Revisiting Reduction of CO2 to Oxalate with First-Row Transition Metals: Irreproducibility, Ambiguous Analysis, and Conflicting Reactivity.

Construction of higher C≥2 compounds from CO2 constitutes an attractive transformation inspired by nature's strategy to build carbohydrates. However, controlled C-C bond formation from carbon dioxide using environmentally benign reductants remains a major challenge. In this respect, reductive dimerization of CO2 to oxalate represents an important model reaction enabling investigations on the mechanism of this simplest CO2 coupling reaction. Herein, we present common pitfalls encountered in CO2 reduction, especially its reductive coupling, based on established protocols for the conversion of CO2 into oxalate. Moreover, we provide an example to systematically assess these reactions. Based on our work, we highlight the importance of utilizing suitable orthogonal analytical methods and raise awareness of oxidative reactions that can likewise result in the formation of oxalate without incorporation of CO2. These results allow for the determination of key parameters, which can be used for tailoring of prospective catalytic systems and will promote the advancement of the entire field.

Introduction

Nature’s ability to convert carbon dioxide (CO2) into higher carbon compounds, mainly via photosynthesis, has inspired researchers for decades to evaluate the feasibility of constructing value-added C≥2 materials from CO2 as the starting material.[1,2] Various catalytic approaches enabling carbonylation,[3,4] carboxylation,[5−8] or copolymerization reactions[9,10] were developed based on CO2 as a C1 building block (Figure ).[11] Moreover, the direct conversion into C≥2 compounds, such as ethane,[12,13] ethylene,[14] ethanol and higher alcohols,[15−17] hydrocarbons,[18−20] and even aromatic compounds[21] has been achieved. Nonetheless, the selective formation of a C≥2 product utilizing CO2 under mild reaction conditions remains a significant challenge. This lack of development contrasts the significant progress for the CO2 reduction to C1 products, namely, methanol, carbon monoxide, formic acid, and methane via thermal, electrochemical, and photochemical approaches.[22−27] One of the major difficulties in the development of selective processes toward C≥2 products resides in the C–C bond forming step.[18,19,21] Prior reduction of CO2 to CO or methanol (MeOH) and subsequent conversion thereof to the final C≥2 products circumvents this step.[16−19] However, a comprehensive understanding of the C–C bond formation starting from CO2 might facilitate the development of tailored catalytic systems which could enable more selective transformations.

Figure 1

Selected examples highlighting the progress in CO2 utilization as a C1 building block or reduction into C1 and C≥2 products as well as potential products derived from oxalic acid.

Dimerization of two CO2 molecules upon two-electron reduction resulting in oxalate, coined CO2 reductive coupling, provides an ideal model reaction to investigate the C–C bond formation of interest. In addition, oxalic acid was recently suggested as a platform chemical for the preparation of sustainable polymers and its conversion into reduced C2 compounds, such as glycolic acid and ethylene glycol.[28,29] Moreover, it is frequently applied in the purification and recycling of actinides by the nuclear industry.[30]

Early reports highlighted the formation of oxalate during electrochemical CO2 reduction on inert electrodes (Hg and Pb).[31−34] Moreover, aromatic esters or nitriles,[33] as well as transition metal complexes,[35−41] were reported to be efficient electrocatalysts for this transformation.[25]

Nearly 40 years ago, a well-defined oxalate complex was isolated by Fröhlich and Schreer from the reaction of a titanium alkyl complex with CO2.[42] Since this pivotal discovery, various oxalate complexes based on Mg,[43] Ti,[42,44] Sc,[45] Sm,[46−49] Yb,[48,50,51] Lu,[52] U,[53−55] Th,[56] Ni,[57,58] Fe,[59,60] and Cu (e.g., 2 and 3)[39,61,62] were obtained from reaction with CO2 (Figure ). Moreover, CO2 reductive coupling with the help of strong chemical reductants, such as KC8, was reported for complexes based on U,[55] Yb,[50] Fe (1 and 5),[63,64] Ni (4),[58] and Cu (6a,b).[65] Although all these studies were of limited practical relevance, they allowed for the identification of intriguing reaction pathways, e.g., dimetalloxycarbenes undergoing nucleophilic attack onto CO2.[44,48,56] In addition, their homogeneous nature facilitated evaluation of the impact of steric and electronic parameters of the corresponding catalytic system onto oxalate formation.[47,48,60,65] However, the influence of reaction parameters, such as employed solvents or ligand properties, displayed major variations between different examples.

