The over consumption of fossil fuels in our modern society has brought about a serious energy crisis and the global warming problem. Many scientists have therefore dedicated themselves to discovering new science and technology for storing solar energy in solar fuels through water splitting [1-4], photocatalytic CO₂ reduction [5-7], and electrocatalytic CO₂ reduction in water [8, 9]. In this context, Wang's group [5] reported a new C/N material with an apparent quantum efficiency of 4.8%, and catalysts containing metal-N sites for electrochemical CO₂ reduction have been well documented by Bao et al. [8]. In fact, the plentiful solar energy is mainly stored by trees and grasses via natural photosynthesis, in which CO₂ and H₂O are converted to sugars, cellulose, etc. Recently, catalytic CO₂ reduction has emerged as one of the hotspots in the interdisciplinary fields of environment, energy, catalysis, and materials [10-23]. To achieve efficient CO₂ reduction as in the case of photosynthesis in plant cells, three processes need to be considered [24, 25]: CO₂ adsorption, quick reduction action, and product desorption. For the photoelectrocatalytic reduction of CO₂, a semiconductor with a suitable conduction band (CB) is the basic requirement for stably hosting the active photoelectrons.

Multi-electron transfer in semiconductor electrodes is a longstanding target because of the potential for application in solar-to-chemical energy conversion [26]. One of the major issues is that most semiconductors possessing a suitable band gap for solar energy absorption, such as Cu₂O [27, 28] and CuFeO₂ [29], are thermodynamically unstable in aqueous solutions. Furthermore, the surfaces of some semiconductors are not catalytically active for multi-electron transfer reactions, especially for CO₂ reduction [30]. Therefore, appropriate cocatalysts are required [31, 32].

Although TiO₂ nanoparticles have been intensively investigated in terms of their stability and economy, their performance in CO₂ reduction is poor, owing to the weak adsorption of CO₂ on the catalyst surfaces and inefficient harvesting of the solar radiation [33]. To enhance the absorption of sunlight, molecular porphyrin is commonly utilized in DSSC devices [34]. CO₂ adsorption can be realized by introducing ligands on the surfaces of photocathodes [35-37].

As is well known, porphyrin structures widely exist in nature, such as chlorophyll, hemoglobin, and vitamin B₁₂. Porphyrins have been extensively investigated in DSSCs as sensitizers owing to their brilliant light absorption ability [38-40]. Furthermore, metalloporphyrins have been increasingly applied to homogeneous and heterogeneous catalysis [41-46], including electrocatalytic CO₂ reduction [47, 48]. Recently, Yang and his coworkers [49] developed an electrocatalytic system composed of covalent frameworks and cobalt metal that has been applied in CO₂ reduction to produce CO with high selectivity. Here, metalloporphyrins were anchored through covalent bonds onto the surface of TiO₂ as the solar absorber and catalyst to improve both the light absorption and efficiency of CO₂ reduction in an artificial photosynthesis (APS) cell (Fig. 1).

Porphyrins and metalloporphyrins were prepared according to the methods reported in the literature [38, 50], and named as TCPP (D1, meso-tetra-p-carboxyphenylporphyrin), CoTCPP (D2), NiTCPP (D3), CuTCPP (D4), and PdTCPP (D5). The FTO glass plates (15 Ω/cm²) used were purchased from Nippon Sheet Glass Co. Ltd.. ¹H NMR spectra were recorded on a Varian Mercury 300/400 spectrometer by using tetramethylsilane as an internal standard. Fourier transform infrared (FTIR) spectra were recorded on a Nicolet NEXUS 670 instrument. UV-vis absorption spectra were obtained by using a UV-3600 spectrophotometer in absorption spectra of a 1 × 10^–5 M (mol/L) solution of D1–D5. DMSO:EtOH = 1:10 at room temperature. Fluorescence spectra were obtained by using an F-7000 fluorescence spectrophotometer. The lifetimes of the porphyrin derivatives were obtained with Edinburgh Instruments FS920. XRD patterns of the electrodes were recorded by using a Philips X'Pert Pro Super diffractometer with Cu K_α radiation (λ = 1.5406 Å). The SEM images and EDX maps were collected with MIRA3 XMU. TEM was carried out on a Tecnai F30 instrument.

