an electron transfer rate of 2.94 Â 10 À3 cm s À1 . The CO 2 conversion to formate was confirmed to be an instantaneous proton-coupled electron transfer process, originating from the rapid photoelectrochemical activation of bpy and dppz in Ru(bpy) 2 dppz as well as the synergic effect of the promoted CO 2 adsorption and the applied molecular catalysis.

Introduction
Natural photosynthesis converts CO 2 into a series of intermediate products (e.g., malate and pyruvate) and finally to glucose with the assistance of solar light. 1 In such a carbon recycling process, CO 2 is fixed by ribulose-1,5-biphosphate carboxylase/oxygenase (RuBisCo). Solar light is harvested, and the electrons are generated by cytochrome. A proton-coupled electron transfer is then regulated and mediated by nicotinamide adenine dinucleotide phosphate hydride, leading to the photochemical reduction of CO 2 . 1 Unfortunately, such a process has a low conversion efficiency. Processes mimicking natural photosynthesis, so-called artificial photosynthesis, have been developed to increase this conversion efficiency and provide more pathways to produce valuable chemicals using CO 2 as the carbon stock. [2][3][4][5][6][7][8] CO 2 adsorption is of the greatest concern during the CO 2 reduction process because the reduction kinetics are highly correlated with the CO 2 concentration. However, the free CO 2 concentration in aqueous solution is only approximately 0.034 M, which severely deteriorates the aqueous heterogeneous reduction. Although various CO 2 absorptive materials have been proposed, [9][10][11] their CO 2 adsorption performance and mechanism during catalytic processes have not been investigated in-depth. Moreover, the obtained efficiency for the CO 2 reduction still remains to be further increased.
Semiconductor-based photoelectrochemical methods are currently popular and have been successful for CO 2 reduction in that the synergetic conjunction of photoelectrocatalysis (namely, electrocatalysis and photocatalysis) facilitates the separation of photo-induced electrons and holes under applied electric fields. 6,7 Light irradiation-induced band bending compensates the required overpotential for CO 2 reduction, causing the CO 2 conversion to occur at relatively positive potentials. Surprisingly, the majority of semiconductors exhibited nonspecific CO 2 surficial binding. In other words, the CO 2 concentration on the surface of the semiconductors was rather low. Therefore, only slow reduction kinetics were reported. 6,7 In addition, the majority of semiconductors have poor conductivities, which deteriorated the electron transfer rate as well. Thus, photoelectrochemical CO 2 reduction still occurred at relatively negative potentials in these systems, for example, at À1.0 V for MoS 2 /TiO 2 and InP/TiO 2 (vs. normal hydrogen electrode, NHE, the same as below), 12,13 at À1.1 V for ZnTe, 5 at À1.2 V for p-GaP and GaAs, 14 and at À0.9 V for Mg-CuFeO 2 . 3 Furthermore, inevitable hydrogen evolution was observed, leading to relatively low Faradaic efficiency.
As an alternative to those semiconductors, molecular catalysts, such as Ru(II) or Re(II) bipyridyl complexes, have been employed because they present unique properties in CO 2 binding affinity, and high CO 2 surface concentrations are expected. 15,16 Furthermore, the formation of a chemical bond between CO 2 and the metal centre will lead to a high turnover rate and high selectivity of CO 2 conversion during a catalytic cycle. 15,16 The main drawback of these catalysts is that additional pathways are required to activate these molecular catalysts. In most cases, an extremely negative potential is required. For example, potentials of À1.52, À1.6, and À1.73 were applied for Ru(bpy)(tpy)(NCCH 3 ), 17 Re(tBu-bpy)(CO) 3 Cl, 18 and Re(pbn)(CO) 3 Cl, 19 respectively, to achieve the catalytic reduction of CO 2 . These molecular catalysts encountered destabilization (e.g., dimerization or dissolution) during a long-term electrolysis process, resulting in rapidly reduced photoelectrochemical activity for CO 2 conversion. 20,21 The complementary combination of a semiconductor and a molecular catalyst was then developed for photoelectrochemical CO 2 conversion. 14,22-24 These combinations accelerated electron transfer generation and lowered the activation barriers for both molecular catalysts and CO 2 binding. The highly efficient and highly selective generation of hydrocarbon fuels, such as methanol and formic acid, has been reported on these systems. 14,22-24 However, these systems have shown significant flaws, such as high overpotentials (e.g., À1.0 V on H-Si-Re(tBu-bpy)(CO) 3 Cl), 14 poor selectivity of reduction products (e.g., the simultaneous formation of HCOOH, H 2 and CO on Ag/TaON-Ru(bpy) 3 -(CH 2 ) 2 -Ru(bpy)(CO) 2 Cl 2 , while HCOOH occupies only 56.5%), 23 and low Faradaic efficiency (e.g., only 62.3% on [Ru(dcbpy) 2 (CO) 2 ] n -p-InP-Zn). 24 Thus, the aim of this work is to construct an adsorptive photoelectrochemical interface for CO 2 reduction with better performance. The constructed interface is schematically shown in Scheme 1. In our concept, CO 2 fixation is promoted by utilizing an adsorptive substrate, a conductive, micropore-dominated three-dimensional carbon aerogel (CA) with a high surface area (up to 1815 m 2 g À1 ); solar light is efficiently harvested by Co 3 O 4 micro-flowers, a visible-light driven and robust photoelectrocatalyst. These highly index-faceted Co(III)-enriched {12% 1} structures are grown epitaxially inside CA networks. An enzyme-mimicking molecular catalyst, Ru(bpy) 2 dppz, is immobilized on Co 3 O 4 /CA to accelerate and regulate the electron transfer for CO 2 reduction. Using such an interface, the CO 2 conversion to formate has been achieved at an onset reduction potential under photoelectrochemical conditions as low as À0.45 V, with a yield of approximately 110 mmol cm À2 h À1 , a selectivity of 99.95%, and a Faradaic efficiency of 86% at À0.60 V.

