Carbon dioxide binding at a Ni/Fe center: synthesis and characterization of Ni(η¹-CO₂-κC) and Ni-μ-CO₂-κC:κ² O,O′-Fe

A heterobimetallic Ni-μ-CO₂-κC:κ² O,O′-Fe species reminiscent of the CODH active site was synthesized, helping to elucidate the role of the unique iron.

Abstract

The degree of CO₂ activation can be tuned by incorporating a distinct electronic coordination environment at the nickel center. A mononuclear nickel carboxylate species (Ni–CO₂, 3) and a dinuclear nickel–iron carboxylate species (Ni–CO₂–Fe, 5) were prepared. The structure of 3 reveals a rare η¹-κC binding mode of CO₂, while that of 5 shows bridging CO₂ binding (μ₂-κC:κ² O,O′) between the nickel and iron, presented as the first example of a nickel-μ-CO₂-iron species. The structural analyses of 3 and 5 based on XRD and DFT data reveal a higher degree of CO₂ activation in 5, imparted by the additional interaction with an iron ion.

Introduction

Activation of carbon dioxide is currently receiving much attention due to its relevance to environmental and energy related issues.¹ In the area of transition metal catalyzed reactions, one of the main challenges is selective reduction of CO₂ to a product such as formate, carbon monoxide, methanol or methane.² In a 2-electron process, the binding mode of the CO₂ may determine the eventual product formation, e.g. formate vs. carbon monoxide.³ When the initial metal–oxygen interaction occurs to form a metal CO₂ adduct M-η¹-CO₂-κO, subsequent hydride transfer via CO₂ addition to a M–H bond generates a metal-formate species. Alternatively, the metal–carbon bond formation can produce a metallacarboxylate species (M-η¹-CO₂-κC), followed by C–O bond cleavage to generate CO. In the latter case, an additional Lewis acid interaction can stabilize the negative charges at the oxygen atoms of the bound CO₂.^4,5 Therefore, CO₂ activation with a bimetallic system can be one way to guide the selectivity of the CO₂ catalyst and is receiving much attention.^5,6 In fact, an excellent example of a bimetallic center utilized in an efficient catalytic conversion of CO₂ can be found in the active site of carbon monoxide dehydrogenase (CODH).⁷ According to recent studies, CO₂ coordination at a heterobimetallic nickel–iron active site can be found in CODH’s intermediate species, which possesses a Ni-μ-CO₂-Fe moiety, Scheme 1.⁸ Although X-ray analysis provides a structural snapshot of the CO₂ reduction sequence, the role of the unique iron ion is currently not well-understood.⁷ Thus, acquiring an understanding of iron assisted CO₂-nickel coordination is of fundamental interest and is crucial for gaining mechanistic insight into this and other enzymatic reactions.

Scheme 1. The active site of carbon monoxide dehydrogenase (CODH, left), and 4- and 5-coordinate nickel CO₂ adducts (right).

In organonickel chemistry, there are few mononuclear Ni-η²-CO₂ adducts possessing both M–C and M–O bonds.⁹ In 1975, Aresta and co-workers reported the first structurally characterized nickel–CO₂ adduct (PCy₃)₂Ni(η²-CO₂), Scheme 1.^9a An analogous complex, (dtbpe)Ni(η²-CO₂) (dtbpe = 1,2-bis(di-tert-butylphosphino)ethane) was recently reported by Hillhouse and co-workers.^9d More recently, our group reported a similar but unique five-coordinate nickel–CO₂ adduct (PP^MeP)Ni(η²-CO₂) (PP^MeP = PMe(2-PⁱPr₂-C₆H₄)₂).^9f According to its structural analysis, the five coordinate nickel CO₂ species supported by three neutral P donors has a weak Ni–O bond available for electrophilic attack.^9f Additionally, by utilizing an anionic tridentate PNP ligand (PNP^– = N[2-PⁱPr₂-4-Me-C₆H₃]₂ ^–), our group reported the nickel hydroxycarbonyl species (PNP)NiCOOH (1), (PNP)NiCOONa (2) and {(PNP)Ni}₂-μ-CO₂-κ² C,O (4), the first examples of Ni–CO₂ complexes that reveal a Ni–CO₂-κC binding mode, Scheme 2.¹⁰ The carboxylate group in these species is stabilized by a Lewis acid such as a proton, sodium or another nickel ion. Our interest then moved to comparing (PP^MeP)Ni–CO₂ and (PNP)Ni–CO₂ to evaluate their fundamental differences in CO₂ activation. The different geometries favored with a PP^MeP or PNP ligand affect the identity of the nickel–CO₂ moiety, which can be Ni(ii)–CO₂ ^2– or Ni(0)–CO₂ or an open-shell Ni(i)–CO₂˙^–, vide infra. Furthermore, by isolating the native nickel–CO₂ species, we can further study the effect of the second iron ion. Although several nickel carboxylate species are already known, iron has never been introduced synthetically into a Ni–CO₂ moiety.

