Silylation of Ni−CO₂ ²⁻—Liberation of CO

Reaction of I with three equivalents of TMS‐OTf (Scheme 1) instantly led to [L^tBuNiOSiMe₃] (1), which was isolated as a microcrystalline green solid in 99 % yield. A single crystal X‐ray analysis revealed, that the siloxide ligand and the [L^tBu]⁻ co‐ligand form a Y‐shaped coordination sphere for the Ni center of 1 (Figure 1).

Figure 1

Molecular structure of 1. ^[59] H atoms are omitted for clarity. Selected bond lengths [Å] and angles [°]: Ni‐N1 1.893(1), Ni‐N2 1.906(1), Ni‐O 1.772(1), O‐Si 1.607(2); N1‐Ni‐O 136.25(5), N2‐Ni‐O 127.20(5).

Concomitant liberation of CO has been verified (qualitatively) via GC‐MS (21 % yield from headspace). Surprisingly, unlike in case of the reactions of [L^tBuNi(OCO‐κ ² O,C)]Li₃[N(SiMe₃)₂]₂, I, with H⁺ (or CO₂), ^[18] formation of [L^tBuNi^I−CO], III, was not observed to occur, although TMS‐OTf likewise is an electrophilic reagent, that should react similarly. It thus may be inferred that III is formed in a secondary process, for example, through the reaction of unconsumed I with CO in the systems I/H⁺ or I/CO₂, which is suppressed in case of TMS‐OTf, as the latter reacts with I at sufficiently high rate (Scheme 2). Indeed, we were able to show that I reacts with CO to give III (ν_CO=2014 cm⁻¹, see S6). It is thus reasonable to assume that the principal reaction of I with electrophiles E⁺ consists of the formation of Ni‐OE units with simultaneous generation of CO (Scheme 2).

Scheme 2

1) Formation of CO in the reaction of [L^tBuNiCO₂]⁻ with electrophiles (Eⁿ⁺). 2) Reaction of [L^tBuNiCO₂]⁻ with in situ liberated CO.

Liberation of CO from Ni centers after reduction of CO₂ has been discussed quite controversially in the last years, especially with regards to the functioning of the Ni,Fe‐CODH. It is still not clear, in which step of the mechanism CO is released and what the oxidation state of the Ni center is then. Scheme 3 shows two possible mechanisms, where mechanism E1 (Scheme 3, left) involves release of CO from a Ni^II state with subsequent reduction of Ni before CO₂ enters again and mechanism E2 (Scheme 3, right) first involves reduction to a Ni⁰‐CO state, followed by substitution of CO by CO₂.

Scheme 3

Possible pathways for CO liberation in the reaction of a Ni‐CO₂ complex with electrophiles E⁺ (blue; for example, H⁺, Me₃Si⁺).

In recent model studies with Ni‐PNP‐pincer complexes it was found that subsequent to CO₂ reduction at a Ni⁰ center protonation of the resulting Ni^II−CO₂ ²⁻ complexes led to Ni^II−CO complexes and only after consecutive reduction to Ni⁰ and addition of CO₂, CO was released (Scheme 3, middle right circle).[ ²⁴ , ²⁵ ] Recent DFT studies on the enzyme also support a mechanism where CO liberation takes place only when CO₂ enters the cycle (Scheme 3, E2), however, in this study Ni remained in the +1 oxidation state (not Ni⁰) upon reduction and CO was generated from this Ni^I−CO state. ^[26] Our findings contrast the abovementioned model studies in that CO is released directly from the Ni^II complex (Scheme 3, middle left) and thus supports the notion that in the Ni,Fe‐CODH reduction takes place after CO elimination or simultaneously. This matches the results of an X‐ray diffraction analysis performed for crystals of the reduced active state of CODH grown in the absence of CO, which led to a structure that was suggested as an appropriate model for this state. ^[27] Obviously, the ligand environment plays a crucial role in these transformations.