Experimental section
Fabrication of the Ru(bpy) 2 dppz-Co 3 O 4 /CA photocathode Monolith bulky CA was synthesized via an ambient pressure resorcinol-formaldehyde (RF) drying method. The detailed synthesis process of CA is shown as Scheme S1 in the ESI. † A solvothermal reaction was utilized to synthesize Co 3 O 4 /CA. In a 50 mL acetone-water mixture (V acetone : V water = 5 : 45), 1.4552 g of Co(NO 3 ) 2 Á6H 2 O, 0.2593 g of NH 4 F, and 1.4019 g of hexamethylenetetramine were added. Magnetic stirring of this mixture for 10 min led to the formation of a pink transparent solution. The solution was then transferred to a Teflon-lined stainless steel autoclave, where one slide of CA (4 Â 1.5 cm 2 ) was located. The solvothermal reaction was conducted at 95 1C for 24 h. After cooling down to room temperature, deposition (pink to violet colour) occurred on the CA. The as-prepared sample was carefully rinsed with acetone and then dried under vacuum at 60 1C for 2 h, followed by a calcination process in a N 2 atmosphere at 450 1C for 2 h. The ramping rate was 10 1C min À1 . After such a calcination process, the Co 3 O 4 /CA sample was synthesized. The amount of Co 3 O 4 /CA loading, averaged over the geometric area of the CA, was typically 8 mg cm À2 . A similar procedure was applied for the preparation of the control sample (Co 3 O 4 /FTO). The loading density of Co 3 O 4 on FTO was 5 mg cm À2 . In this case, a piece of fluorine-doped tin oxide glass electrode (FTO, 4 Â 1 cm 2 ) was put into the autoclave instead of CA.
Chemical polymerization was used to decorate the Co 3 O 4 /CA with Ru(bpy) 2 dppz (the detailed synthesis procedure for Ru(bpy) 2 dppz is provided in the ESI †). Briefly, 0.0040 g of Ru(bpy) 2 dppz (5.75 mmol) was dissolved in 1 mL of acetonitrile solution, denoted as solution A. Next, 0.0650 g of Fe(NO 3 ) 3 Á9H 2 O and 8 mL of pyrrole were dissolved in 1 mL of ethanol, and the mixture was denoted as solution B. The stock solution for the electrode coating was obtained by mixing solutions A and B (with equal volumes) and shaking. One slide of Co 3 O 4 /CA was then coated with 200 mL of the stock solution in a dropwise manner, and the composite was dried at 60 1C for 5 min. After repeating the coating procedure 10 times, Ru(bpy) 2 dppz-Co 3 O 4 /CA was obtained.

Electrochemical quartz crystal microbalance
To calculate the amount of CO 2 adsorbed onto the photocathodes, electrochemical quartz crystal microbalance (EQCM) experiments were conducted on a CHI440A (CH Instruments Inc., USA) with a Au-coated AT-cut quartz crystal (a fundamental frequency of 8 MHz) as the working electrode.
Approximately 10 mg of ground CA powder was ultrasonically dispersed in 1 mL of H 2 O along with 20 mL of 2% Nafion-117 film solution (Alfa Aesar). After intense ultrasonication for 5 min, 10 mL of the dispersed solution was dip-coated onto the Au electrode and dried in air. The CA-modified electrode was further used in the EQCM studies.

In situ IR spectroelectrochemical experiments
For the in situ IR spectroelectrochemical experiments, the mercury-cadmium-telluride detector was cooled down to 77 K by liquid nitrogen. A self-made four-bottleneck cell with a CaF 2 window (2 mm thickness, 25 mm in diameter) was used as the electrochemical cell. A Pt disk electrode (5 mm in diameter), a Ag/AgCl (filled with saturated KCl) electrode, and a Pt foil (1 Â 1.5 cm 2 ) were used as the working, reference and counter electrodes, respectively. A mixture of 0.010 g mL À1 Ru(bpy) 2 dppz and 2% Nafion-117 was stirred in a sonication bath for 1 min. This mixture (10 mL) was dipcoated onto the Pt disk electrode and dried under ambient conditions. One hundred IR spectra were collected with a spectrum resolution of 8 cm À1 and subsequently averaged at each potential. For these experiments, the solution was purged with CO 2 for at least 30 min to completely remove the dissolved oxygen and saturate the solution with CO 2 .

Photoelectrochemical characterization and reduction of CO 2
To evaluate the performance of the Ru(bpy) 2 dppz-Co 3 O 4 /CA catalyst, constant potential photoelectrolysis of CO 2 was conducted in order to evaluate the performance of Ru(bpy) 2 dppz-Co 3 O 4 /CA, which was conducted in a home-made H-type cell with a maximum volume of 100 mL. The as-prepared Ru(bpy) 2 dppz-Co 3 O 4 /CA working electrode and the Ag/AgCl (filled with saturated KCl) reference electrode were placed in the cathodic chamber, while the counter electrode, a graphite plate (4 Â 1 cm 2 ), was placed in the anodic chamber. The two chambers were connected with 0.1 M NaHCO 3 but separated with a Nafion-117 proton exchange membrane (Dupont). Prior to the experiments, the electrolyte in the cathodic chamber was purged with high-purity CO 2 (99.99%) gas for more than 30 min at a flow rate of 20 mL min À1 . Negative potentials (0.0, À0.2, À0.4, À0.6, À0.8, À1.0 V) were applied to the photocathode through the electrochemical workstation. An APLS-SXE300 xenon lamp with a UV cutoff (l 4 420 nm, light intensity at 9 mW cm À2 ) was used as the light source and illuminated on the Ru(bpy) 2 dppz-Co 3 O 4 /CA photocathode upon the addition of negative potential.
