Supramolecular multi-electron redox photosensitisers comprising a ring-shaped Re(i) tetranuclear complex and a polyoxometalate

Redox photosensitisers (PSs) play essential roles in various photocatalytic reactions. Herein, we synthesised new redox PSs of 1 : 1 supramolecules that comprise a ring-shaped Re(i) tetranuclear complex with 4+ charges and a Keggin-type heteropolyoxometalate with 4− charges. These PSs photochemically accumulate multi-electrons in one molecule (three or four electrons) in the presence of an electron donor and can supply electrons with different reduction potentials. PSs were successfully applied in the photocatalytic reduction of CO2 using catalysts (Ru(ii) and Re(i) complexes) and triethanolamine as a reductant. In photocatalytic reactions, these supramolecular PSs supply a different number of electrons to the catalyst depending on the redox potential of the intermediate, which is made from the one-electron-reduced species of the catalyst and CO2. Based on these data, information on the reduction potentials of the intermediates was obtained.

Generally, the photoexcitation of a molecule can induce only one-electron transfer. Hence, typical PSs can initiate only a oneelectron transfer from the electron donor to the acceptor. Since, on the other hand, one-electron reduction of CO 2 requires very high amounts of energy (E 0 = −1.9 V vs. NHE), the two-electron reduction of CO 2 coupled with another chemical reaction, such as proton addition can be applied to lower the required energy to produce stable products such as CO (E 0 = −0.53 V) and HCOOH (E 0 = −0.61 V). Hence, catalysts that completely accept two electrons through redox-photosensitised reaction(s) and reduce CO 2 should be used with PSs to achieve efficient photocatalytic CO 2 reduction. 7 In such two-component photocatalytic systems, an intermediate produced by the chemical reaction(s) of the oneelectron-reduced species (OERS) of the catalyst must rapidly accept one more electron because the side reactions of the active intermediate induce decomposition of the catalyst and lower the durability of the photocatalytic system. For avoiding this decomposition process of the catalyst, a signicantly higher number of PSs than catalysts has been used in many reported photocatalytic systems to suppress these side reactions because the usual PSs initiate only a one-electron transfer. 7 In these systems, however, the decomposition of PSs cannot be avoided owing to the photochemical decomposition of the OERSs of the excess PSs, which accumulate in the reaction solution.
Therefore, developing PSs that can accumulate multielectrons in one molecule and donate them to the catalytic reaction with suitable timing should inspire a new direction in photocatalytic redox reactions. However, only a limited number of PSs can accumulate multi-electrons. For example, Ru(II) complexes with quinone or pyridinium-cation moieties integrated into ligands 22 or viologen moieties attached to ligands 23 have been reported. However, such PSs have weak reduction power, and their application in photocatalytic reactions has been limited to low-energy reactions, such as H 2 evolution. [23][24][25][26][27] Another challenge associated with multi-electron-accumulating PSs is that the accumulated electrons in PSs with multiple numbers of the same or similar photosensitiser units in one molecule, such as ring-shaped Re(I) multinuclear complexes, 28,29,30a,37 have the same or similar reduction powers. It cannot accumulate multi-electrons in the presence of a catalyst because its OERS rapidly passes an accepted electron to the catalyst, i.e., it works only as a one-electron transfer photosensitiser. 28,29 In photocatalytic systems for CO 2 reduction, the reduction potentials of the catalyst and the intermediate derived from the reaction of the reduced catalyst with a substrate such as CO 2 should be different. Since the reduction potential of the intermediate, which is very important information for developing efficient photocatalytic systems, is oen more positive than the rst one owing to the following chemical reaction or reactions, the reduction potential cannot be determined using ordinary electrochemical methods. To the best of our knowledge, there have been no reports of a multi-electron-accumulating PS that can precisely supply two electrons to the catalyst and intermediate in photocatalytic CO 2 reduction reactions.
