Donor-Acceptor Activation of Carbon Dioxide

Graphical Abstract

The cooperative activation of carbon dioxide by donor-acceptor interactions have become a highly popular strategy whose chemical designs span across the whole periodic table. This perspective categorizes the wide variety of systems reported and assesses their limitations and prospects to develop catalytic functionalization protocols.

Abstract

The activation and functionalization of carbon dioxide entails great interest related to its abundance, low toxicity and associated environmental problems. However, the inertness of CO₂ has posed a challenge towards its efficient conversion to added-value products. In this review we discuss one of the strategies that have been widely used to capture and activate carbon dioxide, namely the use of donor-acceptor interactions by partnering a Lewis acidic and a Lewis basic fragment. This type of CO₂ activation resembles that found in metalloenzymes, whose outstanding performance in catalytically transforming carbon dioxide encourages further bioinspired research. We have divided this review into three general sections based on the nature of the active sites: metal-free examples (mainly formed by frustrated Lewis pairs), main group-transition metal combinations, and transition metal heterobimetallic complexes. Overall, we discuss one hundred compounds that cooperatively activate carbon dioxide by donor-acceptor interactions, revealing a wide range of structural motifs.

1 Introduction

Carbon dioxide is the central anthropogenic greenhouse gas emitted on a continuous basis by the combustion of fossil fuels, and its atmospheric accumulation entails a myriad of environmental issues whose ultimate consequences are not yet fully understood.¹ Therefore, there is a long-standing interest in developing efficient protocols for the functionalization and valorization of this molecule.² This is further encouraged by its high availability and low toxicity, important aspects to overcome cost-efficiency limitations. Although there is already a number of industrial processes for CO₂ upgrading,^{3, 4} there are yet many drawbacks to surpass and plenty of highly desired transformations that remain great challenges in catalysis. For instance, its efficient conversion to form C1-products such as formic acid, methanol or methane, or towards C2 or higher products as ethylene, acrylic acid or aryl carboxylic acids, among others. We have already witnessed seminal discoveries and tremendous advances in the area,⁵ though we believe those are just the tip of the iceberg for the next decades to come.

The challenging character of the aforesaid transformations mainly relies on the inert nature of carbon dioxide. It is a linear and non-polar molecule, although it contains two polar C^δ+−O^δ− bonds. To increase the reactivity of CO₂ molecules, C−O bonds must be further polarized and elongated, causing bending of the O−C−O angle, and therefore a net non-zero dipolar moment. This process, known in a simplified general notion as activation of CO₂, is typically the first step when chemically converting CO₂ into high-value products.

Many approaches for the activation of CO₂ have been described⁶ and we do not intend to be comprehensive in this review work. We will rather focus on one of the strategies that we believe holds great prospects for the future development of functionalization protocols, more precisely, the activation of CO₂ by donor-acceptor interactions through rationally designed systems.⁷ This strategy applied to the activation of CO₂ was pioneered by Floriani in the 1970s, who already pointed out at that time that ‘CO₂ fixation involves concerted attack of a base on the CO₂ carbon atom together with coordination of the oxygen to a ‘hard’ metal ion’,⁸ something that was generally accepted by then for metalloenzymes.⁹ In the artificial synthetic version, a Lewis acidic (LA) and a Lewis basic (LB) fragments cooperate for the synergistic activation of the CO₂ molecule. The acidic site interacts with the electronegative oxygen atoms and the Lewis basic moiety with the electrophilic carbon atom producing both elongation and bending of the carbon dioxide molecule. Clearly, this notion is in many ways equivalent to the concept of frustrated Lewis pairs (FLPs)^{10, 11} and, in fact, the latter systems have been amply used for the activation of carbon dioxide.¹²

Despite the apparent simplicity of the model, many potential structures may emerge by push-pull or FLP-type activations (Figure 1), with particular features that are of relevance for subsequent CO₂ transformations. For instance, the LA can coordinate to only one or to both oxygen atoms in a mono- or bidentate manner, respectively, or two LA sites may be used to bind the two oxygen centres. In terms of spatial arrangement, the CO₂ may be inserted between the LA and LB sites or the two active sites may be present at the same plane, usually held together by a bridging fragment. Alternatively, the LA and LB moieties can be bound by a LA−LB bond or weak interaction, which may later participate in the subsequent transformation of CO₂.

Figure 1

Most common general simplified structures for the cooperative activation of CO₂ by push-pull interactions (LA=Lewis acid, LB=Lewis base).

Naturally, this donor-acceptor approach is not exclusive of CO₂ activation. It has been exploited by numerous research groups for the activation of many other substrates,^{11, 13} the isolation of otherwise fleeting species¹⁴ or the design of sensors.¹⁵ In addition, by a somewhat related approach, it has served as a convenient access to radical chemistry¹⁶ or for the synthesis of new materials,¹⁷ among other applications. Besides, it is interesting to remark, as already pointed out by Floriani, that this type of cooperative CO₂ activation holds some reminiscence to biological systems. Structural studies of the active sites of metalloenzymes that are active in the conversion of CO₂ contain multimetallic cores, as for example [NiFe] or [CuMo] CO dehydrogenases,¹⁸ in which the two metals present opposite Lewis character as a key feature of CO₂ binding.

In this review we analyse inter- and intramolecular Lewis acidic and basic pairs that have successfully activated carbon dioxide, causing both bending and elongation of the C−O bonds, and in some examples even C−O bond cleavage. We have divided this review in three general sections based on the nature of the active sites. Therefore, these pairs include metal-free examples (mainly formed by FLP systems), main group-transition metal partnerships and transition metal heterobimetallic complexes where the polarized metal-metal bond is crucial to activate CO₂. The structural information by X-ray diffraction of the CO₂-cooperative activation products for these three types of systems has been collected in the Supporting Information as Tables S1–S3. In addition, these tables include, when available, experimental values of infrared spectroscopy vibration frequencies associated to the CO₂-activated fragments. We analyse the structural parameters of the acid/base systems that contain an activated molecule of CO₂, focusing on how push-pull interactions bring about the elongation and bending of the bonds via formation of bridging LA−O and LB−C bonds. Cooperative polar effects associated to the activation of CO₂ by f-elements lie out of the scope of this work.¹⁹ For further reading on the topic of bimetallic activation of CO₂, including alternative strategies based on homobimetallic compounds, we refer the reader to the enlightening recent review work of Mankad and co-workers.²⁰

2 Metal-Free Acid/base Pairs

2.1 Phosphine/borane Pairs

Intra- and intermolecular systems based on phosphine/borane fragments have been widely investigated as frustrated Lewis pairs for the fixation and activation of carbon dioxide. Figure 2 collects a summary of the molecular structures reported to date involving CO₂ fixation through this approach. In all cases the Lewis basic phosphorus atom coordinates to the central CO₂ carbon, while the Lewis acidic boron centre binds to one oxygen atom (compounds 1–12, 15). Additionally, there are examples where a bis-borane is connected to the two oxygen atoms of CO₂ in a bidentate fashion (compounds 13 and 14). In any case, CO₂ sequestration involving this type of frustrated Lewis pairs always generates zwitterionic compounds.

Figure 2

CO₂ activation by phosphine/borane Lewis pairs.

Among the structures depicted in Figure 2, compound 1, described by Stephan and co-workers in 2009, is the first example of CO₂ activation by phosphine/borane frustrated Lewis Pairs and established the ground rules over which much chemistry has later been developed.²¹ The combination of P^tBu₃ and B(C₆F₅)₃ in the presence of a CO₂ atmosphere afforded compound 1, which reversibly loses CO₂ at 80 °C under vacuum. An equivalent outcome was observed by using the ambiphilic compound (Me₃C₆H₂)₂PCH₂CH₂B(C₆F₅)₂, but the stability of the resulting adduct 2 is lower and it decomposes in dichloromethane or toluene solutions even at −20 °C.²¹ The same research group has subsequently investigated other systems comprised of different phosphine/borane moieties, isolating the expected zwitterionic adducts in most cases (compounds 3–10).^{22, 23} Similarly to compound 2, most of these species release CO₂ under mild conditions (above −15 °C), which matches with their calculated formation energy values.²³

