Rhenium(i) trinuclear rings as highly efficient redox photosensitizers for photocatalytic CO2 reduction† †Electronic supplementary information (ESI) available: Franck–Condon analysis; photochemical one-electron-reduced species formation and characterisation; photophysical, electrochemical and quench

Trinuclear Re(i)-rings were applied as redox photosensitizers in visible light-driven CO2 reduction in tandem with various catalysts, i.e., Re(i)-, Ru(ii)- and Mn(i)-diimine metal complex. The quantum yields for the Ru(ii) and Mn(i) catalysts were the among highest reported.


Introduction
Redox photosensitizers (PSs) have been widely used in various photocatalytic reactions such as for organic synthesis, dyesensitized solar cells, photoinduced H 2 or O 2 production from water and reduction of CO 2 . [1][2][3][4][5][6][7][8] The rst step of the photosensitization is the photoexcitation of the PS, followed by a reductive or oxidative quenching reaction with a substrate or semiconductor particles and electrodes. The produced oneelectron-reduced or one-electron-oxidized species (OERS or OEOS, respectively) donates an electron or hole, respectively, to another substrate in the nal process of the photosensitization. Therefore, PSs are required to have the following properties: (1) stability of the excited state, (2) stability of the OERS and/or OEOS, (3) strong oxidation and/or reduction power in the excited state and (4) strong reduction or oxidation power of the OERS or OEOS.
Some transition-metal complexes are frequently used as PSs not only because they full the aforementioned requirements but also because they have a strong absorption in the visible region, which is an important feature for solar energy conversion. Most reported PSs are mononuclear metal complexes, and their types have been limited mostly to Ru(II)-diimine and cyclometalated-Ir(III) complexes and their derivatives. Although some metal-porphyrins and metal-phthalocyanines as well as Pt(II)-, Os(II)-, Re(I)-and Fe(II)-diimine complexes have also been investigated as PSs, they are used only in limited types of reactions. [8][9][10][11][12][13][14] On the contrary, the possibilities of using multinuclear metal complexes as PSs have only been scarcely investigated to date. 15 We recently reported the photochemical synthesis of ringshaped multinuclear Re(I) complexes with cis, trans-[Re(bpy)(CO) 2 (P-P) 2 ] + (bpy ¼ 2,2 0 -bipyridine, P-P ¼ PPh 2 -(C n H m )-PPh 2 ) as repeating units. [16][17][18] They exhibited outstanding photophysical and electrochemical properties, such as high emission quantum yields, along with long lifetimes of the 3 MLCT excited states, even in solution at room temperature, and stability in the excited state and stronger oxidation power in the excited state compared to the corresponding mononuclear Re complex. They can also photochemically accumulate multiple electrons in one molecule. In the rst report on these Re-rings, we also briey introduced the idea that the Re-ring could be used as an extremely efficient PS for photocatalytic CO 2 reduction in tandem with a Re(I) catalyst under visible light irradiation.
Herein, we report the potentialities of these Re-rings as PSs in detail. Newly designed and synthesized trinuclear Re-rings, R(X), where each Re(I) unit is connected with p-bis(diphenylphosphino)benzene (Chart 1), were applied to photocatalytic CO 2 reduction with three kinds of typical catalysts, namely fac-[Re(bpy)(CO) 3

Results and discussion
Design and synthesis of the Re-rings The design of the new Re-rings was based on our previous work, which clearly demonstrated the relationship between the ring structures (i.e., the size of the ring and the type of the bridging bisphosphine ligand P-P) and their photophysical and electrochemical properties. [16][17][18] The trinuclear Re-ring connected with a phenylene spacer in the bisphosphine ligand exhibited both a long lifetime and a strong oxidation power in the excited state. The corresponding OERS, which are important intermediates in the redox-photosensitized reactions, were relatively stable. In order to tune these properties in a more desirable fashion, further structural modication can be performed on the bpy ligand. Namely, we prepared a series of new trinuclear Re-rings bearing 2,2 0 -bipyridine ligands with electron-donating substituents and connected with a phenylene spacer, as shown in Chart 1.
