The hunger for energy is unbroken and will not even stagnate in the face of a growing human population. The amount of CO₂ emitted in human activities was recently estimated to be about 5.5 Gt carbon per year [1]. With a growing CO₂ concentration, the effects of greenhouse gases to nature enforce recurring discussions on how to handle this interference with effects on the climate that are already noticeable and will even rise. Beside direct heat emission (e.g., via combustion gases) these greenhouse gases are responsible for the greenhouse effect and, according to reports of the Intergovermental Panel on Climate Change (IPCC), cause a slow but increasing global warming [2]. In the context of climate aspects, the discussions are not so much about the existence of global warming, but more about the influence of CO₂ on this process. In 1990, Lindzen et al. reported that 98% of the greenhouse effect is due to water vapor and stratiform clouds with CO₂ contributing less than 2% [3]. However, these studies do not differentiate between condensing and noncondensing greenhouse gases. Unlike water, gases like CO₂, N₂O, CH₄ and chlorofluorocarbons do not condense and precipitate at current climate temperatures. Therefore, those gases provide the stable temperature structure that sustains the current levels of atmospheric water vapor and clouds. The direct influence of water is, in current studies, still estimated to be 75% on the climate but without the radiative forcing supplied by noncondensing gases like CO₂ the earth temperature would be strongly reduced [4].

Even for CO₂, as a noncondensing gas, measurements and reconstructions of the atmospheric CO₂ history reveals that less than half of these emissions remain in the atmosphere. The anthropogenic CO₂ that did not accumulate in the atmosphere has been taken up by the ocean, by the land biosphere, or by a combination of both [5]. This leads to a measurable acidification of the oceans and changes the saturation state of the oceans with respect to calcium carbonate (CaCO₃) particles [6], with a huge negative impact on especially corals and plankton [7,8,9]. On the other hand, the amount of fossil carbon sources are limited. Due to rising oil prices, CO₂ gains more and more interest as cheap and abundant C1 feedstock, although CO₂ is the lowest in energy of all carbon-containing binary neutral species. Reactions that generate reduced forms of CO₂ (e.g., CO, formates) require energy [1]. However, routes towards organic base chemicals have the potential of high volume CO₂ conversions, although the energy problem of these uphill reactions need to be solved. In terms of sustainability the photocatalytic reduction of CO₂ is a very interesting pathway, in which the activation energy originates from sunlight. Within one hour the irradiation power of sunlight on the earth would be sufficient for the human energy consumption of a complete year. However, the intensity of sunlight strongly depends on the wavelength with a maximum at 550 nm. Thus, with sunlight as the energy source, a visible light absorbing photosensitizer is required. These requirements for an effective chromophore, as well as the general principles of light promoted homogeneous redox catalysis in terms of applications in CO₂ conversions, are summarized in this review. Mechanistic aspects, as well as the products, are discussed in the face of possible applications. Especially the question about a replacement of the sacrificial donor which is found in most of the current publications is an important focus of this review.

Homogenous redox photocatalysts, in principle, consist of a light harvesting unit (photosensitizer) and two catalytic sites [10]. One site is for the oxidation process, where a donor provides the electrons, and the other is the reduction site, where the electrons are transferred to an acceptor. For clarity, this review covers only processes where CO₂ acts as acceptor (Figure 1). In many cases the photosensitizer has dual activity, e.g., as photosensitizer and reduction site, as will be discussed later on (especially in rhenium(I) catalysts).

Visible light absorption by the photosensitizer unit leads to the excited state. Concerning CO₂ reduction, the most commonly used photosensitizers are ruthenium (polypyridyl)- and (bipyridyl)-rhenium(I) complexes (2.1 Chromophores). In this case, the light absorption gives rise to a singlet metal to ligand charge transfer (¹MLCT) which produces a triplet excited state (³MLCT) by inter system crossing (Figure 2). This ³MLCT state has different redox properties than the ground state. Due to the reducing properties of this state, direct interaction with an acceptor can lead to the one-electron oxidized ground state of the photosensitizer via oxidative quenching, as it was reported for the interaction of *Ru(bpy)₃²⁺ with methylviologen (Scheme 1) [11,12].