Figure 2

Selected examples for the formation of oxalate complexes from CO2 and CO2 reductive coupling with strong reductants.[58,61−65] The reactions with complexes 1–3 are discussed herein.

Likewise, the reaction mechanism for C–C bond formation appears to be rather specific for a respective system. Moreover, analysis of oxalate in the presence of additional CO2 reduction products via NMR or IR spectroscopy can be challenging[66] and common analytic techniques for oxalate quantification[67−70] have scarcely been adopted within this field. These obstacles hamper systematic development of novel and refinement of existing protocols for the coupling of CO2. Hence, the small number of reports on CO2 reductive coupling compared to well-established CO2 reduction to CO, formic acid, and MeOH[22−25] comes as no surprise. Further complication arises from undesired side reactions and irreproducibility of experimental results.[66,71,72] As an example, in situ analysis by Knope et al. on the formation of a Nd-based oxalate coordination polymer suggested that oxalate formation under hydrothermal conditions arose from oxidative decomposition of 2,3-pyrazinedicarboxylic acid without previously suggested incorporation of CO2.[72] Similarly, in a joint cooperation with Maverick and co-workers, the oxidative decomposition of ascorbate, utilized as reductant for the formation of a CuI species, was identified as the source of the observed oxalate.[66] In this case, isolation of a dinuclear Cu oxalate complex in combination with a misleading difference IR spectrum and air contamination of reactions under CO2 atmosphere caused misinterpretation of the experimental data. Such conflicting reactions provide the basis for incorrect interpretation of data and exacerbate the unsteady progress within this field.

Here, we present a guideline for investigations on the simplest C–C bond formation from CO2 and showcase common pitfalls associated with this reaction. For this, three exemplary protocols based on first-row transition metals that have been reported to facilitate the reductive coupling of CO2 were revisited in detail.

As a first example, the reaction of carbon dioxide and [Fe(tmtaa)] (1, tmtaa = 4,11-dihydro-5,7,12,14-tetramethyldibenzo[b,i]-[1,4,8,11]-tetraazacyclotetradecine) in combination with strong reductants (Na, K, NaC10H8) is reexamined (Figure ).[63] [FeI(tmtaa)Na(THF)3] (7), obtained by reduction of [FeII(tmtaa)] with Na (−3.04 V vs Fc/Fc+)[73] in THF, was reported to show a decisive solvent effect for its reaction with CO2.[63] Reductive disproportionation of CO2 to [FeII(tmtaa)(CO)(THF)] (8) and Na2CO3 was described for the reaction in THF, while in toluene, the formation of oxalate (alongside 1) was claimed based on titration with KMnO4.

The coupling of CO2 to oxalate in the absence of strong chemical reductants is evaluated on the example of a triazacyclononane-derived CuI complex. The allylated tridentate ligand L1 was reported to facilitate the formation of dinuclear complex [(L1)Cu(μ-C2O4)Cu(L1)](BPh4)2 (2) upon reaction of the in situ formed [Cu(L1)]BPh4 complex with exhaled air, CO2 (21%), and CsHCO3 (53%) (Figure ).[61]

Finally, the incorporation of CO2 into oxalate starting from a CuII α-ketocarboxylate complex [Cu(Tp)(O2CC(O)CH(CH3)2)] (9) under oxidative conditions with air, O2, or CO2/O2 via a proposed CO2•– is reviewed.[62] Incorporation of CO2 into the product 3 (Figure ) was reported based on a reaction with 13CO2/O2 and vibrational spectroscopy.

In all these cases, incorporation of CO2 into oxalate proved to be irreproducible.

Results

Reactions with [Fe(tmtaa)]

Based on the previous work, we initiated our investigations by studying the reaction of in situ formed 7 with CO2 in toluene (Scheme ). In our hands, we were unable to identify oxalate by 13C NMR spectroscopy (Figures S52 and S53) and capillary electrophoresis (<0.5% yield based on [Fe]; Figures S349 and S350) after aqueous extraction of the reaction mixture, despite the utilization of excess sodium (Table S1, detailed procedures and results are presented in the Supporting Information). To rule out the insufficient solubility of Na causing the lack of C2O42– formation, further reactions were conducted utilizing NaC10H8 (−3.10 V vs Fc/Fc+)[73] in THF as the reductant for in situ formation of 7. Again, no C2O42– was observed by NMR spectroscopic (Figures S50 and S51) and CE analysis (Figures S347 and S348). To complement the data obtained with Na-based reductants, analogous reactions were performed with K.