Scheme 2. Preparation of mononuclear- and dinuclear-CO₂ adducts.

Here, we present a nickel carboxylate species {Na(12-C-4)₂}{(PNP)Ni-η¹-CO₂-κC} (3), in which the nickel–CO₂ moiety does not have any Lewis acid interactions. We also prepared a dinuclear nickel–iron carboxylate species (PNP)Ni-μ-CO₂-κC:κ² O,O′-Fe(PNP) (5), reminiscent of the NiFe-binuclear active site of CODH. This is an unprecedented example of a nickel–iron hetero-bimetallic complex possessing a bridging CO₂ ligand. The levels of CO₂ activation in compounds 3 and 5 are compared with other Ni–CO₂ adducts and the Ni-μ-CO₂-Fe moiety found in CODH.

Results and discussion

Synthesis and characterization of the Ni-η¹-CO₂-κC complex

The coordination of a hydroxycarbonyl moiety via a Ni–C bond was previously realized at a divalent nickel center supported by a PNP ligand.¹⁰ Following deprotonation of (PNP)NiCOOH (1), its anionic congener (PNP)NiCOONa (2) was also prepared and recently reported by our group.¹⁰ The X-ray structure reveals that two molecules of 2 are oriented to form a pair with ionic interactions with two sodium ions in the crystal lattice, Fig. 1.¹¹ The corresponding CO₂ ligand coordinates to the nickel center in a μ₃-κ¹ C:κ² O,O′:κ¹ O′ mode with a Ni–C1 bond distance of 1.882(1) Å. There are additional bonds of the CO₂ moiety to sodium ions with Na–O bond distances of 2.352(1), 2.217(1) and 2.459(1).¹¹ To obtain a sodium-free adduct, 2 equiv. of 12-crown-4 was added to a solution of 2, resulting in the formation of {Na(12-C-4)₂}{(PNP)Ni-η¹-CO₂-κC} (3). The crystal structure of 3 revealed the successful generation of a mononuclear nickel adduct possessing an η¹-κC coordinated carbon dioxide species with a Ni–C bond distance of 1.911(2) Å, as shown in Fig. 1 and Table 1. The oxidation state of the nickel ion in 3 can be assigned as 2+ based on its similar structural features to previously known nickel(ii) species such as (PNP)NiCOOH (1) and {(PNP)Ni}₂-μ-CO₂-κ² C,O (4), vide infra. The geometry of 3 is square planar (τ ₄ = 0.12¹²) with a similar but slightly elongated Ni–C bond distance in comparison to those of 1 and 4 (d _Ni–C = 1.866(2) and 1.888(2) Å, respectively, Table 1). This is probably due to a lower degree of π back-bonding between the nickel and CO₂. In fact, π back-donation from the nickel center to a CO₂ ligand in such nickel carboxylate species is indicated by shorter Ni–C distances (1.858–1.911 Å) than those of the nickel alkyl species (PNP)NiR (R = Me, Et, ⁿPr) (1.963–2.004 Å).