However, in terms of CO₂ reduction, reductive quenching is the more important process. With an electron that is provided from the oxidation site, a ground state of the one-electron reduced species is formed, which is capable of transferring one electron to the reduction site to recover the d⁶ neutral ground state of the photosensitizer (Figure 3) [13,14,15].

Due to the high reduction potential of the one-electron reduction (−1.9 V vs. normal hydrogen electrode), multielectron reduction processes are more favorable [81]. Scheme 6 shows potential multielectron reduction products of CO₂. From the beginning of homogeneous photocatalytic reduction of CO₂, this topic was associated with the formation of CO.

CO is the feedstock for several synthetic processes, such as d-metal catalyzed Fischer Tropsch (production of hydrocarbons), Monsanato and Cativa (both acetic acid) processes [112]. Carbon monoxide also has a significant fuel value (ΔH_c⁰ = −283 kJ/mol) and can readily be converted into methanol (e.g., by the CuO/ZnO/Al₂O₃-catalyzed ICI process) for use as a liquid fuel [113]. The most active and selective catalysts for the photoreduction of CO₂ to CO are Re(I) type catalysts, but as already mentioned in section 2.2, formic acid can also be obtained via homogeneous photocatalytic CO₂ reduction [13,14,15]. Formic acid has been used as a preservative and an insecticide, as well as a reducing agent and a source of carbon in synthetic chemical industry. However, only a few photocatalysts for selective formation of formic acid from CO₂ have been reported. Yanagida et al. published systems based on oligo(p-phenylenes) with quantum yields of formic acid formation up to 0.084. Another system developed by the same group involves phenazine as photosensitizer, Co^III-cyclam (cyclam = 1,4,8,-11–tetraazacyclotetradecan) as an electron mediator and triethylamine as electron donor. In this case, quantum yields reach values up to 0.07. Both systems require UV-light activation [70,114]. In contrast, Tanaka et al. were able to selectively produce HCOO⁻ via visible light irradiation using [Ru(bpy)₃]²⁺ and [Ru(bpy)₂(CO)₂]²⁺ in a CO₂-saturated TEOA/DMF solution (1:4). The maximum quantum yield in this case is 14% [95,115,116].

According to Scheme 6, the reduction potential for the proton-assisted formation of methane is even smaller than in case of CO formation. Therefore, from a thermodynamic perspective, these photoreduction catalysts should also be capable of producing methane. The reduction of CO₂ to methane, commonly known as the sabatier process, is of substantial industrial importance. This reaction proceeds at high temperatures and hydrogen pressures and is catalyzed by metal catalysts such as Ru, Mo and Ni [117,118,119,120]. However, to date, through strictly homogeneous photocatalysis, this process has never been realized, which is probably due to the requirement of eight electrons that need to be transferred. The only report of photocatalytic reduction of CO₂ to methane employed tris(bipyrazine)ruthenium(II) as the photosensitizer and a Ru metal colloid as the reduction catalyst. In aqueous solution in the presence of TEOA, this system selectively formed methane (no CO formation) in a quantum yield of 0.04% [121].

As amine sacrificial donors react to undesirable site products during the catalytic cycle, one of the most challenging goals is to develop catalytic systems which are able to quench the excited state (³MLCT) without sacrificial agents. In this context, Neumann et al. published an organometallic-polyoxometalate hybrid complex that is capable of replacing sacrificial amines in the photoreduction of CO₂ by hydrogen (Figure 8) [122]. A first important reaction step is the oxidation of hydrogen forming two protons and two electrons, which has been realized by using the Keggin-type polyoxometalate H₃PW₁₂O₄₀ as oxidizing agent. This polyoxometalate was linked to a photocatalyst by coordination of 5,6-(15-crown-5)-1,10-phenanthroline (L) and thus proton and electron transfer between the H₂ oxidative site and the CO₂ reductive site is supported. However, quantum yields and turnover frequencies in the reduction of CO₂ in this system are lower than in some previous rhenium and ruthenium catalyzed photoreductions using amines as sacrificial electron donors.

