Introduction

Artificial photosynthesis (AP) represents an appealing strategy for the conversion of solar energy into stored chemical potential energy1,2. Inspired by natural photosynthesis, photocatalytic reduction of CO2 to value-added chemicals (such as CO, CH3OH, and C≥2 products) is a viable solution to obtain sustainable energy in closing the anthropogenic carbon cycle3,4,5,6,7. Several approaches are available for photochemical reduction of CO2 to access fuels or chemical precursors8. Homogeneous AP systems usually offer distinct advantages with respect to structural optimization and product selectivity9. For example, the reduction potentials of all components can be finely tuned through functional group manipulation to facilitate efficient light-driven electron transfer10. In a typical homogeneous AP system, a photosensitizer (PS) is a key component responsible for light-harvesting and photochemical reactions10,11. To achieve a high activity in photocatalytic CO2 conversion, it is essential to develop satisfactory photosensitizers that have large absorption cross sections in the visible region and are capable of promoting efficient multiple electron transfer processes. Recent studies on precious metal-based PS (such as Ru, Ir, and Re) have shown good activity for CO2 reduction12,13,14,15,16,17,18,19,20,21,22,23,24,25. However, molecular chromophores comprising earth-abundant elements usually give relatively low turnover numbers (TONs), which is a significant limitation in developing inexpensive AP systems26,27,28,29. Thus, developing highly efficient noble-metal-free PS for CO2 conversion remains a persistent challenge for the scientific community.

Anthracene organic dyes, which is an emerging group of inexpensive PS with high molar absorption coefficients in the visible region, have demonstrated to have good stability in photocatalytic CO2 reduction30. However, the activity of these photocatalytic systems is much lower than the ones using metal-based photosensitizers, due to the inefficiency in multiple electron transfer to the catalyst11,31. The system containing 9-cyanoanthracene as the PS and an iron(0) tetrapenylporphyrin as the catalyst produces CO with a turnover frequency (TOF) of ~1.5 h−132. Since 2016, purpurin (PP) (the main coloring matter in natural madder) has been studied as one of the most active organic photosensitizers for CO2 reduction when coupled with a variety of metal catalysts33,34,35, achieving TONCO as high as 1365 (TOF ~ 300 h−1) with respect to the catalyst33. Recently, because of the highly luminescent and tunable redox properties, copper diamine photosensitizers have shown promising activities in driving the reduction of CO212,36. Ishitani et al. have shown that dimeric Cu complexes with heteroleptic diimine and phosphine ligands are effective photosensitizers to exhibit 273 turnovers of CO in 12 h37. The systems employed various copper diimine diphosphine complexes as the photosensitizers, Fe37,38,39, Co40, Ni41, or Mn42,43 metal complexes as the catalysts, and 1,3-dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole (BIH) as the sacrificial electron donor that produced CO up to 2680 turnovers under optimal conditions40. While Cu diimine photosensitizers exhibit relatively higher activities in photocatalytic CO2 reduction than organic PS, they require much higher energetic light irradiation owing to their absorption spectra centered at the UV region36. To facilitate fast electron transfer between an electron donor and acceptor, various PS-catalyst dyads have been studied8,44,45. This strategy has shown an increase in activities from the linked systems, with TONco as high as 434746. However, these studies were performed with precious metals, and fastback electron transfer from the catalyst to the PS is usually observed in these dyads, which could hinder efficient charge separation.

In this work, we report a highly active copper purpurin PS for visible-light-driven CO2 reduction to CO in a noble metal-free system (Fig. 1). By coordinating a redox-active Cu center directly to the redox noninnocent catecholate motifs47 on natural organic dyes, we create a simple yet highly efficient PS to facilitate challenging multielectron reductions. Electrochemical studies reveal that the incorporation of Cu in purpurin produces a much more reducing PS for catalysis. When coupled with a Fe porphyrin as the catalyst and BIH as the sacrificial electron donor, this Cu PS significantly boosts the photocatalytic activity (>50-fold) of the system for CO2 reduction compared with its organic dye component. This homogeneous system achieves over 16,100 turnovers of CO and a TOFmax of 7650 per hour with a 95% selectivity (CO vs H2), which is the highest for homogeneous noble metal-free systems. Importantly, this system demonstrates a >fourfold increase in TONco compared to the most active noble metal-based systems using metal porphyrin catalysts48.