After such constant potential photoelectrolysis for 8 h, the reduction products were collected and quantitatively determined by HPLC and GC using the same procedure as described previously. 25 For the products in the aqueous phase, 0.2 mL of the liquid sample was collected and transferred into a 10 mL test-tube. The pH of the sample was adjusted to neutral by adding 0.2 mL of pH 7.6 phosphate buffer solution. Subsequently, 2,3,4,5,6-pentafluorobenzyl bromide (20 g L À1 , 1.0 mL) was added. The mixture was shaken for 1 min and then kept at 60 1C for another 1 h. The esterification product was extracted with 2.0 mL of n-hexane and centrifuged at 3000 rpm for 5 min. The upper layer was the organic phase, which was filtered through a 0.45 mm membrane. A C18 column was used with a mobile phase consisting of 65% methanol and 35% H 2 O at a flow rate of 1.0 mL min À1 . The detection wavelength was 225 nm. For gaseous products, 1.0 mL of the gas sample was collected through a syringe. The detection conditions were an injection inlet temperature of 130 1C, an oven temperature of 80 1C, a detector temperature of 150 1C, N 2 carrier gas, and a gas flow rate of 0.2 L min À1 .
Prior to an isotopic 13 C experiment, 13 CO 2 ( 13 C enrichment 98%) was purged into 0.1 M NaH 13 CO 3 ( 13 C enrichment 98%) electrolyte solution for at least 30 min in order to fully expel oxygen and other impurity gases. The photoelectrochemical reduction of 13 CO 2 saturated NaH 13 CO 3 (0.1 M) was identical to the procedure described in the Experimental section, holding the constant potential at À1.2 V vs. Ag/AgCl. Blank experiments using nitrogen purged Na 2 SO 4 (0.1 M) were also conducted using an identical procedure. After reduction, 0.5 mL of catholyte solution was mixed with 0.1 mL of D 2 O (Sigma Aldrich) containing 0.5 mL of DMSO as the internal standard. A one-dimensional 1 H nuclear magnetic resonance (NMR) spectrum was recorded with water suppression using a pre-saturation method.

Enhanced CO 2 adsorption
For photoelectrochemical CO 2 conversion, the surface concentration of CO 2 on the photocathode, namely, CO 2 fixation, is a View Article Online chief concern. A high surface concentration of CO 2 on the photocathode accelerates the photoelectrochemical kinetics. Herein, the surface concentration of CO 2 (G ads ), the normalized amount of adsorbed CO 2 with the electrochemical active surface area (S EASA ), was adopted as the parameter to evaluate the efficiency of CO 2 fixation on the photocathodes. The results in Table S1 (ESI †) demonstrate the advantage of using a CA substrate for CO 2 fixation. This statement is further supported by the EQCM results of CA in a CO 2 -saturated electrolyte under negative potentials, as shown in Fig. S1 (ESI †). The mass addition is 6-10 times heavier on activated CA than that on an Au electrode. For example, at À0.4 V, a mass addition of 174 ng cm À2 was obtained on a CA substrate, whereas it was barely observed on an Au quartz substrate. At À0.6 V, the mass addition on the CA substrate was 8.2-fold larger than that on the Au quartz substrate. At À0.9 V, the mass addition reached 143.5 ng cm À2 on the CA substrate. When the applied potentials were more negative than À0.9 V, the mass addition on the CA substrate was even larger than that on the Au quartz substrate. CA is not capable of reducing CO 2 electrochemically; thus, the mass addition on the CA substrate is mainly ascribed to the promoted electro-sorption of carbonaceous species (e.g., CO 2 and HCO 3 À ) from the electrolyte. The carbonaceous species are significantly promoted because the micropores contributed a high Brunauer-Emmett-Teller surface area (S BET ) of up to 1815 m 2 g À1 (Fig. S2, ESI †). This high S BET offers numerous sites for CO 2 adsorption, as indicated by Fujishima et al. and Yaghi et al., 21,26 as CO 2 molecules tend to be adsorbed in the micropores of a material. This value is also higher than that of common porous carbon materials, such as ordered mesoporous carbon (812.3 m 2 g À1 ), commercially available Vulcan (237.9 m 2 g À1 ) and carbon nanocoil (233 m 2 g À1 ). 27 The superior performance of CA supports the notion that CO 2 adsorption on other high-surfacearea carbon materials (e.g., activated carbon, carbon nanotubes) might result from the following aspects. First, the microporous feature of CA (Fig. S2, ESI †) offers numerous sites for CO 2 adsorption. Second, the activated CA possesses a relatively high surface area (1815 m 2 g À1 ), which provides many sites for Co 3 O 4 loading. According to our measurements, the Co 3 O 4 loading amount was increased to 8 mg cm À2 compared to 5 mg cm À2 on FTO. The S BET of Co 3 O 4 /CA of 713 m 2 g À1 (Fig. S2a, ESI †) is considerably higher than that of commonly designed porous inorganic semiconductor electrodes. 28,29 This result is further supported by the well-maintained micropore domination of the electrode (Fig. S2b, ESI †). Based on the voltammetry (Fig. S3, ESI †) and chronocoulometry (Fig. S4, ESI †) for Co 3 O 4 /CA, G ads exhibits a 20-fold increment (0.25 pmol cm À2 ) compared to that (0.01 pmol cm À2 ) of Co 3 O 4 /FTO. The electrochemically active surface area (S EASA ) was determined to be 8015 cm 2 (Table S1, calculated from Fig. S5, ESI †) with respect to that of Co 3 O 4 /FTO (316 cm 2 ). Experimental tests on the effects on CO 2 fixation of other porous carbon-based substrates are currently in progress.
