Photocatalytic CO2 reduction with high turnoverfrequency and selectivity of formic acid formationusing Ru(II) multinuclear complexesYusuke Tamakia, Tatsuki Morimotoa, Kazuhide Koikeb, and Osamu Ishitania,c,1

aDepartment of Chemistry, Graduate School of Science and Engineering, Tokyo Institute of Technology, O-okayama 2-12-1-NE-1, Meguro-ku, Tokyo152-8551, Japan; bNational Institute of Advanced Industrial Science and Technology, Onogawa 16-1, Tsukuba, 305-8569, Japan; and cAdvanced LowCarbon Technology Research and Development Program (ALCA), Japan Science and Technology Agency, Sanbancho, Chiyoda-ku, Tokyo 102-0075, Japan

Edited by Thomas J. Meyer, University of North Carolina at Chapel Hill, Chapel Hill, NC, and approved July 24, 2012 (received for review February 6, 2012)

Previously undescribed supramolecules constructed with variousratios of two kinds of Ru(II) complexes—a photosensitizer and acatalyst—were synthesized. These complexes can photocatalyzethe reduction of CO2 to formic acid with high selectivity and dur-ability using a wide range of wavelengths of visible light andNADH model compounds as electron donors in a mixed solution ofdimethylformamide–triethanolamine. Using a higher ratio of thephotosensitizer unit to the catalyst unit led to a higher yield offormic acid. In particular, of the reported photocatalysts, a trinuc-lear complex with two photosensitizer units and one catalyst unitphotocatalyzed CO2 reduction (ΦHCOOH ¼ 0.061, TONHCOOH ¼ 671)with the fastest reaction rate (TOFHCOOH ¼ 11.6 min−1). On theother hand, photocatalyses of a mixed system containing twokinds of model mononuclear Ru(II) complexes, and supramoleculeswith a higher ratio of the catalyst unit were much less efficient,and black oligomers and polymers were produced from the Rucomplexes during photocatalytic reactions, which reduced the yieldof formic acid. The photocatalytic formation of formic acid usingthe supramolecules described herein proceeds via two sequentialprocesses: the photochemical reduction of the photosensitizer unitby NADH model compounds and intramolecular electron transferto the catalyst unit.

Recently, global warming and shortages of fossil fuels and car-bon resources have become serious issues. The development

of technologies to convert CO2 into useful organic compoundsusing sunlight as an energy source would serve as an ideal solutionto these problems.

Formic acid, which is the two-electron reduction product ofCO2, has recently attracted attention as a storage source of H2

(1, 2). Formic acid itself is an important chemical. It has beenemployed as a preservative and an insecticide and is also a usefulacid, reducing agent, and source of carbon in synthetic chemicalindustries.

Only a few photocatalysts for the selective formation of formicacid from CO2 have been reported (3–8). Although oligo(p-phe-nylenes) (3) or a mixed system of phenazine and Co cyclam (4)successfully photocatalyzed the reduction of CO2 to formic acid,these systems cannot work with visible light. It has been reportedthat ½RuðbpyÞ2ðCOÞ2�2þ (bpy ¼ 2,2′-bipyridine) acted as a cata-lyst for reducing CO2 (5–8). Under basic conditions, a mixed sys-tem of this complex with ½RuðbpyÞ3�2þ as a redox photosensitizerphotocatalyzed the reduction of CO2 to formic acid with highselectivity (6, 8). However, this photocatalytic system is limited byinstability as evidenced by the fact that the catalyst decomposedfollowing prolonged irradiation and generated black precipitates.

We have recently developed a unique architecture for con-structing visible-light-driven supramolecular photocatalysts,consisting of a ½RuðN∧NÞ3�2þ (N∧N ¼ a diimine ligand)–typecomplex as a photosensitizer and a Re(I) diimine complex asa catalyst (9–12). These supramolecules can selectively photo-catalyze the reduction of CO2 to CO with high efficiency. Therequirements for constructing efficient supramolecular photo-

catalysts with Ru(II) and Re(I) complexes involves two keyprinciples: First, in the triplet metal-to-ligand-charge-transfer(3MLCT) excited state of the Ru(II) unit, the excited electronshould be located at the bridging ligand. Second, a nonconjugatedbridging ligand should be used because a conjugated system in thebridge ligand lowers the reducing power of the catalyst unit (9).The efficiencies and the stability of these supramolecular photo-catalysts (ΦCO ¼ 0.12–0.15, TONCO ∼ 200) are much greaterthan those of the other reported supramolecular-type photocata-lysts for both CO2 reduction (13–16) and H2 evolution (16–19).

We sought to apply this architecture to construct unique supra-molecular photocatalysts for the reduction of CO2 to selectivelyform formic acid (which has not been reported to date). We suc-cessfully developed such photocatalysts having high efficiency,high selectivity for formic acid formation, high stability, and a fastreaction rate.

The structures and abbreviations of the synthesized supra-molecular complexes are shown in Chart 1 and include½RuðdmbÞnðBLÞ3−n�2þ [dmb ¼ 4,4′-dimethyl-2,2′-bipyridine;BL ¼ 1,2-bis(4'-methyl-[2,2'-bipyridin]-4-yl)ethane] as the photo-sensitizer unit; and ½RuðdmbÞmðBLÞ2−mðCOÞ2�2þ as the catalystunit. The abbreviations x and y in ðx; yÞ indicate the numbers ofeach unit in one molecule (x: photosensitizer, y: catalyst). Theirmodel mononuclear complexes are also abbreviated; (1,0) and(0,1) represent the photosensitizer and catalyst, respectively.