Fig. 1: Illustrative scheme of the photocatalytic CO2 reduction system.
figure 1

BIH is the sacrificial reductant, and CuPP is the photosensitizer, and FeTDHPP is the catalyst investigated in this study.

Full size image

Results and discussion

Preparation and characterization of CuPP

As outlined in Fig. 2, Cu purpurin complex was prepared by heating Cu(NO3)2, PP (2.0 equiv), and NaHCO3 (4.0 equiv) in dimethylformamide (DMF) at 100 °C for 24 h. X-ray diffraction (XRD, vide infra) quality single crystals of the isolated compound have not been obtained due to its low solubility in most organic solvents. Electrospray ionization mass spectrometry (ESI-MS) and elemental analysis of this product suggest the formation of a Na2Cu(PP)2 complex (Supplementary Fig. 1a). Cation exchange of this compound was carried out in water with tetrabutylammonium bromide (TBABr) (2.0 equiv). Recrystallization of the resulting (TBA)2Cu(PP)2 (CuPP) by slow diffusion of diethyl ether into an acetonitrile solution gave dark-red crystals suitable for single-crystal XRD studies.

Fig. 2: Synthetic route for Cu purpurin complexes and crystal structure of CuPP.
figure 2

Thermal ellipsoids are shown at the 50% probability level. Parts of hydrogen atoms and TBA cations are not shown for clarity.

Full size image

Polyhydroxy-9,10-anthracenediones have been used for the detection of various transition metal ions49,50; however, solid-state structures of these complexes are very rare51,52. Figure 2 shows the crystal structure of a Cu purpurin complex. In this structure, the Cu center displays a square planar geometry and is coordinated with the 1- and 2-hydroxyl groups on the 9,10-anthracenediones, which is different from the 2,3-oxygen chelates proposed in the literature based on theoretical calculation53. The Cu–oxygen distances (Cu–O1 = 1.927(2) and Cu–O2 = 1.917(2) Å) are similar to the Cu–O bonds reported for Cu(II) catecholate compounds in the literature (Supplementary Table 1). This evidence along with the electronic charge balancing, and paramagnetic proton resonances in the 1H NMR spectra (Supplementary Figs. 3 and 4) are all consistent with the assignment of a Cu(II) center in CuPP. The catecholate carbon–oxygen distances (C2–O2 = 1.295(3) and C1–O1 = 1.303(3) Å) are identical to those found for a Ru(CO)2(PBu3)2(1,2,3-trihydroxy-9,10-anthracenedione) complex, suggesting they are single C–O bonds51. In the anthraquinone moieties, the C4–O3 distance (1.239(3) Å) is typical a double bond, while the other double bond C11–O4 (1.274(3) Å) is elongated due to the hydrogen bonding interaction with its adjacent hydroxyl.

The UV–vis absorption spectra of CuPP and PP were measured at room temperature in DMF (Fig. 3a). CuPP displays a much stronger visible absorption profile than PP, with intense absorption that tails to almost 650 nm (Fig. 3a). The significant red shift of the maximum absorption band at 478 nm (ε = 9.94 × 103 M−1 cm−1) of PP to 566 nm (ε = 3.85 × 104 M−1 cm−1) of CuPP is probably due to the generation of dianion of hydroxyanthraquinone and an induction effect from the Cu(II) center. Linear Beer’s law at two wavelengths (278 and 566 nm) applied for the range of concentrations used in CuPP (Supplementary Fig. 10), suggesting negligible solute–solute interactions.

Fig. 3: Absorption and emission profiles of CuPP and PP.
figure 3

a UV–vis analysis of CuPP (20 μM, orange) and purpurin (40 μM, blue) in DMF; excitation and emission spectra of PP (b) and CuPP (c) at different wavelengths at 298 K in DMF. Source data are provided as a Source Data file.