As shown in the scanning electron microscopic (SEM) images of Co 3 O 4 /CA (Fig. S6, ESI †), the CA backbone is clearly visible even after the solvothermal growth of Co 3 O 4 microflowers. The XRD patterns of both CA and Co 3 O 4 in Co 3 O 4 /CA can also be clearly observed (Fig. S7, ESI †). In other words, such an epitaxial growth pattern fully exposes the adsorption sites to CO 2 , which results in an increase in the G ads for CO 2 . Moreover, the 3D structure of CA allows for the penetration of electrolytes into the pores of the electrode, leading to an increase in the interfacial area between the electrode and the electrolyte. Finally, CA possesses a high conductivity (electrical resistivity o 40 mO), similar to those of carbon nanotubes and graphene.
G ads further increases to 3.79 pmol cm À2 when Co 3 O 4 /CA is coated with a molecular catalyst, Ru(bpy) 2 dppz, which is 15.2 and 300 times larger compared to those on Co 3 O 4 /CA and Co 3 O 4 /FTO, respectively. However, the S EASA of Co 3 O 4 /CA is approximately 4 times larger than that of Ru(bpy) 2 dppz-Co 3 O 4 /CA, likely due to the blockage of Co(III) active sites by immobilized Ru(bpy) 2 dppz during the chemical polymerization. Nonetheless, the active area of this electrode is maintained at 1.64 Â 10 À2 cm 2 (Table S1 and Fig. S3, ESI †). The existence of the Ru(bpy) 2 dppz molecular catalyst promotes CO 2 fixation on the photocathode with an enhancement factor of 300. G ads was also normalized by the catalyst weight (Table S1, ESI †) and exhibited the same trends as that normalized by S EASA . Hence, the enhancement of CO 2 adsorption and photoelectrochemical density is in fact the synergic effect of multiple factors: the electrosorption from the CA substrate, the Co(III) sites from the high index faceted Co 3 O 4 on CA and the electrochemical activity of Ru(bpy) 2 dppz. These results confirm that CA, Co 3 O 4 and Ru(bpy) 2 dppz co-promote CO 2 fixation and lead to a high CO 2 surface concentration on such an interface. Thus, accelerated kinetics are expected for photoelectrochemical CO 2 conversion.

Photoelectrochemical CO 2 conversion
The reactivity of Ru(bpy) 2 dppz as a homogenous catalyst toward electrochemical CO 2 conversion was first studied in CO 2saturated NaHCO 3 solution. As shown in Fig. S8 (ESI †), the CO 2 reduction potential remains at À0.40 V in the current aqueous solution, which is at least 950 mV more positive than those reported in organic solvents. 30 This is the first time that a molecular catalyst, polybipyridyl Ru(II), was observed to convert/ reduce CO 2 at such a low potential, indicating the high electron transfer ability of Ru(bpy) 2 dppz toward CO 2 conversion. CO 2 conversion on Ru(bpy) 2 dppz-Co 3 O 4 /CA was investigated in detail using a potentiodynamic mode. As seen in Fig. 1a, CO 2 conversion/reduction occurs at an onset potential of À0.45 V under photoelectrochemical conditions (red line). This potential is an ''underpotential'' of approximately 160 mV with respect to the thermodynamic redox potential for CO 2 to formic acid. 31 The peak potential is approximately À0.61 V, considerably lower than other reported values, e.g., À1.33 V on p-Cu 2 O immobilized by Re(tBu-bipy)(CO) 3  This higher k s value suggests an accelerated and enhanced electron transfer on Ru(bpy) 2 dppz-Co 3 O 4 /CA. Such a method to evaluate the efficiency of CO 2 conversion is entirely new. According to our knowledge, this method has not been presented in the literature to date. Considering the CO 2 reduction potential and k s on both electrodes, the catalytically active centre for CO 2 reduction is Ru(bpy) 2 dppz-Co 3 O 4 /CA, not Co(III). This statement is further demonstrated from the Tafel plots described below.
The current density reaches approximately 8.1 mA cm À2 on Ru(bpy) 2 dppz-Co 3 O 4 /CA, which is considerably more intense than sole Ru(bpy) 2 dppz and those reported using similar molecular catalyst-semiconductor composites. For instance, on Ru(dcbpy) 2 (CO) 2 /p-InP and CO-dehydrogenase/p-NiO, the photocurrent density reached only the level of mA cm À2 at similar potentials. A photocurrent density larger than À2.0 mA cm À2 was obtained on a Re(tBu-bpy)(CO) 3 Cl/Cu 2 O photocathode, but a potential of À2.0 V was applied. 32 So, photoelectrochemical CO 2 conversion on Ru(bpy) 2 dppz-Co 3 O 4 /CA occurs at a low overpotential but with a high photocurrent density.