We successfully developed a new one-pot synthetic method for these Re-rings without using photochemical reactions. In the previously reported synthesis (Scheme S1, ESI †), the photochemical reaction was an essential procedure for removing the edge CO ligands from the corresponding linearshaped Re(I) multinuclear complexes. In the synthesis of the R(4$5) and R(OMe) rings, however, this reaction could not be used due to the fast photochemical decomposition of the linearshaped Re(I) trimers. In the new synthetic method, the oligomerization of the mononuclear Re(I) complex and the cyclization process can be completed simultaneously. A tricarbonyl Re(I)-diimine complex was reacted with Me 3 NO, which is reported to be an effective decarbonylation reagent for metal carbonyl complexes, 19,20 affording the biscarbonyl-Re(I) mononuclear complex as a building block with a labile ligand (Re(X)ph-L, L ¼ solvent molecule, Scheme 1). These reactions were instantaneous even under mild conditions and proceeded quantitatively. Prolonged reux without any additional reagents led to both coupling and cyclization, affording the corresponding trinuclear Re-ring as the major product, with larger linear and eventually ring-shaped Re(I) multinuclear complexes as minor products (Fig. S1, ESI †). The Re-rings R(4$5), R(OMe) and R(5) were successfully isolated from the reaction mixtures using size exclusion chromatography in ca. 20% yield, which is highly comparable with or better than the total yields of the multi-step synthetic strategy, including the photochemical reaction (Scheme S1, ESI †).

Photophysical properties
Fig . 1A shows the UV-vis absorption spectra of the Re-rings, measured in DMF. In addition to an interligand transition (IL) of the bpy ligands at l abs z 300 nm, a well-dened and intense 1 MLCT absorption band centred at l abs z 400 nm was observed for the Re-rings with methyl-substituted bpy ligands. In contrast, R(OMe) displayed both a strong absorption at l abs z 300 to 350 nm and a shoulder peak at l abs z 380 to 410 nm.
All of the Re-rings emitted strongly in DMF, even at room temperature (Fig. 1B). The emissive state is attributable to the 3 MLCT excited state in the cases of the Re-rings with methylsubstituted bpy ligands because of their broad and non-vibrational shapes. However, the emission spectrum of R(OMe) showed a vibrational structure (l em z 540 nm, 510 nm (sh)), which indicates that the emissive state contained not only 3 MLCT but also 3 IL character. Table 1 summarizes the photophysical properties of the Rerings. Both the emission quantum yields (F em ) and the emission lifetimes (s em ) concomitantly increased in the case of the Re-ring with a higher emission energy. It is noteworthy that the emission quantum yield of R(OMe) is the highest among the reported Re(I) complexes (F em ¼ 0.66), and the lifetime of the excited R(OMe) was also very long (s em ¼ 7.8 ms), mainly due to the contribution of the 3 IL state. These absorption and emission properties were not affected by the presence of CO 2 . Fig. 2 shows the cyclic voltammograms of the Re-rings, where one reversible reduction wave was observed at E red 1/2 ¼ À1.82 to À1.91 V vs. Ag/AgNO 3 . This is attributed to a 3-electron reduction process, 21 except for R(4), where each electron was supplied to the diimine ligand in each Re(I) unit of the Rering (bpy/bpyc À ). This strongly suggests that the Re units in the Re-ring have no strong electronic interaction with each other.

Electrochemical properties
On the contrary, the corresponding wave of R(4) could be deconvoluted with two Gaussian functions with a 2 : 1 intensity ratio (Fig. S2, ESI †). Hence, some electronic interplay between the units may occur in R(4), just as we have also observed in smaller Re rings with shorter bridging ligands. 17,18 The second irreversible waves at E p ¼ À2.4 to À2.5 V vs. Ag/ AgNO 3 correspond to the reduction of the metal centre. Table 2 summarizes the electrochemical properties.

Reductive quenching and formation of one-electron-reduced species
The redox properties of the 3 MLCT excited state, i.e., the values of E * red were calculated on the basis of E red 1/2 ( Table 2) and DG 0 MLCT (these values were obtained using Franck-Condon analysis, see ESI † for a detailed description).
