Experiments with the titanium dioxide-ruthenium tris-bipyridine-nickel cyclam system for the photocatalytic reduction of CO₂

1 Introduction

The increased production of CO₂ is one of the most significant anthropic effects on the environment and causes the notorious greenhouse effect. This has led to extensive investigation of physical and chemical processes that may contribute to the capture, reutilization and valorization of this gas, certainly a key issue of “green chemistry” [1–4]. Potentially, the reduction of CO₂ is a source of CO [5–7] and thus may offer a feedstock for various synthetic processes, such as the Fischer-Tropsch reaction, the Monsanto, and the Cativa protocols for the synthesis of acetic acid and the ICI process for the production of methanol. An approach is based on the electrochemical reduction of CO₂ on metallic cathodes. Although this method requires a very negative potential, the use of a metal complex catalyst such as [Ni(cyclam)]²⁺ [8] and its derivatives [9] was shown to allow the conversion of CO₂ into CO at a reasonable voltage [approximately -1.25 vs. saturated calomel electrode (SCE)]. However, some limitations remain, since at negative potentials the use of mercury electrodes is essential to optimize the selectivity of the process, by suppressing or, at least, minimizing, the competitive reduction of proton to hydrogen occurring at the electrode (recently, the use of a glassy carbon electrode has been proposed [10]). The latter process is obviously pH dependent and occurs significantly at low pH (below pH 4) [8]. Photochemistry offers an alternative approach, as first reported by Hawecker et al. [11]

A key advancement in this direction was obtained by the Calvin group in the 1990s [12, 13], with the demonstration that both CO and H₂ were generated by visible light irradiation (1000 W lamp) of a CO₂ saturated aqueous solution containing tris-(2,2′-bipyridyl)ruthenium(II) chloride as the photosensitizer, ascorbic acid as the electron donor, and 1,4,8,11-tetraazacyclotetradecane nickel(II) chloride as the catalyst. A related system was reported by Mochizuki shortly afterwards [14] and further studies have been carried out in the frame of homogeneous photocatalyzed reduction of CO₂ [15, 16]. Of the two metal complexes used in those experiments, the convenient properties of ruthenium derivatives as a visible light absorbing photosensitizer, taking part in redox processes, are well known [17]. These and related complexes have been increasingly applied in different fields, ranging from solar energy conversion by producing hydrogen via hydrolysis of water [18, 19], to organic synthesis under mild (“green”) conditions [20, 21]. It should be noted that reduction of CO₂ to formate was the main process observed if the nickel catalyst was omitted and only the ruthenium complex was used [11]. Closely related systems have been exploited also for photoreduction processes on substrates different from CO₂, such as the reductive dehalogenation of organic halides [22].

In essence, this bimolecular strategy involves the use of a photoactive complex used for light harvesting and of another molecule acting as a catalyst. A more elaborated version involved hybrid systems, where the Ru and the Ni ligands were covalently linked [23], which resulted in some improvement, such as [CO]/[H₂] ratios higher than those observed using separate units. In a more complex embodiment of this principle, Woolerton et al. recently reported an efficient device for the photoreduction of CO₂ to CO, where the light harvesting moiety [a Ru(bpy) derivative] and the catalytic unit [an enzyme with a (Ni₄Fe-4S) active site] were wired by TiO₂ nanoparticles [24].

More generally, a semiconductor such as TiO₂ may act as the catalyst upon either high-energy irradiation (UV) for band gap excitation [25, 26], or visible light irradiation sensitization by using a dye capable of injecting electrons into the conduction band [27]. In particular, surface functionalization of semiconductor nanoparticles has been applied for dye-sensitized solar cells and photocatalysis, and leads to interfacial electron transfer into the conduction band of the semiconductor. In a Grätzel cell, for example, TiO₂ is functionalized with Ru polypyridyl complexes attached via carboxylate substituents [28]. Phosphonate [29, 30] and hydroxamate [31] groups have likewise been used for this aim. We thus deemed it appropriate to test the application of such semiconductor-linked photosensitizers to the reduction of carbon dioxide.