The current densities shown in this paper were obtained using the geometric areas of the electrodes. Due to their porous structures, CA-based photoelectrodes will have a higher S EASA value. Their S EASA values were determined using surface-sensitive redox probes of Fe(CN) 6 3À/4À (Table S1 and Fig. S3 To compare our results with those obtained using other porous materials presented in the literature, the current densities shown throughout the paper were then calculated using the geometric areas of the photoelectrodes, the most frequently applied approach for electrochemical and photoelectrochemical CO 2 reduction. Fig. 1b shows the LSVs on Ru(bpy) 2 dppz-Co 3 O 4 /CA in 0.1 M NaHCO 3 solution purged with N 2 or CO 2 . The pH value of 0.1 M NaHCO 3 is in the range of 8.3-8.5 after being purged with N 2 . Once it was saturated with CO 2 , the pH value decreased to the range of 6.5-7.0. 34 Such a weak acidic environment provides a favourable protic environment for CO 2 reduction using a Ru(bpy) 2 dppz catalyst. [2][3][4][5] Notably, the peak potential for CO 2 reduction is 50 mV more positive than that of its N 2 -purged counterpart, with the peak current density doubled. These results clearly confirm the involvement of protons in CO 2 reduction. Fig. 1b shows the effect of light irradiation on CO 2 conversion as well. On Ru(bpy) 2 dppz-Co 3 O 4 /CA without light irradiation, the peak current density of the CO 2 reduction decreases by a ratio of 35% along with a 50 mV negative shift in the peak potential compared to the case when light irradiation is applied. Similar tendencies were observed on Co 3 O 4 /CA and Co 3 O 4 /FTO (Fig. S9, ESI †). This phenomenon could be due to the photoelectrocatalytic properties of the Ru(bpy) 2 dppz-Co 3 O 4 /CA electrode.
An amperometric photocurrent response was further investigated at a fixed potential of À0.40 V. As shown in Fig. 1c, the steady photocurrent density reaches approximately 1.5 mA cm À2  The potentiodynamic voltammograms in 0.1 M Na 2 SO 4 also reflected that Ru(bpy) 2 dppz-Co 3 O 4 /CA showed a distinguished photoelectrochemical activity. The difference between the photocurrent ( j PEC ) and dark current ( j EC ), denoted as j PEC À j EC , is summarized in Fig. S10 (ESI †). j PEC À j EC is approximately 10 times higher on Ru(bpy) 2 dppz-Co 3 O 4 /CA than on FTO. This result could be explained by several reasons. First, the {12% 1} crystal facet of Co 3 O 4 (with an inter-planar spacing of 0.286 nm) being finely exposed (Fig. S11, ESI †), which enriches triply uncoordinated Co(III) sites. Such a high-index-facet Co 3 O 4 facilitates the photoelectrochemical reduction of CO 2 , as our previous work has indicated. 4 Second, the epitaxial growth manner of high-index-facet Co 3 O 4 on CA provided a direct electron transfer channel. Photo-induced holes could rapidly transfer to the highly conductive CA network. Electrons and holes are efficiently separated. Finally, the ingenious merging of photocatalysis and electrocatalysis on a single surface has been found to efficiently separate the photo-induced carriers, resulting in a reduction of the overpotential for CO 2 conversion. 2,4 The applied negative potential creates a more upward bending for Co 3 O 4 , a p-type semiconductor. Then, the driving force for photoelectrons to cross the semiconductor-electrolyte junction is enlarged, resulting in an enhanced photocurrent density. In the meantime, the light-induced upward band-bending of Co 3 O 4 /CA lifts the Fermi level, which compensates part of the required applied negative potential under dark conditions, yielding a reduced overpotential for CO 2 reduction. All of these effects reveal the essential role of Co 3 O 4 as the photoelectrocatalyst in a photoelectrochemical strategy for CO 2 conversion. Comparing the cyclic voltammograms of different photoelectrodes, the lowest overpotential and the highest photocurrent density for CO 2 conversion were obtained on Ru(bpy) 2 dppz-Co 3 O 4 /CA. Apart from the respective contributions from the enhanced adsorption on CA and distinguished photoelectrochemical properties of Co 3 O 4 /CA, the molecular catalyst Ru(bpy) 2 dppz also exhibits excellent electrochemical reductive activity toward CO 2 reduction, as suggested by the above-mentioned k s . The synergistic effect among those effects exerted on Ru(bpy) 2 dppz-Co 3 O 4 /CA has superiority in magnifying the photocurrent density, reducing the overpotential, and accelerating and regulating the electron transfer pathway for the photoelectrochemical CO 2 reduction process.
The photoelectrochemical CO 2 conversion on Ru(bpy) 2 dppz-Co 3 O 4 /CA was further studied in CO 2 -saturated 0.1 M NaHCO 3 by varying the applied potential. Its variation with the potentials applied is summarized in Fig. 2a. The amplitude of j(CO 2 ) À j(N 2 ) increases when the potential is more negative, following the order of Co 3 O 4 /FTO o Co 3 O 4 /CA o Ru(bpy) 2 dppz-Co 3 O 4 /CA. A sharp increase in the amplitude of j(CO 2 ) À j(N 2 ) occurs at À0.4 V. At a reduction potential of À0.6 V, the net photocurrent reaches À0.35 mA cm À2 . This value is twice as high as that on Co 3 O 4 /CA and nearly 10 times higher than that on Co 3 O 4 /FTO.