Scheme 1

Reduction of [Fe(tmtaa)] to Complex 7 and Observed Products of Subsequent Reaction with CO2 Highlighting the Absence of Oxalate

In these cases, reaction in THF and toluene also provided no evidence for the reductive coupling of CO2. However, a signal in the 13C NMR spectra located between 162 and 164 ppm was observed for all reactions (Figures S50–S59), which was assigned to CO32– by adjustment of the pH value of the NMR sample (with 1 M NaOH) and comparison with Na2CO3 (Figure S49). The formation of carbonate in both THF and toluene suggests an accompanied CO2 reduction to CO regardless of the solvent. This was indeed confirmed by the presence of the distinct CO vibration[63] (1911–1912 cm–1) of iron carbonyl complex 8 in the IR spectra of the solid product from the reaction of 1 with K in THF and subsequent treatment with CO2 in THF or toluene (Figures S190–S196). Interestingly, no gaseous CO was found during gas chromatographic analysis of the reaction headspace (Figure S456). NMR analysis of the residue after treatment with K and CO2 revealed the presence of 1(63) for the reaction in toluene, while line broadening, presumably due to paramagnetism, prohibited assignment to 1 or 8 for the reaction in THF (Figure S48).

Based on these results, we found no evidence for the formation of oxalate utilizing complex 1 and strong chemical reductants while reductive disproportionation was confirmed in both THF and toluene.

CO2 Reductive Coupling by tacn-Derived Cu Complexes [Cu(L)]X

We initiated our investigations on the oxalate formation with L1 reported by Peacock and co-workers by following the literature protocols.[61] In our hands, CO2 bubbling through a suspension of CuI, L1, and NaBPh4 in MeOH resulted in a color change, indicating oxidation of the CuI complex. However, the residue obtained from this experiment featured an IR band at 1628 cm–1 (Figure S197), significantly shifted compared to the literature value of 1660 cm–1,[61] and recrystallization from MeNO2 did not provide the desired oxalate complex. Interestingly, when CO2 bubbling was paused after an initial saturation of the reaction mixture over 2 h and the mixture was stirred overnight under an Ar/CO2 atmosphere, a decolorization was observed that was repeatable for the same reaction mixture (up to six times, see Figure S1). This color change indicated a reversible oxidation of the CuI complex during CO2 bubbling, possibly accompanied by reduction of a CuII complex by NaBPh4.[74] Reaction with NaPF6 instead of NaBPh4 did not cause a decolorization upon paused CO2 bubbling, thus corroborating the role of BPh4– in the reduction. The color change during reactions with CO2 in the presence of NaBPh4 was accompanied by precipitation of a yellow solid. NMR spectroscopic analysis of this precipitate (in THF-d8) indicated the formation of a CuI complex different from [Cu(L1)I][75] (Figure S61), presumably [Cu(L1)]BPh4 based on the absence of additional signals in the 13C NMR and the integral ratio in the 1H NMR spectrum. Contact of the THF-d8 solution to air resulted in crystallization of [(L1)Cu(μ-OH)2Cu(L1)](BPh4)2 (10, Figure ), which is in good accordance with analogous compounds formed via aerial oxidation of the CuI complexes.[76,77] However, the desired oxalate complex was not obtained from these experiments, and IR vibrations were observed around 1630 cm–1, thus shifted by 30 cm–1.

Figure 3

Compound 10 and the molecular structure of [(L1)Cu(μ-OH)2Cu(L1)]2+ in 10 (thermal ellipsoids displayed at 30% probability and C-bound H atoms omitted for clarity; symmetry operator for generating equivalent atoms: −x + 1, −y + 1, −z + 1).

Inspired by the possibility of removing potentially formed oxalate from the product complex,[62] we attempted the synthesis of complex 2 starting from Na2C2O4. While a reaction utilizing in situ formed [Cu(L1)(NO3)2] (11) provided a mononuclear complex [Cu(L1)(κ2-C2O4)] (12), we were able to obtain blue needles of [(L1)Cu(μ-C2O4)Cu(L1)](BF4)2 (13) in 62% yield utilizing Cu(BF4)2·6H2O and Na2C2O4 in MeOH/H2O. As expected, the coordination geometry in 13 is similar to that of 2 (Figure ). Moreover, its distinct C–O stretching vibration is observed at 1651 cm–1 (Figure S178).

Figure 4

Molecular structure of [(L1)Cu(μ-C2O4)Cu(L1)]2+ in 13 (thermal ellipsoids displayed at 30% probability and H atoms omitted for clarity; symmetry operator for generating equivalent atoms: −x, −y + 1, −z + 1).