Bearing in mind that not many Ni^II‐CO complexes are known to date,[ ²⁵ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ ] due to the limited ability of Ni^II to stabilize such entities via backbonding, we next tested whether a Ni^II‐CO complex with the L^tBu− ligand system would be stable at all and reacted the cationic complex [L^tBuNi^II(OH₂)][B(Ar^F)₄] (2, see S3) with CO gas. A change of color from green to purple could be observed, and evaporation of all volatiles left behind a purple solid. A single‐crystal X‐ray analysis showed that indeed the H₂O ligand had been replaced by CO to generate [L^tBuNi^II(CO)][B(Ar^F)₄], 3, in 82 % yield (Figure 2). The structure determination further revealed that one of the aryl rings of L^tBu undergoes an η² coordination, thus completing a square planar coordination sphere of the Ni center, which also reflects in a diamagnetic ¹H NMR spectrum of a solution of 3 in CD₂Cl₂.

Figure 2

Molecular structure of 3. ^[59] H atoms and the [B(Ar^F)₄]⁻ counter ion are omitted for clarity. Selected bond lengths [Å] and angles [°]: Ni‐N1 1.817(3), Ni‐N2 1.922(3), Ni‐C36 1.836(4), O‐C36 1.125(5), Ni‐C1 2.072(3), Ni‐C2 2.230(3); N1‐Ni‐C36 162.1(1), N2‐Ni‐C36 102.2(1), Ni‐C36‐O 176.6(3).

Isolation of 3 shows that binding of CO to L^tBuNi^II is principally possible, but the initial product formed in the reaction between I and TMSOTf would be L^tBuNi(CO)(OSiMe₃), that is, a complex that can be derived from 3 by replacing the aryl donor in 3 by an anionic siloxide ligand with release of the counteranion. The failure to detect this product thus may suggest that binding of both CO and siloxide beside each other at the Ni center causes an unfavorable situation, leading to the formation of 1 with liberation of CO. To examine this hypothesis 3 was reacted with KOSiMe₃ in CD₂Cl₂. The color of the solution was observed to change from purple to yellow within minutes, and the ¹H NMR spectrum of the solution subsequently showed the complete absence of signals belonging to 3, while new signal sets were observed that indicated the formation of a mixture of products (see S4). One of those exhibited the characteristic signal set of 1, proving, that CO can indeed easily be replaced by ^‐OSiMe₃. Consistently, after removal of all volatiles a ν_CO stretching band ([L^tBuNi^I−CO]) was hardly detectable in the ATR‐IR spectrum of the product mixture, confirming that the reaction of I with TMS‐OTf leads to the generation of CO as a gas. Altogether these investigations thus show, that the contact of I with electrophiles leads to the direct formation of CO, which is of interest also considering that the CO₂ ²⁻ ligand in I is derived from formate: Liberation of CO from HOOCH is known to occur in reactions with strong acids, however, there are only rather few examples, where formate coordinated to early transition metals (Ti,[ ¹² , ²⁰ ] W ^[21] ) releases CO upon deprotonation.

As the CO₂ ²⁻ ligand in I is spanned between various metal centers (and as mentioned already can in fact also be generated by CO₂ activation[ ¹⁸ , ²² , ²³ ]), we were now interested in understanding the parameters that lead to activation.

Influence of Lewis Acids on CO₂ ²⁻ Activation

Having learned that I readily generates CO in contact with electrophiles, the role of the Li₃ shell surrounding the CO₂ ²⁻ ligand in this conversion was explored. Initial in situ ¹H NMR spectroscopic studies had indicated that the synthesis of [L^tBuNi(OCO‐κ ² O,C)]M₃[N(SiMe₃)₂]₂ is possible also with M=Na⁺ or K⁺, ^[18] however, no structural information had been available then. We were now able to access the Na⁺ derivative [L^tBuNi(OCO‐κ ² O,C)]Na₃[N(SiMe₃)₂]₂ (4) and crystallize it so that a single‐crystal X‐ray analysis could be performed, which revealed a structure that is similar to the one of I (Figure 3, left) but also shows some differences. The Ni−C bond lengths in I and 4 are almost identical (1.786(4) Å vs. 1.791(2) Å), but variations can be found for the different C−O bonds. While in 4 the bond of the C atom to the distal O atom is quite short (1.216(2) Å) and the one to the coordinated O atom is long (1.298(3) Å), for I they converge (1.234(5) Å; 1.275(5) Å). The OCO angles appear similar (127.3(2)° for 4; 128.0(4)° for I).