Results and DiscussionFor evaluation of the photocatalysis, we determined both theformation quantum yield (Φ) and the turnover number (TON)

Chart 1. Structures and abbreviations of Ru(II) complexes

Author contributions: O.I. designed research; Y.T. and T.M. performed research; K.K.analyzed data; and Y.T. and O.I. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: ishitani@chem.titech.ac.jp.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1118336109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1118336109 PNAS ∣ September 25, 2012 ∣ vol. 109 ∣ no. 39 ∣ 15673–15678

CHEM

ISTR

YSP

ECIALFEAT

URE

Dow

nloa

ded

by g

uest

on

June

21,

202

0

of the product along with the turnover frequency (TOF) in part.Since suitable irradiation conditions for obtaining these valuesare different, we chose the following reaction conditions (SI Text):

Light Irradiation Condition (LIC) 1 for determining Φ: Forevaluating the number of photons absorbed by the photocatalyst,monochromic light is required, and the absorbance of the reac-tion solution at the wavelength of irradiation should be suffi-ciently high, at least 2 where 99% of the irradiated photons isabsorbed by the photocatalyst. We used 480-nm monochromiclight obtained using a 500-W Xe lamp with a band-pass filter(FWHM ¼ 10 nm); the light intensity was 4.9 × 10−8 or4.9 × 10−9 einstein s−1, which was controlled using neutral den-sity (ND) filters. The concentration of the photosensitizer unit(not the photocatalyst) was 0.3 mM, and the solution was vigor-ously mixed during the irradiation. The incident light flux wascompletely absorbed by the photosensitizer unit because theabsorbance of the reaction solutions was larger than 3 at 480 nmand the light-pass length was 1 cm.

In this reaction condition, it is difficult to determine reason-able TON values because too-large amounts of the products—i.e., formic acid and BNA2s—should be accumulated in the reac-tion solution after long irradiation and the products obstruct thephotocatalytic reaction (see below). The photon flux determinedthat the rates of the photocatalytic reactions—i.e., the ordinarylight source—even using a 500-W Xe lamp cannot supply enoughphoton flux for determining maximum TOF values. For obtainingmore exact values of TON and TOF, we reduced the concentra-tions of the photocatalysts in the reaction solutions and used dif-ferent light sources as follows.

LIC2 for determining TON: >500-nm light was obtained usinga 500-W high-pressure mercury lamp equipped with a uranyl glassand a K2CrO4 (30% w∕w, d ¼ 1 cm) aqueous solution filter,which supplied stronger light flux than that used in LIC1. Theconcentration of the photosensitizer unit was 0.05 mM. TheTONs were determined after 20-h irradiation in all cases, andcalculated as the produced amount of formic acid divided bythe amount of supramolecule added. Using a merry-go-roundirradiation apparatus, we could irradiate up to eight samplessimultaneously with the same light-intensity.

LIC3 for determining TOF and measuring NMR: >420-nmlight was obtained using a 500-W high-pressure mercury lamp ora 500-W Xenon lamp with a cutoff filter. This visible-light sourcecould irradiate the reaction solution with the highest intensity inour laboratory.

The photosensitizer unit can be selectively excited using LIC1and LIC2, while, in the case of LIC3, the irradiated light isabsorbed mainly by the photosensitizer unit and partially by thecatalyst unit where slow photolysis of the photocatalysts wasobserved and TON obtained in LIC3 was slightly lower than thatin LIC2 (Table 1). The reductants (BNAH,MeO-BNAH) and theBNA2 scarcely absorbed >420-nm light (Fig. S1A).

In a typical run for determining TON, a solution of dimethyl-formamide (DMF) and triethanolamine (TEOA) (4∶1 v∕v) con-taining (1,1) (0.05 mM) and 1-benzyl-1,4-dihydronicotinamide(BNAH, 0.1 M) as a sacrificial electron donor was bubbled withCO2 for 20 min and then irradiated (LIC2) to give formic acidwith high selectivity. The TON of formic acid formation exceeded300 after 20-h irradiation with small amounts of CO and H2

(Eq. 1, Fig. 1). On the other hand, formic acid was not producedat all in the absence of BNAH or without irradiation.

[1]

In a labeling experiment using 13CO2, a strong signal attribu-ted to H13COO− was observed at 168.2 ppm in the 13C NMRspectrum of a DMF-d7-TEOA (4∶1 v∕v) solution containing (1,1)(0.5 mM), BNAH (0.1 M), and 13CO2 (532 Torr) after irradiation(LIC3) for 14.5 h (TONHCOOH ¼ 59) (Fig. S2, Upper)*. Also, inthe 1H NMR of the same solution, a doublet (JCH ¼ 188 Hz)attributable to the proton coupled to the 13C of H13COO− wasobserved at 8.52 ppm, but no singlet due to the proton ofH12COO− was detected (Fig S2, Lower)†. These results clearlyshow that the carbon source of formic acid is CO2.