Herein, we report the rst examples of supramolecular redox photosensitisers consisting of a ring-shaped Re(I) tetranuclear complex (Ring 4+ ) and Keggin-type heteropolyoxometalate [XW 12 O 40 ] 4− (XPOM 4− , X = Si, Ge) (Chart 1). Ring 4+ has four positive charges owing to the one plus charge of each Re(I) unit and the space inside the ring structure. 28-30a Its lowest 3 MLCT excited state has a long lifetime even in solution at 25°C (s = 225 and 406 ns), of which the dual phosphorescence property is attributed to stable conformers, 28 and high oxidation power. Its OERS is relatively stable, and its reduction power is high. In contrast, XPOM 4− , in which twelve octahedral tungsten oxyanions surround a central silicate or germanate group, has four negative charges. They can electrochemically accumulate multi-electrons in a single molecule because their reduced states are stable. [31][32][33] We found that strong electrostatic interactions between XPOM 4− and Ring 4+ formed a 1 : 1 ion pair (Ring 4+ )-(XPOM 4− ), as shown in Chart 1. In these supramolecules, the Ring 4+ unit functions as an intramolecular PS that photochemically transfers multi-electrons to the XPOM 4− unit in the presence of triethanolamine in the initial stage. It accumulates one more electron in one of the Re(I) complex moieties, resulting in three electrons with different reduction powers. Using this novel photochemical multi-electron accumulating system as a PS, we developed new photocatalytic CO 2 reduction systems accompanied by Re(I)or Ru(II)-complex catalysts (RuCAT and ReCAT in Chart 1). In addition, a comparison of the photosensitising abilities of the two types of (Ring 4+ )-(XPOM 4− ), where X = Si or Ge, provides information on redox potentials of the intermediates formed from the OERSs of the Re and Ru catalysts.
Since Ring 4+ and XPOM 4− interact with each other even in DMSO solutions, we measured particle sizes in DMSO solutions containing (Ring 4+ )(PF 6 diameters (D) of approximately 1 nm were mainly observed, and these are attributable to solvated monomeric Ring 4+ , SiPOM 4− , and GePOM 4− , respectively, some of which may interact with the counter anions or cations because of the size similarity to the X-ray structure of each ion of (Ring 4+ )-(SiPOM 4− ) (Fig. 1). In contrast, in the solution containing (Ring 4+ )-(SiPOM 4− ), larger particles (2.9 ± 0.4 nm) were mainly observed compared to those of Ring 4+ and SiPOM 4− (Fig. 2a). Therefore, even in the DMSO solution, Ring 4+ and SiPOM 4− maintained the interaction and formed a supramolecule consisting of one molecule of Ring 4+ and one molecule of SiPOM 4− . In the DMSO solution of (Ring 4+ )-(GePOM 4− ), particles of similar size (3.2 ± 0.5 nm) were observed (Fig. 2b). Therefore, it can be deduced that strong electrostatic interactions maintain supramolecular interactions between Ring 4+ and XPOM 4− even in DMSO solutions and mainly form a 1 : 1 supramolecule consisting of one ion each. Fig. 3 shows UV-vis absorption spectra of (Ring 4+ )-(XPOM 4− ) and their constituent ions in DMSO solutions. The similarity in the spectra of (Ring 4+ )-(XPOM 4− ) and the summation of those of (Ring 4+ )(PF 6 − ) 4 and (TBA + ) 4 (XPOM 4− ) indicates that there was no strong electric interaction between the Ring 4+ and XPOM 4− units of (Ring 4+ )-(XPOM 4− ) in the ground states, although they exist closely in the solutions as described above. In both spectra of (Ring 4+ )-(XPOM 4− ), a broad peak was observed at l max = 408 nm, which is attributed to singlet metal-to-ligand-chargetransfer ( 1 MLCT) absorption of the Ring 4+ unit.