Interestingly, the reaction between PⁱPr₃ and B(C₆F₅)₃ did not result in the formation of the corresponding CO₂ adduct. Conversely, a nucleophilic aromatic substitution by the phosphine takes place at the para position of one of the perfluorinated aryl rings, yielding compound (ⁱPr)₃P(p-C₆F₄)B(F)(C₆F₅)₂. This outcome is not observed when the fluorine at the para position is substituted by a proton in compound B(C₆F₄H)₃, giving rise to adducts 3 and 4 (Figure 2). The same occured when other groups were placed at that position instead of a proton (n-hexyl, cyclohexyl, norbornyl, phenyl, chloride, triflate; compounds 5–9), which again had no impact on the stability nor the CO₂ activation energy profile according to the experimental and computational data and, therefore, no clear trends could be established.²³ In all cases, the thermal instability of these species likely prevents any potential derivatization, as it has been proven by attempting to reduce these compounds with dihydrogen.²³ A marked difference was found instead for a geminal phosphine/borane FLP described later on by Slootweg and co-workers, who envisioned the use of a geminal intramolecular FLP leading to compound 11 after CO₂ activation (Figure 2).²⁴ This system has shown to be more reactive due to the ideal orientation of the two sites for optimal interaction with the CO₂,²⁵ which therefore is activated without the need of strongly electron-withdrawing substituents on the borane. Moreover, and in contrast with the above adducts, the five-membered heterocycle generated is thermally stable and does not show release of CO₂ even at 100 °C under vacuum.²⁴

A range of diboranes have also been studied as Lewis acids for this process, allowing to isolate and structurally characterize adducts 12–15 (Figure 2). Stephan and co-workers investigated the reaction of P^tBu₃ with O(B(C₆F₅)₂)₂ in the presence of CO₂ and could obtain compound 12 as crystals at low temperature.²⁶ CO₂ is coordinated to just one boron atom since the large B−O−B angle inhibits chelation by the two boron centres. In solution, NMR data revealed facile exchange of the O₂C-P^tBu₃ fragment between the two boron atoms. In order to reduce the bite angle and favour the chelation, the authors used a bis-borane with a C(sp²)-bridge as the Lewis acid, which permitted the isolation of compounds 13 and 14, in which CO₂ is chelated by the two boron centres of the bis-borane. Nonetheless, despite this bidentate interaction, NMR essays evidenced loss of CO₂ even at low temperatures.²⁶ At variance, the same group has succeeded in reacting the FLP formed by P^tBu₃ and 1,2-C₆H₄(BCl₂)₂ with CO₂, achieving the highly stable compound 15 (Figure 2).²⁷ In this case, CO₂ is strongly coordinated in an asymmetric fashion and no decomposition is observed by heating this species to 80 °C for 24 h, in contrast to previous observations in compounds 13 and 14. The stability of this adduct enabled further stoichiometric reactions with reducing agents. Thus, treatment of this species with the amino-borane Me₂NHBH₃ followed by quenching with D₂O led to formation of deuterated methanol.²⁷

2.2 Amine/borane Pairs

Although phosphine/borane pairs are the most common motifs in FLP chemistry, amines have also been employed as Lewis bases to fix and activate CO₂ cooperatively with boranes (Figure 3). In compounds 16–19, CO₂ is trapped intermolecularly, but there are also examples in which it is activated via intramolecular FLP systems affording six-membered ring compounds 20–22. For instance, Stephan and Erker reported the synthesis of different amine/borane adducts and their reactivity towards CO₂.²⁸ Compound 16 was obtained by the combination of PhCH₂NMe₂ and B(C₆F₅)₃ in the presence of CO₂, but this adduct showed to be unstable in solution and decomposed even at low temperatures. By using a secondary amine, PhⁱPrNH, the authors could isolate the zwitterionic salt 17 and the use of the diamine 1,4-C₆H₄(CH₂NH^tBu)₂ led to species 18. Additionally, the combination of the latter diamine with B(C₆F₅)₃ and CO₂ but in the presence of two equivalents of 1,2,2,6,6-pentamethylpiperidine as well, led to isolation of compound 19, in which the more basic piperidine gets protonated.²⁸

Figure 3

CO₂ activation by amine/borane Lewis pairs.

Intramolecular CO₂ activation via an amine/borane pair was also reported by Stephan.²⁹ The boron amidinate species resulting from the reaction of Piers’ borane HB(C₆F₅)₂ and isopropyl carbodiimide was reacted with CO₂ achieving compound 20 as a six-membered heterocycle (Figure 3). In this case, the authors suggest an initial equilibrium between the strained four-membered boron amidinate and its open-chain form, which facilitates CO₂ activation via an FLP mechanism.²⁹ In a similar manner, Cantat and co-workers used the FLP system comprised by the basic guanidine 1,5,7-triazabicyclo[4.4.0]dec-5-ene (TBD) and the Lewis acid 9-borabicyclo[3.3.1]nonane (9-BBN) to afford compound 21 through reaction with CO₂.³⁰ Compound 21 is stable under reduced pressure, in contrast to the previously described TBD-CO₂ adduct, which decarboxylates under otherwise identical conditions.³¹ Interestingly, NMR analysis of the reaction crude indicated the concomitant formation of formate and acetal derivatives as reduced species, thus pointing out to the potential to perform metal-free catalysts following this cooperative activation of CO₂. In fact, compound 21 is catalytically active in the hydroboration of CO₂ to methoxyborane, while mechanistic investigations proved that this LA/LB combination promotes the hydride transfer from the boron to the carbon atom of CO₂ (at variance to an alternative mechanism based on borane activation).³¹ Finally, reaction of compound 21 with 9-BBNI (the iodine-version of BBN) resulted in isolation of the unprecedented adduct 22, in which the CO₂ molecule is coordinated to two boron atoms (Figure 3), confirming the capacity of the donor-acceptor CO₂ structure to bind an additional borane molecule.³⁰

2.3 Phosphine/alane Pairs

The isolation of compounds 23–32 reveals that aluminium can alternatively be used as an efficient Lewis acid to activate CO₂ together with phosphines (Figure 4). As for borane-based systems, this molecule can be fixed both through intra- and intermolecular pathways. In addition, it can bind to either one or two aluminium centres through the oxygen atoms in a mono- or bidentate manner, respectively, affording in all cases zwitterionic compounds. The Stephan group reported the first examples of CO₂ trapped by P/Al Lewis pairs by donor-acceptor interactions.³² Compounds 23 and 24 were obtained by the exposure of 1 : 2 mixtures of PMes₃/AlR₃ (R=Cl, Br) to 1 bar of CO₂ atmosphere. As expected, the phosphorus atom in these adducts is bonded to the carbon atom of CO₂ while aluminium centres are linked to the oxygen atoms. The incorporation of the two aluminium centres in these species is responsible for their high stability, and consequently they do not release CO₂ even at 80 °C under high vacuum. Remarkably, the abovementioned Lewis pairs proved to be competent in reducing CO₂ to methanol (after hydrolysis) by using ammonia borane as the hydrogen source.²² The same research group isolated compound 25 by the reaction of ^tBu₃P(CO₂)B(C₆F₅)₂Cl (compound 8, Figure 2) with one equivalent of Al(C₆F₅)₃ at room temperature. In this case, and in contrast to compounds 23 and 24, the resulting FLP involves the aluminium counterpart in a 1 : 1 ratio relative to the Lewis base, though no further studies on the capability of this species for reducing CO₂ were disclosed.³³

Figure 4

CO₂ activation by phosphine/alane Lewis pairs.

This Lewis pair combination has also been employed to trap CO₂ in an intramolecular fashion. Thus, Slootweg and co-workers used a geminal phosphorus/aluminum FLP to synthesize adduct 26 as a five-membered heterocycle (Figure 4). The authors observed that this process is reversible and could regenerate the starting FLP by heating compound 26 in the solid state at 135 °C under vacuum. Additionally, computational analysis for this transformation revealed that CO₂ is activated by initial interaction with the acceptor site, contrasting with previous observations in the case of P/B-based FLP where CO₂ interacts first with the Lewis base.³⁴ Later on, the same research group explored the reaction between the dimeric compound [^tBu₂PCH₂AlMe₂]₂ and CO₂. The combination of these species under ambient conditions resulted in the formation of adduct 27 as an oil, which could be isolated and characterized by spectroscopic methods, and showed to be thermally stable even by heating at 70 °C. Surprisingly, this species proved to be unstable in the reaction mixture and its rearrangement led to a carboxylate dimer. Authors hypothesized that an excess of CO₂ could be the origin of this remarkable divergence. Computational studies confirmed the cooperative action of two CO₂ molecules to cleave the Al−P bonds of the initial heterocyclic dimer.³⁵

More recently, Kaupp and Limberg reported that a xanthene-based P/Al FLP intramolecularly traps CO₂.³⁶ Thus, compound 28 was synthesized by adding CO₂ to the bimetallic species Xant(PPh₂−AlMe₂Cl−AlMeCl), that is, an interesting P/Al xanthene FLP with a AlMe₂Cl fragment trapped by donor-acceptor interactions. The latter aluminium fragment participates as well in the push-pull activation of CO₂. More precisely, the latter molecule is bound via the carbon atom to the phosphine function and the two oxygen atoms are linked to the two aluminium centres (Figure 4).³⁶ Following the same approach the authors were able to isolate the structurally similar adducts 29 and 30. However, while formation of adduct 28 is a reversible CO₂-pressure dependent process, no decomposition is observed for compounds 29 and 30 when the excess pressure of CO₂ is released. According to the authors, the reversibility of this process is highly dependent on the Lewis acidity of the aluminium centres, which is tuned by modifying their substituents. Thus, the stability of adduct 29 is higher than 28 and this results in a larger barrier for the back reaction. For compound 30, bearing more electron-withdrawing C₆F₅ groups, this energy barrier is even higher due to further stabilization as a result of a low reorganization energy.