Fig. 2 shows the IR spectra of (0,1) in DMF (a) and inDMF-TEOA (b). Two peaks observed in DMF are attributedto the symmetric and asymmetric stretching bands of the two COligands (2094 and 2040 cm−1, respectively). On the other hand,only one peak appeared in the νCO region in the DMF-TEOA

Table 1. Photocatalytic properties of the supramolecules and the model system*

entry photocatalyst reductant

products∕μmol †

ΦHCOOH‡ TONHCOOH

§ Φem¶ τem

∥ ns kq** 107 M−1 s−1 ηq

†† %HCOOH CO H2

1 (2,1) BNAH 30.4 1.8 1.8 0.041 562 0.085 726 1.97 592 MeO–BNAH ‡‡ 36.8 2.4 1.9 0.061 671 (396) §§ 4.72 773 (1,1) BNAH 26.9 2.8 0.9 0.038 315 0.083 745 1.64 544 (1,2) BNAH 8.4 7.5 1.0 0.030 353 0.089 755 2.33 645 (1,3) BNAH 5.3 8.1 0.7 0.017 234 0.082 733 2.82 676 (1,0)+(0,1) BNAH 5.3 1.9 0.5 − 316 0.087¶¶ 766¶¶ 1.02¶¶ 44¶¶

*A 4-mL CO2-saturated dimethylformamide–triethanolamine (DMF–TEOA) (4∶1 v∕v) solution containing reductant (0.1 M) and complexes was irradiated.The concentrations of all the photocatalysts were 0.3 mM for “ΦHCOOH” and “Products” and 0.05 mM for “TONHCOOH,” except for (2,1), of whichconcentrations were adjusted to half of the other complexes because (2,1) has two photosensitizer units. Therefore, the concentrations of thephotosensitizer units were same in all the photocatalytic systems in the series of the experiments.

†Irradiated in the Light Irradiation Condition (LIC) 1 for 5 h (light intensity: 4.9 × 10−8 einstein s−1).‡Irradiated in the LIC1 (light intensity: 4.9 × 10−8 einstein s−1).§Irradiated in the LIC2 for 20 h. TONHCOOHs were calculated as [produced HCOOH]/[added supramolecule].¶Emission quantum yield of the complexes in DMF–TEOA (4∶1 v∕v) (Excitation wavelength: 480 nm).∥Emission lifetime of the complexes in DMF–TEOA (4∶1 v∕v) (Excitation wavelength: 456 nm).**Quenching rate constant of emission from the complexes by the reducing agent.††Quenching fractions of emission from the Ru(II) complexes by 0.1 M of the reducing agent, calculated as 0.1kqτem∕ð1þ 0.1kqτemÞ.‡‡1-(4-methoxybenzyl)-1,4-dihydronicotinamide (25).§§Irradiated in the LIC3 for 20 h.¶¶Determined by emission from (1,0).

*It is reasonable that the TONHCOOH (59 for 14.5-h irradiation) in the case of the NMRmeasurement was much smaller than that obtained in the LIC2 (TON ¼ 315 for 20-hirradiation) because the concentration of the photocatalyst in the reaction solutionwas 10 times higher.

†The peak of H13COOH in DMF-d7 was observed at 163.2 ppm in 13C-NMR spectrum andhad shifted to 166.2 ppm after addition of TEOA (1.4M). Lehn et al. also reported that thepeak of H13COO− was observed at 167.4 ppm in a DMF-DMF-d7-TEOA (3∶1∶1 v∕v∕v)mixed solution (8).

15674 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1118336109 Tamaki et al.

Dow

nloa

ded

by g

uest

on

June

21,

202

0

(4∶1 v∕v) solution (1958 cm−1). This clearly indicates that (0,1)was converted to a complex with only one carbonyl ligand in thepresence of TEOA. A candidate is ½RuðdmbÞ2ðCOÞðCOOHÞ�þ,which could be produced by addition of OH− from contaminatedwater. However, this should not be the main complex in the re-action solution because the νCO of ½RuðdmbÞ2ðCOÞðCOOHÞ�þ inDMF-TEOA (1;955 cm−1) was similar to but clearly differentfrom that of (0,1) in the same solvent system. The structuraltransition of (0,1) in DMF-TEOA is attributable to the additionof deprotonated TEOA to the CO ligand giving½RuðdmbÞ2ðCOÞfCOOC2H4NðC2H4OHÞ2g�þ, (0,1)′, as shownin Eq. 2. The formation of (0,1)′ was also confirmed by anESI-MS spectrum of the reaction solution in which the parentpeak of (0,1)′ (m∕z ¼ 674) was observed as one of the mainsignals (Fig. S3). A similar conversion of one of the CO ligandsto −COOC2H4NðC2H4OHÞ2 should occur in all multinuclearcomplexes because similar IR spectral changes were observedin DMF-TEOA in all cases.

[2]

The UV–visible light (UV-vis) absorption spectra of (1,0),(0,1), and (0,1)′ are shown in Fig. 3. The metal-to-ligand-charge-transfer (MLCT) band of (0,1)′ was red-shifted comparedwith that of (0,1) because of the weak ligand field of the−COOC2H4NðC2H4OHÞ2 ligand compared with that of the COligand. However, because the MLCTabsorption band of (1,0) wasobserved at a much longer wavelength than that of (0,1)′, it can beconcluded that the irradiated lights absorbed only by the photo-sensitizer unit(s) in LIC1 (480 nm) and LIC2 (>500 nm).