2 Experimental section

Nano-to-microsecond transient absorption experiments were performed using a nanosecond LP-921 (Edinburgh Instruments, Livingston, UK) laser flash photolysis apparatus. Excitation pulses at 355 nm (20 ns, 1 mJ) were provided by a 20 Hz Nd:YAG laser (Lumonics, Cologno Monzese, Italy). The probe light was provided by a Xe flash lamp (XBO 150 W/CR OFR). Samples were placed in a quartz cell (10×10 mm² section) at a concentration adjusted to obtain an optical density (OD) value of 0.5 at 355 nm. Ion chromatography analyses were performed by means of a Dionex GP40 instrument equipped with a conductimetric detector (Dionex 20 CD20, Thermo Scientific, Rodano, Italy) and an electrochemical suppressor (ASRS Ultra II, 4 mm), by using the following conditions: chromatographic column IONPAC AS23 (4 mm×250 mm), guard column IONPAC AG12 (4 mm×50 mm), eluent: NaHCO₃ 0.8 mm+Na₂CO₃ 4.5 mm, flux: 1 ml/min; current imposed at detector: 50 mA. The formation of CO and H₂ was detected by gas chromatography thermal conductivity detector analyses and their amounts determined by means of a calibration curve.

Compounds Ru(bpy)₃Cl₂ (1a), sodium ascorbate and triethanolamine (TEOA) are commercially available and were used as received, whereas 1d (obtained as the chloride salt) [32] and 2 (as the triflate salt) [33] were synthesized by a known procedure.

2.2 Synthesis of Ru(II)(bpy)₂(4,4′-(PO₃Et₂)₂bpy) complex (1b)

4,4′-Diethyl phosphonato-2,2′-bipyridine (Scheme 1) was obtained via a multistep synthesis according to literature procedures [34–39], followed by complexation with Ru(bpy)₂Cl₂, in order to get the desired complex 2b. In a typical reaction, Ru(bpy)₂Cl₂ (40 mg, 0.082 mmol) was treated with silver nitrate (28 mg, 0.164 mmol) in methanol for 3 h at room temperature. The solution was filtered in order to remove silver salts and the filtrate was added to a round bottom flask containing 4,4′-diethyl phosphonato-2,2′-bipyridine (31 mg, 0.082 mmol). The solution was refluxed for 3 h in the dark under an argon atmosphere. At this time, the reaction mixture was allowed to go to room temperature and the solvent was evaporated under reduced pressure. The obtained solid was dissolved in a minimum amount of MeOH and precipitated as a PF₆ salt, by adding a saturated aqueous solution of ammonium hexafluorophosphate. The flask was kept at -20°C overnight to obtain complete precipitation and filtered. After repeated washings with water and a copious quantity of diethyl ether, the obtained product was kept under vacuum overnight to yield a reddish brown solid (79 mg, 84%).

Scheme 1

Photocatalytic reduction of carbon dioxide.

Scheme 2

Complexes investigated in this paper.

Spectroscopic data are in accordance with the literature [40].

ESI-MS m/z=987.1317 [M+PF₆]⁺ Calcd. for C₃₈H₄₂F₆N₆O₆P₃Ru 987.1321.

3 Results

As a first step in this direction, we report an investigation into the photosensitized CO₂ reduction in the presence of different functionalized Ru(bpy)₃ complexes (1a, 1d) and Ni(cyclam) (2). In the first part of the study the effect of introducing functional groups, such as acids, that are suited for connecting the ligands used for the light harvesting unit to a support was tested. Thus, besides parent 1a, three Ru complexes were chosen, viz. (bpy)₂Ru(bpy(PO(OEt)₂)(2PF₆), (1b), (bpy)₂Ru(bpy(COOEt)₂)(2PF₆) (1c) and (bpy)₂Ru(bpy(COOH)₂)(2Cl) (1d) (see Scheme 2). The key photophysical parameters, as well as the redox behavior, were determined and are shown in Table 1. The substituted complexes 1b–1d exhibited a significant emission, although with a Φ_em value lower than that of parent 1a. Moreover, the presence of an electron withdrawing substituent on the bipyridyl ring shifted λ_em to lower energy values with respect to the unsubstituted 1a. A dependence on the proticity of solvent was observed, with the emission lifetime significantly shortened in water (down to 185 ns for 1c).