All of the Re-rings have sufficiently positive values of E * red to use triethanolamine (TEOA) as a reductant, which has oen been used in various photocatalytic reactions. However, the quenching efficiencies of the excited states of some other PSs, such as Ru(II)-trisdiimine complexes, are low because of the relatively weak reducing power of TEOA (E ox ¼ 0.51 V vs. Ag/ AgNO 3 ), 22 or the too short a lifetime of the metal complex. Fig. 3 shows the linear Stern-Volmer plots of the emission intensity of the Re-rings in the presence of increasing concentrations of TEOA, which allowed us to obtain the Stern-Volmer constants, K SV , and the quenching rate constants, k q , listed in Table 2. All of the Re-rings with the phenylene chain exhibited higher quenching fractions (h q ) in a DMF-TEOA (5 : 1 v/v) solution compared to the corresponding mononuclear Re(I)-bisphosphine complexes and even compared to other rings with a saturated alkyl spacer in P-P. For example, the k q of R(5) was 7 times larger than that of the corresponding Re-ring with ethylene chains (R(5)-e) instead of phenylene chains.
Irradiation of an Ar-saturated DMF-TEOA (5 : 1 v/v) solution containing the Re-rings at l ex ¼ 436 nm caused changes in the UV-vis absorption spectra. Fig. 4A shows the case of R(5), where the change is attributed to the formation and accumulation of the OER species R(5)c À in the solution because the differential spectrum before and aer the irradiation (Fig. 4B) was very similar to that of R(5)c À obtained using bulk electrolysis (Fig. 4D). The formation quantum yield of R(5)c À (F OER ) was 1.2, and further irradiation induced an accumulation of about 1.1 electrons into one molecule of R(5) (Fig. 4C).
In the cases of the other Re-rings, the corresponding OERS were also accumulated in the solution using irradiation in the presence of TEOA ( Fig. S3 and S4, ESI †). These results strongly indicate that the photochemical formation of the OERS of the Re-rings efficiently proceeds using TEOA as the reductant, and the OERS are relatively stable in anaerobic solution even under Scheme 1 Synthetic route for the Re-rings.   16   irradiation. It is noteworthy that the OERS also accumulated even under a CO 2 atmosphere, with similar F OER in the absence of the catalysts that were used for CO 2 reduction, as described below.
Photocatalytic reaction with fac-[Re(bpy)(CO) 3 (CH 3 CN)] + . Because the excited states of the Re-rings with the phenylene chains have stronger oxidation powers compared to the previously reported ring with ethylene chains, which requires the usage of 1,3-dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole (BIH) as an electron donor, 39 we could employ triethanolamine (TEOA), which is commonly used in various photocatalytic reactions but is a weaker reductant than BIH. fac-[Re(bpy)(CO) 3 (CH 3 CN)] + (Re-ACN) was used as a catalyst; it is quantitatively converted into the CO 2 adduct (Re-OCO(O)NR 2 ) with the aid of TEOA (eqn (1)) under the photocatalytic reaction conditions. 40 Although Re-ACN is not the real catalyst, we use this nomenclature in the following text because it was used as the starting complex. (1) In a typical run (eqn (2)), a CO 2 saturated DMF-TEOA (5 : 1 v/ v) mixed solution containing both R(5) and Re-ACN in an equimolar ratio (0.05 mM) was irradiated at l ex ¼ 436 nm (5.7 Â 10 À9 einstein per s), where 87% of the irradiated photons were absorbed by R(5), giving CO selectively. (2)    The quantum yield (F CO , eqn (3)) was 0.61 and the TON (eqn (4)) of CO formation (TON CO ) was 71 aer 15 h of irradiation (entry 2, Table 3).

F product ¼ product [mol]/total absorbed photons by the solution [einstein]
(3) The use of other PSs, R(4$5) and R(OMe), instead of R(5) gave similar F CO values; however, TON CO decreased (entries 4 and 5, Table 3). In the case of using R(4), F CO was lower, while TON CO was slightly higher than the others (entry 1, Table 3). Note that the photocatalyses of the systems using these new Re-rings as the PS were higher than that using R(5)-e (entry 6, Table 3), which was the most efficient PS for CO 2 reduction using BIH, which is a much stronger reductant than TEOA, in the reported system. The lower quenching efficiency of the 3 MLCT excited state of R(5)-e by TEOAin other words, its weaker oxidation powershould be a main reason for this lower F CO .
Irradiation at a slightly shorter wavelength with a lower light intensity (l ex ¼ 405 nm, instead of 436 nm), and a half-molar ratio of the catalyst (0.025 mM) increased the F CO up to 0.74 (Table 3), mainly because of the reduced inner-lter effect of the catalyst, which affects the number of photons absorbed by the photocatalytic mixture (Table S4, ESI †). In addition, lower light intensity may sometimes have an inuence on the quantum yield. This phenomenon was observed in the cases using 436 nm-irradiation light, for example, the R(5)/Re-ACN photocatalytic system gave slightly higher F CO (0.68), under irradiation with a 3.5-fold lower light intensity (1.65 Â 10 À9 einstein per s).