The emission quantum yields of (1,0) and the multinuclearcomplexes observed in DMF-TEOA are summarized in Table 1.They were very similar to the emission quantum yield of (1,0)observed in a DMF solution (Φem ¼ 0.088). Therefore, in themultinuclear complexes, the intramolecular quenching of theexcited state of the photosensitizer unit by the catalyst unitwith the −COOC2H4NðC2H4OHÞ2 ligand(s) does not occur inDMF-TEOA (Scheme 1)‡.

Under the photocatalytic reaction conditions, BNAH shouldfunction as a reductant, but TEOA cannot, as suggested by theobservation that the emission from the 3MLCT excited state ofthe photosensitizer unit was mostly quenched by BNAH (Table 1),while quenching by TEOA was negligible (Scheme 1). However,TEOA should have other important roles in the photocatalyticreaction. It has been reported that TEOA can act as a base thatcaptures a proton from the one-electron oxidation product ofBNAH (i.e., BNAH ·þ) and suppresses the back-electron transferfrom the reduced photosensitizer unit to BNAH ·þ (process 1 inScheme 2) (9). Another potential role for TEOA as a baseinvolves the control of product distribution in CO2 reduction.Tanaka et al. have reported that in the case of the electrochemicalreduction of CO2 using ½RuðbpyÞ2ðCOÞ2�2þ as a catalyst, the pro-duct distribution changed depending on the pH of the reactionsolution: Formic acid was the main product under basic condi-tions, but CO and H2 were produced with higher yields underacidic conditions (5–7). Actually, in the case of the photocatalyticreaction using (1,1), the formation rate of formic acid was 4.7times lower in the absence of TEOA compared with that in thepresence of TEOA, and the formation selectivity of formic acidwas also lower in the absence of TEOA (Fig. S5A).

The reducing agent BNAH can potentially act either as atwo-electron or one-electron donor (20–22). The two-electronprocess gives BNAþ via an electrochemical, chemical, and elec-trochemical (ECE) sequence (process 2 in Scheme 2), while theone-electron process gives 4,4′- and 4,6′-BNA2 via the coupling

Fig. 1. Photocatalytic formation of formic acid (red filled circle), CO (greenfilled square), and H2 (orange filled triangle) as a function of irradiation time.The (red dotted line) shows the amount of formic acid generated by photo-catalysis, under the assumption that the side reaction effect by BNA2s and theeffect of the decrease of BNAH can be neglected: A CO2 saturated DMF-TEOA(4∶1 v∕v, 4 mL) solution containing BNAH (0.1 M) and (1,1) (0.05 mM) wasirradiated by >500-nm light.

Fig. 2. IR spectra of (0,1) in DMF (blue line) and in DMF-TEOA (4∶1 v∕v)(red line).

Fig. 3. UV-vis absorption spectra of (1,0) (black line) and (0,1) (red line),which was converted to (0,1)′ (see the main text), in DMF-TEOA (4∶1 v∕v)and (0,1) in DMF (blue line).

‡The intramolecular quenching of emission from (1,1) was observed in a DMF solutionwithout TEOA as shown in Fig. S4A. This is probably due to electron transfer from theexcited state of the photosensitizer unit to the catalyst unit because the reductionpotential of (0,1) (−1.52 V vs. Ag∕AgNO3) was more positive than that of (0,1)′(−1.92 V) (Fig. S4B).

Tamaki et al. PNAS ∣ September 25, 2012 ∣ vol. 109 ∣ no. 39 ∣ 15675

CHEM

ISTR

YSP

ECIALFEAT

URE

Dow

nloa

ded

by g

uest

on

June

21,

202

0

reaction of BNA · as oxidation products (process 3 in Scheme 2).To clarify the role of BNAH, we used high performance liquidchromatography to quantify both the decrease of BNAH andthe formation of its oxidation product(s) in the photocatalyticreaction solutions (LIC2: Fig. 4). For example, 20-h irradiationinduced the consumption of 141 μmol of BNAH, and the produc-tion of 63 μmol of formic acid, 4 μmol of H2, and 3 μmol of CO.This reaction solution also contained 29 μmol of 4,4′-BNA2

§ (23),and 25 μmol and 14 μmol of the diastereomers of 4,6′-BNA2. Onthe other hand, BNAþ was not detected in the reaction solution.The formation of all reduction products (HCOOH, H2, and CO)requires two electrons per molecule, and the formation of oneBNA2 molecule from two BNAH molecules donates two elec-trons. Thus, we conclude that BNAH acted only as a one-electrondonor. The electron balance in the photocatalytic reduction ofCO2 to formic acid using (1,1) is shown in Eq. 3, below.