¹³ The molecular orbitals generated from DFT calculations also show the presence of π back-donation from the Ni d_xz to the CO₂ π* orbital (see ESI†). Due to the absence of Lewis acid interactions in 3, a lower π-accepting ability of the CO₂ ligand is expected. Its structural data also revealed that the plane of the CO₂ ligand is perpendicular to that of the square planar (PNP)Ni moiety. One of the oxygen atoms (d _Ni1–O2 = 2.614(1) Å) is slightly closer to the nickel center than the other (d _Ni1–O1 = 2.776(1) Å), Table 2. These Ni–O distances are much longer than those for other known Ni-η²-CO₂ adducts (1.9–2.2 Å, Table 2), suggesting that neither of the oxygen atoms are bound.⁹ The DFT analysis also supports minimal interaction between the nickel and oxygen atoms (Wiberg index = 0.1358 for Ni1–O2 and 0.1679 for Ni1–O1, see Table 2). The two C–O bond distances are nearly identical (d _C1–O1 = 1.247(2) Å, d _C1–O2 = 1.248(2) Å, Table 1) and slightly shorter than in the analogous carboxylate complexes 2 and 4 (Table 1), due to the absence of a Lewis acid, Na or Ni. According to the DFT analysis, the HOMO of 3 possesses contributions from both a nickel d_x²–y² orbital and a CO₂ π* orbital, see Fig. 2. Due to additional electron density from CO₂ ^2– being shifted to the nickel, the CO₂ moiety is slightly oxidized compared to the sp² hybridized carboxylate ligands found in 2 and 4. The larger O–C–O angle (128.4(2)°) of 3 compared to others (124.0(1) and 123.7(2)°) also supports this electronic feature, vide infra. Although an η¹-κC CO₂ coordination mode has been proposed for many CO₂ reduction strategies,^2,3 the only example of a crystallographically identified metal η¹-κC CO₂ complex is a rhodium CO₂ adduct, Rh(CO₂)(Cl)(diars)₂ (diars = o-phenylene-bis(dimethylarsine)), reported by the Herskovitz group.¹⁴ According to their C–O bond distances (1.20(2) and 1.25(2) Å) and O–C–O angle (126(2)°), the CO₂ moiety in 3 shares similar structural and electronic features. Thus, compound 3 is a unique example possessing η¹-κC CO₂ binding, since such a binding mode is unknown for 1^st row transition metals and is rare in structurally characterized metal–CO₂ adducts.

Fig. 1. Displacement ellipsoid (50%) representations for (a) (PNP)NiCOONa (2) in a dimeric assembly with co-crystallized THF molecules,¹¹ and (b) {Na(12-C-4)₂}{(PNP)Ni-η¹-CO₂-κC} (3). A co-crystallized 12-crown-4 molecule and hydrogen atoms are omitted for clarity.

Table 1. Selected bond distances and angles for the nickel carboxylate species 1, 2, 3, 4 and 5, and CO₂-bound CODH.

	1 10	2 11	3	4 10	5	CODH8b
d Ni–C (Å)	1.866(2)	1.882(1)	1.911(2)	1.888(2)	1.858(1)	1.805(31)
d M–O (Å)	—	2.352(1) a	—	1.897(2) b	2.204(1) c	2.030(18) c
		2.217(1) a			2.066(1) c
		2.459(1) a
d C–O (Å)	1.269(3)	1.260(1)	1.247(2)	1.240(3)	1.269(2)	1.298(30)
	1.313(3)	1.271(1)	1.248(2)	1.296(3)	1.289(2)	1.316(30)
Δd C–O (Å)	0.044	0.011	0.001	0.056	0.020	0.018
∠O–C–O (°)	119.6(2)	124.0(1)	128.4(2)	123.7(2)	116.5(1)	117.2(26)

^aM = Na.

^bM = Ni.

^cM = Fe.

Table 2. Selected physical parameters and bond indices from the natural bond orbital analysis.