Under identical photoelectrolysis conditions, the yield rate of formate on these four photocathodes is in the order of Ru(bpy) 2 dppz-Co 3 O 4 /CA 4 Co 3 O 4 /CA 4 Ru(bpy) 2 dppz/CA 4 Co 3 O 4 /FTO at the same applied potentials, as shown in Fig. 2b. For example, 869.8 mmol formate was produced at a potential of À0.6 V on Ru(bpy) 2 dppz-Co 3 O 4 /CA. This value is considerably higher than that presented previously (384 mmol) on hierarchical Co 3 O 4 , even at À0.7 V. 5 If the electrode area is considered, the predicted yield rate of formate on Ru(bpy) 2 dppz-Co 3 O 4 /CA is approximately 450 mmol cm À2 at the onset potential of À0.45 V under photoelectrochemical conditions. At À0.7 V, the yield rate of formate reaches approximately 900 mmol cm À2 . Considering the reduction time span further, the yield rate of formate reaches nearly 110 mmol cm À2 h À1 , even at a reduction potential of À0.6 V. The yield rate of formate on Ru(bpy) 2 dppz-Co 3 O 4 /CA is higher than most reported values, e.g., nearly 29-, 5.2-, and 2-fold higher than that on Ru(II)-Re(I) multinuclear complexes, 22 Ru(II) electro-polymerized p-InP-Zn, 24 and Cu 2 ZnSnS 4 , 35 respectively. The estimated Faradaic efficiency for the yield of formate is approximately 86%, which is comparable to that on semiconductor-molecular photocatalytic systems as well as on other semiconductor-based photocatalytic systems, such as Mg-doped CuFeO 2 3 and Au/ZnO/ZnTe. 5 Notably, formate yield rates on the Ru(bpy) 2 dppz/CA were lower than that on the Co 3 O 4 /CA at each investigated potential, the reason is described below.

View Article Online
To confirm the carbon source for the reduction product of formate, isotopic measurements were conducted using 13 CO 2 -saturated NaH 13 CO 3 (0.1 M) electrolyte solution. 36,37 Control experiments were also performed in N 2 -purged Na 2 SO 4 (0.1 M) solutions. These related 1 H-NMR spectra are shown as Fig. 2c. The spectrum obtained from the N 2 -purged Na 2 SO 4 (0.1 M) solutions is featureless (Fig. 2c-I) in the chemical shift range of 9.2 to 7.7, indicating that no formate is produced when no carbonaceous species (e.g., CO 2 , HCO 3 À ) are present. In Fig. 2c-II, a singlet peak at the chemical shift value of 8.42 is detected, likely resulting from the yield of HCOO À from CO 2 -saturated NaHCO 3 . When 13 CO 2 -saturated NaH 13 CO 3 solution is used, a doublet peak is shown at the chemical shift values of 8.64 and 8.25 (Fig. 2c-III). This doublet peak indicates the production of H 13 COO À . 3 In other words, formate is obtained from the 13 C species in the electrolyte rather than from the impurities in the electrolyte or on the surface of the photocathode. Photoelectrochemical CO 2 reduction thus occurs on the adsorption-enhanced molecular catalystsemiconductor hybrid interface.
To investigate the energy conversion efficiency, the turnover number (TON) and turnover frequency (TOF) for photoelectrochemical CO 2 reduction were estimated at different potentials on this interface. These values are tabulated in Tables S2 and S3 (ESI †). Co 3 O 4 /CA has a 3-to 5-fold increment in both TON and TOF compared to Co 3 O 4 /FTO, resulting from the effect of improved CO 2 adsorption. This result further demonstrates the importance of the primary CO 2 fixation process. On the Ru(bpy) 2 dppz molecular catalyst immobilized interface, the magnitudes of both TON and TOF were increased by approximately 1 order. This trend was also proven by the results from [Zn(II)TRP] 4+ /[SiW 12 O 40 ] 4À 38 and COF-367-Co, 21 both of which possess a high surface area, thus resulting in a high TON value. The TON value was increased with the negative shift of the potential. For example, at À0.6 V vs. NHE, the TON value already reached 978.7, with a TOF value of 122. In Table 1, these results are further compared with those shown in the literature. The higher TON value (978.7) in our case compared to those reported (including those of similar semiconductor-molecular complex model systems (e.g., InP/[MCE2-A + MCE4]-TiO 2 /Pt, 39 the NiO-RuRe complex 40 and Ru(dcbpy)/N-Ta 2 O 5 41 )) demonstrates the distinguished photoelectrochemical performance of the concurrent Ru(bpy) 2 dppz-Co 3 O 4 /CA system.
Hydrogen, as the only side product detected, only occupied 0.05% of the total products, which is indicative of the high selectivity of Ru(bpy) 2 dppz-Co 3 O 4 /CA toward CO 2 reduction. On Co 3 O 4 /CA, the hydrogen production is 3 times (Fig. S12b, ESI †) higher, but the yield rate of formate is approximately 1.5 times lower (Fig. 2b). On Ru(bpy) 2 dppz/CA (Fig. S12c, ESI †), lower amounts of H 2 were detected after applying a potential of À0.9 V. On Co 3 O 4 /FTO (Fig. S12d, ESI †), the yield rate of formate is the lowest, although no hydrogen is detected. The selectivity for Ru(bpy) 2 dppz-Co 3 O 4 /CA is higher than those of Ru(II)-electropolymerized p-InP-Zn 24 and Cu 2 ZnSnS 4 , 35 in which non-negligible amounts of H 2 and CO were generated. Therefore, the selectivity of the CO 2 conversion to formate is extremely high on Ru(bpy) 2 dppz-Co 3 O 4 /CA (99.95%).
The hydrogen evolution potentials were further estimated based on the potential-dependent hydrogen evolution profiles shown in Fig. 2b and Fig. S12a-d (ESI †). The potentials on Ru(bpy) 2 dppz-Co 3 O 4 /CA, Co 3 O 4 /CA, Ru(bpy) 2 dppz/CA and Co 3 O 4 / FTO are À0.6 V, À0.7 V and À0.9 V vs. NHE, respectively. However, there are no clear cathodic peaks shown in Fig. 1 for hydrogen evolution because the Ru(II) centre on the molecular catalyst Ru(bpy) 2 dppz has fully coordinated to the nitrogen atoms from two 2,2 0 -bipyridine and dppz ligands. Then, both the bipyridine ligand and dppz are not able to dissociate from Ru(II) during the electrochemical processes. In other words, there will be no opportunity to generate any catalytic wave for hydrogen evolution. These results indicate that this unwanted reaction has not been involved in our system during CO 2 reduction. Thus, an improved conversion efficiency and high selectivity are expected.