Figure 3

Left. Molecular structure of 4⋅C₇H₈. ^[59] H atoms and co‐crystallized toluene are omitted for clarity. Selected bond lengths [Å] and angles [°]: Ni‐N4 1.901(2), Ni‐N3 1.852(2), Ni‐C10 1.791(2), Ni‐O1 1.926(2), C10‐O1 1.298(3), C10‐O2 1.216(2); O1‐C10‐O2 127.3(2), Ni‐C10‐O2 157.5(2). Middle. Side view on Li₃, Na₃ and K₃ complexes. Right. DFT‐optimized structure for the K₃ adduct. H atoms are omitted for clarity. Selected bond lengths [Å] and angles [°]: Ni‐N1 1.9008, Ni‐N2 1.9711, Ni‐C1 1.8243, Ni‐O1 2.0323, C10‐O1 1.2864, C10‐O2 1.2373; O1‐C10‐O2 129.25 Ni‐C10‐O2 151.21.

It thus emerges that the nature of Lewis acidic centers around the CO₂ ²⁻ unit influences its character and it was interesting to investigate how this reflects the spectroscopic properties. Hence, first of all the ¹³C NMR chemical shifts (δ¹³C) were compared for the Li₃, Na₃ as well as the K₃ variants. A δ¹³C of 177.6 ppm was determined for the CO₂ ²⁻ C atom in I, which shifted to 173.2 ppm upon replacement of the Li⁺ ions by Na⁺ ions, that is, in 4. A further shift to 169.5 ppm is observed for the K₃‐adduct. As expected, an inverse trend was found for the ν_CO bands in the IR spectrum of the complexes, with frequencies of 1616 cm⁻¹ for Li₃, 1634 cm⁻¹ for Na₃ and 1637 cm⁻¹ for the K₃ adduct (1568, 1594 and 1594 cm⁻¹ for the ¹³C labelled derivatives). From these data it may be inferred that the activation of CO₂ is strongest in I, which is reasonable as the Li⁺ ions are small and hard acids interacting strongly with the O atoms, while Na⁺ and K⁺ are softer and thus pull electron density to a smaller extent. Furthermore, in 4 the CO₂ appears to be somewhat more activated than in [L^tBuNi(OCO‐κ ² O,C)]K₃[N(SiMe₃)₂]₂ but the difference is not as pronounced as for the transition from I to 4.

This raised the question in how far the differences in activation translate into the reactivity of the NiCO₂M₃ moieties and thus also the Na⁺ and K⁺ derivatives were reacted with three equivalents of TMS‐OTf. Again, 1 was formed in both cases as verified by means of ¹H NMR. However, an ATR‐IR spectrum of the solid crude product isolated after the reaction of the K⁺ derivative revealed that, unlike in cases of the Li⁺ and Na⁺ derivatives, additionally [L^tBuNi^I−CO], III, had been formed. This implicates a slower conversion in the case of K⁺ counterions, leaving enough time for the secondary process 2 as depicted in Scheme 2, that leads to III. While this fits to the expectations, the difference between the Na⁺ and K⁺ derivatives in the grade of CO₂ activation according to IR spectroscopy is not so large, that it should lead to such a sharp contrast in reactivity. Hence, steric factors had to be considered as an origin and for this purpose the structure of [L^tBuNi(OCO‐κ ² O,C)]K₃[N(SiMe₃)₂]₂ was calculated by DFT, using the structure of 4 as a basis for an optimization after replacement of all Na⁺ ions by K⁺ ions. Comparing the resulting structure to those of I and 4 a structural transformation becomes obvious that is most pronounced in the transition from Na⁺ to K⁺ (Figure 3 middle). The trend can be well illustrated, comparing the angles between the planes spanned by the (N)₂Ni(CO₂) core and the plane defined by the three alkali metal cations, respectively. The small Li⁺ ions integrate themselves almost perfectly into the (N)₂Ni(CO₂) plane (3.0° deviation) and in fact even the amide N atoms almost lie in the same plane, too (N to plane distances 0.35 Å and 0.08 Å). For Na⁺, the twist angle between the two planes rises to 8.2°, and now the N atoms deviate from the main plane by 1.10 Å and 0.28 Å. Finally, in case of the large K⁺ ions the distortion becomes dramatic and the twist angle increases to 43.3°. This can clearly be seen in Figure 3, where it also becomes obvious, that in such a structure the CO₂ ²⁻ unit is shielded more strongly by the SiMe₃ groups of the [N(SiMe₃)₂]⁻ ligands than in case of the two other derivatives, thus explaining the comparatively low reactivity.