The TiO₂/FTO electrodes were prepared by a spin coating method. A paste of TiO₂ was obtained by grinding a mixture (4.5 g P25, 0.2 mL Triton X-100, and 0.5 g polyethylene glycol 20000) with 10 mL pure water. The coated FTO glass plates were sintered at 500 ℃ for 1 h. The area of the TiO₂ film was about 1.5 × 1.5 cm², and the average weight was about 1.25 mg. These TiO₂/FTO electrodes were immersed in an aminopropyltriethoxysilane (APTES) ethanol solution (5%) and kept at 70 ℃ for 1 h. These amino-functionalized TiO₂/FTO electrodes (NH₂@TiO₂) were then placed in a 10 mL DMF solution for 24 h at room temperature, which contained 1 mM D1–D5, 2.5 mM dicyclohexylcarbodiimide, and 1 mg 4-dimethylaminopyridine. The electrodes were washed with pure water and dried at 50 ℃ in vacuum. The obtained porphyrin-modified electrodes were named as D1@TiO₂, D2@TiO₂, D3@TiO₂, D4@TiO₂, and D5@TiO₂.

After porphyrin modification of the electrodes, Pd nanoparticles were deposited onto the Dx@TiO₂ films by pulsed electrodeposition technique to yield Pd/Dx@TiO₂ electrodes.

The BiVO₄ electrodes were modified by using FeOOH and NiOOH according to the literature method [50].

Cyclic voltammetry curves of the porphyrins D1–D5 (0.1 M DMSO) were obtained by using a CHI 660E instrument with the traditional three-electrode system by employing ferrocene as the standard. The working electrode, reference electrode, and counter electrode were a glassy C disk electrode, Ag/AgCl, and a Pt wire electrode, respectively. The linear sweep voltammetry (LSV) curves of the porphyrin-functionalized electrodes were recorded by using the same reference and counter electrodes in 0.1 M aqueous NaHCO₃ solution. The j-t curves of the cell were measured by using a two-electrode system, in which BiVO₄ served as both the reference and counter electrodes upon irradiation from a 300 W Xe lamp (PLS-SXE300). The incident light intensity was 100 mW/cm² and calibrated with a standard Si solar cell. The photovoltaic reduction of CO₂ was performed in a 100 mL aqueous NaHCO₃ solution (0.1 M) saturated with CO₂.

The APS cell of Dx@TiO₂||BiVO₄ was set up in a sealed quartz cell equipped with an external Si solar cell, and filled with 100 mL of the electrolyte, 0.1 M aqueous NaHCO₃ solution. The voltage of the Si solar cell was adjusted to -0.45 to -0.9 V to mimic the membrane voltage of plants, and the system was irradiated by using simulated sunlight with the light density of 100 mW/cm².

The liquid products in the electrolyte were quantified by ¹H NMR (JNM-ECS400, JEOL) spectroscopy. 0.7 mL of the electrolyte was mixed with an internal standard solution of 0.035 mL deuterium water that contained dimethyl sulfoxide (10 mM) and phloroglucinol (50 mM). The ¹H spectra were obtained through water suppression technique [24]. The released gas products were detected by using a VARIAN CP-3380 gas chromatograph.

¹³CO₂ (Aldrich, 99%) was directly used for the isotopic labelling experiment to trace the C source by utilizing D1@TiO₂ as the photocathode. After purging the whole system with Ar gas for 30 min, 100 mL of 0.1 M NaHCO₃ was saturated with ¹³CO₂ gas in a sealed photoelectric cell.