The stability of the proposed photoelectrocatalytic interface was examined by recording the XRD patterns and cyclic voltammograms of Ru(bpy) 2 dppz-Co 3 O 4 /CA before and after a long-term photoelectrochemical CO 2 reduction. The XRD patterns of both CA and Co 3 O 4 do not vary after the 8 h photoelectrochemical reduction of CO 2 (Fig. S13, ESI †). The peak current density of Ru(bpy) 2 dppz-Co 3 O 4 /CA is slightly enhanced after 100 cycles of cyclic voltammetry (Fig. S14, ESI †). These facts confirm the high stability of our molecular catalyst-semiconductorassembled photocathode.
The underlying mechanism of such a synergic effect, the high selectivity for formate and its relation to the photoelectrochemical CO 2 reduction is discussed. Prior to this discussion, the interfacial energetics between the two components were analysed. The bandgap (Fig. S15, ESI †) and the flat-band potential (Fig. S16, ESI †) of Co 3 O 4 /CA were estimated to be 1.87 V and 0.37 V, respectively.

View Article Online
Provided that the valence band of Co 3 O 4 is 0.1 V more positive than the flat band potential, the valence band of Co 3 O 4 was estimated to be 0.47 V. So the conduction band of Co 3 O 4 /CA was estimated to be À1.40 V vs. NHE (E g = |E CB À E VB |). In contrast, the energy difference between the HOMO and the LUMO for Ru(bpy) 2 dppz is approximately 2.29 eV, as calculated using the intersection wavelength at 540.8 nm from its normalized UV-vis and fluorescence spectra in acetonitrile (Fig. S17a, ESI †). According to the Ru(III)/Ru(II) redox potential (Fig. S17b, ESI †), the LUMO energy level of Ru(bpy) 2 dppz was calculated to be À0.71 V. Ru(bpy) 2 dppz's LUMO energy level is then 0.69 V more positive than the conduction band of Co 3 O 4 /CA. Therefore, the photo-induced electrons on Co 3 O 4 transfer to the LUMO of the molecular catalyst, as confirmed by various reports. 24,32,35,40,41 Although the value of the conduction band and the redox potential of the molecular complex vary with the pH value of the electrolyte, the shift of the redox potential of a similar molecular complex [Ru(phen) 2 (ptpbb)] 2+ by a rate of À62 mV pH À1 49 is in line with the typical À59 mV pH À1 for semiconductors.
Electrochemical characterization data can also support such an electron transfer process. The onset potential of electrochemical CO 2 reduction using Ru(bpy) 2 dppz was at À0.4 V vs. NHE, which is similar to that for photoelectrochemical CO 2 reduction using Ru(bpy) 2 dppz-Co 3 O 4 (À0.45 V vs. NHE). Moreover, the cathodic current density of electrochemical CO 2 reduction on Ru(bpy) 2 dppz (ca. 0.63 mA cm À2 ) was far smaller than that on Ru(bpy) 2 dppz-Co 3 O 4 (ca. 15.0 mA cm À2 ). From in situ IR spectroelectrochemical spectra (Fig. 3a), the upward IR peaks at 1419 and 1446 cm À1 , assigned to the A1 mode of the C-C-H deformation bending vibration on bpy À and dppz À , 50,51 validate such a statement. A relatively strong broad upward peak at 1716 cm À1 displays the same trend. This peak likely arises from the stretching of CQO from the as-formed formate. 52,53 These facts suggest that the electrons are transferred from the excited Co 3 O 4 to Ru(bpy) 2 dppz, and then take part in photoelectrochemical CO 2 reduction. As a result of such a photo-induced electron transfer to the LUMO of Ru(bpy) 2 dppz, CO 2 conversion/reduction occurs in the adsorptive substrate CA, the photocatalyst Co 3 O 4 and the molecular catalyst Ru(bpy) 2 dppz-Co 3 O 4 /CA, exactly as illustrated in Scheme 1. Such an electron transfer process is the core of such an interface. Note that Ru(bpy) 2 dppz in its activated form is capable of reducing CO 2 into formate with a two-electron process. The potential of its singly reduced form, Ru(bpy) 2 (dppz À ), is À0.95 V vs. NHE in MeCN. 30 Its doubly reduced state, Ru(bpy)(bpy À )(dppz À ), is possible to be obtained by merging Ru(bpy) 2 dppz and Co 3 O 4 . This is because the electrons at the conduction band of Co 3 O 4 bear only an energy of À1.40 V. In the presence of Co 3 O 4 , the formation of a band-alignment interface based on Ru(bpy) 2 dppz and Co 3 O 4 allows the direct injection of photo-induced electrons to the LUMO of Ru(bpy) 2 dppz, which greatly suppresses the quenching of the molecular catalyst in the pure water phase. 49 Eventually, the photocatalytic activity of Ru(bpy) 2 dppz is remarkably improved. This is further supported by the fact that the rate yield of formate, along with the TON/TOF values of the Ru(bpy) 2 dppz-Co 3 O 4 /CA after 8 h of reduction was almost 20 times higher than that on the Ru(bpy) 2 dppz/CA (Fig. 2b). And on the basis of the Marcus-Gerischer model, 54,55 the negative bias applied to the Ru(bpy) 2 dppz-Co 3 O 4 system suppresses the back electrons transferred from Ru(bpy) 2 dppz to Co 3 O 4 . The utilization of the electrons is then greatly improved, and a high heterogeneous electron transfer rate of 2.94 Â 10 À3 cm s À1 is reached.