As discussed, the contact with a Lewis acidic center can prepare a CO₂ ²⁻ ligand, formed via CO₂ activation at a reduced transition metal moiety, for the elimination of CO (Scheme 4). However, a M−CO₂ ²⁻ moiety generated by formate deprotonation has a further option if M is redox active: an intramolecular single electron transfer leading to CO₂ ^.−‐derived products and it seems likely that this process will also be influenced by LAs. While there is no report on a system where both pathways were accessible so far, notably, this is the case in the system described here, controlled by the incorporated Lewis acidic center, as will be outlined in the next paragraph.

Scheme 4

Pathways for generation and reactivity of reduced CO₂ fragments with coordination of LAs.

Deprotonation with K[N(SiMe₃)₂]—Nickel Formate to Oxalate

Having shown that the nature of M⁺ influences the degree of activation, we hypothesized that the number of ions M⁺ will have an effect as well and started an investigation employing less than three equivalents of M[N(SiMe₃)₂]. For M=Li⁺ or Na⁺ this led to incomplete conversions to the corresponding CO₂ adducts I and 4 only. While, as described above, it is possible to also generate the K₃‐analogue of I and 4, namely by slow addition of [L^tBuNiOOCH], II, to three equivalents of K[N(SiMe₃)₂], the deprotonation of II with just one equivalent of K[N(SiMe₃)₂] in C₆D₆ leads to a mixture of paramagnetic and diamagnetic products, according to a ¹H NMR spectrum recorded subsequently. In order to accomplish a more uniform reaction, deprotonation was carried out then with only 0.5 equivalents of K[N(SiMe₃)₂] in C₆D₆, which proved sufficient to convert all [L^tBuNiOOCH], II, employed, as evidenced by the complete disappearance of all of its signals in the ¹H NMR spectrum. The latter revealed the concomitant formation of minor amounts of 1, as well as the signal sets for two paramagnetic complexes that formed as the main products. One of those was identified as the known Ni oxalate complex [L^tBuNi(C₂O₄)NiL^tBu], IV. ^[22] The second species shows paramagnetically broadened signals in the ¹H NMR spectrum, commonly observed for Ni^I compounds and the presence of Ni^I was indeed confirmed by EPR spectroscopy, showing a rhombic signal (best simulated for a 1:1 ratio of two very similar species, Figure 4, left). Hence, under these conditions deprotonation obviously triggers an electron transfer from the CO₂ ²⁻ ligand (generated by deprotonation of half of the [L^tBuNi^IIOOCH], II, present) to a Ni^II center (which presumably belongs to the second half equivalent of II, considering that all II is consumed). This would lead to a Ni^I species and formally a CO₂ ^.− radical anion that likely represents the origin of the oxalate found in the second product IV. The electron transfer mentioned could either occur directly from a Ni^II‐CO₂ ²⁻ unit, or, subsequent to an initial intramolecular electron transfer, from a Ni^I‐CO₂ ^.− unit. Two equivalents of the Ni^II‐CO₂ ^.− entities thus generated can then dimerize to give IV, leaving formally behind two equivalents of K[L^tBuNi^IOOCH], the fate of which is investigated in the following.

Figure 4

X‐band EPR spectrum (26 K, 9.39 GHz) of the reaction mixture resulting from the reaction between II and 0.5 equivalents K[N(SiMe₃)₂] in C₆D₆ (black) and powder simulation for g=2.47, 2.14, 2.07 and g=2.48, 2.13, 2.08 (red). Right. EPR spectrum (26 K, 9.46 GHz) of 5 in C₆D₆ (black) and powder simulation for g=2.47, 2.11, 2.09 and g=2.46, 2.11, 2.06 (red).