Full size image

Because the UV–vis spectrum of CuPP exhibits identical features as compared to the dianion form of PP in the visible region54, the absorption band at 566 nm is assigned to the charge transfer band on the PP ligands. The incorporation of Cu(II) enhances this charge transfer process by showing a much higher molar extinction coefficient (Supplementary Fig. 10). The photoluminescence quantum yield for CuPP (8.2 × 10−3) is lower than that of PP (2.7 × 10−2) (Supplementary Table 2). Instead, Grazia et al. observed an increase of the fluorescence intensity when adding Al(III) to PP55. Short stokes shifts of emissions were observed for both Al-PP and PP, and their excitation spectra are comparable with the absorption spectra (Supplementary Fig. 11a)55,56, suggesting that these compounds undergo relatively low structural reorganization between the ground and excited states. In contrast, CuPP displays structureless emission at 693 nm with a large bathochromic shift. The distinct absorption and emission spectra of CuPP (Supplementary Fig. 12b) indicate a significant structural change in its excited state. Since Cu(I) complexes tend to adopt distorted geometries compared to their Cu(II) analogs40,57,58,59,60, the nature of the emission of CuPP is likely from a Cu(I) excited state generated from a LMCT process. Excitation–emission spectra (Fig. 3b, c) show that the emission of CuPP is most intense when excited at 375 nm. The excited-state lifetimes are on the nanosecond timescale for both CuPP (τ = 0.98 ns) and PP (τ = 1.1 ns)35,54.

Electrochemical analysis of CuPP

Figure 4 shows the cyclic voltammograms (CVs) of CuPP and PP in DMF solution containing 0.1 M tetrabutylammonium hexafluorophosphate (TBAPF6). Indeed, the incorporation of a Cu(II) center into PP creates additional reduction events. CuPP displays three irreversible reduction potentials at −1.05, −1.50, and −1.69 V vs SCE. The most negative redox couple of CuPP at −1.75 V is 540 mV more negative than the PP2−/PP− couple (Fig. 4a)33. Square wave voltammetry (SWV) reveals that there are four successive reductions, which correspond to 2, 1, 1, and 2 electron processes (Supplementary Fig. 17).

Fig. 4: Electrochemical data of CuPP and purpurin under N2 or CO2.
figure 4

Experiments were performed in DMF containing 0.1 M TBAPF6 at scan rate 0.1 V/s: a 1 mM CuPP (red) and 1 mM PP (black) under N2; b 1 mM CuPP under N2 (black) and CO2 (red); c 1 mM CuPP (red) and 1 mM PP (black) under CO2. Source data are provided as a Source Data file.

Full size image

To explore the electron requirements for photocatalytic systems, further CV experiments were performed to study the electrocatalytic activity for CO2 reduction. Under an atmosphere of CO2, CuPP shows a quasi-reversible reduction wave at −0.97 V and a catalytic wave that starts to appear below −1.89 V as compared to no observation of current enhancement in N2 (Fig. 4b), suggesting that CuPP could act as a CO2 reduction catalyst. CV control experiments for PP and electrolyte solution show little current enhancement at the same potentials in the presence of CO2 (Fig. 4c). The integrals of SWV show a total of four-electron reduction at −1.13 V prior to the catalytic current (Supplementary Fig. 17). The different electrochemical profiles observed for CuPP in the presence of CO2 and N2 indicate that the four quinone moieties on the PP ligands undergo four-electron reductions and four complexations of CO2 to generate carbonates. In fact, a similar observation for 9,10-anthracenediones has been reported in the literature61. Controlled potential coulometry experiments, performed at −1.89 V (vs SCE) in CO2-saturated DMF, show that CO is the main gas product with a faradaic yield of 17% and only a negligible amount of H2 generated (Supplementary Fig. 20). In a control experiment without CuPP, there is little CO or H2 produced. These data confirm that the observed current enhancement of CuPP corresponds to CO2 reduction, not proton reduction.

Photocatalytic CO2 reduction

The electrocatalytic CO2 reduction properties of CuPP (see Fig. 4) together with its strong absorption in the visible spectral region (Fig. 3) provide strong evidence that CuPP may serve as both photosensitizer and catalyst for CO2 reduction. Photocatalytic experiments were performed by irradiating CuPP (0.1 mM) in DMF with a white light-emitting diode (λ > 400 nm) in the presence of BIH (0.1 M) as the sacrificial donor. The amount of gaseous products were monitored in real time using gas chromatography (GC). Figure 5a shows that CuPP-containing system generates CO upon irradiation with an initial turnover frequency of 4.3 h−1, while there is no CO observed in the control system using PP (2.0 equiv) instead of CuPP (Table 1). However, the rate of CO evolution decreases dramatically after one hour. The disappearance of the color of the reaction mixture after irradiation indicates the decomposition of CuPP.