These high values of F CO can be attributed to several factors. One important reason is likely the high formation yields of the OERS of the Re-rings via reductive quenching of the excited Rerings with TEOA because of their long lifetimes and strong oxidation powers in the excited state, as described in the previous section. Another reason is the rapid electron transfer process from R(X)c À to the catalyst, which suppresses the accumulation of the OERS of the Re-rings. Since accumulation of the OERS induces the inner-lter effect owing to their strong absorption in the visible region (Fig. 4), it should decrease the number of photons absorbed by the (not-reduced) Re-rings in the photocatalytic mixture. This makes the apparent quantum yield lower. Fig. 5 shows the UV-vis absorption spectral changes of the photocatalytic reaction solution containing the R(5) PS with the Re-ACN catalyst, where accumulation of the OERS of R(5) was not observed. Note that most of the R(5) was converted to the corresponding OERS within 3 min of irradiation in the absence of the catalyst (Fig. 4). These results clearly indicate that electron transfer from the OERS of R(5) to the catalyst was not the rate-limiting step. Since the reduction potential of R(5) ( Table 2) was more negative than the onset potential of Re-OCO(O)NR 2 (more positive than À1.8 V vs. Ag/AgNO 3 in DMF-TEOA, Fig. S5, ESI †), the electron scavenging process from the OERS of R(5) using the catalyst has an exergonic character.
Similar results were obtained in the photocatalytic reactions using the other PSs except R(4) (Fig. S6 to S8, ESI †). A very small amount of the OERS of R(4) accumulated within 2 min of irradiation, which suggests a slower electron transfer process for the catalyst compared to the other Re-rings, probably because of the more positive redox potential of R(4). This may be a reason why F CO decreased in the case using R(4) compared to the other Re-rings; another reason is likely the lower quenching ratio of the excited state of R(4) using TEOA (Table 2).
Until reaching 1 h of irradiation, R(5) functioned stably as the PS; however, a longer irradiation time induced the  decomposition of R(5) aer most of the catalyst was already decomposed (see the result aer 2 h of irradiation as an example in Fig. 6B). Therefore, the Re-rings should be durable PSs in the presence of the catalyst.
In the absence of the Re-ring, only a trace amount of CO was observed (entry 7, Table 3); hence the catalyst without the Rering could not photocatalyze CO 2 reduction under these reaction conditions. In the case using solely R(5) without the catalyst, a certain amount of CO was produced (entry 3, Table 3: TON CO ¼ 6 for 1 h irradiation, and 8 even for 15 h irradiation). Note that the Re-ring has 6 CO ligands, the decomposition of R(5) should yield some CO molecules. Although the decomposition products of R(5) might work as catalysts for CO 2 reduction with the residual R(5) as a PS, their contribution to the photocatalytic reaction should be very low because the CO formation mostly stopped aer 1 h of irradiation.
In the photocatalytic system of R(5)/Re-ACN, usage of BIH instead of TEOA as the reductant did not affect the efficiency of the photocatalytic reaction, i.e., F CO did not change in the presence of various concentrations of BIH (F CO $ 0.6, Fig. 7A). However, TON CO drastically increased with higher concentrations of BIH, as illustrated in Fig. 7B. The total amount of produced CO was similar or even greater than the initial amount of BIH aer a long irradiation, e.g., the amount of evolved CO was 50 mmol aer 15 h of irradiation in the presence of 40 mmol of BIH, where both BIH and TEOA should work as reductants. It has been reported that BIH is a quantitative 2e À donor (eqn (5)), and its oxidation product (BI + ) did not impede the photocatalytic reduction of CO 2 . 39 Therefore, accumulation of the oxidation products of TEOA in the reaction solution may disturb the photocatalytic reaction in some process(es) because CO formation levelled off aer TON CO reached 70 in the presence of only TEOA as the reductant.