In an attempt to synthesize a system with a 1 : 1 phosphine/alane stoichiometry, the authors explored the use a rigid biphenylene linker. Based on a geometric rationalization, this should avoid the interaction between the phosphorus and aluminium centres due to a larger spatial separation between them, thus facilitating access to a CO₂ adduct with the mentioned stoichiometry. Following this approach, compound 31 could be isolated as the eight-membered heterocyclic species shown in Figure 4, which proved to be stable, not releasing CO₂ even under high vacuum.³⁷

Lastly, Mitzel and co-workers have recently reported the use of an oxygen-bridged aluminium/phosphorus FLP to activate CO₂.³⁸ Thus, the reaction of ^tBu₂P−O−AlBis₂ (Bis=CH(SiMe₃)₂) with CO₂ afforded adduct 32 (Figure 4). In this compound, CO₂ binding is stable at room temperature contrasting with similar P−O−B systems, in which an equilibrium process is observed.³⁹ Computational studies supported this observation and, additionally, confirmed that the formation of this compound takes place in a concerted manner.

2.4 Silicon-based pairs

Silicon has also been successfully used as a Lewis acid counterpart for the fixation and activation of CO₂. Müller and co-workers combined bulky trialkylphosphines with triarylsilylium ions to fix CO₂ in the form of the silylacylphosphonium borates 33 and 34 (Figure 5). These compounds are stable species only under ambient conditions, because the phosphonium ions decompose upon heating in toluene before CO₂ is released. Therefore, the CO₂ fixation is not reversible in this case. Moreover, although these adducts were spectroscopically well-characterized, their structure could not be elucidated by single-crystal X-ray diffraction studies.⁴⁰

Figure 5

CO₂ activation by silicon-based FLP-type systems (OTf=[CF₃SO₃]⁻).

Later on, Stephan and co-workers reacted silyl triflates and 2,2,6,6-tetramethylpiperidine (TMP) in the presence of 1 bar of CO₂ atmosphere to afford silyl carbamates 35–38 (Figure 5) and [TMPH][OTf] as the by-product.⁴¹ While the molecular structure of compound 35 reveals a five-coordinate silicon centre with the CO₂ moiety binding to this atom in a κ² chelating fashion, for adducts 36–38 bearing respectively methyl, phenyl and pentafluorophenyl groups on the silicon atom, the CO₂ molecule is coordinated in a κ¹ manner. Once more, this observation shows the influence of the Lewis acid electrophilicity on the precise type of preferred CO₂ fixation. Regarding their stability, these compounds do not release CO₂ even under high vacuum. This could be due to the separation from the by-product salt which prevents the inverse reaction.⁴¹

In the same work, the authors investigated the sequestration of CO₂ by silicon/phosphine Lewis pairs. Thus, the combination of silyl triflates with the trialkylphosphines PEt₃ and P^tBu₃ in the presence of CO₂ yielded adducts 39–42 (Figure 5). Nevertheless, these species are unstable and release CO₂ under dinitrogen atmosphere. When using Ph₂Si(OTf)₂ as the Lewis acid partner, compounds 43–46 were obtained depending on the stoichiometric equivalents of phosphine employed. For instance, adduct 43 is the major species when employing 0.5 equivalents of PEt₃; however, bis-CO₂ activation product 45 exists as predominant species in equilibrium with free phosphine when 4.0 equivalents of PEt₃ are used. Furthermore, it was observed that the equilibrium of the CO₂ insertion depends on the phosphine employed and, accordingly, the generation of the bis-CO₂ product is favoured by using PEt₃ rather than P^tBu₃. According to the authors, the steric hindrance of P^tBu₃ is probably less favourable to lead to a suitable arrangement of the Lewis groups.⁴¹

The latest examples of CO₂ activation by silicon-based Lewis pairs have been reported by Cantat and co-workers using guanidine TBD systems related to those discussed before in section 2.2.⁴² Thus, the combination of TBD-based silyl compounds with CO₂ afforded adducts 47 and 48 (Figure 5), in which the carbon atom of the CO₂ molecule coordinates to a nitrogen atom on the heterocycle backbone and one oxygen atom binds to the silylium centre. Interestingly, the formation of compound 47 is more favoured in dichloromethane than in THF, which agrees with the more facile cleavage of Si−Cl bonds observed in this solvent.⁴³ Regarding their stability, adduct 48 showed to be more stable than the analogous compound 47.

Thus, while 47 decomposed to the decarboxylated form after solvent removal, compound 48 requires up to 7 days under vacuum to complete CO₂ release. It is interesting to note that the formation of these adducts seems to be highly dependent on the steric hindrance around the silicon centre, since no reaction is observed when analogous Si/N Lewis pairs bearing isopropyl or phenyl groups are employed as starting materials. As in their prior work on TBD/borane pairs,³¹ these silicon-based species showed to be active in the catalytic hydroboration of CO₂ to the methoxide level with different boranes.⁴²

2.5 Other donor-acceptor pairs based on main group elements

The previous examples describe the most common combinations of main group elements with contrasting Lewis acidic and basic character that have been used for the push-pull activation of CO₂. Nonetheless, the landscape of main group partnership is rather rich, and other matches, particularly associated to bimetallic structures, have shown particularly relevant results. This sub-section will briefly describe some of these advances with emphasis on structures that have been authenticated by X-ray diffraction studies.

Besides phosphorus and nitrogen, group 16 elements have been employed as Lewis basic centres in combination with aluminium for CO₂ activation (compounds 49–51, Figure 6).^44-46 Thus, the reaction of sulfido and selenido complexes K[Al(NON^Dipp)(E)] (E=S, Se) with CO₂ afforded compounds 49 and 50 via a [2+2]-cycloaddition pathway with Al=E bond reduction. Interestingly, while the reaction with the sulphur compound took two days at room temperature for completion, the selenium analogue 50 was formed in only fifteen minutes under the same conditions. Furthermore, it is important to remark that, in both cases, the molecule exists as a non-symmetric dimer with each CO₂ moiety bridging to a potassium cation of the second unit.⁴⁴ Following a similar approach, the tellurodicarbonate species 51 could be synthesized by the reaction of the aluminium tellurido compound K[Al(NON^Dipp)(Te)(THF)] with 1 bar of CO₂. NMR data is consistent with the initial CO₂ cycloaddition to the Al=Te bond to afford a tellurocarbonate intermediate, which evolves to the tellurodicarbonate compound 51 by the insertion of a second CO₂ molecule between aluminium and tellurium centres. Moreover, as in compounds 49 and 50, potassium cations stabilize the crystal structure via C=O⋅⋅⋅K interactions.⁴⁵ In another example, exposure of aluminium telluride complex (IMe₄)₂Al(Tipp)=Te (where IMe=1,3,4,5-tetramethylimidazol-2-ylidene; Tipp=C₆H₂-2,4,6-ⁱPr₃) to 1 bar of CO₂ resulted in the formation of complex 52, which is thermally unstable and decomposes even at low temperatures. In this case a triple CO₂ insertion takes place, in which the pentacoordinate aluminium centre binds to two CO₂−NHC moieties and one CO₂Te fragment in a κ²O,O′ fashion. Computational studies suggest that the reaction most likely occurs through two initial CO₂ insertions into the Al−C^NHC bonds, with the third insertion into the Al=Te bond proceeding afterwards. In addition, calculations reveal that the CO₂Te ligand in compound 52 can be described as the first example of a tellurocarbonate [CO₂Te]²⁻.