In the much shorter wavelength region, singlet ligand-to-metalcharge-transfer ( 1 LMCT) absorption of the XPOM 4− unit and   singlet p-p* absorption of the Ring 4+ unit were observed. Therefore, only the Ring 4+ unit of the supramolecule can be selectively excited by light at l ex $ 400 nm in photochemical reactions. Fig. 4 shows emission spectra of (Ring 4+ )(PF 6 − ) 4 and (Ring 4+ )-(XPOM 4− ) dissolved in DMSO solutions using excitation light at l ex = 400 nm. This outcome is due to the emission from Ring 4+ or the Ring 4+ unit of the supramolecules because the shapes of the emission spectra were very similar in all the solutions. However, the strengths of both (Ring 4+ )-(SiPOM 4− ) and (Ring 4+ )-(GePOM 4− ) were signicantly weaker than that of (Ring 4+ )(PF 6 − ) 4 . The photophysical properties of the samples are summarised in Table 1. The emission quantum yields (F em ) of the Ring 4+ units in (Ring 4+ )-(SiPOM 4− ) and (Ring 4+ )-(GePOM 4− ) were only 15% and 13% compared to that of (Ring 4+ )(PF 6 − ) 4 , respectively. Therefore, it can be inferred that most of the 3 MLCT excited states of the Ring 4+ units in the supramolecules were quenched by the XPOM 4− unit.  free Ring 4+ , i.e., the former was much weaker than the latter when the number of integrations was unied. This result clearly indicates that static quenching of the excited state of the Ring 4+ unit by the SiPOM 4− unit rapidly proceeded within the time resolution of the apparatus (200 ps). Based on careful investigation of these emission decay data as described in the ESI section, † we can conclude that the percentage of dissociation of (Ring 4+ )-(SiPOM 4− ) was only 3.9%, and more than 85% of the excited state of the Ring 4+ unit was statically quenched by the SiPOM 4− unit in the DMSO solution dissolving 0.05 mM of (Ring 4+ )-(SiPOM 4− ).
Although the exact reduction potential of the 3 MLCT excited state of Ring 4+ has not been determined because of a lack of information on the excitation energy of Re complexes, it is expected as approximately −1.0 to −1.1 V vs. Ag/AgNO 3 . 30 This is more negative compared to the rst reduction potential of SiPOM 4− and GePOM 4− (E 1/2 = −0.93 V and −0.88 V, respectively). Considering that there is no strong electric interaction between the Ring 4+ and XPOM 4− units, we can conclude that the emission quenching observed in the supramolecule (Ring 4+ )-(XPOM 4− ) proceeds via electron transfer from the excited Ring 4+ unit to XPOM 4− (eqn (5)). The reduction potentials described above are listed in Table 2.
The UV-vis absorption spectra of the reduced XPOM 4− were obtained using the ow electrolysis of (TBA + ) 4 (XPOM 4− ). Fig. 8 shows the results for SiPOM 4− measured in DMSO. It indicates a two-step reduction; the rst one started at −0.8 V and nalised at −1.1 V and the second one started at −1.4 V and nalised at −1.7 V. They should be attributed to a rst one-electron reduction (SiPOM 4− /SiPOM 5− ) and the second one-electron reduction (SiPOM 5− /SiPOM 6− ), of which redox potentials are E 1/2 = −0.93 V and −1.40 V as described above (Table 2). The number of electrons injected into one molecule (n) at each step is calculated using eqn (6) with a ow rate (v = 0.15 mL min −1 ) and the currents to give n = 1.0 at E = −1.0 V and n = 1.1 at E = −2.3 V, respectively: is concentration of (TBA + ) 4 (SiPOM 4− ), i.e., 0.5 mM, and F is the Faraday constant. Fig. 8a shows the difference in the UV-vis absorption spectra of the electrolysis solution at various potentials and at E = −0.6 V where the reduction of SiPOM 4− did not proceed. New broad absorptions with absorption maxima at l = 480 nm and 730 nm were observed during the rst reduction process. These maxima shied to l = 500 nm and 650 nm aer the second reduction. Since these absorption changes were well synchronised with the  In the case of (TBA + ) 4 (GePOM 4− ), 4-electrons reduction proceeds at a more positive potential than the reduction of Ring 4+ , and this reduction process couples with proton addition (PCET processes). Since the addition of TEOA to DMSO made the viscosity of the solution too high for use in ow electrolysis, ethanol (EtOH) was used as the proton source. A similar CV was obtained in the DMSO-EtOH (5 : 1 v/v) mixed solution containing (TBA + ) 4 (GePOM 4− ) to that in the DMSO-TEOA (5 : 1 v/v) mixed solution (Fig. S5 †). Fig. 9b shows the I-V curve for the ow electrolysis of (TBA + ) 4 (GePOM 4− ) in a DMSO-EtOH (5 : 1 v/v) solution, where the current mainly changed in three steps: the current drastically increased at E = −0.8 V, −1.45 V, and −1.8 V. In each step, GePOM 4− accepts 1.1, 1.0, and 1.8 electron(s), respectively. These results indicate that the three steps of total four-electron reduction proceeded as shown in eqn (2)-(4). Fig. 9a shows the UV-vis absorption spectral changes in the solution during electrolysis.