Recently, trans-Ru(N^N)(CO) 2 Cl 2 (N^N ¼ diimine ligand) has been oen used as a catalyst for CO 2 reduction, giving CO under neutral and acidic conditions or HCOOH under basic conditions. 25,41,42 Since this type of complex usually suffers undesirable reductive polymerization during the photocatalytic reaction, which slows or disables the reaction progress, we chose an Ru(II) catalyst with bulkier tert-butyl groups on the diimine ligand, trans(Cl)-Ru(dtbb)(CO) 2 Cl 2 (Ru(tBu)-Cl 2 , Chart 2). Its absorption and electrochemical properties in DMA-TEOA are shown in the ESI (Fig. S10 and S14B †).
A CO 2 -saturated DMA-TEOA (5 : 1 v/v) mixed solution containing an equimolar ratio (0.05 mM) of Ru(tBu)-Cl 2 catalyst and R(5) PS was irradiated under 436 nm light (4.2 Â 10 À9 einstein per s), where 98% of the photons were absorbed by R(5). Formic acid was evolved as a major product, accompanied by H 2 and CO as minor products (Fig. 8). The turnover number (TON) of HCOOH reached 290, and the TONs for H 2 and CO were around 70 and 20, respectively, aer 23 h irradiation (Table 4).
In the photocatalytic reaction, R(5) was very durable, and ca. 80% of R(5) remained even aer 23 h of irradiation (Fig. 9, bottom). The quantum yield of HCOOH (F HCOOH ) formation was 58% and was independent of the light intensity. This quantum yield is the highest value reported to date for photocatalytic systems for the reduction of CO 2 to HCOOH.   It should be noted that most of the reported photocatalytic systems comprising mononuclear PSs and a catalyst usually contain a much larger amount of PS than the catalyst to supply electrons to the catalyst more rapidly and/or to prevent light absorption by the catalyst. Although the same concentration of the PS R(5) as the catalyst (Ru(tBu)-Cl 2 ) was used in this photocatalytic reaction, only a small amount of dimer [Ru I (dtbb)(CO) 2 L] 2 , possibly including OER of the Ru(II) catalyst, was observed at a very early stage (Fig. 9, top). Further irradiation caused a decrease in dimer concentration (Fig. 9, middle and bottom), and the polymer of the Ru complex was not detected during the photocatalytic reaction. Hence, Ru-polymerization was efficiently suppressed. However, steric hindrance of the tert-butyl groups was not the main factor for the inhibition of polymerization because a similar experiment under Ar atmosphere instead of CO 2 induced quantitative formation of the dimer of the Ru complex at a very early stage (Fig. S11, ESI †), followed by its polymerization. Conversely, during the photocatalytic reaction under a CO 2 atmosphere, neither the OERS of the Re-ring (R(5)c À ) nor the Ru polymer were detected, suggesting that the reaction of R(5)c À with the catalyst (and/or the dimer) and the subsequent reaction with CO 2 were very efficient and fast.

(6)
Photocatalytic reaction with fac-[Mn(dtbb)(CO) 3 (CH 3 CN)] + . As an earth-abundant metal catalyst, fac-Mn(N^N)(CO) 3 Br has attracted attention mainly in electrochemical systems for the reduction of CO 2 to CO. 44,45 Recently, the Mn complex was also applied to photocatalytic CO 2 reduction with a [Ru(N^N) 3 ] 2+ PS, where the main product was HCOOH. 38,46 However, the efficiencies of the photocatalytic systems are not very high (F HCOOH ¼ 0.05 to 0.14).
In a typical run, a CO 2 -saturated DMA-TEOA (5 : 1 v/v) mixed solution containing both R(5) as the PS and Mn(tBu)-ACN as a catalyst in an equimolar ratio (0.05 mM) was irradiated at l ex ¼ 436 nm (5.3 Â 10 À9 einstein per s), of which 72% was absorbed by R(5) in at least the starting stage of the photocatalytic reaction. Formic acid was evolved as the major product, with an overall TON of 85, and accompanied by CO (Fig. 10).