Figure 6

CO₂ activation by other donor-acceptor pairs based on main group elements.

Gallium has been explored as well as a Lewis acid for intramolecular CO₂ sequestration.^{47, 48} Goicoechea and co-workers reported the quantitative formation of adduct 53 by the reaction of CO₂ and a phosphagallene featuring a gallium-phosphorus double bond. At variance with other donor-acceptor systems, the negative charge is located between the gallium centre and the more electronegative phosphorus atom, as supported by a particularly low ³¹P NMR frequency. Moreover, adduct formation is not reversible under mild conditions, which the authors attribute to the new strong σ-bonds formed at the cost of a weak P−Ga π-bond in the starting phosphagallene species.⁴⁷ Later on, Schulz reported the synthesis of complexes 54 and 55 by the cycloaddition reaction of CO₂ to the four-membered metallaheterocycles LGa(Cl)P[μ-C(O)NR]GaL (where L=HC[C(Me)N(2,6-ⁱPr₂C₆H₃)]₂, and R=ⁱPr, Cy). Interestingly, the process is reversible in this case, forming the corresponding starting species with CO₂ release at high temperatures.⁴⁸ Regarding the use of germanium, only a few compounds have so far been described for the cooperative activation of CO₂.^{49, 50} The germanimine [(HMDS){(SiMe₃)(Mes)N}Ge=N(SiMe₃)] (where HMDS=hexamethyldisilazane and Mes=2,4,6-trimethylphenyl) was reacted with CO₂ to afford compound 56 via [2+2] cycloaddition, in which CO₂ is trapped as part of a four-membered metallacycle.⁴⁹ However, while in that case germanium acts as a Lewis acid, in compounds 57 and 58 it behaves as a nucleophile in the addition of CO₂, bonding to this molecule in a η¹-CO₂-κC manner. These adducts were synthesized by the reaction between the corresponding potassium germanylidene salts and 1 bar of CO₂, obtaining unidentified by-products as well. It is worth highlighting that the μ-CO₂-κC : κO coordination mode displayed in these compounds is rare for low-valent germanium species and it is not accessible for the neutral analogues.⁵⁰

Lastly, it is important to remark that some tin-based compounds have also been described to cooperatively activate CO₂. Reaction between the geminal frustrated Lewis pair (F₅C₂)₃SnCH₂P^tBu₂ and CO₂ led to the formation of adduct 59 upon cooling the solution to −70 °C, in a process characterized as a temperature-dependent equilibrium. In the abovementioned zwitterionic compound, CO₂ is trapped intramolecularly in a five-membered heterocycle.⁵¹ More recently, Aldridge and co-workers have reported the synthesis of compound 60 by the combination of the stannaimine ^MesTerSn(N(SiMe₃)₂)=NMes [^MesTer=C₆H₃-2,6-(C₆H₂-2,4,6-Me₃)₂] and 1 bar of CO₂, which represents the first example of [2+2] cycloaddition product derived from Sn=N bond-based compounds.⁵²

3 Donor-Acceptor Activation of CO₂ by Transition Metal-Main Group Pairs

The combination of a transition metal and a main group element in cooperative pairs for the activation of CO₂ encompasses advantages associated to the partly filled d orbitals of the transition metal, which can partake in subsequent reactions for the catalytic functionalization of CO₂. Although in these cases the notion of donor-acceptor interactions is not always unequivocal, we believe that it is of interest to discuss all kind of heterobimetallic and related systems in which the CO₂ molecule is cooperatively activated between a transition metal and a main group element. The first well-characterized example found in the literature that involved a push-pull heteronuclear activation of a carbon dioxide molecule performed by a transition and a main group metal (Co/K) was reported in 1978 by Floriani and co-workers,⁵³ following its prior spectroscopic identification in 1974.⁸ Complex 61 (Figure 7) presents a structure with a nucleophilic site, the cobalt centre, in a nearly square pyramidal geometry, capable of interacting with the electrophilic carbon of the carbon dioxide molecule. Meanwhile, the acidic role played by the potassium centres is more complicated. There are two different potassium cations that act as acidic partners interacting with up to six basic oxygen atoms each. The first, labelled as K¹ in Figure 7, binds one CO₂ molecule in a bidentate fashion, two THF units and two oxygen atoms from the ligand. In turn, K² interacts with two CO₂ molecules in a κ¹-manner, as well as with the two salen ligands in a bidentate mode. As a result of the interaction of each CO₂ molecule with three metal centres, the C−O bonds are elongated (1.24 and 1.22 Å; cf. 1.155(1) Å in free CO₂),⁵⁴ and the carbon dioxide is bent (O−C−O 135°). This foremost example of donor-acceptor CO₂ activation was followed by other related systems in which the Floriani group could investigate the reversibility of CO₂ fixation and some ground rules for further functionalization.⁵⁵ Another early example of CO₂ activation via transition metal/main group pairs was described by Komiya et al. in the 90s.⁵⁶ A CO₂ complex of Fe(0) reacted with R₃SnCl (R=Me, Ph) forming stannyl ferracarboxylate complexes. Complex FeCl(CO₂-SnPh₃)(depe)₂ (62; depe=1,2-bis(diethylphosphino)ethane), was characterized by X-ray diffraction analysis showing coordination of the CO₂ in a 1κC¹ : 2κO¹ : O² mode with C−O bond distances of 1.32(1) and 1.28(1) Å, slightly elongated with respect to the parent Fe−CO₂ adduct (1.28(2) and 1.25(3) Å), which highlights the higher activation capacity of the bimetallic cooperative approach. This type of coordination was analogous to that shown previously by Gladysz and co-workers for a bridging-CO₂ complex based on a Re/Sn core, though in this case the activated CO₂ did not emerge from the free gas, but through deprotonation of a formate ligand.⁵⁷ Somewhat similar results were obtained from a polynuclear species described by Adams that led to an Os(μ-CO₂)Sn moiety⁵⁸ and for Fe(μ-CO₂)Sn structures reported by Cutler.⁵⁹

Figure 7

Early examples of heterobimetallic CO₂ activation.

More recently, Hayton and co-workers have reported another interesting example in which potassium is again capable of stabilizing a CO₂ adduct by acting as a Lewis acid. In particular, the authors described how the nickel centre in the same complex may behave as an acidic or basic site for the cooperative activation of CO₂ depending on its oxidation state. Thus, in the Ni(II) sulphide complex [K(18-crown-6)][L^tBuNi^II(S)] (L^tBu={(2,6-ⁱPr₂C₆H₃)NC(^tBu)}₂CH), exposure to carbon dioxide forms the diamagnetic complex 64 (Figure 8) as the first structurally characterized transition metal complex containing a thiocarbonate [SCO₂]⁻ ligand.⁶⁰ Formally, both the Ni(II) and K(I) metals would serve as Lewis acidic sites coordinating each of the two oxygen atoms of CO₂, while the basic S²⁻ site binds the central carbon. In stark contrast, addition of CO₂ to the reduced sulphur-free Ni(0)−N₂ complex, which was previously reported by Limberg and co-workers,⁶¹ led to complex 63 (Figure 8), in which now the Ni(0) centre acts as the Lewis base coordinating the central carbon of CO₂, while the potassium centre binds one of the oxygen atoms. Nonetheless, the compound is better described as the oxidative insertion of the CO₂ towards a square planar Ni(II) with a bridging [CO₂]²⁻ ligand.

Figure 8

Cooperative CO₂ activation by Ni/K pairs. Dipp=2,6-diisopropylphenyl.

The donor-acceptor approach emphasized in this review is better exemplified, as discussed in the previous section, by systems that can be described as metallic frustrated Lewis pairs.⁶² For instance, in 2016, Bourissou's group observed the irreversible reaction of CO₂ with a bimetallic platinum-aluminium complex FLP, leading to the formation of complex 65 (Figure 9).⁶³ According to ¹³C labelling studies (using ¹³CO₂), they could unequivocally confirm the direct Pt−CO₂ connectivity by NMR studies (supported by a large ¹J_CPt coupling constant). The coordination mode was corroborated by X-ray diffraction analysis revealing the insertion of CO₂ into the constrained Pt→Al bond, resulting in a T-shape Pt complex. The compound has two phosphines in trans disposition (P−Pt−P, 176.46(8)°) and the CO₂ molecule is κ¹-coordinated to the basic Pt(0) centre through the C atom (Pt−C bond distance of 1.96(1) Å). In turn, one of the oxygen atoms interacts with the acidic Al(III) centre (Al−O distance of 1.833(7) Å) promoting the elongation of the corresponding C−O bond (1.30(1) vs 1.22(1) Å for the unbound CO).