Fig. 5: Photocatalytic CO2 reduction.
figure 5

Experiments were performed in CO2-saturated DMF solution containing a 0.1 mM CuPP (red) or 0.2 mM PP (black), and 100 mM BIH; b 0.2 mM PP (black) or 0.1 mM CuPP (red), 0.2 µM FeTDHPP, and 10 mM BIH; c 0.1 mM CuPP, 0.2 μM FeTDHPP, and 100 mM BIH. Error bar (95% confidence interval based on three runs) is included at the 1-h point, which is the typical error in the measurement (14%). Source data are provided as a Source Data file.

Full size image
Table 1 Control and other photocatalytic CO2-reduction experiments.
Full size table

The activity and stability of photocatalytic CO2 reduction can be significantly improved with the addition of chloro iron (III) 5, 10, 15, 20-tetrakis (2’,6’-dihydroxyphenyl)-porphyrin (FeTDHPP) as the cocatalyst (Fig. 1), which has been shown to be an active electrocatalyst for CO2 reduction62. In a typical experiment, the system containing 0.1 mM CuPP, 0.2 µM FeTDHPP, and 10 mM BIH, produces 4779 TONs of CO/mol of catalyst in 7 h in CO2-saturated DMF solution (Fig. 5b and Table 1). For comparison, an experiment performed using PP as the PS under the same conditions gives a TON of 88 for CO2 (Fig. 5b and Table 1), underscoring the importance of CuPP for photocatalytic CO2 reduction. The CuPP PS can also be generated in situ with the addition of 1:2 ratio of Cu2+ and PP. The amount of CO produced with the in situ-generated PS is ~25% lower than that of the system using isolated CuPP (Supplementary Fig. 21a). UV–vis spectra confirm that CuPP can be generated in minutes when mixing Cu2+ and PP in the presence of BIH (Supplementary Fig. 21b). Control experiments show that the absence of PS leads to no CO or H2 production (Supplementary Table 4).

An induction period observed at low BIH concentration disappears as [BIH] increased above 0.1 M (Supplementary Fig. 22), which suggests that the rate of CO production is limited by the concentration of sacrificial donor. At 100 mM concentration of BIH, the activity of CO2 to CO conversion is further improved, achieving 16,109 TONs in 23 h with an initial TOF of 7650 h−1 with respect to [FeTDHPP] (Fig. 5c and Table 1), which is more than ten times more active than previously reported systems using PP as the PS (Supplementary Table 6). The activity of the photochemical systems was optimized further by varying the [CuPP] and [FeTDHPP]. At fixed concentrations of BIH (100 mM) and FeTDHPP (2 µM), increasing the CuPP concentration increases the overall rate of CO production (Supplementary Fig. 23). However, the rate of CO generation does not increase above 0.1 mM concentration of CuPP. When the CuPP concentration is fixed at 0.1 mM and [FeTDHPP] is varied (Supplementary Fig. 24), the catalyst is most active (on a TON basis) at low concentrations. The total amount of CO evolved increases up to 40 µM FeTDHPP, suggesting that the activity of the system becomes limited by electron transfer to the catalyst. For a system containing 2.0 µM FeTDHPP/100 mM BIH/0.1 mM CuPP, the initial 1 h quantum yield of CO2 to CO conversion at 450 nm was determined at 6.0 ± 0.6%.

Quantification of the gaseous products shows an overall selectivity of 95% for CO and 5% for H2. In the solution, 1.4 µmol formic acid was quantified by high-performance liquid chromatography (HPLC). However, a similar amount (1.5 µmol) of formic acid was detected from the experiment performed under N2, suggesting that formic acid is not generated from CO2 reduction. In fact, formic acid generated from hydrolysis of DMF has been previously reported63. To confirm that CO2 is the source of carbon in photolysis, isotopic labeling experiments were performed under an atmosphere of 13CO2. The GC–MS analysis of the gaseous products shows a diagnostic peak of 13CO (Fig. 6). It reveals that CO is derived from CO2 reduction rather than decomposition of other components in the photocatalytic system.

Fig. 6: GC/MS chromatograms of carbon monoxide.
figure 6

Photocatalytic CO2 reduction in 13CO2-saturated (a) and 12CO2-saturated (b) DMF solutions containing 1.0 μM FeTDHPP, 100 mM BIH, and 0.1 mM CuPP.

Full size image

To evaluate the homogeneity of the system, excess of Hg (0.02 mL, 2700 equiv of CuPP) was added to the reaction vessel at the beginning of the experiments, the activity of CO generation remains identical over the course of photolysis (Supplementary Fig. 25). In addition, dynamic light-scattering (DLS) experiments show there is no detectable nanoparticle generated in the reaction mixture before and after CO2 reduction (Supplementary Fig. 26). These control experiments suggest that there are no Cu colloids generated and the CuPP/FeTDHPP/BIH system stays homogeneous during photocatalytic CO2 reduction.