In the initial stage, the UV-vis absorption spectral changes accurately indicated dimerization of the Mn complex, giving [Mn(dtbb)(CO) 3 ] 2 with a strong absorption in a wide range of the visible region (Fig. 11A). During the dimer formation, which increased up to 4.5 min of irradiation, the OERS of R(5) was not detected. These results clearly suggest that electron transfer from R(5)c À to the Mn complex proceeded rapidly, followed by dimerization of the OERS of the Mn complex. The absorption of the Mn dimer disappeared within 1 h of irradiation (Fig. 11B). The CO formation overlapped with the prole of the dimer lifetime, i.e., the CO formation temporarily reached a plateau  aer the Mn-dimer decomposed (TON CO ¼ 4 at 1 h). Therefore, the Mn-dimer may function as a catalyst for CO formation. The formation of CO restarted aer further irradiation. The time prole of the HCOOH formation was different from that of the CO formation (Fig. 10). The quantum yield of the HCOOH formation was 48% despite the inner lter effect of the dimer absorption. This is the highest value among reported photocatalytic systems using a Mn complex as the catalyst so far. The photocatalytic reaction in the absence of the photosensitizer produced only 0.5 mmol of CO, and no HCOOH was observed, while the known photochemical fac-to mer-isomerisation with negligible Mn-dimer formation 47 and simultaneous decomposition of the Mn(I) complex occurred rapidly (Fig. S16, ESI †).
The durability of the photocatalytic system was not improved by the addition of the stronger electron donor BI(OH)H. The formation of the Mn dimer and its persistence in solution were very similar to those in the system without BI(OH)H. Since the oxidized product, BI(O À ) + , also caused an inner lter effect (Fig. S17, ESI †), the apparent quantum yield was lower when using BI(OH)H than when not using it.
Employing R(4$5), which has a more negative reduction potential, did not improve the photocatalytic performance. The TONs for HCOOH and CO decreased compared to R(5), owing to the lower stability of this ring (Fig. S18 and Table S5, ESI †).
Photochemical reactions were performed with a 500 W highpressure mercury lamp EHBWI (Eikosha) with a uranium glass lter (>330 nm) in a Pyrex doughnut-form cell, with bubbling of N 2 gas. During irradiation, both the reaction vessel and the light source were cooled with tap water. Separation of the larger complexes was achieved using size exclusion chromatography (SEC) using a pair of Shodex PROTEIN KW-2002.5 columns (300 mm Â 20.0 mm i.d.) with a KW-LG guard-column (50 mm Â 8.0 mm i.d.) and a JAI LC-9201 recycling preparative HPLC apparatus (Japan Analytical Industry Co.) with a JASCO 870-UV detector (the detection wavelength was chosen as 360 nm). The eluent was MeOH-CH 3 CN (1 : 1 v/v) containing CH 3 COONH 4 (0.15 M) and the ow rate was 6.0 mL min À1 . 56 For analytical SEC, we used two sequentially connected Shodex KW-402.5-4F columns (300 mm Â 4.6 mm i.d.), a KW400G-4A guard-column (10 mm Â 4.6 mm i.d.), a JASCO 880-51 degasser, a 880-PU pump, a MD-2010 Plus UV-vis photodiode-array detector (l det ¼ 360 nm) and a Rheodyne 7125 injector. The column temperature was maintained at 40 C using a JASCO 860-CO columnoven. The eluent was MeOH-CH 3 CN (1 : 1 v/v) containing CH 3 COONH 4 (0.5 M) and the ow rate was 0.2 mL min À1 . For LC analysis of the photocatalytic reactions, we used a SHIMAZU UPLC Nexera X2 apparatus with a Waters Acquity SEC column (150 mm Â 4.6 mm i.d.), a Shimadzu DGS-20A degasser, a LC-30AD pump, a SPD-M30A UV-vis photodiode-array detector and a Rheodyne 7125 injector. The column temperature was maintained at 40 C using a JASCO 860-CO column-oven. The eluent  was MeOH-CH 3 CN (1 : 1 v/v) containing CH 3 COONH 4 (0.5 M), and the ow rate was 0.2 mL min À1 . 