	Ni(PCy3)2(η2-CO2)9a	(dtbpe)Ni(η2-CO2)9d	(PPMeP)Ni(η2-CO2)9f	3	5
Structural parameters
d Ni–C (Å)	1.84(2)	1.868(2)	1.904(1)	1.911(2)	1.858(1)
d Ni–O (Å)	1.99(2)	1.904(2)	2.191(1)	2.614(1)	2.718(1)
				2.776(1)	2.792(1)
d C–O (Å)	1.17(2)	1.200(3)	1.218(2)	1.248(2)	1.269(2)
	1.22(2)	1.266(3)	1.252(2)	1.247(2)	1.289(2)
∠O–C–O (°)	133	138.0(2)	135.1(1)	128.4(2)	116.5(1)
ν CO2 (cm–1)	1740	1724	1682	1620	1510
Wiberg bond indices a
Ni–C	—	0.5766	0.5286	0.6143	0.6277
Ni–O	—	0.4300	0.3117	0.1679	0.0798
				0.1358	0.0644
C–O	—	1.6927	1.6384	1.5112	1.3993
		1.4080	1.4701	1.4949	1.2933

^aWiberg bond indices were calculated using single-point calculations, for which geometries were obtained from the XRD data.

Fig. 2. Combination of the Ni d_x²–y² and CO₂ π* orbitals provides the DFT calculated HOMO of {Na(12-C-4)₂}{(PNP)Ni-η¹-CO₂-κC} (3).

Synthesis and characterization of the heterobimetallic nickel–iron CO₂ complex

To gain a better understanding of the role of the second metal ion in the CODH active site, we prepared a heterobimetallic nickel–iron carboxylate species possessing a Ni–CO₂–Fe fragment by addition of {(PNP)Fe}⁺ to a Ni-η¹-CO₂-κC species. To a yellow solution of {Na(12-C-4)₂}{(PNP)Ni-η¹-CO₂-κC} (3) in toluene, a purple solution of (PNP)FeCl was added. The immediate formation of a new orange species (PNP)Ni-μ-CO₂-κC:κ² O,O′-Fe(PNP) (5) was confirmed, using the ¹H NMR spectrum, from the absence of peaks for 3 and (PNP)FeCl and the presence of new paramagnetically shifted signals. The same product was also prepared by substitution of the sodium ion of (PNP)NiCOONa (2) with (PNP)FeCl. The solid-state structure of 5 clearly revealed a dinuclear nickel–iron complex with a bridging CO₂ ligand in the μ₂-κC:κ² O,O′ mode (Fig. 3). The Ni and Fe ions are separated by a distance of 4.3690(3) Å. The two C–O bond distances are 1.269(2) and 1.289(2) Å, revealing that a significant elongation has occurred due to the iron interaction compared to 3 (Table 1). The bond distances between the iron and both oxygen atoms are 2.204(1) and 2.066(1) Å. The nickel center possesses a square planar geometry (τ ₄ = 0.10¹²). The geometry around the iron is distorted square pyramidal (τ = 0.13,¹⁵ Fig. 3). The O1–C1–O2 angle (116.5°) reflects the sp² hybridization of the carboxylate ligand in 5. In fact, recent crystallographic data of CODH at atomic resolution (d _min = 1.03 Å) revealed that the bound CO₂ molecule (∠O–C–O = 117.2(26)°) is a carboxylate anion (CO₂ ^2–).^8b,8c Regarding the similarity between these angles, the carboxylate moiety in 5 might be close to CO₂ ^2–. The asymmetric vibration for CO₂ observed at 1510 cm^–1, which is similar to that observed for the dinickel carboxylate species (4) at 1518 cm^–1, also indicates a reduced state of CO₂. The effective magnetic moment of 5 was determined using the Evans' method (μ _eff = 4.95 μ _B in C₆D₆), which indicated an S = 2 spin state.¹⁶ According to DFT calculations, most of the spin density is located on the iron center (see ESI†). For CODH, the unique iron, Fe_u, was assigned as a high spin iron(ii) (ferrous component II, FCII) using Mössbauer spectroscopy,¹⁷ and a low-spin nickel(ii) was demonstrated using X-ray absorption spectroscopy (XAS).¹⁸ The current structural and spectroscopic analyses suggest that 5 might share a similar electronic structure to that found in CODH. Gibson classified the μ₂-κC:κ² O,O′ binding modes of CO₂ into two types according to the difference between the two C–O distances.¹⁹ Due to the two similar C–O distances of the CO₂ moiety, compound 5 (Δd _C–O = 0.020 Å) can be assigned as a class I complex.²⁰ In the dinickel CO₂ species (4), the CO₂ molecule is coordinated in a μ₂-κC:κO mode with the absence of a Ni2–O2 interaction (d _Ni2–O2 = 3.14(7)) and two different C–O bond distances (Δd _C–O = 0.056 Å). In CODH, CO₂ is coordinated in a μ₂-κC:κO fashion between the nickel and iron ions, but the two C–O bond distances are quite comparable (Δd _C–O = 0.018 Å), akin to the μ₂-κC:κ² O,O′ mode. This might be due to hydrogen bonding with the protein matrix, since both the CO₂ oxygens are hydrogen bonded to His93 and Lys563, respectively.⁸