Electron Transfer—Nickel(I) Formate

To simulate this situation, II, was reacted with KC₈, which led to the formation of a vibrant red solution. Its ¹H NMR spectrum exhibited a set of paramagnetically broadened signals, which indeed matched the ones observed beside the signal set of IV after the reaction with K[N(SiMe₃)₂]. Moreover, in the EPR spectrum a rhombic signal appeared, which was very similar to one observed before (Figure 4, right). After filtration and evaporation of the volatiles [L^tBuNi^I(μ‐OOCH)Ni^IL^tBu]K (5) could be isolated as a vibrant red solid in 53 % yield. X‐ray analysis of single crystals revealed a structure where two L^tBuNi^I moieties are bridged by a formate ligand and a K⁺ ion (Figure 5). Thus, the decrease of the oxidation state from +1 to +2 has led to a change of the binding mode of the formate ligand from η² to η¹ and thus to a change of the coordination sphere from square planar to trigonal planar, which is plausible. The ATR‐IR spectrum of pure 5 exhibits an asymmetric CO₂ stretching vibration at 1565 cm⁻¹ (Figure 6 c), which shifts to 1525 cm⁻¹ when the ¹³C labelled formate precursor is utilized. This absorption band can also be found in the ATR‐IR spectrum of the product mixture isolated after the reaction between [L^tBuNiOOCH] with 0.5 equiv. of KHMDS in C₆D₆ (Figure 6 a) beside the characteristic CO stretching vibration of [L^tBuNi(C₂O₄)NiL^tBu], IV (Figure 6 b). Altogether the findings lead to a stoichiometry as shown in Scheme 5 for the reaction of II with 0.5 equiv of K[N(SiMe₃)₂].

Figure 5

Molecular structure of 5⋅2(Et₂O). ^[59] H atoms and Et₂O molecules are omitted for clarity. Selected bond lengths [Å] and angles [°]: Ni1‐N1 1.861(2), Ni1‐N2 1.886(2), Ni2‐N5 1.866(2), Ni2‐N4 1.896(2), Ni1‐O2 1.894(3), Ni2‐O1 1.905(3), C1‐H1 1.03(3), K‐O1 2.672(4), K‐O2 2.681(2), C1‐O1 1.245(5), C1‐O2 1.237(6); O1‐C1‐O2 121.4(4), Ni1‐C1‐Ni2 169.6(2).

Figure 6

ATR‐IR spectra of the reaction mixture resulting from the reaction between II and 0.5 equiv. K[N(SiMe₃)₂] (a, black), pure oxalate complex IV (b, purple) and pure Ni^I formate complex 4 (c, red). Dotted lines indicate the correlation of the characteristic bands between a and b (1648 cm⁻¹, dotted purple line) and between a and c (1565 cm⁻¹, dotted red line).

Scheme 5

Reaction scheme for the deprotonation of L^tBuNiOOCH with 0.5 equivalents of K[N(SiMe₃)₂].

Reaction of [L^tBuNi^I(μ‐OOCH)Ni^IL^tBu]K, 5, with one equivalent of K[N(SiMe₃)₂] in C₆D₆ does not lead to a significant decrease of the signals of 5 in the ¹H NMR spectrum. Accordingly, the η² coordination as in II appears to be a prerequisite for formate deprotonation. With this knowledge the reaction of II with more than half an equivalent of K[N(SiMe₃)₂] was revisited. Small amounts of K[N(SiMe₃)₂] were added stepwise to a solution of ¹³C‐labelled II and the proceedings monitored by means of ¹H and ¹³C NMR spectroscopy (see S22). As outlined above, addition of 0.5 equiv. of K[N(SiMe₃)₂] led to the vanishing of all characteristic signals of II, while the signals for oxalate complex IV and Ni^I formate 5 reached a maximum intensity. Adding more K[N(SiMe₃)₂] led to the decrease of the signals of complex IV and the evolution of a new signal set that could be assigned to [L^tBuNi−N(SiMe₃)₂] (6, see S7), until one equivalent of K[N(SiMe₃)₂] had been added in total. Accordingly, K[N(SiMe₃)₂] preferably deprotonates II, forming IV and 5, however, after completion of this conversion, K[N(SiMe₃)₂] displaces the oxalate ligand, with concomitant precipitation of K₂C₂O₄. The formation of 1 as a minor side product can also be observed and may be rationalized by a minor side reaction of in situ generated 6 with traces of CO (see S8).