Several additional experiments were carried out to probe the mechanism of decomposition of the system. When the rate of CO production decreases significantly, fresh solutions containing the same amounts of CuPP, FeTDHPP, BIH, or a mixture of them were added to the reaction mixtures at 23 h. Addition of one or two components leads to much less recovery of the original activity, compared to the one that added all three components (Supplementary Fig. 28). This suggests that the decomposition of all three components occurs during CO generation. Furthermore, absorption of the reaction mixture containing CuPP, FeTDHPP, and BIH in DMF was monitored by UV/vis spectroscopy during photolysis (Supplementary Fig. 29). The intensity of the absorption band at 400 nm, which corresponds to the reductions of PP ligands35, decreases gradually on the course of irradiation. This confirms that CuPP decomposes during CO2 reduction.

Mechanistic studies

Further experiments were performed to study the photochemical mechanism of CO2 reduction. UV/vis studies show there is no change of the CuPP spectrum with the addition of BIH or FeTDHPP under CO2 (Supplementary Fig. 30), suggesting that the first electron transfer step is initiated by photon absorption. Photochemical steps were studied through fluorescence quenching of CuPP in DMF. The reductive quenching rate constant (using BIH as the quencher) is 9.23 × 109 M−1 s−1 based on Stern–Volmer equation (Supplementary Fig. 31). A linear plot for the oxidative quenching could not be obtained due to the fact that FeTDHPP has a nonnegligible absorption at both the excitation and emission wavelengths (Supplementary Figs. 32 and 33), which is in agreement with the literature32,34. Because [BIH] used in most of our photocatalytic experiments is more than 5 orders of magnitude higher than [FeTDHPP] and the kq for reductive quenching is at the diffusion-controlled limit, reduction of CuPP* by BIH is expected to be the first electron-transfer step.

CO generation was observed even at [BIH] as low as 0.1 mM. UV–vis spectra show no generation of the additional absorption band in the visible region during CO2 reduction (Fig. 7a), indicating CuPP is the active PS at this condition. Since [L2CuII]– (L = PP2–) has a much higher reduction potential (−1.05 V) than that of Fe(I)TDHPP (E1/2 = −1.55 V), CO2 reduction has to proceed through oxidative quenching of (L2)+CuI* (−1.61 V) (Supplementary Table 3) in the photochemical step (Fig. 8). At high [BIH] (100 mM), a yellow species with an absorption band at 400 nm is generated under visible-light irradiation (Fig. 7b). When the reaction mixture is exposed to the air, the solution returns to purple color and over 80% of the original CuPP is recovered based on UV–vis spectra (Supplementary Fig. 34). Electrochemical studies in the presence of CO2 (Fig. 4b) suggest that CuPP could undergo four steps of reduction and complexation of CO2 to generate a [((CO2)2L)2CuII]4– species (Fig. 8). In fact, PP has been shown to proceed a two-electron reduction to generate a species at 398 nm during photocatalytic CO2 reduction34. Thus, we conclude that the intermediate with maximum absorption at 400 nm is a [((CO2)2L)2CuII]4– species. [((CO2)2L)2CuI]5– (−1.75 V), generated by further reduction of [((CO2)2L)2CuII]4– through reductive quenching, is corresponding to electron transfer to the Fe catalyst (Fig. 8).

Fig. 7: UV–vis absorption spectra of systems.
figure 7

a In all, 0.01 mM CuPP, 0.5 µM FeTDHPP, 0.1 mM BIH in a quartz cuvette (10-mm path length), and b 0.1 mM CuPP, 0.2 µM FeTDHPP, and 100 mM BIH upon irradiation with white LED light. Solutions for (b) at different times were transferred and diluted five times with DMF to a quartz cuvette (2-mm path length) under N2.

Full size image
Fig. 8: Proposed reaction scheme.
figure 8

Different pathways for photocatalytic CO2 reduction are proposed for low [BIH] (left) and high (BIH) (right).