Another possible method to circumvent the use of sacrificial agents has been described by Carpenter et al. [123]. Instead of replacing an external reductant, the idea was to recycle the sacrificial agent by hydrogenation. To realize this approach it was necessary to find amine structures which do not form typical reactive species such as amino radicals, enamines and iminium ions from the radical cation, but an alkene product [124]. One suggested structure, with complaints about these requirements, is shown in Figure 9. The tricyclic structure blocks the formation of undesired reactive intermediates and reacts to the corresponding olefinic compound, which can then be converted back to the starting material with a Pd/C catalyst in EtOAc. It has to be mentioned that there is a counterion- and solvent-dependent side reaction of the olefin, as well, which is shown in Figure 10.

One of the most desired reactions in the field of photocatalytic CO₂ reduction is the coupling to water oxidation. There are some reports of photocatalytic water oxidation in homogeneous systems such as molecular ruthenium complexes [125,126,127,128,129,130,131,132,133,134,135], as well as earth abundant metal complexes [136,137,138] in the presence of sacrificial electron acceptors. In this case, a recently published mononuclear ruthenium complex shows TOF values up to 0.32 s⁻¹ [139]. The main structure motif is a ruthenium center with a tridentate ligand and three picoline ligands (Figure 11).

By varying the tridentate ligand site, Tong et al. could show that anionic tridentate ligands compared to neutral ligands demonstrate an increased picoline/water substitution rate which results in higher activity of the catalysts. These results demonstrate a correlation between the activity of WOCs and ligand properties, providing important information for future catalyst design. However, water oxidation in the absence of sacrificial acceptors to date was only realized in heterogeneous systems. For example Kudo et al. developed an Ag cocatalyst-loaded ALa₄Ti₄O₁₅ (A = Ca, Sr, Ba) photocatalyst, which reduces CO₂ to CO and HCOOH in aqueous suspension without any additional sacrificial reagents [140]. Furthermore Kajino et al. have been able to combine reduction of CO₂ to HCOO^- and water oxidation with a p-type InP/Ru-complex polymer hybrid photocatalyst connected to TiO₂. In this case, water is the donor of electrons, as well as the source of protons. The conversion efficiency (solar energy to chemical energy) of this catalytic process is 0.03–0.04%. Therefore, an electrical bias-free reaction approaching photosynthesis was found. Further optimization in selectivity, as well as efficiency, can be achieved by configuration of the energy band, variation of catalyst structure, as well as conjugation conformation [141]. An alternative heterogeneous approach was developed by Yoneyama et al. in 1992 (Scheme 7) [142,143]. In this case, malic acid was synthesized from lactic acid and CO₂ using a multicomponent system containing a cadmium sulfide suspension as photocatalyst, methylviologen (MV) as electron relay, and malic enzyme (ME) and ferredoxin-NADP⁺ reductase (FNR). Firstly, photogenerated electrons are transferred from the cadmium sulfide photocatalyst to MV²⁺ and the reduced MV provides the electrons to reduce NADP⁺ and H⁺. The resulting hole in the conduction band of the semiconductor particles leads to a potential which is adequate to regenerate the photocatalyst and to oxidize lactic acid. The resulting pyruvic acid is then converted to malic acid by reaction with CO₂ and NADPH in a ME-catalyzed process. This catalytic cycle therefore consists of two coupled photoredox reactions. First, the reduction of MV²⁺, and, second, the oxidation of lactic acid. The C-C bond formation between CO₂ and pyruvic acid takes place in the absence of light with the quantum efficiency of this system being around 2.5% [142,143].