Figure 9

Representative examples of CO₂ activation by donor-acceptor interactions by metallic frustrated Lewis pairs (FLPs). Mes=2,4.6-trimethylphenyl.

Alternatively, and as a matter of fact, on a more frequent basis, the transition metal tends to act as the Lewis acid in metallic FLP designs, instead of as base, as seen for the Bourissou system. An early example from Stephan involves the substitution of the well-known Lewis acid B(C₆F₅)₃ in adduct ^tBu₃P(CO₂)(B(C₆F₅)₃) (1 in Figure 2) by the titanium cation [Cp₂TiMe]⁺, which proceeds cleanly to yield compound [^tBu₃P(CO₂)TiCp₂Cl]⁺ after Me/Cl metathesis.³⁴ Nonetheless, transition metal FLPs were shortly after more thoroughly explored by the groups of Wass and Erker using Zr/phosphine combinations, as Lewis acidic and basic sites, respectively. Among those systems, Wass and co-workers reported in 2016 complexes 66 to 69 (Figure 9)⁶⁴ after trapping CO₂ by several intermolecular FLPs based on a zirconocene species and various phosphines. They provided the X-ray structure of complex 67 proving that the Zr−OMes bond distance (1.962(2) Å) is slightly longer than the distance found in the zirconocene precursor (1.937(2) Å), which can be explained by the CO₂ coordination. The C−O bond distance for the oxygen attached to the metal is 1.278(4) Å, while it is shorter for the unbound oxygen (1.221(5) Å), though both are longer than in free CO₂ (1.155(1) Å).⁵⁴

Early transition metals have also been involved in cooperative activation of CO₂ in partnership with boranes. Piers and co-workers reported deoxygenative hydrosilation of carbon dioxide using a Sc/B(C₆F₅)₃ pair.⁶⁵ In another interesting example described by Erker's group in 2017, a CO₂ molecule was trapped as a five-membered cyclic boratacarbonate product (complex 70, Figure 9), in which an oxygen atom acts as the base forming a new covalent bond with the central carbon atom of CO₂, while both the zirconium and boron sites behave as Lewis acids.⁶⁶ The X-ray structure shows that both Zr−O−C angles are almost linear (168.9(2)° and 166.9(2)°) and that the five-membered carbonate ring features a distorted twist conformation with three different C−O bond lengths (1.245(3), 1.278(3) and 1.317(3) Å). DFT analysis indicates that the formation of compound 70 can be regarded as a concerted CO₂ addition to the O/B pair of a geminal FLP intermediate. Overall, the examples depicted in Figure 9 and discussed along the previous lines are paramount systems that can clearly be described as metallic FLPs, but not the only ones presented in this section. Some other donor-acceptor structures discussed in the following paragraphs have similarly been synthesized and exploited in CO₂ activation processes following the same FLP approach, and in those cases the analogy will be drawn.

In the previous examples, the Lewis acidic or basic behaviour of the transition and main group metals was rather clear and defined by the well-known chemistry of each fragment. However, in a recent study Goicoechea, Aldridge and co-workers succeeded in trapping CO₂ by an Al/Au complex in which the notorious Lewis electrophilicity of gold was reversed, acting instead as the donor centre for the push-pull activation of CO₂ (complex 71).⁶⁷ In this work, they described the formation of an Al^δ+−Au^δ− complex due to the transfer of polarity between a “naked” (NON)Al aluminyl(I) anion and a gold(I) (NON=4,5-bis(2,6-diisopropylanilido)-2,7-di-tert-butyl-9,9-dimethylxanthene). This umpolung entails an unusual nucleophilic reactivity at the gold centre which leads to the reductive insertion of the heteroallene. Compound 71 presents similar bond distances for the bridging oxygen atoms (C−O bonds of 1.284(8) and 1.2899(8) Å) and for the Al−O bonds (1.880(5) and 1.861(5) Å). This result set a precedent for a whole set of bimetallic compounds with related structures. For example, Aldridge's group studied analogous structures with zinc fragments as the nucleophile (complex 72), once more a highly unusual behaviour of zinc.⁶⁸ The AlO₂C four membered ring resulted from the exposure of the starting Al/Zn adduct to CO₂, which resembles the reductive insertion attained with the Al/Au complex, exhibiting similar bond lengths and angles (see Table S2).

On a subsequent work, they described a similar set of bimetallic precursors employing silver and copper.⁶⁹ The Al/Ag complex was exposed to CO₂ atmosphere producing the analogous silver dioxocarbene complex 73 after the reductive insertion of the gas into the metal-metal bond. When complex 73 is heated up to 80 °C, it evolves to compound 74 (Figure 10), which is a dimer in the solid state with two oxygen and two silver atoms in a planar diamond-shape. In a similar manner, the isostructural Al/Cu analogue reacts under CO₂ atmosphere to form directly the same carbonated product (complex 75, Figure 10), though no traces of a stable dioxocarbene intermediate were detected. All these examples demonstrate the prowess of bimetallic structures with reverse polarity for the activation and further functionalization of carbon dioxide, an approach whose development is still at a very early stage.

Figure 10

CO₂ activation by Al/M (M=Au, Cu, Ag and Zn) pairs (examples from Goicoechea and Aldridge's works). Dipp=2,6-diisopropylphenyl.

The groups of McMullin, Hill and Coles have built on the same principles of using highly unusual nucleophilic alumanyl synthons to access bimetallic structures with reversed polarity and thus remarkable reactivity, including the donor-acceptor activation of CO₂. First, Hill, McMullin and co-workers⁷⁰ used a seven-membered heterocyclic diamidoalumanyl species to prepare two heterobimetallic species featuring an Al−Cu bond with the same unusual nucleophilic character at the copper centre. Compounds 76 and 77 (Figure 11) result from the insertion of CO₂ into the corresponding Al−Cu bonds, where the copper binds the electrophilic carbon centre. Compound 77 was further characterized by X-ray diffraction studies exhibiting two bridging oxygen atoms (C−O bond distances of 1.301(3) and 1.307(3) Å) and an acute O−C−O angle of 111.5(2)°. More recently, McMullin and Coles groups described other two related examples based on a related alumanyl motif as precursor towards Al−Zn and Al−Mg heterobimetallic complexes. Once more, these compounds insert CO₂ into the metal-metal bonds to produce the corresponding dioxocarbene structures.⁷¹ However, the two compounds exhibit a peculiar divergent reactivity. Compound 78 could be isolated and fully characterized, and alike 76 and 77, it presents two bridging oxygen atoms that coordinate in a bidentate manner to the Al centre with two, almost equal, C−O bond distances of 1.2917(18) and 1.2970(18) Å. In stark contrast, the analogous species based on an Al−Mg core could not be detected and immediately evolved towards the carbonate structure 79 (Figure 11). After a series of experimental and computational mechanistic investigations, the authors could support a μ-oxo intermediate (from CO extrusion from the dioxocarbene), as the central intermediate towards the formation of the carbonate structure by insertion of a second molecule of CO₂, a process that does not take place for the analogous Al/Zn species 78.

Figure 11

Other CO₂ insertion products in Al/M (M=Cu, Zn, Mg) complexes. Dipp=2,6-diisopropylphenyl; Mes=2,4,6-trimethylphenyl.

All prior systems operate through conventional two-electron pathways for the donor-acceptor activation of CO₂. Nonetheless, in a very interesting work Mankad and co-workers recently disclosed a cooperative Al/Fe system capable of activating CO₂ by an alternative and intriguing radical pair mechanism.⁷² They described an heterobinuclear system with an Fe−Al bond that dissociates homolytically generating metalloradical intermediates which cooperatively reduce carbon dioxide by donating one electron per metal. The authors used as their heterobimetallic precursor the adduct L^Dipp(Me)Al-FeCp(CO)₂ (L^Dipp=HC{(CMe)(2,6-i-Pr₂C₆H₃N)}₂, formed by a β-diketiminate-supported Al centre bonded to a FeCp(CO)₂ fragment. Complex 80 forms after the addition of 1 bar of CO₂, exhibiting a penta-coordinated aluminium centre with distorted square pyramidal geometry (Figure 12). The C−O bond lengths for the CO₂ moiety in 80 (1.271(2) and 1.298(2) Å) fall between the typical ranges for C−O single and double bonds. Although this species is thermally stable and the reaction is not reversible by applying vacuum, it reacts under UV light. The resulting product 81 (Figure 12) was characterized by X-ray diffraction studies, revealing now a tetracoordinated aluminium site due to methyl migration to the β-diketiminate backbone, hence making it a dianionic ligand.