Intermediate of Oxalate Formation

The formation of IV from complex II is remarkable, as the direct transformation of formate to oxalate so far has only been achieved via thermal decomposition, calcination or at elevated temperatures and high pressures from bulk CO₃ ²⁻ (via formate intermediates) but is unknown in coordination chemistry.[ ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ ] There are only a few examples for oxalate formation directly from CO₂[ ²² , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵² ] and one precedent case, where the reaction of formate derived CO₂ ²⁻ with CO₂ led to oxalate. ^[19]

Thus, having clarified the stoichiometry of the reaction between [L^tBuNiOOCH], II, and K[N(SiMe₃)₂], the mechanism of oxalate formation shifted into our focus. As discussed at the end of the last but one section a conceivable intermediate is, for instance, [L^tBuNi^I(OCO)^.−]K, A, namely the product of an intramolecular electron transfer within [L^tBuNi^II(CO₂)²⁻]K (see Figure 7). This does not occur within [L^tBuNi(OCO‐κ ² O,C)]K₃[N(SiMe₃)₂]₂ as there two further K⁺ ions pull electron density. As many attempts to trap an intermediate failed, we pursued information through computational studies and modelling of the binding situation in this species. As the starting structure for a geometrical optimization of intermediate A we used the structural data of [L^tBuNi^INCO]K (8, see S9), which is isoelectronic with a Ni^I(CO₂) complex, and exchanged the cyanate ligand by CO₂ ^.−. The resulting structure (Figure 7, left) is stable with a Ni^I ion binding a bent CO₂ ^.− ligand (135.8°; 134° for CO₂ ^.−[53]), supported by a K⁺ ion, which interacts with the C atom and the Ni bound O atom of the CO₂ ^.− ligand. A very similar coordination mode with an exposed C atom has been proposed as the initial intermediate for CO₂ reduction to oxalate catalyzed by Cu complexes. ^[47] In the system discussed here, intermediate A would transfer an electron (and K⁺) to unconsumed II, forming half an equivalent of IV.

Figure 7

Left. Possible intermediate A leading to oxalate formation. Right. DFT‐optimized structure for A. H atoms are omitted for clarity.

Nickel Mesoxalate

In order to quantify the formation of the oxalate complex, crystallization attempts from benzene and hexane were conducted. In case of benzene (ca. 0.03 M) this gave solely purple crystals of the oxalate complex [L^tBuNi(C₂O₄)NiL^tBu] (IV, 12 % absolute yield, 48 % relative yield). However, when working at high concentrations (ca. 0.15 M), from hexane solutions repeatedly and reproducibly also small orange crystals grew, that is, a further product was generated at high concentrations. Extraction with Et₂O and subsequent crystallization led to the enrichment of the orange crystals (see S10) and single‐crystal X‐ray crystallographic analysis revealed the structure of the unique multinuclear complex [(L^tBuNi)₆(C₂O₄)₂(C₃O₆)₂K₆] (9, Figure 8). The quality of the crystals did not allow an anisotropic refinement of all atoms during the solution of the structure, and hence the latter does not permit a discussion of metric parameters, but the molecular structure determined and the constitution are without doubt. 9, which was formed in up to 9 % yield based on formate (6 % based on L^tBuNi), contains, as IV, oxalate ligands, but also ligands with the configuration C₃O₆. Accordingly, three formate‐derived CO₂ units have been coupled, which to our knowledge has never been reported so far. However, while the formation of oxalate (C₂O₄ ²⁻) from two CO₂ ^.− species (vide supra) is straight forward, a threefold coupling would give the unknown radical C₃O₆ ^.3−. Therefore, in a more realistic scenario two CO₂ ^.− equivalents and one CO₂ ²⁻ entity are combined yielding in the fourfold deprotonated form of dihydroxymalonic acid (C₃O₆ ⁴⁻), which indeed is known as the hydrate of mesoxalic acid and represents a stable compound; for its formation routes via reaction of oxalate with CO₂ ²⁻ or of a C₂O₄ ^.3− intermediate with CO₂ ^.− are conceivable. 9 then contains two quadruply charged mesoxalate and two doubly negatively charged oxalate anions as well as six [L^tBu]⁻ ligands. This leads to 18 negative charges which are compensated by six K⁺ ions and six Ni^II ions, partly with tetrahedral and partly with square planar coordination spheres, which is reasonable (note that the oxalate complex IV has tetrahedrally coordinated Ni centers, while II features a square planar coordination).