The UV-vis absorption spectra of TCPP (D1), CoTCPP (D2), NiTCPP (D3), CuTCPP (D4), and PdTCPP (D5) are shown in Fig. 2a. Light absorption in these porphyrins occurs around 416 nm, which is assigned to the B band, and the several peaks observed in the range 500–700 nm are assigned to the Q bands of the porphyrin ring. These data, as well as the molar absorption coefficients, are listed in Table 1. Their ground-state oxidation potentials (E_ox) were determined by cyclic voltammetry (Fig. 2c) and assigned to their HOMO. The values of E_ox vary from 1.0 to 1.16 eV. The excitation energies (E_0-0 = E_LUMO – E_HOMO) were determined to be 2.03 V for TCPP, 1.97 V for CoTCPP, 1.98 V for NiTCPP, 1.95 V for CuTCPP, and 2.01 V for PdTCPP. Therefore, their excited-state potentials (E_S/S^*) could be calculated from the normalized absorption and photoluminescence spectra (E_ox − E_0-0 = E_S/S^*) [40], and were found to be higher than the potential for CO₂ reduction (-0.38 V for methanol). The energy-level diagrams of porphyrin, metalloporphyrins, and anatase (TiO₂) are presented in Fig. 2d.

The morphologies of these photocathodes were obtained by SEM and TEM technologies and are presented in Fig. 3a and 3b, respectively. It can be seen that the TiO₂ nanoparticles are uniformly compacted onto the FTO plates and their surfaces are very flat. The APTES and porphyrin modification of the TiO₂ nanoparticles via covalent bond formation can be confirmed by the absorptions corresponding to the ν_Si–O stretching vibration (1125 cm^–1) and ν_C=C vibration of the aryl ring (1539 cm^–1). On the other hand, the IR spectrum of D1@TiO₂ displays peaks in the region of 1705 to 1380 cm^–1, which are attributed to ν_C=O and ν_C–N stretching vibrations (Fig. 3c). The elemental distribution map of Pd/D3@TiO₂ electrode is presented in Fig. 3d, in which the elements C, Pd, Ti, O, N, and Ni are well distributed on the surface of the electrode, which also confirmed the successful modification.

Compared with those of pure TiO₂ and APTES-modified TiO₂ (NH₂@TiO₂), the solid UV-vis absorbances of the Dx@TiO₂ photocathodes (Fig. 4a) are evidently enhanced in the range of visible light wavelengths. The band gap values can be obtained from these UV-vis data (Fig. 4b). Furthermore, the flat-band potentials of the photocathodes were determined from the Mott-Schottky (MS) plots; positive slopes in the linear region indicate that TiO₂ is an n-type semiconductor. Because the flat-band potential of the n-type semiconductor is close to the bottom edge of the CB, these flat-band potentials can also be considered as the CB potentials (Fig. 4c). These CB potentials of the photocathodes are in the range -0.65 to -0.79 V, that is, they are more negative than the potential (-0.62 V vs. SCE) required for the conversion of CO₂ to methanol. The valence band (VB) potentials of these photocathodes can be calculated and are shown in Fig. 4d. These results reveal that porphyrin molecules could shift the CB and VB potentials of the electrodes to more negative values and thus narrow their band gaps, so that they can harvest more visible light photons and have enough power to reduce CO₂ to methanol.

In order to investigate the photoelectrocatalytic activities of the new cells, the LSV curves of the APS cells Dx@TiO₂||BiVO₄ were recorded on an electrochemical workstation at a scan rate of 100 mV/s in 0.1 M aqueous NaHCO₃ solution saturated with CO₂ (Fig. 5). The applied voltage varied from 0.0 to -1.0 V for this new APS system. We can clearly see that the zero current of the cell is observed at -0.43 V for D1@TiO₂, and positive current is measured between the voltages of 0.0 and -0.43 V, which indicate that oxidation reaction occurs at the working electrode; the negative current observed between -0.43 and -1.0 V suggests the occurrence of a reduction reaction at the working electrode. Therefore, the voltages of -0.45, -0.60, -0.75, and -0.90 V were chosen as the external voltages of the APS cell for CO₂ reduction.

In this new APS cell of D1@TiO₂||BiVO₄, when the voltage was set to -0.45 V for the photocathode, the only methanol production could be obtained at a rate of 14 μM h^–1 cm^–2; when the voltage was set to -0.6 V, the formation rate of methanol increased to 51.2 μM h^–1 cm^–2 (Fig. 6). When the voltage crossed -0.75 V, hydrogen evolved, the amount of which could be determined and enhanced through an increase in the external voltage (Fig. 7). That is consistent with its LSV curve, in which the slope of the current evidently increases when the voltage is up to -0.75 V, due to the facilitation of the reaction involving H₂ release, which competes against the decrease in CO₂ reduction (Fig. 8a). Therefore, -0.6 V was chosen as the best voltage for mimicking the membrane voltage of plants.