Figure 12

Cooperative CO₂ activation by an Al/Fe system (Dipp=2,6-diisopropylphenyl).

In some cases, the donor-acceptor activation of CO₂, characterized by the elongation of at least one C−O bond, is a prelude of its subsequent reductive cleavage towards CO and a, typically bridging, oxo ligand. A recent example was reported by Camp and co-workers in the reaction between a highly polarized Al(III)^δ+−Ir(III)^δ− complex and CO₂, which resulted in the decarbonylation of carbon dioxide obtaining CO, which was trapped at the Ir centre (acting here as a nucleophile) forming a Cp*Ir(H)₂(CO) (Cp*=C₅Me₅) species. Another Ir compound was detected in the reaction (Cp*IrH₄), as well as the alkylaluminum oxo coproduct [(ⁱBu)(OAr)Al(Py)]₂(μ-O) (Oar=3,5-di-tert-butyl-4-hydroxytoluene; Py =NC₅H₅).⁷³ The reaction mechanism was investigated by DFT and involves the participation of intermediate species exhibiting metal-metal interactions, with compound 82 in Figure 13 being a key transition state. It is important to highlight that unlike previous examples that imply changes of the oxidation state of low-valent aluminium atoms, the obtention of an Al-oxo compound without having to isolate the low-valent aluminium derivative is highly unconventional.

Figure 13

Reductive C−O bond cleavage through cooperative CO₂ activation by Al/Ir and Al/Fe bimetallic pairs (OAr=3,5-di-tert-butyl-4-hydroxytoluene).

Another example of the scarce group of bimetallic compounds based on Fe and Al capable of activating CO₂ was recently published by our group following an intermolecular bimetallic FLP approach. In this work we described the first example of a compound with an oxo-bridge between these metals. Complex 83 was obtained by the addition of 1 bar of CO₂ to a solution of Fe(depe)₂N₂ (depe=1,2-bis(diethylphosphino)ethane) and Al(C₆F₅)₃,⁷⁴ achieving both the cleavage of the C−O bond and the formation of a Fe(CO) unit and a Fe(μ-O)Al fragment with the metals in an almost ideal trans disposition. The Fe−O−Al bond angle is almost linear, 171.9(4)°, and the O−Al and Fe−O bond distances are 1.685(7) and 2.007(7) Å, respectively, being the latter larger than the distances reported for oxo-bridged diiron complexes, which could be explained by the high acidity of the alane fragment. Computational studies support the initial formation of a push-pull activated CO₂ adduct by the Lewis basic Fe(0) centre and the acidic Al(III) site, which weakens one of the C−O bonds to the extent of being readily cleaved without intermediate detection.

From the examples described above it is obvious that there are plenty of opportunities to explore the still nascent chemistry of aluminium/transition metal systems for the donor-acceptor activation of CO₂ and its further functionalization. Despite this remarkable reactivity, stronger focus has so far been given to transition metal/borane pairs, most likely due to the wider availability and synthetic amenability of boron-based reagents compared to their aluminium counterparts. In fact, the commercially available and well-known B(C₆F₅)₃ has been amply used for these endeavours. One example is the work published by Lee and co-workers in 2014 where they described the addition of CO₂ to a dinuclear nickel-dinitrogen species, releasing the N₂ molecule and forming a monometallic complex with one η²-bound CO₂ molecule.⁷⁵ After the addition of tris(pentafluorophenyl)borane, compound 84 (Figure 14) was obtained where an elongation of the C−O bond is observed (1.340(4) Å; cf. 1.252(2) Å in the Ni precursor) due to the presence of the Lewis acid that interacts with one of the oxygen atoms of the CO₂ unit, while the electrophilic carbon of the heteroallene forms a Ni−C bond (1.923(3) Å) giving a tetracoordinated Ni centre.

Figure 14

Donor-acceptor activation of CO₂ by B(C₆F₅)₃/TM (TM=Ni, Re, Pt, Zr, Mo, Fe) systems.

In another example that also follows a metallic FLP approach Berke's group reported the reaction of a Re-hydride compound with tris(pentafluorophenyl)borane under CO₂ atmosphere, which generates complex 85 (Figure 14).⁷⁶ Here the Lewis acid is interacting with one of the oxygen atoms of the heteroallene and the Re−H, acting as the Lewis base, binds to the CO₂ fragment with η²-coordination. The stability of this complex is limited and even under 1 bar of CO₂ it gradually evolves and, therefore, characterization was only performed by spectroscopic techniques in solution and no X-ray data is available. The interaction between carbon dioxide and the Re−H/borane pair is weak and the capture of CO₂ is reversible as exposure to N₂ or vacuum regenerates the starting precursors. Importantly, this system was used for the catalytic reduction of CO₂ with either Et₃SiH or H₂ in the presence of bases, highlighting the potential of using donor-acceptor combinations involving a transition metal for the catalytic functionalization of challenging substrates such as CO₂.⁷³ Moreover, the effect of adding a Lewis acid in transition metal catalysed reduction of CO₂ has been well recognized in the past,⁷⁷ in many cases involving the stabilization of key intermediates by push-pull forces.

Once more, following a metallic FLP approach, the Wass group described in 2015 the synthesis of complex 86 (Figure 14) by adding tris(pentafluorophenyl)borane to a basic Pt(0) compound under CO₂ atmosphere (1 bar) after 3 days.⁷⁸ As anticipated, the Pt centre coordinates to the electrophilic carbon through an η²-binding mode with the C−O bond, while the borane links the external oxygen centre. Comparing the C−O bond lengths in the product (1.262(3) and 1.273(3) Å) with the free CO₂ distances (1.155(1) Å),⁵⁴ an elongation of more than 0.1 Å occurs due to the cooperative behaviour of the two FLP fragments. In another example, Agapie and co-workers followed the alluded CO₂ activation strategy with a Mo/B pair reaching complex 87 by two different molecular pathways:⁷⁹ by addition of tris(pentafluorophenyl)borane to the already formed Mo(0) precursor with a carbon dioxide molecule attached to the metal centre; or by adding CO₂ to a Mo/B complex with a dinitrogen molecule coordinated between the molybdenum and boron centres that is released after exposure to carbon dioxide. The final product (87) presents a bridging carbon dioxide coordinated in a 1κO¹ : C¹ : 2κO² fashion. The C−O bond length of the carbonyl interacting with the metal is 1.246(1) Å, whereas the other oxygen binds to the boron centre with a B−O distance of 1.554(1) Å, with concomitant elongating of the corresponding C−O bond distance to 1.275(1) Å.

Our group has also contributed to this set of examples involving transition metal-boron FLPs for the push-pull activation of CO₂. We were able to confirm by X-ray analysis the formation of product [(depe)₂Fe(μ-CO₂)B(C₆F₅)₃] where the activated CO₂ is again defined by a μ-CO₂-1κO¹ : C¹ : 2κO² coordination (complex 88, Figure 14).⁷⁴ The push-pull activation causes the elongation of the boron-bound C−O bond (1.305(4) Å) compared to the moiety connected to the metal (C−O, 1.249(3) Å). At variance with the analogous iron/alane system (compound 83 in Figure 13), this complex does not evolve by means of C−O bond cleavage, as supported by DFT analysis. Although both the alane and borane share the same push-pull intermediate, only in the former case the subsequent transformation is exergonic, evidencing the importance of tuning the Lewis acidic partner in CO₂ donor-acceptor activation systems to modulate reactivity. Moreover, we described an additional CO₂ donor-acceptor activation product using instead Zn(C₆F₅)₂ as the Lewis acid, which resulted in a dissimilar configuration. Complex 89 (Figure 15) exhibits a CO₂ molecule bridging the two metals, but now the electrophilic Zn centre binds the two oxygen atoms of the heteroallene, forcing an acute O−C−O angle of 114(2)° and the C−O bonds elongates to an average value of 1.29 Å.⁷⁴ This structure is highly unusual as a molecule of N₂ also coordinates to the Fe centre, being the first organometallic species structurally characterized in which both N₂ and CO₂ molecules are linked to the same metal.

Figure 15

Push-pull activation of CO₂ by a Fe(0)/Zn(II) FLP.