Fig. 3. Displacement ellipsoid (50%) representation for (PNP)Ni-μ-CO₂-κC:κ² O,O′-Fe(PNP) (5). Hydrogen atoms are omitted for clarity.

Compound 5 is the first example of a dinuclear nickel–iron–CO₂ complex. While dinuclear CO₂ complexes mostly employ 2^nd and 3^rd row transition metals,¹⁹ several bimetallic iron carboxylates (Fe–CO₂–M, M = Ti, Zr, Sn, Re) have been reported.^5a,5c–e,21 However, such complexes typically possess an Fe–C bond rather than an Fe–O bond with CO₂. There have been numerous examples of nickel–iron bimetallic complexes reported for synthetic model studies of NiFe hydrogenase,²² but a bimetallic complex possessing a Ni-μ-CO₂-Fe moiety closely related to CODH chemistry is not known. The Holm group constructed a series of [NiFe₃S₄] cubanes as structural model complexes for the [NiFe₄S₄] core in CODH.²³ However, the installation of an iron species corresponding to the Fe_u in CODH and reactions involving CO and CO₂ have not yet been investigated. More recently, the Holm group also reported a bimetallic complex containing nickel and iron supported by a binucleating macrocycle.²⁴ With respect to CODH chemistry, bridging hydoroxo, cyanido and formato species have been generated, however, a Ni-μ-CO₂-Fe fragment had not yet been isolated.

Activation of CO₂ in 3 and 5

Previously known 4-coordinate Ni–CO₂ complexes possessing an η²-CO₂ binding mode have a formally zero-valent nickel center, Scheme 1.⁹ This suggests limited CO₂ activation in such species. Interestingly, the 5-coordinate nickel CO₂ adduct (PP^MeP)Ni(CO₂) also has a similar level of activation of CO₂ based on the C–O bond distances and the O–C–O angle, Table 2. Regarding CO₂ binding and activation, {Na(12-C-4)₂}{(PNP)Ni-η¹-CO₂-κC} (3) is a unique example. It is striking that a neutral pincer-type ligand PP^MeP (PP^MeP = PMe[2-PⁱPr₂-C₆H₄]₂) favors 5-coordinate η²-CO₂ coordination at a single nickel center, while 3 remains as a 4-coordinate species with η¹-CO₂ coordination.^9f Although the total number of Ni d-electrons and CO₂ π*-electrons in both 3 and (PP^MeP)Ni(CO₂) is the same, (PP^MeP)Ni(CO₂) can be considered as formally Ni(0)–(CO₂) while 3 can be better described as a Ni(ii)–(CO₂ ^2–) species. The asymmetric CO₂ stretching frequency for 3 is significantly shifted to a lower vibration, 1620 cm^–1, compared to those of the Ni-η²-CO₂ complexes (Table 2), which is evidence of a reduced CO₂ moiety in 3.⁹ This may be due to the influence of the trans atom: an anionic amide nitrogen vs. a neutral phosphorus atom. The anionic nitrogen in 3 electrostatically favors a divalent nickel center, while the neutral π-acidic P atom in the PP^MeP ligand favors a Ni(0) center. In fact, the PNP ligand typically stabilizes a square planar geometry while the PP^MeP ligand favors a pseudo-tetrahedral geometry. Thus, 3 prefers to accommodate a divalent nickel center while (PP^MeP)Ni(CO₂) prefers Ni(0). However, the reduction state of the CO₂ moiety in 3 is a little ambiguous according to the O–C–O angle. The O–C–O angle in 3 of 128.4(2)° is larger than those of an ideal sp² hybridized carbon (120°) and the other nickel(ii) carboxylate species 1, 2 and 4 (119.6(2)°, 124.0(1)° and 123.7(2)°, respectively, Table 1). The O–C–O angle of a CO₂ radical anion (CO₂˙^–) is suggested to be 133°,²⁵ which is fairly similar to those of the previously reported Ni-η²-CO₂ complexes, Table 2. Thus, the geometry of the CO₂ moiety in 3 may be thought of as being between a CO₂ radical anion and a carboxylate.