[3]

Although BNA2s are recognized as good electron donors (24),they did not act as reducing agents in this photocatalytic reactionand were the final oxidation products of BNAH. Back electrontransfer is expected to occur too fast for the one-electron oxidizedproducts of BNA2s (BNA2 ·þ) to be irreversibly decomposed[Eq. 4]. Since the oxidation potentials of BNA2s [for example,E ∘

oxð4; 4 0-BNA2Þ ¼ 0.26 V vs. SCE] (24) are more negative thanthat of BNAH [E∘

oxðBNAHÞ ¼ 0.57 V vs. SCE] (25), it is possi-ble that the accumulation of BNA2s obstructs the photochemicalelectron transfer from BNAH to the excited state of (1,1)[kqðBNAHÞ ¼ 1.64 × 107 M−1 s−1] in view of the much fasterquenching rate constant of the emission from the excited (1,1)by BNA2s: kqð4; 4 0-BNA2Þ ¼ 3.09 × 108 M−1 s−1. For example,after 8-h irradiation, the reaction solution contained 291 μmol ofBNAH and 51 μmol of BNA2s where only 18% of the excitedstate of (1,1) was quenched by BNAH, whereas 54% could bequenched in the first stage of the photocatalytic reaction. Webelieve that this is the main reason why the photocatalytic effi-ciency decreased after the 8-h irradiation. If the effects of theside reaction and the decrease of BNAH were prevented, theformation of HCOOH would increase by 1.85-fold after 8-hirradiation (red dotted line in Fig. 1). This increased rate shouldbe approximately the same as that in the first stage of the photo-catalytic reaction.

[4]

The photocatalytic ability of (1,1) was much greater than thatof a mixed system of the corresponding mononuclear modelcomplexes (ð1; 0Þ þ ð0; 1Þ; 1∶1) (Fig. 5). The irradiation of thereaction solution containing 0.3 mM of (1,1) by 480-nm mono-chromatic light (LIC1: light intensity: 4.9 × 10−8 einstein s−1)gave formic acid with ΦHCOOH ¼ 0.038. On the other hand, the

irradiation of a solution containing (1,0) and (0,1) (0.3 mM each)instead of (1,1) gave a much smaller amount of formic acid withan induction period in the initial stage, but formation of formicacid ceased after irradiation for only 1 h. Furthermore, the irra-diation of the mixed system with lower light intensity (LIC1:4.9 × 10−9 einstein s−1) produced no formic acid at all, whereas(1,1) still functioned as a photocatalyst with ΦHCOOH ¼ 0.045(Fig. S6). It should be emphasized that ΦHCOOH was similar evenwhen using a one-order difference in light intensity. During thephotocatalytic reaction of the mixed system of (1,0) and (0,1), thesolution changed color from red to black: This color change wasmore significant at lower light intensity after the same quantity ofphotons was introduced into the reaction solution (Fig. 6A).However, no such color change was observed when (1,1) was usedas a photocatalyst (Fig. 6B).

The ratio between the photosensitizer unit(s) and the catalystunit(s) in the supramolecular system strongly influenced the out-come of photocatalysis. A higher ratio of the photosensitizer unitto the catalyst unit generated a higher yield of formic acid (LIC1:Fig. 5). In the cases in which the reaction solutions had the sameabsorbance at 480 nm [this light can be absorbed only by thephotosensitizer unit(s)], the results of the photocatalytic reac-tions are summarized in Table 1. As an example, (2,1) has twophotosensitizer units and one catalyst unit and generated formicacid with the highest selectivity and yield. In contrast, using (1,3)with one photosensitizer unit and three catalyst units, CO was the

Scheme 1. Initial process of the photocatalytic reaction using (1,1).

Scheme 2. Oxidation processes of BNAH.

Fig. 4. Photocatalytic production of the reduction and oxidation products—i.e., HCOOHþ COþ H2 (red filled circle) and BNA2s (blue filled diamond)—and consumption of BNAH (black filled circle): a CO2 saturated DMF-TEOA(4∶1 v∕v, 4 mL) solutions containing BNAH (0.1 M) and (1,1) (0.05 mM)was irradiated at >500-nm light.§It has been known that the yield of another isomer of 4,4′-BNA2 is very low.

15676 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1118336109 Tamaki et al.

Dow

nloa

ded

by g

uest

on

June

21,

202

0

major product, and the photocatalytic activity was much lowerwith this photocatalyst than with (2,1), even though the reactionsolution of (2,1) contained only 1/6 the number of catalyst unitspresent in the (1,3) solution. All complexes produced only a smallamount of H2. The photocatalytic activities of (1,2) and (1,3) de-creased rapidly (Fig. 5), and the color of the solution changedfrom red to black. For CO formation with (1,2) and (1,3), therewas an induction period, and the production of CO acceleratedafter the color change (Fig. S5B). However, no such phenomenonwas observed with (1,1) and (2,1). These results suggest that theactual photocatalyst for CO formation with (1,2) and (1,3) isthe black product(s) produced during the photocatalytic reaction.Table 1 summarizes the emission quantum yields of (1,0) andthe supramolecules using 480-nm light for excitation. Althoughthe emission yields from the supramolecules were slightly smallerthan that from (1,0), there was no correlation between the emis-sion yield and the photocatalytic ability. Table 1 also shows thefractions of the emission quenched by 0.1 M BNAH (ηq). Clearly(1,1) and (2,1) are better photocatalysts, although emission from(1,2) and (1,3) was more efficiently quenched by BNAH than thatfrom (1,1) and (2,1).