From these results and investigations, we determined the molar extinction coefficients of the reduced GePOM 4− species, that is, We can use the absorption spectrum of the OERS of the model mononuclear complex cis-(CO)-[Re(bpy)(CO) 2 (PEtPh 2 ) 2 ] + (blue line in Fig. 10c: l max = 362 nm, 490 nm, 515 nm) as that of the OERS of Ring 4+ because the spectrum of the photochemically reduced Ring 4+ was well tted by using the spectra of cis-(CO)-[Re(bpy)(CO) 2 (PEtPh 2 ) 2 ] + and its OERS. 30a Photochemical reduction of (Ring 4+ )-(XPOM 4− ) A DMSO-TEOA (5 : 1 v/v) solution containing (Ring 4+ )-(SiPOM 4− ) was irradiated at l ex = 436 nm with a light intensity of 5 × 10 −9 einstein s −1 . Fig. 10a and b show the spectral changes during irradiation (irradiation times of 5 min, 30 min, and 60 min) and differential spectra before and aer the irradiation, respectively. Immediately aer irradiation, a new broadband with absorption maxima was observed at l max = 480 nm and 730 nm. This result was attributed to the formation of OERS of SiPOM 4− as it had a spectrum similar to that of SiPOM 5− obtained by ow electrolysis (orange line in Fig. 8a). Further irradiation causes a change in the absorption shape. For example, aer irradiation for 30 min, the absorption maxima became l max at 500 nm and 650 nm, which are attributed to the two-electron reduced species (TWERS) of the SiPOM 4− unit because of the similarity of the spectrum to that of SiPOM 6− (green line in Fig. 8a). Irradiation   periods longer than 30 min caused another spectrum change with new absorption maxima at l max = 360 nm, 490 nm (sh), and 515 nm (60 min irradiation: purple line in Fig. 10b).
This outcome was due to the additional formation of OERS of the Ring 4+ unit (blue line in Fig. 10c), 30a that is, the three-electron reduced species of the supramolecule ([(Ring 3+ )-(SiPOM 6− )] 3− ) was produced. All the observed spectra during irradiation can be well tted by the spectra of Ring 4+ , SiPOM 4− , SiPOM 5− , SiPOM 6− , and/or Ring 3+ . Fig. 10c shows a typical example of the tting result of the spectrum aer 60 min of irradiation, in which SiPOM 6− , Ring 4+ , and Ring 3+ were used.
In contrast, the second reduction of TWERS, [(Ring 4+ )-(XPOM 6− )] 2− , should be produced via intermolecular reductive quenching of the excited Ring 4+ unit in [(Ring 4+ )-(XPOM 5− )] − by TEOA, followed by intramolecular electron transfer from the Ring 3+ unit to the XPOM 5− unit (eqn (16)), because the reduction potential of XPOM 5− (E red 1/2 = −1.40 V) is much more negative than the oxidation potential of the 3 MLCT excited state of Ring 4+ (E * ox = −1.0 ∼ −1.1 V). Intramolecular oxidative quenching is a highly endothermic process. (14) (15) (16) The third reduction process to form [(Ring 5+ )-(SiPOM 6− )] 3− and [(Ring 4+ )-(H 2 GePOM 6− )] 2− should proceed via intermolecular reductive quenching of the excited Ring 4+ unit by TEOA for a similar reason. The quantum yields of the second and third reduction processes were lower than those of the rst one. This result is reasonable because the intramolecular electron transfer from the reduced POM unit to the excited Ring 4+ unit, which is an energy-consuming process owing to the subsequent backelectron transfer, should compete with the intermolecular electron transfer from TEOA.