4 Activation of CO₂ via Polarized Transition Metal Heterobimetallic Systems

In this section we will describe the still limited number of heterobimetallic transition metal complexes with polar metal-metal bonds that are capable of activating CO₂ by polarising the C−O bond via push-pull interactions. Besides, we will also comment on a series of heterobimetallic structures in which the CO₂ molecule is stabilized by similar donor-acceptor bonding between the metallic fragments whose origin does not involve the addition of gaseous CO₂, but the latter instead forms in situ by ligand transformations or by the addition of other external substrates. At variance, other polar heterobimetallic systems in which one of the transition metals acts as a metalloligand without direct involvement in the bonding with CO₂, despite being highly interesting, lie out of the scope of this review as they rely on a different approach.⁸⁰ Relevant structural parameters of the compounds mentioned below have been included in Table S3. The first two examples of CO₂ activation by polar heterobimetallic species appeared in the 90s with the work of Cutler et al.⁸¹ and Bergman et al.⁸² who independently reported the formal insertion of CO₂ into highly polarized metal-metal bonds. In the first example (90, Figure 16) the reaction of a M−Zr (M=Fe, Ru) compound with CO₂ leads to a new complex with a proposed bidentate coordination where both O atoms bind to the Zr centre whereas the carbon atom forms a new M−C bond with the nucleophilic metal. No X-ray diffraction structure data was reported to confirm the proposed configuration. ¹³C labelled NMR and IR spectroscopic techniques confirmed the presence of the carboxylate group, although the ¹³C labelled analysis suggests an equilibrium between the CO ligands and the carboxylate group. The stabilization of these species was attributed to the push-pull effect of the electron-rich Fe/Ru fragment and the oxophilic Zr moiety. In terms of mechanism, the authors considered both the prior ionization into [M]⁻ and [Zr]⁺ ions, which would then act as a bimetallic FLP for CO₂ activation or, alternatively, a direct insertion of CO₂ into the M−Zr bond.⁸⁰ Spectroscopic and reactivity studies prompted the authors to favour the latter mechanism.

Figure 16

Carbon dioxide insertion into Fe−Zr and Ru−Zr bonds.

In another key example, Bergman's group succeeded in authenticating by X-ray diffraction studies the insertion of a C=O bond of CO₂ across a strained and polarized Ir−Zr bond, which yielded compound 91 (Figure 17).⁸² The coordination differs from that in 90 and shows one Ir–C bond and one Zr−O bond, maintaining one terminal CO moiety that does not directly interact with any of the metals. IR stretching absorptions at 1569 and 1015 cm⁻¹ were recorded for the bridging CO₂ group, although no NMR studies were reported. The C−O bond lengths are 1.23(1) Å for the terminal CO, while a longer C−O bond length of 1.31(1) Å was measured for the bridging CO group. The angle of the O−C−O moiety is bent to 122.3(9)° and the Ir−Zr bond elongated to 2.9022(4) Å (cf. 2.598(2) Å in the Ir/Zr precursor). Although the authors did not conduct further experiments to determine the mechanism of CO₂ activation, it was suggested that it proceeds through initial coordination to the electrophilic Zr site followed by subsequent intramolecular insertion of the C=O bond into the strained Ir−Zr bond.

Figure 17

Addition of carbon dioxide across a Zr−Ir bond.

Although the donor-acceptor approach is not as evident as in the two previous systems, the group of Caulton reported slightly earlier another highly significant result in the area of heterometallic activation of CO₂ based on a Rh/Os bimetallic precursor.⁸³ Thus, exposure of (COD)RhH₃Os(PMe₂Ph)₃ (COD=1,5 cyclooctadiene) to CO₂ atmosphere (1 bar) led to the formation of the trimetallic compound 92, along with water and H₂Os(CO)(PMe₂Ph)₃ (Figure 18). The latter species derives from the reduction of one molecule of CO₂ towards carbon monoxide, as demonstrated later by isotopically labelling experiments,⁸⁴ while a second molecule of CO₂ is trapped by donor-acceptor interactions in 92. The carbon atom is bound to the Os site featuring an Os−C bond distance of 2.06(2) Å, while each oxygen centre binds one rhodium atom (2.06(1) and 2.06(2) Å). Interestingly, a fourth metal could be incorporated by adding ZnBr₂, with the final tetrametallic structure 93 being corroborated by X-ray diffraction studies. The most distinctive variation upon this extra electrophilic activation of CO₂ is the shortening of the Os−C bond, which the authors attribute to the more carbenic character of the carbon site, as well as the higher bending of the O−C−O angle from 116.3(16)° (92) to 112.2(10)° (93).

Figure 18

Addition of carbon dioxide across a Zr−Ir bond.

The previous examples of this section are based on heterometallic structures based on singly bonded metals. Nonetheless, it is also possible to access highly polarized multiply bonded structures with potential reactivity towards CO₂. In fact, the Thomas’ group reported in 2011 the first C−O bond cleavage of CO₂ over this kind of motif by using a multiply bonded Zr/Co bimetallic complex to yield compound 94 (Figure 19).⁸⁵ Later, they expanded this reactivity to other Zr/M pairs, where M is an alkali metal atom (M=Li, K) and where the activation of CO₂ was also examined.⁸⁶ Exposure of the phosphinoamide-bridged heterobimetallic Zr/Co complex to CO₂ atmosphere results in the oxidative addition of CO₂ across the multiple metal-metal bond, generating a terminal CO and a bridging Zr−O−Co oxo unit. The final paramagnetic compound was characterised by X-ray diffraction studies, naturally exhibiting an elongated Co⋅⋅⋅Zr distance of 2.89 Å compared to the parent heterobimetallic complex (2.14(1) Å),⁸⁷ thus revealing almost no interaction between the metals. Other relevant distances are: Zr−O, 1.8764(18) Å and Co−O, 1.9161(19) Å, for the bridging oxo group; and 1.783(3) Å for the terminal CO. A characteristic IR signal was also observed at 1926 cm⁻¹ attributed to the terminal CO group. The highly polarized Co→Zr bond facilitates the extreme activation of the CO₂ molecule which leads to the cleavage of the C−O bond.

Figure 19

CO₂ activation by a heterobimetallic Zr/Co complex. Mes=2,4,6-trimethylphenyl.

The CO₂ to CO interconversion is well known to be performed in nature by the CO dehydrogenase (CODH) enzymes which present, among others, Ni/Fe metallic cores.¹⁸ There has been a long-standing interest in mimicking the natural systems by artificial approaches.⁸⁸ One of these attempts led by Lee allowed the isolation of the first genuine heterobimetallic complex featuring a Ni(μ-CO₂)Fe core (compound 95 in Figure 20),⁸⁹ thus emulating the active site of the aforesaid CODH enzymes. In this and the rest of the examples discussed in this section, the trapped CO₂ molecule has not been added as an exogenous substrate. Instead, it has been generated chemically by reaction between the ligands and other added substrates. Nonetheless, although these systems cannot be considered as proficient per se in the activation of CO₂, we believe it is of interest to still discuss them as they represent key structures in which a molecule of carbon dioxide is trapped by heterobimetallic push-pull forces. In Lee's work, a preformed Ni-formate complex was reacted with a pincer PNP-type iron(II) chloride compound releasing NaCl and yielding the aforesaid bridging-CO₂ structure 95. The activation of CO₂ effected by the Lewis acidic Fe(II) centre is evidenced by the elongation of the C−O bonds to 1.269(2) and 1.289(2) Å, compared to those distances in the parent Ni-formate precursor (1.247(2) and 1.248(2) Å). Other geometric parameters, spectroscopic analysis and computational studies support the notion of a high similarity of this bimetallic core with that found in CODH enzymes.

Figure 20

Transition metal heterobimetallic complexes exhibiting donor-acceptor trapping of CO₂ in which the latter molecule is not added as a substrate but chemically produced within the bimetallic structure. Tol=4-methylphenyl.