Upon addition of iron to compound 3, the CO₂ is further reduced to carboxylate (CO₂ ^2–). The C–O bond distances and O–C–O angle in 5 clearly show a 2-electron reduced state of the CO₂ moiety, Table 2. This was also indicated by the asymmetric CO₂ vibration observed at 1510 cm^–1, which is significantly lower than those of other CO₂ species and 3. The Wiberg bond indices nicely agree with the bond distances, Table 2. These analyses of a series of nickel–CO₂ compounds demonstrate how the degree of CO₂ activation can be tuned by incorporating a distinct electronic coordination environment at the metal center, and may have parallels to the efficient CO₂ conversion found in CODH.

In order to study further activation of the bound CO₂ via C–O bond cleavage, protonation of 3 and 5 was attempted. Our group previously reported that reversible C–O bond cleavage/formation occurs with a nickel hydroxycarbonyl species (1).¹⁰ From reaction of 3 with 1 equiv. of HBF₄·Et₂O, a nickel hydroxycarbonyl species (1) was produced and isolated with a 74% yield. A similar reaction of 3 with 2 equiv. of HBF₄·Et₂O resulted in the formation of a carbonyl species {(PNP)NiCO}{BF₄} in 75% yield, revealing that two sequential protonations can occur with compound 3 possessing a Ni-η¹-CO₂-κC moiety, which are key steps in the transformation of CO₂ to CO. Although protonation of compound 5 seems to produce 1 and {(PNP)NiCO}{BF₄}, unfortunately, their yields were not clear due to thermal decomposition of 5 and the generation of multiple products. Demetallation of the iron seems to be one of the decomposition processes.

Conclusions

In conclusion, the generation of unprecedented nickel–carbon dioxide adducts possessing a Ni–C bond accommodated by a (PNP)Ni scaffold was accomplished. A mononuclear CO₂ adduct {Na(12-C-4)₂}{(PNP)Ni-η¹-CO₂-κC} (3) and a dinuclear nickel–iron carboxylate species (PNP)Ni-μ-CO₂-κC:κ² O,O′-Fe(PNP) (5) were synthesized successfully. While the solid state structure of 3 revealed a rare η¹-κC binding mode, compound 5 was structurally characterized to reveal a unique class I type μ₂-κC:κ² O,O′ binding mode. This heterobimetallic CO₂ adduct is the first example of a nickel–iron carboxylate species, of which the structural and electronic features are reminiscent of those of the Ni-μ-CO₂-Fe fragment found in the C-cluster of CODH. Comparison of the η¹-CO₂-κC species 3 and dinuclear Ni-μ-CO₂-Fe species 5 with previously reported Ni–CO₂ adducts suggested that the CO₂ ligand can be stabilized and activated by interaction with the second metal. Protonation of 3 produces a nickel carbonyl species {(PNP)NiCO}{BF₄} via C–O bond cleavage, while the reactivity of 5 is limited. Further studies on incorporating a stable iron species and the subsequent reactivity toward protonation are currently underway.