1 H-NMR and 31 P-NMR spectra were acquired with a JEOL AL400, JEOL ECX400 or JEOL ECA400II spectrometer. Chemical shis (d/ppm) were referenced to the residual 1 H-signals of the deuterated solvent (1.94 ppm for CD 3 CN and 2.05 ppm for CD 3 COCD 3 ) and the 31 Psignal of PF 6 À (À145 ppm), respectively. All NMR spectra were recorded at room temperature. Electrospray ionization mass spectrometry (ESI-MS) was conducted on a Shimadzu LCMS-2010A mass spectrometer. Electrospray ionization time-of-ight mass spectrometry (ESI-TOFMS) was undertaken with a Waters LCT Premier instrument. FT-IR spectra were recorded with a JASCO FT/IR-610 spectrometer at 1 cm À1 resolution with a TGS detector. Electrochemical voltammetric techniques were performed using a CHI720D electrochemical analyzer (CH Instruments, Inc.) with a glassy-carbon working electrode (3 mm i.d.), an Ag/AgNO 3 (10 mM) reference electrode and a Pt counter electrode. The supporting electrolyte was Et 4 NBF 4 (0.1 M), which was dried under vacuum at 100 C for 1 day prior to use. Bulk electrolysis was performed in a quartz UV-vis OTTLE cell (1.0 mm optical path length) equipped with a Pt-mesh working electrode, an Ag/AgNO 3 (10 mM) reference electrode, and a Pt counter electrode. In situ UV-vis spectral changes were measured using a Photal MCPD-9800 spectrometer. The applied potential was controlled with a CHI720D electrochemical analyzer (CH Instruments, Inc.). The supporting electrolyte was Et 4 NBF 4 (0.1 M), which was dried in vacuum at 100 C for 1 day prior to use. UV-vis absorption spectra were recorded with a JASCO V-565 or V-670 spectrophotometer. Emission spectra were recorded at 25 C using a JASCO FP-6500 spectrouorometer and were corrected for PMT response. Emission quantum yields were determined with a calibrated integrating sphere (Absolute PL Quantum Yield Measurement System C9920-01, Hamamatsu Photonics k.k.), comprising a Xe lamp as an excitation source and a multichannel spectrometer (C10027). 57 Emission lifetimes were obtained at 25 C using a HORIBA FluoroCube timecorrelated single photon counting system. The excitation light source was a NanoLED-405L (401 nm, <200 ps). All measurements were performed in a quartz cubic cell (1 cm optical path length); the absorbances were adjusted to z0.1 at l ex , and the solutions were degassed with Ar prior to the measurements.
Photochemical OERS formation experiments were performed in Ar-saturated DMF-TEOA (5 : 1 v/v) solution (4 mL) containing R(X) in a quartz cubic cell (1 cm optical path length). Photocatalytic reactions were performed in a CO 2 -saturated DMF-TEOA or DMA-TEOA (5 : 1 v/v) mixture containing R(X) and a catalyst in a quartz cubic cell (1 cm optical path length; 11 mL volume). In the case of the Re(I) or Mn(I) catalysts, a DMF or DMA solution, respectively, was prepared one night prior, starting from the corresponding acetonitrile complex. The sample solutions were irradiated using a high-pressure Hg lamp (Ushio USH-500SC) combined with a CuSO 4 $5H 2 O aqueous solution lter (250 g L À1 , 5 cm pass length), a 436 or 405 nm band-pass lter (fwhm ¼ 10 nm, Asahi Spectra Co.) and neutral density glass lters (Chuo Precision Industrial Co.) were used to adjust to the desired light intensity. In situ measurements of the UV-vis absorption spectra were conducted using a Photal MCPD-9800 or Photal MCPD-2000 spectrometer. The temperature of the reaction solution during the irradiation was maintained at 25 C using an IWAKI CTS-134A constant-temperature system. The incident light intensity at 436 or 405 nm was determined using a K 3 [Fe(C 2 O 4 ) 3 ] chemical actinometer. 58 The gaseous reaction products (CO and H 2 ) were analysed using GC-TCD (GL Science GC323) with an active carbon column. HCOOH was analysed using a capillary electrophoresis system (CAPI-3300I, Otsuka Electronics Co.). As pretreatment for HCOOH quantication, the photocatalytic reaction solution was diluted by 10 times with H 2 O.

Syntheses
General procedure for the synthesis of Re(X)-ph. A CH 2 Cl 2 solution of fac-Re(N^N)(CO) 3 Br and AgOTf (1.05 eq.) was reuxed until the complete substitution of the Br À ligand by OTf À (3 to 7 h) as veried using TLC. The solution was ltered through a pad of Celite and evaporated to dryness. The crude fac-Re(N^N)(CO) 3 OTf and p-bis(diphenylphosphino)benzene (5 eq.) were reuxed in THF under an inert atmosphere in dim light for 2 d. The solvent was removed under reduced pressure; the crude solid was puried using column chromatography (SiO 2 , CH 3 CN-CH 2 Cl 2 1 : 5). Re(X)-ph was converted into PF 6 À salt for the NMR, FT-IR and ESI-MS characterisation as follows. The solid was dissolved in a small amount of MeOH; then, a few drops of a saturated aqueous solution of NH 4 PF 6 were added, leading to precipitation of the product.