Earlier examples of bridged CO₂ complexes, where the carbon dioxide moiety comes from a substrate different than an external CO₂ molecule, were reported in the 90s. Gibson and coworkers described Re/Fe examples of μ₂-κ² and μ₂-κ³-CO₂ bridged compounds where the carbon dioxide unit comes from the parent [(C₅H₅)Fe(CO)(PPh₃)(CO₂)]⁻K⁺. Reaction of the latter with rhenium cations, Re(CO)₄(L)(BF₄) (L=PPh₃, P(OPh)₃, CO) results in the formation of μ₂-κ² complexes eliminating KBF₄. Structure 96 (L=PPh₃) was characterized by X-ray diffraction studies (Figure 20) showing C−O distances of 1.298(3) and 1.226(3) Å.⁹⁰ Thermolysis of 96 afforded a new compound with a μ₂-κ³ coordination mode for the CO₂ moiety. Characterization by X-ray diffraction analysis was obtained for 97 (L=P(OEt)₃), exhibiting longer C−O bond lengths (1.322(8) and 1.274(8) Å) compared to the μ₂-κ² species and a O−C−O angle of 111.3(6)°.⁹¹ Similar bimetallic compounds based on Re/M (M=Fe, Mo, W) cores and originating from rhenium metallocarboxylates have also been synthesized and characterized.⁹²

The μ₂-κ³ coordination showed in 97 had already been described by Gibson for an iron-tin pair⁹³ and by Geoffroy and coworkers for rhenium-tungsten compounds.⁹⁴ Oxo complexes, (C₅H₅)₂M=O (M=Mo, W), reacted with cationic species [Cp'M(CO)₂(NO)]⁺ (Cp’=C₅H₄Me, C₅Me₅; M=Mn, Re) or [Cp'M(CO)₃]⁺ (M=Fe, Ru), to form a μ₂-κ³ CO₂ ligand via a [2+2] cycloaddition of the M=O bond across a C=O bond of the cationic carbonyl complex.⁹⁵ Compound 98, (C₅Me₅)(CO)(NO)Re(μ₂-κ³-CO₂)W(C₅H₅)₂][BF₄], was fully characterized including X-ray crystallography indicating that the carbon dioxide moiety is best described as a dimetalated dioxycarbene ligand with a Re=C double bond (2.04(4) Å) and with single C−O (av. 1.33 Å) and W−O (av. 2.09 Å) bonds. Other [2+2] cycloaddition reactions of a Ti=O bond across a C=O bond of a carbonyl group forming similar bimetallic compounds to 98 were also described.⁹⁶ Heterobimetallic bis(carbon dioxide) complexes with a Re₂Rh₂-(μ³-CO₂)₂ core were established by X-ray structure determination. In compound 99, [(η⁵-C₅Me₅)(CO)(NO)Re(CO₂)Rh(η⁴-COD)]₂, each square planar Rh centre is linked to two COD (COD=1,5-cyclooctadiene) ligands and two oxygen atoms, one from each Re carboxylate group.⁹⁷ Highly polarized Zr/Ru complexes exhibit oxygen transfer reactivity from sulfoxides to a carbonyl ligand giving the corresponding thioether and a bridging CO₂ group between the two transition metal atoms (100, Figure 19).⁹⁸ Reaction of Me₂SO with [MeSi(SiMe₂NTol)₃Zr−Ru(C₅H₅)(CO)₂] (Tol=4-methylphenyl) yields complex [MeSi(SiMe₂NTol)₃Zr(μ-O₂C)RuCp(CO)(SMe₂)] that reacts with a second molecule of the sulfoxide to give a higher crystalline product, 100, in which the second Me₂SO molecule is bonded to the Zr centre. The geometrical characteristics of the bridging carboxylate groups (C−O distances: 1.276(11) and 1.297(11) Å; O−C−O angle: 115.9(8)°) are similar to those found for other Zr/Ru examples previously reported by Gibson.⁹⁹ Similar reactivity was observed for highly polarized ketones (1,2-diphenylcyclopropenone) instead of sulfoxides, where the oxygen transfer occurs from the C=O bond of the ketone to the CO of a carbonyl group. The obtained product is a metallocarboxylato Zr/Fe complex, [HC{SiMe₂N(2,3,4-F₃C₆H₂)}₃Zr(μ-CO₂)FeCp(CO)(C₃Ph₂)], with an analogous bridging μ-CO₂ to the one found for structure 100. Nonetheless, it exhibited shorter Zr−O bond distances compared to the Zr/Ru example (99), probably due to the higher degree of electronic unsaturation of the Zr in the Zr/Fe pair.¹⁰⁰

5 Summary and Outlook

The exploitation of donor-acceptor interactions has proven to be a successful approach in overcoming the inertness of carbon dioxide. This approach is not limited to artificial synthetic designs, as enzymes have evolved to synergistically combine Lewis acidic and basic sites for CO₂ activation. Both natural and artificial systems involve a nucleophilic fragment binding to the electron-deficient central carbon of CO₂, while one or two electrophiles coordinate to the electron-rich oxygen centre(s), causing elongation of the C−O bonds and bending of the O−C−O angle. The resulting structural motifs offer a diverse range of possibilities for further functionalization of the activated CO₂ molecule.

This review highlights the versatility of this approach, as both metals and non-metal atoms from across the periodic table can serve as Lewis acidic and basic fragments to accommodate donor-acceptor binding of CO₂. Importantly, the roles of Lewis acid and base are not exclusive to any element, as evidenced by reverse reactivity for CO₂ activation, such as gold(I) or copper(I) atoms acting as the nucleophilic sites.

While thermodynamically favorable, the high stability that is often found for CO₂ donor-acceptor adducts presents a challenge for achieving catalytic turnover. Nonetheless, examples of catalytic conversion following donor-acceptor activation have been reported, including metal-free systems through frustrated Lewis pair designs. Nevertheless, we believe that the incorporation of at least one transition metal with its set of partly filled d-orbitals and elementary organometallic reactions is expected to facilitate the development of catalytic processes relying on the initial push-pull activation of CO₂.

The fact that adding external Lewis acids enhances the catalytic performance of numerous CO₂ reduction catalysts encourages further research to fully understand these effects and hence for the design of better catalytic cooperative partnerships. Moreover, important synthetic efforts will be required to better balance the equilibrium between too weak activation of CO₂ and the formation of too stable donor-acceptor CO₂ adducts, as well as to understand the factors that influence further reactivity such as C−O bond cleavage or subsequent C−C bond coupling. Computational methods can assist in designing and optimizing donor-acceptor systems to facilitate catalyst development.

Overall, the activation of CO₂ by donor-acceptor interactions has already revealed a landscape of opportunities and certainly holds great promise for the development of sustainable technologies based on CO₂ functionalization. Nonetheless, further research in this area will be crucial in advancing efficient catalytic protocols for transforming CO₂ into value-added products. The next decade promises to be an exciting period for witnessing these developments.

Acknowledgments

This work was supported by the European Research Council (ERC Starting Grant, CoopCat, Project 756575), the Spanish Ministry of Science and Innovation (PID2019-110856GA−I00) and Junta de Andalucía (P18-FR-4688). M. P.-J. and F. de la C.-M. acknowledge the Spanish Ministry of Universities for Margarita Salas postdoctoral fellowships. We also thank Dra. M. Roselló for helpful discussions.

Conflict of interest

The authors declare no conflict of interest.

Open Research

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.

Biographical Information

Marina Pérez Jiménez completed her PhD in Organometallic Chemistry in 2021, under the supervision of Ernesto Carmona and Jesús Campos, focusing on the reactivity of dimolybdenum complexes with quadruple metal-metal bonds. In 2019, she joined the group of T. Don Tilley in University of Berkeley, California, for a visiting stay. Since 2022, she works as a postdoctoral researcher at Imperial College London, under the supervision of Mark R. Crimmin, funded by a Margarita Salas Fellowship.

Biographical Information

Helena Corona García de Leaniz studied for a Chemistry Degree and Master at the University of Seville until 2022. Since this same year, she is currently doing her PhD in Organometallic Chemistry under the supervision of Jesús Campos, focusing on small molecule activation via cooperative interaction of bimetallic compounds.

Biographical Information

Felipe de la Cruz-Martínez studied Chemistry at University of Extremadura. He moved to the University of Castilla-La Mancha where obtained his PhD on chemical valorization of CO₂ in the group of Prof. Agustín Lara. He also did a PhD stay at the University of York in the group of Prof. Michael North focused on the synthesis of CO₂-based carbamates. In 2021, he joined the group of Dr. Jesús Campos as a postdoctoral fellow working on the development of bimetallic pairs for the activation of small molecules. Since September 2022, he is based at the University of Castilla-La Mancha as an Assistant Professor.

Biographical Information

Jesús Campos graduated from the University of Sevilla and carried out his Master research at the University of Manchester. He obtained his PhD (2012) in the group of E. Carmona, spending a visiting stay with M. Brookhart (UNC). As a postdoctoral researcher he joined the universities of Yale (R. Crabtree) and Oxford (S. Aldridge). In 2016 he moved back to Sevilla and a year later became tenured scientist at CSIC. Since then, he has focused on developing organometallic systems with a strong focus on unusual cooperative processes under the umbrella of an ERC Starting Grant.