Full size image

On the basis of a much faster turnover of CO2 to CO with CuPP compared to PP, electron transfer from the reduced PS to the catalyst is proposed to be a turnover-limiting step (TOLS) (Fig. 8). Although CuPP is active for photocatalytic CO2 reduction without an additional catalyst, the significantly lower TOF observed in the CuPP/BIH system clearly shows that this process is much slower than the TOLS of the CuPP/FeTDHPP/BIH system (Fig. 8). Electrochemical studies show that CuPP exhibits a more negative reduction potential (−1.75 V vs SCE) than that of PP (−1.21 V vs SCE). In order to access a previously reported Fe(0) porphyrin intermediate for CO2 reduction (Fig. 8)62, CuPP is expected to have a larger driving force for reduction and hence a faster rate for electron transfer to the catalyst than PP.

This work shows an example where an active photosensitizer for CO2 reduction has been constructed through direct coordination of a redox-active metal center to natural organic dyes. This Cu photosensitizer displays strong visible-light-harvesting and multiple electrochemical properties. In the AP systems using metal porphyrins as the catalysts, CuPP exhibits great photocatalytic activity for CO2 reduction when compared with previously reported homogeneous photosensitizers, even including the noble metal ones (Supplementary Tables 5–8). For example, the CuPP/FeTDHPP system achieves over 16,100 turnovers of CO with a maximum TOF of 7650 h−1, which is two orders of magnitude more than a reported Ir(ppy)3/FeTDHPP system (TON = 140 in 55 h)32. This work not only paves the way to improve CO2 to CO conversion in homogeneous light-driven systems, but also provides perspectives for the rational design of efficient but low-cost photosensitizers for solar fuel production.

Methods

Materials

All solvents and chemicals were commercially purchased and used as obtained without further purification, except for purpurin. Purpurin was recrystallized twice from absolute ethanol before usage.

Preparation of BIH

BIH was synthesized from a modified method based on the literature64. In total, 6.0 g of 2-phenylbenzimidazole was treated with 16 g of methyl iodide in 30 mL of methanol containing 1.28 g of NaOH. The reaction mixture was heated at 110 °C for 24 h to give a faint yellow solid (BIH+I−). The crude product was washed with H2O to decolorize. The yield is over 80%. Then, a solution of BIH+I− in methanol was reduced by NaBH4 (2 eq.) under N2 to give a white solid (BIH). The white solid was purified by washing with EtOH/H2O (5:1, v/v). The yield is over 90%. The synthetic intermediates and target complexes were evidenced by 1H NMR.

Preparation of chloro iron (III) 5,10,15,20-tetrakis (2′,6′-dihydroxyphenyl)-porphyrin

Chloro iron (III) 5,10,15,20-tetrakis (2′,6′-dihydroxyphenyl)-porphyrin was prepared following a reported procedure62. A solution of 2′−6′-dimethoxybenzaldehyde (1 g, 6.02 mmol) and pyrrole (0.419 mL, 602 mmol) in chloroform (600 mL) was degassed by argon for 20 min, then BF3·OEt2 (0.228 mL, 0.87 mmol) was added via a syringe. The solution was stirred at room temperature under an inert atmosphere in the dark for 1.5 h, and 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ) (1.02 g, 4.51 mmol) was added to the reaction. The mixture was stirred for an additional 1.5 h at reflux, cooled to room temperature, and 1 mL of triethylamine was added to neutralize the excessive acid. Then the solvent was removed, and the resulting black solid was purified by column chromatography (silica gel, dichloromethane) affording 5, 10, 15, 20-tetrakis(2′,6′-dimethoxyphenyl)-21H,23H-porphyrin as a purple powder (290 mg, 23%). To a solution of 5, 10, 15, 20-tetrakis(2′,6′-dimethoxyphenyl)-21H,23H-porphyrin (200 mg, 0.235 mmol) in dry dichloromethane (10 mL) at 0 °C was added BBr3 (1000 μL, 10.38 mmol). The resulting green solution was stirred for 24 h at room temperature, then placed in ice water. Ethyl acetate was added to the suspension, and the mixture was washed with NaHCO3. The organic layer was separated, washed twice with water, and then dried over anhydrous Na2SO4. The resulting solution was evaporated. The residue was purified by column chromatography (silica gel, 2:1 ethyl acetate/dichloromethane) to yield 5, 10, 15, 20-tetrakis(2′,6′-dihydroxyphenyl)-21H,23H-porphyrin as a purple powder (150 mg, 87%). Finally, a solution of 5, 10, 15, 20-tetrakis(2′,6′-dihydroxyphenyl)-21H,23H-porphyrin (100 mg, 0.135 mmol), FeCl2·4H2O (270 mg, 1.35 mmol), and 2,6-lutidine (39 μL, 0.335 mmol) was heated at 50 °C and stirred for 3 h under an inert atmosphere in dry methanol. After methanol was removed, the resulting solid was dissolved in ethyl acetate, washed with 1.2 M HCl solution until pH was neutral. The crude product was purified by column chromatography (silica gel, ethyl acetate) to give chloro iron (III) 5,10,15,20-tetrakis (2′,6′-dihydroxyphenyl)-porphyrin as a brown solid (100 mg, 89%). The synthetic intermediates and target complexes were evidenced by 1H NMR and HRMS spectroscopy.