The photocatalytic activities of (1,2), (1,3), and the mixedsystem of (1,0) and (0,1) deteriorated rapidly and the reactionsolution turned black even in the early stages of irradiation, asdescribed above. As a typical example, Fig. 6A shows the changesin the UV-vis absorption spectrum of the mixed system duringirradiation (LIC1). Tanaka et al. observed the precipitation ofblack solids during the photocatalytic reaction of the mixed sys-tem of ½RuðbpyÞ3�2þ and ½RuðbpyÞ2ðCOÞ2�2þ; they suggested thatthe precipitation causes the deactivation of the photocatalyst (6).Deronzier and Ziessel et al. reported that the electrochemicalreduction of ½RuðbpyÞ2ðCOÞ2�2þ caused the dissociation of thebpy ligand to give a black polymer with ruthenium-rutheniumbonds (i.e., ½RuðbpyÞðCOÞ2�n, Eq. 5) (26). As the absorptionspectrum of this polymer was similar to that in Fig. 6A, the de-activation of the photocatalysts ð1; 0Þ þ ð0; 1Þ, (1,2), and (1,3) ispresumably due to the formation of similar oligomers and/orpolymers from the reduced catalyst or catalyst unit produced dur-ing irradiation¶. Lower photocatalytic abilities of ð1; 0Þ þ ð0; 1Þ,(1,2), and (1,3) in the LIC1 were observed compared with thosein the LIC2 (“products” vs. “TONHCOOH” in Table 1). As de-

scribed above, the formation of the oligomers and/or polymerswas enhanced with the lower light intensity. Therefore, this de-activation process should be accelerated with either higher con-centration of the complex(es) in solution or lower irradiated lightintensity. It should be emphasized that (2,1) showed the highestphotocatalytic ability in any photocatalytic reaction condition.

[5]

In the cases of (1,2) and (1,3), an additional absorption bandwas observed with a maximum at 600 nm during the irradiation(LIC1: Fig. S8B). This can be attributed to the reduced catalystunit, because this absorption alone disappeared rapidly on intro-ducing air into the solution. The remaining spectra after the in-troduction of air were very similar to that of the black polymer(Fig. 6A). The formation of formic acid involves a two-electronreduction, which could occur by the twofold reductive quenchingof the excited photosensitizer unit by BNAH. Probably, the largernumber of catalyst units in the supramolecule gives rise to electrondonation from the one-electron reduced species of the photosen-sitizer unit to the different catalyst units. This should increase thelifetime of the reduced catalyst unit and trigger the oligomerizationof the catalyst units. This is a major reason why (1,2) and (1,3) havelow photocatalytic ability. Solutions of (1,1) and (2,1) did not un-dergo any similar color change during irradiation (Fig. 6B andFig. S8A). An increase in the number of photosensitizer units inthe supramolecule should increase the likelihood of injection ofa second electron into the catalyst unit, thereby suppressing theprocess of deactivation of the catalyst unit. We checked thedependence of the photocatalytic efficiency on the concentrationof (2,1) (0.02–0.15 mM) in the LIC1 (light intensity: 4.9 ×10−8 einstein s−1) as shown in Fig. S9A. This clearly shows thatthe efficiency is not affected by the concentration of (2,1) at all.

Only 59% of the emission from the photosensitizer unit of (2,1)was quenched by BNAH even in the first stage of the photocata-lytic reaction (Table 1, entry 1). If this process of reductive quench-ing of the excited state of (2,1) can be improved, the efficiency andstability of the photocatalyst should also be improved. For this pur-pose, the stronger reducing agent 1-(4-methoxybenzyl)-1,4-dihy-dronicotinamide (MeO-BNAH)∥, having the redox potential0.50 V vs. SCE (25), was used as a sacrificial electron donor insteadof BNAH [redox potential ¼ 0.57 V (25) and ηq ¼ 59%]. As aresult, ηq increased to 77% (Table 1, entry 2: Fig. S1B), the photo-catalytic ability of (2,1) was substantially improved (i.e.,ΦHCOOH ¼

Fig. 5. Photocatalytic formation of formic acid as a function of irradiationtime: CO2 saturated DMF-TEOA (4∶1 v∕v, 4 mL) solutions containing BNAH(0.1 M) and (2,1) (0.15 mM: purple filled circle), (1,1) (0.3 mM: red filled circle),(1,2) (0.3 mM: blue filled circle), (1,3) (0.3 mM: orange filled circle), or a mix-ture of (1,0) and (0,1) (0.3 mM each: black filled circle), were irradiated by480-nm light of intensity 4.9 × 10−8 einstein s−1.

Fig. 6. UV-vis absorption spectral changes of solutions containing the mix-ture of (1,0) and (0,1) (0.3 mM each: A) or (1,1) (0.3 mM: B) during irradiation(0–10 h at 1-h intervals): CO2 saturated DMF-TEOA (4∶1 v∕v, 4 mL) solutionscontaining BNAH (0.1 M) and the complex(es) were irradiated by 480-nmlight of intensity 4.9 × 10−9 einstein s−1.

¶A unique IR absorption band attributable to νCO of the oligomers was observed at2;029 cm−1 in the case of (1,3) (Fig. S7). ∥The UV-vis absorption spectra of MeO-BNAH, BNAH, and 4,4′-BNA2 are shown in Fig. S1A.

Tamaki et al. PNAS ∣ September 25, 2012 ∣ vol. 109 ∣ no. 39 ∣ 15677

CHEM

ISTR

YSP

ECIALFEAT

URE

Dow

nloa

ded

by g

uest

on

June

21,

202

0

0.061 in the LIC1 and TONHCOOH ¼ 671 in the LIC2) comparedwith that using BNAH (i.e., ΦHCOOH ¼ 0.041 and TONHCOOH ¼562). The TOF of formic acid formation (TOFHCOOH) reached11.6 min−1 in the LIC3 (7.8 min−1 in the case using BNAH)**.To the best of our knowledge, this system affords the fastest photo-catalytic reduction of CO2 reported to date.