With regards to the electrochemical properties, four polarographic waves were detected for the ruthenium complexes in MeCN. As shown in Table 1, the Ru(III)/Ru(II) oxidation potential increased only slightly (by up to 0.1 V) when an electron-withdrawing substituent was introduced in the bipyridine ligand. By contrast, the less negative reduction potential (E_RED1) in Table 1 dropped by up to 0.6 V, with respect to 1a, when either a carboxylic or a phosphonic ester substituent was introduced. The Weller equation [E_red*=E_red-E_oo-e²/εd] allowed estimation of the redox potential in the excited state of the complex (E_red*) from that in the ground state and the excited state energy (E_oo) [43]. E*[Ru(II)/Ru(I)] was found to range from 0.84 to 1.45 V.

The photochemical study was carried out either in DMF or in water, depending on the solubility of the components considered, with sodium ascorbate as the sacrificial donor. Nanosecond laser flash photolysis (see experimental section) was used to characterize the intermediates. Thus, flashing an Ru(bpy)₃²⁺ solution (5×10^-5m) in DMF in the presence of ascorbate (0.01 m), caused the formation of the well-known spectrum of the reduced Ru(I) complex with λ_max at 510 nm [44]. This transient absorption was quenched by Ni(cyclam)²⁺ at a rate 3×10⁷ mol^-1 sec^-1 (see Figure 1). Analogous results were observed when changing the solvent from DMF to water (not reported).

Figure 1

Transient spectrum of a 5×10^-5m Ru(bpy)₃²⁺ solution in DMF in the presence of 0.01 m sodium ascorbate 81 μs after the flash. Inset: Profile of the absorbance change at 510 nm (upper curve, measured τ=31 μs); with the further addition of 5×10^-3m Ni(cyclam)²⁺ (2) (lower curve, τ=7 μs).

The photoinduced electron transfer process took place analogously with the complexes 1b–1d. However, the introduction of electron-withdrawing groups led to much longer-lived Ru(I) complexes, with a lifetime ranging from 93 μs of the dicarboxy complex (1d,Figure 2B, Table 2) to 2.3 ms of the corresponding ethoxycarbonyl complex (1c). By contrast, quenching of these complexes by Ni(cyclam)²⁺ was less effective, approximately 2×10⁷ mol^-1 sec^-1 (an example in Figure 2B for complex 1d).

In the second part of the study, the efficiency of carbon dioxide reduction in the presence of Ru and Ni complexes by irradiation at 450 nm was tested. In order to avoid the formation of oxalate that disturbed analysis, triethanolamine was used in the place of ascorbate ion as the donor and DMF as the solvent. The results obtained are shown in Table 2.

Figure 2

Transient spectrum of: (A) a 1b solution (5×10^-5) in DMF in the presence of 0.01 m sodium ascorbate. Inset: profile of the absorbance change at 510 nm (upper curve, measured τ=960 μs) and in the presence (lower curve) of Ni(cyclam) (2) 5×10^-3 m (τ=92 μs). (B) An analogous experiment with 1d (τ=93 μs upper curve and τ=80 μs in the presence of 2, lower curve).

The evolution of hydrogen and carbon monoxide, as well as the generation of formic acid, were monitored. As shown in Table 1, under our conditions, photocatalysis by Ru(bpy)₃²⁺(1a)caused the generation of formic acid as the main process from CO₂, as well as of some CO, while a little hydrogen was formed after a lag time. However, the reaction soon ceased. Conversely, the irradiation of a Ni(cyclam)²⁺ solution caused no reaction on this time scale. When the Ni complex was added in equimolar amount to the Ru(bpy)₃²⁺ solution, only a slight improvement of the yield of all of the photoproducts was observed (entry 5, 64 h irradiation). However, when present at a higher concentration [10 times that of Ru(bipy)₃], Ni(cyclam)²⁺ allowed the process to continue up to 120 h, with a [CO]/[H₂] value of 0.72 (entry 9). The evolution of the UV-visible spectrum was likewise monitored. Thus, in the absence of the Ni complex, the Ru(II) spectrum progressively bleached, probably undergoing photosolvolysis, with substitution of a bipyridyl ligand [12, 13] (judging from the spectrum in Figure 3), through two subsequent steps. By contrast, adding 5×10^-3m2 caused a smaller shift and a much reduced decrease, demonstrating that 1a was partially or totally preserved.

Figure 3

Absorbance change by irradiation of a 5×10^-3m solution of 1a in DMF in the presence of 0.2 m triethanolamine (TEOA): (A) in the absence; and (B) in the presence of 2 (5×10^-3m). Notice that the photolyzed solution was diluted by a factor ten.