Photocatalytic reduction of CO 2 with RuCAT
Photocatalytic CO 2 reduction was conducted using (Ring 4+ )-(XPOM 4− ) or Ring 4+ as PS together with RuCAT, which has been frequently used as a catalyst for CO 2 reduction in photocatalytic and electrocatalytic systems. 7,42,43 A DMSO-TEOA (5 : 1 v/v) solution containing one of the PSs and RuCAT (0.05 mM each) was irradiated at l ex = 436 nm, with a high light intensity of 2.5 × 10 −7 einstein s −1 under a CO 2 atmosphere, giving HCOOH as the main product with CO and H 2 as minor products in all cases (Fig. 12). In the reaction using (Ring 4+ )-(SiPOM 4− ) as PS, HCOOH continuously formed for up to 6 h, and the turnover number and the selectivity of the HCOOH formation were TON HCOOH = 480 and S HCOOH = 86%, respectively (Fig. 12a). In the photocatalytic reaction using (Ring 4+ )-(GePOM 4− ) as PS, similar results were obtained; TON HCOOH = 446 (Fig. 12b). Notably, the use of Ring 4+ , a wellknown PS as described in the Introduction section, instead of (Ring 4+ )-(XPOM 4− ) induced the lowest durability of photocatalysis (Fig. 12c, TON HCOOH = 357), although the formation speed of HCOOH was faster in the initial stage compared to that of (Ring 4+ )-(XPOM 4− ) as PS. The slower formation of HCOOH in the cases using (Ring 4+ )-(XPOM 4− ) could be caused by the slower reduction processes of their TWERS ([(Ring 4+ )-(XPOM 6− )] 2− ), compared to that of free Ring 4+ .
The CV of RuCAT measured in a DMSO-TEOA solution containing (TBA + )(PF 6 − ) under a CO 2 atmosphere showed a broad irreversible wave at E p = −1.42 V vs. Ag/AgNO 3 (Fig. 7). Therefore, although the OERS of the Ring 4+ unit (Ring 3+ ) and protonated four-electron reduced species of the GePOM 4− (H 2 GePOM 6− ) unit can transfer an electron to RuCAT, the other reduced species of the SiPOM 4− and GePOM 4− units cannot perform electron transfer to RuCAT because of their more positive redox potentials (Table 2). 40 Upon irradiation with high light intensity, certain amounts of various species with different reducing powers are accumulated in the photocatalytic reactions, and we cannot separately investigate the roles of each reduced species of the supramolecular photocatalysts (ESI † [41][42][43] ). Therefore, the photocatalytic reactions were conducted using a lower light intensity (5.0 × 10 −9 einstein s −1 ) to investigate more details of the photocatalytic systems using either (Ring 4+ )-(SiPOM 4− ) or (Ring 4+ )-(GePOM 4− ) as PS by observing the exact amounts of the reduced species of the photosensitisers, as described below. HCOOH was photocatalytically produced as the main product in both cases (Fig. 13). Similar photoreactions without PS or RuCAT produced only small amounts of HCOOH (Table 3). These results indicate that (Ring 4+ )-(XPOM 4− ) and RuCAT functioned as PS and catalyst, respectively. Using (Ring 4+ )-(SiPOM 4− ) as PS, HCOOH continuously formed even aer irradiation for 12 h, and the turnover number and the selectivity of the HCOOH formation were TON HCOOH = 12 and S HCOOH = 97% aer irradiation for 12 h (Fig. 13a). Although in the photocatalytic reaction using (Ring 4+ )-(GePOM 4− ) as PS, HCOOH was also produced as the main product, an induction period was observed in the initial stage, and the formation rate of HCOOH slowed down aer irradiation for 6 h (Fig. 13b), in which TON HCOOH was 11 aer 12 h irradiation (S HCOOH = 94%). 44 We investigated the electron-accumulation behaviour of (Ring 4+ )-(XPOM 4− ) during photocatalytic reactions with low light intensity (5.0 × 10 −9 einstein s −1 ). Fig. 14a shows the UV-vis absorption spectra of the reaction solution with (Ring 4+ )-(SiPOM 4− ) as the PS during irradiation. In the initial stage, OERS ([(Ring 4+ )-(SiPOM 5− )] − ) was formed with a similar time scale to the photoreaction without RuCAT (Fig. 10a). The formation yield of the OERS was ∼100% aer 18 min of irradiation (Fig. 14b). Although longer irradiation times slowly induced the formation of a small amount of TWERS ([(Ring 4+ )-(SiPOM 6− )] 2− ), the yield did not change for up to 12 h. In other words, most of the produced TWERS were consumed during the photocatalytic reaction. Three-electron reduced species, that is, [(Ring 3+ )-(SiPOM 6− )] 3− was not observed at all. Based on these results and the similar formation potentials of the TWERS and the OERS of RuCAT (Table 2), we can conclude    that the TWERS can slowly supply one electron to RuCAT (eqn (17)) and can efficiently supply another electron to intermediate(s) as well, which should be produced from the OERS of RuCAT and CO 2 and/or H + , giving HCOOH (eqn (18) and (19)). This second supply of electrons from the TWERS should be sufficiently fast to suppress the decomposition of the Ru catalyst, which should be polymerization of reduced Ru complexes 41-43 (eqn (20), ESI †), and recover the OERS. In other words, TWERS (E 1/2 red = −1.40 V) has sufficient reduction potential to reduce the intermediate(s) of the formation of HCOOH with high efficiency.