It is noteworthy that the formation rate of formic acid using(2,1) as a photocatalyst was strongly dependent on the irradiatedlight flux as shown in Fig. S9B. Even in the reaction condition withthe highest light flux (LIC3), TOF was still increasing, and, there-fore, the rate-limiting process of the photocatalytic CO2 reductionusing (2,1) was the excitation of the photosensitizer unit; i.e., higherTOFHCOOH should be achievable by using a stronger light source.

Finally, the reaction mechanism of formation of formic acidusing the good photocatalysts; i.e., (1,1) and (2,1), is discussed.Formation of formic acid requires two-electron reduction ofCO2. It is clear that the initial one-electron transfer to the catalystunit proceeds from the one-electron-reduced species (OER) ofthe photosensitizer unit, which is produced via the reductivequenching process of the excited photosensitizer unit by BNAH.The second-electron reduction of the catalyst unit, which is prob-ably converted to the CO2-adduct form via chemical processesafter the first reduction, should also proceed through the samesequence—i.e., photochemical formation of the OER of thephotosensitizer unit and then its donation of another electron tothe catalyst unit—because the reductant BNAH works only as aone-electron donor. Therefore, the maximum quantum yield offormic acid formation (ΦHCOOH) should be 0.5 using the photo-catalytic systems (27). Since the quantum yield of formation offormic acid was not dependent on both the light intensity andthe concentration of the photocatalyst, the sequential two-stepphotochemical electron transfer processes from the OER of thephotosensitizer unit to the catalyst unit should proceed intramo-lecularly; namely, the photocatalytic formation of formic acidproceeds with participation of only one supramolecule and does

not require participation of two molecules of the photocatalyst.The disproportionation of one-electron reduced supramoleculesis not the main process for the formation of formic acid.

Materials and MethodsDetails of synthesis of ruthenium(II) complexes, general experimental condi-tions, and photocatalytic reaction conditions, and analysis of formic acid areprovided in SI Text.

ConclusionWe successfully developed unique supramolecular photocatalystsconstructed with various ratios of two kinds of ruthenium(II)complexes for reduction of CO2 to formic acid with high selec-tivity and durability. The ratio between photosensitizer units andcatalyst units strongly affected the photocatalytic activities—i.e.,the higher ratio of the photosensitizer unit caused a higheryield of formic acid—and (2,1) exhibited the highest photocata-lytic ability (ΦHCOOH ¼ 0.061, TONHCOOH ¼ 671, TOFHCOOH ¼11.6 min−1). The following sequence takes place twice for the for-mation of one formic acid molecule in the photocatalytic reaction:First, the excitation of the photosensitizer unit; second, the reduc-tive quenching of the excited state of the photosensitizer unit byBNAH; and third, the intramolecular electron transfer from the re-duced photosensitizer unit to the catalyst unit. The addition of CO2

into the catalytic site probably occurs between the two sequences.

ACKNOWLEDGMENTS. This work was partially supported by Japan Scienceand Technology Agency (Research Seeds Quest Program). Y.T. thanks theJapan Society for the Promotion of Science (JSPS) for a Research Fellowshipfor Young Scientists.

1. Enthaler S (2008) Carbon dioxide—The hydrogen-storage material of the future?ChemSusChem 1:801–804.

2. Joó F (2008) Breakthroughs in hydrogen storage—Formic acid as a sustainable storagematerial for hydrogen. ChemSusChem 1:805–808.

3. Matsuoka S, et al. (1992) Photocatalysis of ollgo(p-phenylenes): Photochemical reduc-tion of carbon dloxlde with trlethylamine. J Phys Chem 96:4437–4442.

4. Ogata T, et al. (1995) Phenazine-photosensitized reduction of CO2 mediated by a cobalt-cyclam complex through electron and hydrogen transfer. J Phys Chem 99:11916–11922.

5. Ishida H, Tanaka K, Tanaka T (1987) Electrochemical CO2 reduction catalyzed by½RuðbpyÞ2ðCOÞ2 �2þ and ½RuðbpyÞ2ðCOÞCI�þ : The effect of pH on the formation ofCO and HCOO−. Organometallics 6:181–186.

6. Ishida H, Terada T, Tanaka K, Tanaka T (1990) Photochemical CO2 reduction catalyzedby ½RuðbpyÞ2ðCOÞ2�2þ using triethanolamine and 1-benzyl-1,4-dihydronicotinamide asan electron donor. Inorg Chem 29:905–911.

7. Ishida H, et al. (1990) Ligand effects of ruthenium 2,2′-bipyridine and 1 ,10-phenan-throline complexes on the electrochemical reduction of CO2. J Chem Soc Dalton Trans2155–2160.

8. Lehn J-M, Ziessel R (1990) Photochemical reduction of carbon dioxide to formate cat-alyzed by 2,2′-bipyridine- or 1,10-phenanthroline-ruthenium(II) complexes. J Organo-met Chem 382:157–173.

9. Gholamkhass B, et al. (2005) Architecture of supramolecular metal complexes forphotocatalytic CO2 reduction: ruthenium-rhenium bi- and tetranuclear complexes.Inorg Chem 44:2326–2336.