Preparation of CuPP

A solution of Cu(NO3)2·3H2O (672 mg, 2.7 mmol) in DMF (5 mL) was added to a solution of purpurin (1380 mg, 5.4 mmol) in DMF (50 mL). The orange solution quickly turned into a brown suspension upon addition. Then a solution of NaHCO3 (907 mg, 10.8 mmol) in H2O (10 mL) was added to the reaction mixture with stirring for 24 h at 100 °C. After the mixture cooled down to room temperature, insoluble solids were removed by suction filtration. The filtrate was concentrated by rotary evaporation to ~10 mL, then excess diethyl ether was added, and the mixture was allowed to stir overnight at room temperature to precipitate out the product. A black precipitate (Na2Cu(PP)2) was collected under suction filtration, washed with diethyl ether (3 × 50 mL), and dried under vacuum (1.0 g, 60% yield). Anal. Calcd. for C28H12CuO12Na2 · 2H2O: C, 51.43; H, 2.47; found: C, 51.26; H, 2.47. Cation exchange was carried out with the addition of tetrabutylammonium bromide (TBABr) (522 mg, 1.62 mmol, 2.0 equiv) to Na2Cu(PP)2 (500 mg, 0.81 mmol) in water at 60 °C for 6 h. The mixture was filtered and a brown solid was washed with water (3 × 10 mL) and then collected. Recrystallization of this compound from vapor diffusion of diethyl ether into an acetonitrile solution gave red crystals amenable to X-ray diffraction studies. The overall yield was 22%. 1H MMR (400 MHz, d6-DMSO, 25 °C): δ = 17.21, 8.18, 7.70, 7.54, 7.18, 6.64, 3.20–3.13, 1.57, 1.32–1.23, 0.93 ppm. All resonances are broad. Anal. Calcd. for C60H84N2O10Cu: C, 68.19; H, 8.01; N, 2.65; found: C, 67.94; H, 8.16; N, 2.60.

Characterization

1H and 13C NMR spectra were recorded at a Bruker advance III 400-MHz NMR instrument. HRMS spectra were obtained on a Thermo Scientific Orbitrap Q Exactive ion trap mass spectrometer and QTOF-MS (Bruker Daltonics, times TOF). UV–vis spectra were determined on a Thermo Scientific GENESYS 50 UV–visible spectrophotometer. Elemental analyses were performed at an Elementar Vario EL analyzer. Dynamic light-scattering experiments were performed with a Brookhaven Elite Sizer zata-potential and a particle-size analyzer. Photoluminescence quantum yields were determined using a FLS 980 (Edinburgh instruments) absolute photoluminescence quantum yield measurement system and integrating sphere as a sample chamber. The steady-state fluorescence of solid samples was measured with FLS 920 fluorescence spectrometer (Edinburgh instruments).

X-ray crystallography

X-ray diffraction data were collected on SuperNova single-crystal diffractometer using the CuKα (1.54184 nm) radiation at 100 K. Absorption correction was carried out by a multiscan method. The crystal structure was solved by direct methods with SHELXT65 program, and was refined by full-matrix least-square methods with SHELXL65 program contained in the Olex2 suite66. Weighted R factor (Rw) and the goodness of fit S were based on F2, conventional R factor(R) was based on F (Supplementary Table 9). Hydrogen atoms were placed with the AFIX instructions and were refined using a riding mode. Figures were drawn with Diamond software. Details can be obtained from the Cambridge Crystallographic Data Centre at www.ccdc.cam.ac.uk for CCDC accession number 2017326.