DLS was applied to the photocatalytic reaction solutions (irradiated for 90 min) to detect the Ru polymer. In the case of (Ring 4+ )-(GePOM 4− ), not only particles with D = several nanometres, which are attributed to (Ring 4+ )-(GePOM 4− ) and partially to the accumulated oligomers, but also much larger particles with D = 140 nm, which are attributable to the Ru polymer (Fig. 15a). However, when using (Ring 4+ )-(SiPOM 4− ), such large particles were not observed aer the photocatalytic reaction (Fig. 15b). Therefore, under these reaction conditions (low light intensity), Ru polymer formation was suppressed during the photocatalytic reaction using (Ring 4+ )-(SiPOM 4− ) as the PS but not entirely in the system using (Ring 4+ )-(GePOM 4− ).
In the photocatalytic reaction using (Ring 4+ )-(GePOM 4− ) as PS and ReCAT as a catalyst, all (Ring 4+ )-(GePOM 4− ) were converted into the TWERS ([(Ring 4+ )-(GePOM 6− )] 2− ); that is, the TWERS were fully accumulated not only with an irradiation light intensity of 5.0 × 10 −9 einstein s −1 (l ex = 436 nm) but also with a lower intensity (1.0 × 10 −9 einstein s −1 ) ( Fig. 18a and b). This result indicates that the intermediate produced from the OERS of ReCAT cannot be reduced by TWERS [(Ring 4+ )-(GePOM 6− )] 2− , whose redox potential is E 1/2 = −1.33 V. This result is consistent with the results obtained using (Ring 4+ )-(SiPOM 4− ) as the reduction of the intermediate proceeded by [(Ring 4+ )-(SiPOM 6− )] 2− , with a redox potential of E 1/2 = −1.40 V, only at a slow rate. Therefore, in this photocatalytic reaction using (Ring 4+ )-(GePOM 4− ) as the PS, the intermediate must also be reduced only by [(Ring 4+ )-(H 2 GePOM 6− )] 2− (and partially by an a-amino radical produced by deprotonation of the oxidised TEOA 40 ). Notably, the formation rate of [(Ring 4+ )-(H 2 GePOM 6− )] 2− was slightly higher than that of [(Ring 3+ )-(SiPOM 6− )] 3− under these reaction conditions ( Fig. 11a and S7a). This effect is probably the reason for the similar CO formation rates in the two photocatalytic systems. Based on the results and the investigations, we can deduce that, in  the photocatalytic reduction of CO 2 using ReCAT, the reduction potential of the intermediate produced from We recently claried that the following process of ReCAT − to the intermediate is a unimolecular reaction with a rate constant of k = 1.8 s −1 at 298 K. 48 Although several assumptions of the structure of the intermediate were reported, this is the rst information about a redox potential of the intermediate in the photocatalytic reactions using Re(I)-complex catalysts to the best of our knowledge.