10. Sato S, Koike K, Inoue H, Ishitani O (2007) Highly efficient supramolecular photoca-talysts for CO2 reduction using visible light. Photochem Photobiol Sci 6:454–461.

11. Koike K, Naito S, Sato S, Tamaki Y, Ishitani O (2009) Architecture of supramolecularmetal complexes for photocatalytic CO2 reductionIII: Effects of length of alkyl chainconnecting photosensitizer to catalyst. J Photochem Photobiol A 207:109–114.

12. Tamaki Y, et al. (2012) Development of highly efficient supramolecular CO2 reductionphotocatalysts with high turnover frequency anddurability. Faraday Discuss 155:115–127.

13. Kimura E, Bu X, Shionoya M, Wada S, Maruyama S (1992) A new nickel(II) cyclam(cyclam ¼ 1,4,8,11-tetraazacyclotetradecane) complex covalently attached toRuðphenÞ3 2þ (phen ¼ 1,10-phenanthroline): A new candidate for the catalytic photo-reduction of carbon dioxide. Inorg Chem 31:4542–4546.

14. Kimura E, Wada S, Shionoya M, Okazaki Y (1994) New series of multifunctionalizednickel(II)-cyclam (cyclam ¼ 1,4,8,11-tetraazacyclotetradecane) complexes: Applicationto the photoreduction of carbon dioxide. Inorg Chem 33:770–778.

15. Komatsuzaki N, Himeda Y, Hirose T, Sugihara H, Kasuga K (1999) Synthesis and photo-chemical properties of ruthenium-cobalt and ruthenium-nickel dinuclear complexes.Bull. Chem. Soc Jpn 72:725–731.

16. Rau S, Walther D, Vos JG (2007) Inspired by nature: Light-driven organometalliccatalysis by heterooligonuclear Ru(ii) complexes. Dalton Trans 915–919.

17. Ozawa H, Haga M-A, Sakai K (2006) A photo-hydrogen-evolving molecular devicedriving visible-light-induced EDTA-reduction of water into molecular hydrogen. JAm Chem Soc 128:4926–4927.

18. Rau S, et al. (2006) A supramolecular photocatalyst for the production of hydrogenand the selective hydrogenation of tolane. Angew Chem Int Ed 45:6215–6218.

19. Arachchige SM, et al. (2009) Design considerations for a system for photocatalytichydrogen production from water employing mixed-metal photochemical moleculardevices for photoinitiated electron collection. Inorg Chem 48:1989–2000.

20. Martens FM, Verhoeven JM (1981) Photo-induced electron transfer from NADH andother 1,4-dihydronicotinamides tomethyl viologen.Recl Trav Chim Pays Bas 100:228–236.

21. Pac C, et al. (1984) Redox-photosensitized reactions. 11: RuðbpyÞ3 2þ-photosensitizedreactions of 1-benzyl-1,4-dihydronicotinamide with aryl-substituted enones, deriva-tives of methyl cinnamate, and substituted cinnamonitriles—electron-transfer me-chanism and structure reactivity relationships. J Org Chem 49:26–34.

22. Ishitani O, Yanagida S, Takamuku S, Pac C (1987) Redox-photosensitized reactions. 13.RuðbpyÞ3 2þ-Photosensitized reactions of an NADH model, 1-benzyl-1,4-dihydronicoti-namide, with aromatic carbonyl-compounds and comparison with thermal-reactions.J Org Chem 52:2790–2796.

23. Kano K, Matsuo T (1976) Photoinduced one-electron reduction of 1-benzyl-3-carba-moylpyridinium chloride and 3,5-bis(ethoxycarbonyl)-2,6-dimethylpyridine. Bull ChemSoc Jpn 49:3269–3273.

24. Patz M, Kuwahara Y, Suenobu T, Fukuzumi S (1997) Oxidation mechanism of NADdimer model compounds. Chem Lett 6:567–568.

25. Fukuzumi S, Koumitsu S, Hironaka K, Tanaka T (1987) Energetic comparison betweenphotoinduced electron-transfer reactions from NADH model compounds to organicand inorganic oxidants and hydride-transfer reactions from NADH model compoundsto para-benzoquinone derivatives. J Am Chem Soc 109:305–316.

26. Chardon-Noblat S, Collomb-Dunand-Sauthier M-N, Deronzier A, Ziessel R, Zsoldos D(1994) Formation of polymeric [fRu0ðbpyÞðCOÞ2gn] films by electrochemical reductionof ½RuðbpyÞ2ðCOÞ2 �ðPF6Þ2 : Its implication in CO2 electrocatalytic reduction. Inorg Chem33:4410–4412.

27. Yui T, Tamaki Y, Sekizawa K, Ishitani O (2011) Photocatalytic reduction of CO2 : Frommolecules to semiconductors. Photocatalysis, Topics in Current Chemistry. Vol 303, edCA Bignozzi (Springer Berlin/Heidelberg), pp 151–184.

**Since the corresponding one-electron-oxidized and dimerized products—i.e.,ðMeO-BNAÞ2—was detected by ESI-Mass spectrum of the irradiated reaction solution(m∕z ¼ 488), the photocatalytic reaction using MeO-BNAH probably proceeded via asimilar mechanism to that using BNAH.

15678 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1118336109 Tamaki et al.

Dow

nloa

ded

by g

uest

on

June

21,

202

0