The different photocathodes (D2-D5) were then assembled with BiVO₄ to form the new APS cells, which worked at -0.6 V under optimal conditions. The results are presented in Fig. 8b. The cell of D3@TiO₂||BiVO₄ with NiTCPP as the light harvester and catalyst showed the best performance towards CO₂ reduction at a rate of 55.5 μM h^-1 cm^-2 and 0.92% light quantum efficiency, which are attributed to their better ability to capture protons. In contrast, the cells D2, D4@TiO₂||BiVO₄ showed lower formation rates of methanol owing to the good oxidative properties of CoTCPP and CuTCPP. The cell D5@TiO₂||BiVO₄ also exhibited a lower formation rate of methanol owing to a higher polarization potential (0.64 V) of the photocathode (Fig. 5).

The stability and recyclability of electrodes are important factors in practice. A long-term test over 8 h was conducted for D3@TiO₂ photocathode by repeating the test every 2 h. The results are presented in Fig. 9. The catalytic activity diminished to half the original value because the porphyrin molecules serve as not only a catalyst but also a sensitizer in this APS cell.

In order to trace the C source, ¹³CO₂ labelling experiments were carried out. After purging the whole system with Ar, ¹³CO₂ was bubbled into the electrolyte. The ¹³C-labeled methanol was detected by using GC-MS (Fig. 10).

On the basis of our experimental results, a possible mechanism has been proposed in Scheme 1. The electrons in the HOMO of porphyrin harvest photons and jump into the LUMO, then, the generated photoelectrons are quickly injected into the CB of TiO₂ and transferred to TCPPs, where they can be captured by the protons adsorbed on the metal and N atoms of the Lewis base center to produce H atoms. The CO₂ molecules activated by the N of the amide or the N of metalloporphyrin are then reduced immediately to methanol by these highly active H atoms, like in a Calvin cycle. When the electrons of OH^- groups are transferred to the BiVO₄ photoanode and then flow through an external circuit, O₂ would be released. The external electrons fill the HOMOs of the porphyrin molecules under the action of voltage, thus mimicking the membrane voltage of leaves and completing the recovery of the porphyrin molecules. When the OH^- groups directly donate electrons to the HOMOs of porphyrin, the photocatalytic CO₂ reduction can proceed without involving the electrons of the external circuit, which results in a high quantum efficiency for this APS cell.

To better understand the advantages of our new APS system, the recently reported results for the use of metalloporphyrin as catalysts are summarized in Table 2. The formation rate of methanol could be changed to 8333 μmol h^-1 g^-1 in the case of an average weight of 1.25 mg of TiO₂. This result is really superior to the results reported in the literature.

A new APS cell that consists of a photocathode of metalloporphyrin-functionalized TiO₂ and a modified BiVO₄ photoanode was designed for CO₂ reduction in water. This PEC system can mainly produce methanol at a high efficiency without using a sacrificing agent. The excellent results imply that the metalloporphyrin molecules covalently linked on the surface of anatase (TiO₂) could play the roles of chlorophyll, NADP, and Calvin cycle in the CO₂ reduction. The highest optical quantum efficiency of the conversion of CO₂ and water to methanol without any side-product by using nickel porphyrin-modified TiO₂ photocathode in D3@TiO₂ can be up to 0.92%, which is about two times that of a plant cell. A new mechanism of APS was proposed, in which metalloporphyrins could behave as a sensitizer, light absorber, and catalyst. Accordingly, this new APS system obviously demonstrates the synergetic effect of electrocatalysis and photocatalysis, which leads to excellent APS performance of cells, thus providing a new way to enhance the light absorbance and improve the efficiency of PEC CO₂ reduction to value-added chemicals.

Acknowledgments

This work was funded by the Natural Science Foundation of Gansu Province (17JR5RA212) and the State Key Laboratory of Coal Conversion (J19-20-913-1).