It also noteworthy that the absence of the stretching vibrations of Ru-H, N-H, and C-H in the IR spectra indicates that the protons are not directly bonded to Ru(bpy) 2 dppz. 56,57 A broad peak at approximately 3691 cm À1 emerges from the featureless IR spectra when the applied potential is higher than À0.6 V vs. NHE (Fig. 3a).

View Article Online
This newly emerged upward peak is assigned to the stretching vibration of a non-hydrogen bonded hydroxyl. Water molecules are attracted and nearly dissociate from their bulky form due to the existence of localized electrons on the bpy ligand in Ru(bpy) 2 dppz, thus causing the appearance of the peak at 3691 cm À1 . Hydrodynamic voltammograms using rotating glassy carbon disk electrodes were recorded using the electrode coated with Ru(bpy) 2 dppz (Fig. S18a, ESI †) and Co 3 O 4 (Fig. S18b, ESI †). Fig. 3b shows the related Koutecky-Levich plot for the electrode coated with Ru(bpy) 2 dppz, indicating its non-linear features. This non-linearity is quite helpful for CO 2 conversion in that it indicates that CO 2 and the surrounding water fix into the outer sphere of Ru(bpy) 2 dppz, apart from its diffusion and surface reaction with Ru(bpy) 2 dppz under hydrodynamic conditions. 58 Taking a Langmuir-Hinshelwood bimolecular reaction mechanism as a model, 59 these results clearly confirm the enzymemimicking role of the molecular catalyst, Ru(bpy) 2 dppz, for CO 2 conversion. In contrast, a linear reaction is observed on the electrode coated with Co 3 O 4 (Fig. S18c, ESI †), indicating different electrokinetics between Co 3 O 4 and Ru(bpy) 2 dppz. The rate-determining step of the CO 2 conversion was estimated through Tafel analysis (Fig. 3c). For Ru(bpy) 2 dppz-Co 3 O 4 /CA, a Tafel slope of 145 mV dec À1 is determined. Due to the porous feature of CA, it is slightly deviated from its classical value of 118 mV dec À1 . 60,61 Therefore the rate-determining step of photoelectrochemical CO 2 conversion on Ru(bpy) 2 dppz-Co 3 O 4 /CA is the single electron transfer step rather than the proton transfer step. This result is in agreement with our previous study regarding the photoelectrochemical CO 2 conversion on hierarchical Co 3 O 4 . 4 This finding also supports the notion that the protonation of the reaction intermediate occurs rapidly in CO 2 -saturated solution. Namely, faster electrode kinetics for CO 2 conversion were obtained with the help of Ru(bpy) 2 dppz. In contrast, a Tafel slope of 191 mV dec À1 is observed for Co 3 O 4 /CA, which originates from the porous structure of CA and is indicative of its slow photoelectrochemical kinetics in the absence of Ru(bpy) 2 dppz. In addition, the activation energy of the CO 2 reduction in our system, defined as the energy required to activate CO 2 into its radicals, 62-64 was calculated to be 1.047 kJ mol À1 (see the ESI † for details). This value is presented for the first time in this study and therefore further density functionalbased theoretical calculations are required to confirm this value.
Based on the above analysis, the activation of Ru(bpy) 2 dppz is accomplished only by the photoelectrochemical process from the irradiated Co 3 O 4 /CA. Ru(bpy) 2 dppz is actually ''catalysed'' by Co 3 O 4 /CA rather than merely an electron conductor. Therefore, the reduction performance is markedly enhanced with the assistance of the photoelectrocatalyst of Co 3 O 4 . Therefore, as schematically shown in Scheme 1, the possible pathways for CO 2 conversion on Ru(bpy) 2 dppz-Co 3 O 4 /CA involve the transfer and mediation of electrons, protons, and the activated form of Ru(bpy) 2 dppz.
The CO 2 adsorption is remarkably increased through the utilization of a high-surface-area and micropore-dominated carbon aerogel substrate. The epitaxial growth of high-index-facet Co 3 O 4 along with the intrinsic highly conductive network of the carbon aerogel enables an elevated electron transfer rate with the assistance of a molecular catalyst of Ru(bpy) 2 dppz. Eventually, such an interface bears a unique proton-coupled electron-transfer reactivity and rapid electron kinetics toward CO 2 reduction. Through the simultaneous activation of the bpy and dppz ligands of the Ru(II) molecular catalyst, the photoelectrochemical conversion of CO 2 to formate has been realized on such a photocathode with an onset potential as low as À0.45 V vs. NHE. At a potential of À0.6 V vs. NHE, the yield rate of formate reaches approximately 110 mmol cm À2 h À1 , with a selectivity of 99.95% and a Faradaic efficiency of 86%. The mechanism of such a rapid photoelectrochemical electron transfer process is explained as the synergistic effect of the photoelectrochemical activation of bpy and dppz inside Ru(bpy) 2 dppz as well as the remarkably promoted CO 2 adsorption on Co 3 O 4 /CA. Although future work on the effect of the morphology of Co 3 O 4 /CA (e.g., the size, facets, and shape of Co 3 O 4 on CA) and the loading amount of Co 3 O 4 /CA (e.g., distribution and density) as well as the light density on the efficiency of the CO 2 conversion must be conducted, such an adsorption-enhanced molecular catalytic photoelectrocatalytic interface has potential for application in the production of useful chemicals from CO 2 carbon stock in the future.