Figure 8

Left: Molecular structure of 9⋅2 Et₂O. ^[59] Atoms that could not be refined as ellipsoids are depicted as balls. Middle. Formate and CO₂ and C−C coupling products. Right: Ball and Stick structure of the core of 9⋅2 Et₂O. H atoms, Et₂O molecules and [L^tBu]⁻ C atoms are omitted for clarity.

Separation of 9 from residual IV proved possible by extraction with MeCN that dissolves 9 to give a clear orange solution, from which an orange solid was isolated. However, storage of such solutions for crystallization leads to the precipitation of a white solid, indicating a certain instability of IV in MeCN. Thus, all analytical methods had to be applied within a short period of time after contact with MeCN.

The ATR‐IR spectrum of 9 shows bands at 1666, 1617, 1588 and 1572 cm⁻¹ in the ν_CO region, originating from the different carbonyl groups in the oxalate and mesoxalate ligands (1648 cm⁻¹ for IV).

The ¹H NMR spectrum displays a combination of diamagnetic and paramagnetically shifted signal sets, as one would expect due to the different coordination environments of the six Ni centers (four square planar; two tetrahedral) in the crystal structure. While a detailed assignment of the signal sets is not possible, the ratio of the integrals from paramagnetic to diamagnetic signals should be around 1:2 or less (depending on how perfect the square planar coordination modes are maintained in solution), matching the 1:1.6 found for solutions of IV in [D₃]acetonitrile.

In the ¹³C NMR spectrum of MeCN‐d₃ and C₆D₆ solutions of 9 neither characteristic signals for oxalate nor for mesoxalate could be detected (the commercially available dihydroxymalonic acid disodium salt resonates at 171 and 91 ppm in D₂O), not even after ¹³C‐labelling. This is not surprising as also the corresponding ¹³C NMR signal of the oxalate ligand in IV eluded detection due to its connection with the paramagnetic Ni^II centers. However, trying to study the in situ formation of 9 through deprotonation of ^13C II in hexane‐d₁₄ or C₆D₆ in an NMR tube sealed with a J. Young valve, after 4 days a characteristic set of signals at 177.7 ppm (dd, ¹J_CC=57.8 Hz) and 103.9 ppm (t, ¹J_CC=57.8 Hz) could be detected (see S10; compare: ¹J_CC of 1,2–¹³C oxalic acid diethyl ester is 58.5 Hz). This consolidates the formation of mesoxalate, and its detection by ¹³C NMR spectroscopy in this experiment likely became possible, as 9 is not stable in solution over four days, so that the ligands get detached from the paramagnetic Ni centers.

Mesoxalate is found rather rarely in coordination chemistry. There are reports of Pd and Cu complexes with mesoxalate ligands, generated through decomposition (after heating/time) of D‐erythrulose (C₄H₈O₄; Pd) ^[54] or D‐glucuronate (C₅H₈O₆; Cu). ^[55] Although it does not concern mesoxalate directly, a publication by Meyer and co‐workers is noteworthy in this context, too, as it describes the reaction of CO₂ with a coordinated diketone‐derived enolate (an oxalate equivalent) and based on this finding they envisioned that the coupling of three CO₂ building blocks should be possible. ^[56] The formation of 9, which involves the reduced CO₂ species CO₂ ²⁻ is to our knowledge the first experimental example of a direct threefold coupling of CO₂ entities, illustrating the feasibility of such transformations. This may even be relevant to the prebiotic synthesis of organic compounds,[ ⁵⁷ , ⁵⁸ ] and motivates further attempts to synthesize multi‐carbon compounds directly from formate or CO₂.