In 2011, Zhao et al. published a light-induced hydrocarbon oxygenation coupled to H₂ formation from water splitting (Figure 12) [144]. Starting from the photocatalytic oxidation of hydrocarbons using [Co(NH₃)₅Cl]²⁺ as sacrificial electron acceptor, Zhao’s group has been able to develop a complex system containing a [Ru(TPA)(H₂O)₂]²⁺ photocatalyst, Ru- and Ir photosensitizers as well as the proton reduction catalyst Fe₃(CO)₁₂ for water splitting in a one-pot reaction without sacrificial agents (Figure 13). This kind of electron transfer cascade is believed to be a good initial point for photocatalytic CO₂ reduction combined with desirable oxidation processes.

The bridging ligand in multinuclear tethered complexes affects the properties of the assemblies, such as photoinduced energy transfer [145,146,147,148,149], and intervalence transition [150,151]. There are many examples of multifunctional CO₂ reduction photocatalysts with visible light absorbing antenna (e.g., [Ru(dmb)₃]²⁺ derivatives) and CO₂ catalytic reduction centers (e.g., [Re(dmb)(CO)₃Cl], where the catalytic activity strongly depends on the efficiency of electron transfer from reduced photosensitizers to catalytic centers [74,75,110,152,153]. By direct comparison of the photocatalytic activity of the two Ru(II)–Re(I) binuclear complexes of Figure 14, one comprising a saturated linker and one with a conjugated linker, it was found that the saturated covalently bridged photocatalyst shows higher activity for the photoreduction of CO₂ compared with its conjugated analogue [154]. The nature of the ligand has an impact on the electrochemical properties of the supramolecular systems, which strongly affects the photoreduction activity [155]. In the case of the conjugated ligand with strong electronic communication across the bridging ligand, the reducing power of the OER species is insufficient for efficient reduction of CO₂. Therefore, strong electronic communication between the photosensitizer moiety and the reduction site lessens the photocatalytic activity even though it accelerates the electron transfer between them. [53].

Another consideration for supramolecular systems is adjustment of the distance between the covalently tethered metal centers. Koike et al. investigated a series of covalently linked Ru-Re complexes with a saturated alkyl linker of various lengths and compared their activity in the photoreduction of CO₂ (Figure 15). The most efficient catalyst was that comprising of the shortest alkyl chain in the bridging position. According to the photophysical properties of these complexes, the most obvious difference is found in the luminescence quenching with BNAH. According to Stern-Volmer plot measurements, the quenching of the emissive state to form the OER was most effective for the complex with n = 2. The quenching fraction affects the efficiency of the photocatalysis of the diad as this reductive quenching is the first process of the photocatalytic reaction [156]. The electron transfer rate from the Ru site to the Re site may be another factor, but it could play only a minor role because it is assumed not to be a rate-determining process in the photocatalytic reaction.

In a recent review [111], the photocatalytic ability of these supramolecular Ru-Re catalysts was further modified by phosphorus ligands on the rhenium site, resulting in one of the most active photocatalysts for CO₂ reduction (TOF_CO = 281 h⁻¹). Especially for R = p-FPh, the selectivity towards CO formation and the durability of the catalysts was optimized by the phosphorus ligands (Figure 16).

In contrast to thermal reactions, photocatalytic systems have the potential to conduct uphill transformations, which cannot be realized by thermal reactions. The central building unit of each visible light photocatalyst is the photosensitizer. The most widely investigated systems are Ru(II) catalysts which are known for their high absorption of visible light and their long-lived ³MLCT state. With regards to the reduction site rhenium(I) catalysts with their high efficiencies and high selectivities towards CO formation have been subject to most study. However, other products such as formates can be obtained via photoreduction of CO₂ with special optimized reduction sites. One major task for prospective investigations is the search for alternatives to the sacrificial donors which are used in most cases. To date, there are a limited number of reports on alcohol or olefin oxidation that give defined oxidation products, but they are barely used in photoredox reactions with CO₂. The search for supramolecular catalysts that combine both an effective oxidation and reduction site must continue. Many important parameters (e.g., saturated bridging linkers) for the building principle of supramolecular systems have already been identified and will help to find optimized systems.