Cyclic voltammetry

Cyclic voltammetry (CV) measurements were performed with a CHI-760E electrochemical analyzer using a one-compartment cell with a glassy carbon disk working electrode (diameter 3 mm), Pt auxiliary electrode, and a SCE reference electrode. The electrolyte solutions were 0.1 M tetrabutylammonium hexafluorophosphate in DMF. Solutions were purged with N2 or CO2 over 30 min before measurements. All reported potentials in this work are versus SCE.

Fluorescence quenching

A solution of photosensitizer was degassed by N2 for 15 min in a sealed quartz cuvette with a septum cap. Different concentrations of BIH or catalyst were added to the solution of photosensitizer under N2. The steady-state fluorescence for solution samples was measured by Duetta fluorescence and absorbance spectrometer. The excited-state lifetime (τ0) of CuPP and PP was measured with an FLS 920 fluorescence spectrometer (Edinburgh instruments), in which a picosecond pulsed diode laser (λ = 472 nm) (Edinburgh instruments EPL-470) was used as the excitation source. The quenching rate constant (kq) was calculated by Stern–Volmer equation:

$$I_0/I\;{\mathrm{or}}\;\tau_0/\tau = 1 + k_q \times \tau _0 \times [Q]$$
(1)

where \(I_0\) and \(I\) represent the fluorescence intensity of photosensitizer in the absence and presence of quencher, \(k_q\) is the quenching rate constant, \(\tau _0\) is the lifetime of the photosensitizer, and \([Q]\) is the concentration of quencher.

Photocatalytic CO2 reduction

Photocatalytic experiments were conducted in a closed scintillation vial with rubber plug and magnetic stirring. The headspace of the vial was 53.4 mL. A reaction mixture (5.0 mL) was bubbled with CO2 for 25 min and then irradiated with a white LED light setup (λ > 400 nm, PCX-50B/50 C, Beijing Perfectlight Technology Co., Ltd.). The gaseous products were analyzed by Shimadzu GC-2014 gas chromatography equipped with Shimadzu Molecular Sieve 13 × 80/100 3.2 × 2.1 mm × 3.0 m and Porapak N 3.2 × 2.1 mm × 2.0-m columns. A thermal conductivity detector (TCD) was used to detect H2 and a flame ionization detector (FID) with a methanizer was used to detect CO and other hydrocarbons. Nitrogen was used as the carrier gas. The oven temperature was kept at 60 °C. The TCD detector and injection port were kept at 100 and 200 oC, respectively. 13C isotopic labeling experiments were carried out in a 13CO2 atmosphere and gas products were detected by GC–MS (Agilent 7890A-5975C). HPLC (HP1100, Hewlett Packard) was used to analyze the photolysis products in the solution.

Quantum yield measurement

The difference between the power of light passing through the blank (containing FeTDHPP and BIH) and through the sample (containing CuPP, FeTDHPP, and BIH) was used to calculate the light absorbed by the CuPP. The average intensity of the irradiation P (W/cm2) was measured with a FZ-A Power meter (Beijing Normal University Optical Instrument company). The monochromic light of 450 nm obtained using blue LED light setup (λ = 450 nm, PCX-50B/50C, Beijing Perfectlight Technology Co., Ltd.). Generally, the quantum yield for CO2 reduction to CO was calculated after irradiation for 1 h at 450 nm using the following equation:

$${\Phi} _{{\mathrm{CO}}} = \frac{{2 \times {\mathrm{number}}\,{\mathrm{of}}\,{\mathrm{the}}\,{\mathrm{CO}}\,{\mathrm{molecules}}}}{{{\mathrm{number}}\,{\mathrm{of}}\,{\mathrm{incident}}\,{\mathrm{photons}}}} \times 100\%$$
(2)

that is,

$${\Phi} _{{\mathrm{CO}}} = \frac{{2 \times n({\mathrm{CO}})}}{I} \times 100\%$$
(3)

where n(CO) is the number of molecules of CO produced, I is the number of incident photons; the calculation formula of the incident photon number I is as follows:

$$I = PSt\frac{\lambda }{{hc}}$$
(4)

where S is the incident irradiation area (S = 0.785 cm2), t is the irradiation time, λ is the incident wavelength, h is the Plank constant (6.626 × 10−34 J · s), and c is the speed of light (3 × 108 m s−1)

$${\Phi} = \frac{{2 \times n \times N_A}}{{PSt \times \frac{\lambda }{{hc}}}} \times 100\%$$
(5)

where NA is the Avogadro constant (6.02 × 1023 mol−1).