This method for determining the reduction potentials of the reaction intermediates was applied to another Re(I) complex, fac-[Re(Me 2 bpy)(CO 3

){OC(O)OC 2 H 4 N(C 2 H 4 OH) 2 }]
(ReMeCAT, Me 2 bpy = 4,4 ′ -dimethyl-2,2 ′ -bipyridine, Fig. S10 †), which has a more negative reduction potential (Ep red = −1.64 V, Fig. S11 †) than that of ReCAT (Ep red = − 1.50 V). We can expect the intermediate produced from the OERS of ReMeCAT to have a more negative reduction potential than that of ReCAT if the intermediates have similar structures. In this case, the RingMe 4+ unit with Me 2 bpy ligands instead of the bpy ligands must be used instead of the Ring 4+ unit in the PS because the reduction power of Ring 3+ (E 1/2 (Ring 4+/3+ ) = −1.54 V) is not sufficient to reduce ReMeCAT. In contrast, RingMe 3+ (E 1/2 (RingMe 4+/3+ ) = − 1.78 V, Fig. S11 †) can reduce ReMeCAT. A DMSO-TEOA (5 : 1 v/v) solution containing (RingMe 4+ )-(SiPOM 4− ) and ReMeCAT (0.05 mM each) was irradiated at l ex = 436 nm with a low light intensity (5.0 × 10 −9 einstein s −1 ) under a CO 2 atmosphere, also giving CO selectively (TON CO = 43 aer irradiation for 12 h, Table S4 †). The UV-vis absorption changes of the photocatalytic reaction solution indicate that two electrons were accumulated in (RingMe 4+ )-(SiPOM 4− ) in the photostationary state of the photocatalytic reaction (Fig. S12; † the details of these experimental results are described in the ESI †). This clearly shows that TWERS [(RingMe 4+ )-(SiPOM 6− )] 2− cannot reduce the intermediate produced from the OERS of ReMeCAT. Therefore, we can conclude that the intermediate has a more negative reduction potential than −1.40 V. These results strongly support the reliability of this estimation method for the reduction potential of the reaction intermediate.

Conclusions
We synthesised supramolecular photosensitisers consisting of ring-shaped tetranuclear Re(I) complexes and Keggin-type heteropolyoxometalates in a 1 : 1 ratio. These supramolecular photosensitisers can photochemically accumulate multielectrons in the presence of TEOA as the electron donor. In some combinations of photosensitisers and catalysts, supramolecular photosensitisers can donate two electrons to the catalyst and the intermediate to induce the photocatalytic reduction of CO 2 to CO or HCOOH. In molecular photocatalytic systems, CO 2 reduction generally proceeds via an intermediate produced by the OERS of the catalyst and CO 2 and/or H + . In many cases, since the reduction potential of the intermediate is more positive than the rst reduction potential of the catalyst, it is a challenge to obtain information on the redox properties of the intermediate. Moreover, electrons accumulated in supramolecular photosensitisers have different reduction powers; therefore, we can use these supramolecular photosensitisers to gain insight into the reduction potentials of the intermediates produced from the OERSs of the catalysts. The following conclusions were drawn from this study: (i) The reduction of the intermediate from the OERS of ReCAT does not proceed using an electron donor (reduced PS) with E red 1/2 = −1.33 V but proceeds with E red 1/2 = −1.40 V. (ii) The intermediate from the OERS of ReMeCAT cannot be reduced, even by the reduced PS with E red 1/2 = −1.40 V. (iii) The reduction of the intermediate made from the OERS of RuCAT and CO 2 barely proceeded with the reduced PS with E red 1/2 = −1.33 V. These photochemical methods for supplying multi-electrons with different potentials from one molecule should also aid in clarifying the redox properties of the intermediates of other photocatalytic reactions, such as H 2 evolution and photoredox catalytic reactions, providing multi-electron reduction products. In addition, the supramolecular photosensitizers developed in this study can suppress decomposition of intermediates produced from OERS of the catalyst owing to their rapid second electron donating abilities to increase durability of the photocatalysis.

Data availability
Data supporting the work within this manuscript is included within the full text of the manuscript and the provided ESI. †

Author contributions
MT and TA did most of the experiments and equally contributed. TM and YT guided them. NH made crystals of the samples. YK measured the X-ray diffraction data and analyzed the crystal structures, and KF and MY supported YK. OI conceived and leaded this research, and leaded to write this paper.

Conflicts of interest
There are no conicts to declare.