As for the substituted ruthenium complexes, phosphonate 1b gave formate and hydrogen in the same yield as 1a; 1c and 1d produced a lower yield. CO was not formed in any case.

Building a heterogeneous system was then attempted. The carboxyl derivative 1d was conjugated to the semiconductor under two different set of conditions. In the first case, covalent binding to titania nanoparticles was obtained by stirring in the presence of dicyclohexylcarbodiimide and N,N-dimethylaminopyridine as activating reagents. The brick red particles obtained were analyzed by inductively coupled plasma optical emission spectroscopy and were shown to contain 2.2 mg Ru per gram of catalyst (Cat1). In the second method, the titania nanoparticles were activated by treatment with Piranha solution, washed, suspended in a 1M HCl solution, irradiated at 254 nm, filtered, and stirred with an ethanolic solution of 1c, affording particles (Cat2) containing 3.3 mg Ru per g of titania. In a blank experiment, the second treatment was applied to 1a, not containing reactive groups, giving Ru-free nanoparticles (Cat3). DMF was chosen again in this case as the medium, in order to avoid the instability of carboxylate anchoring groups in protic solvents, such as water [45].

Irradiation of suspensions (2 mg of catalysts in 5 ml) of these materials, under the same conditions as used for homogeneous solutions, caused the reduction of CO₂ to formate with all of the examined nanoparticles, including the blank experiment with Cat3, but CO was formed, in a modest yield, only in the case of Cat2 (see Table 3).

4 Discussion

The processes which occurred are summarily indicated in Scheme 1. The introduction of electron-withdrawing substituents in ruthenium tris-bipyridine complexes 1b–1d, reported in Table 1, does not hinder the first step of the process, oxidation of either ascorbate or triethanolamine (E_ox, 0.783 and 0.328, respectively, compared with the excited state E_red* 0.84–1.45 V vs. SCE) that act as sacrificial donors (D). Flash photolysis experiments showed the formation of the Ru(I) complexes in all cases. Although reduction of CO₂ upon photocatalysis takes place under these conditions, this is accompanied by decomposition of Ru(II) complexes 1, which gives solvates that are unable to catalyze the reaction.

Under these conditions, the maximal turnover number of four is reached (in the case of 1a). On the contrary, when the functions of the sensitizer and catalyst are separated, fast quenching of the Ru(I) intermediate by nickel cyclam 2 precludes photodecomposition, and the conversion proceeds further. In the experiment in entry 9 (Table 2) a turnover number >15 is reached and the reaction still occurs, with no slackening after 120 h of irradiation. Formation of formic acid is largely favored in the process, with a HCOOH/CO ratio of about 5–6, and a significant amount of H₂. The selectivity and the course of the process is qualitatively and quantitatively comparable with that obtained using a similar system in water, with a higher powered lamp (1000 W Xenon arc with a continuous emission) [13], while in this case, low-consumption LEDs are used. As mentioned above, introducing a carboxyl or phosphonyl group in the bipyridine ring of the Ru(II) complexes does not hinder their activity as electron transfer photosensitizers, at least in the present case. A more significant change occurs with the corresponding Ru(I) complexes that are strongly stabilized by the introduction of such electron-withdrawing groups. As one may expect, the resulting longer lifetime is in part compensated for by the lower quenching rate by the Nickel complex, but on balance, this step is also not precluded. However, with 1b–1d, CO₂ reduction proceeds at or below the rate with 1a, and CO is not formed at all. By contrast, heterogenization on titania particles was obtained in the case of 1d; however, this brings no improvement, at least at the low Ru charge achieved under the present conditions. As shown in Table 3, the bonding of a ruthenium complex little changes the activity of untreated titania nanoparticles (compare Cat1, Cat2 with ruthenium free Cat3), the only positive indication being the formation of a small amount of CO upon sensitization by Cat2.

In conclusion, the present work suggests that some elaboration of Ru(bipy)₃ complexes suitable for heterogenization is possible, by inserting acidic groups on one of the bipy rings and bonding to titania nanoparticles. Such complexes are active in CO₂ photoreduction in solution, whereas bonding to titania causes only limited changes in the photobehavior of the semiconductor.