With [(L1)Cu(μ-C2O4)Cu(L1)](BF4)2 in hand, we assessed the removal of C2O42– by stirring in toluene/1 M NaOH (2:1, v:v).[62] To our delight, we detected C2O42– with a yield of 47% by CE for two reactions on 24 μmol scale (1, 48%; 2, 47%). On this scale, the 13C NMR signal of C2O42– at 173 ppm (Figure S60 provides 13C NMR shifts for common CO2 reduction products in D2O) was not observable (however, a minor broadening of the baseline for reaction 1 was noticed).

Having established the possibility of oxalate removal from the desired dinuclear Cu complex, we turned our attention to the reaction with solid bicarbonates, which was supposed to yield higher quantities of complex 2.[61] It is noteworthy that we isolated [Cu(L1)Cl2] (14) from an initial reaction of CuI/L1/NaBPh4 (1:1:1) with CsHCO3 (1.3 equiv) in MeOH following filtration, column chromatographic separation (SiO2, MeOH/MeNO2/2 M NH4Cl, 7:1:2), and recrystallization from hot MeNO2. 14, which can be independently prepared from CuCl2, presumably results from oxidation of the initial CuI complex during the aerial workup and subsequent ligand exchange with NH4Cl during purification.

Additional reactions (Table ) with CsHCO3 or NaHCO3 in combination with CuI or [Cu(MeCN)4]PF6 in MeOH did not result in notable oxalate formation (according to NMR and CE analysis). Notably, a broad signal coinciding with the oxalate signal was observed for the reactions with CuI in the absence of NaBPh4 (Table , entries 5–6; Figures S368 and S369). However, no oxalate was detected by ESI-HRMS analysis for entry 6 (Figure S305), suggesting the presence of another compound with a similar migration time. In this case, most likely the iodide anion was detected as the mobility constants for oxalate and I– are comparable.[78] Indeed, control measurements confirm this behavior (Figure S357).

Table 1

Treatment of In Situ Formed Cu-tacn Complexes with Bicarbonatesa

entry	[Cu] [μmol]	[MHCO3] [μmol]	t [h]	13C NMR (C2O42–)	CE (C2O42– (?)) [%]b
1	CuI (126)	CsHCO3 (133)	16	n.d.	<1
2	CuI (120)	CsHCO3 (126)	117	n.d.	<1
3c	CuI (150)	CsHCO3 (300)	168	n.d.	<0.5
4d	CuI (296)	NaHCO3 (214)	210	n.d.	<1
5e	CuI (126)	CsHCO3 (134)	17	n.d.	11
6e	CuI (122)	CsHCO3 (129)	122	n.d.	7
7f	[Cu] (120)	CsHCO3 (127)	118	n.d.	<0.5
8f	[Cu] (120)	NaHCO3 (127)	118	n.d.	<1

Reaction conditions: [Cu]/L1/NaBPh4 (1:1:1) combinations were utilized in degassed MeOH. n.d. = not detected.

Oxalate yields were calculated for signals coinciding with Na2C2O4 added as internal standard after initial measurement and were found to overlap with the signal of iodide.

NaPF6 (1.2 equiv) instead of NaBPh4 was utilized.

1.7 equiv of CuI was utilized.

Reaction in the absence of NaBPh4.

Reaction with [Cu(MeCN)4]PF6.

Due to the lack of oxalate formation by CO2 bubbling and reaction with MHCO3, we evaluated the exposure of in situ formed [Cu(L1)]X (X = I, BPh4) to air to rule out an oxidative origin of oxalate. However, leaving a suspension of CuI, L1, and NaBPh4 in MeOH or MeOH/EtOH (1:4, v:v) in air over 41 and 1 d, respectively, did not result in substantial C2O42– formation according to CE (Table S4).

Likewise, treatment of the in situ formed CuI complex with O2 did not cause formation of oxalate. Interestingly, peaks likely belonging to I– with integrals corresponding to C2O42– yields of 7% and 6% were observed for the exposure of solutions of in situ formed [Cu(L1)I] to air in the absence of NaBPh4 (Figures S375 and S376).

Due to the absence of oxalate formation in all these experiments, we investigated the impact of reaction parameters to possibly facilitate CO2 reductive coupling (Table ).

Table 2

Treatment of CuI, NaBPh4, and L1 with CO2 under Varying Reactions Conditionsa

entry	CuI [μmol]	parameter	t [h]	13C NMR (C2O42–)	CE (C2O42– (?)) [%]b
1	120	-	24	n.d.	<1
2	120	-	118	n.d.	<0.5
3	182	-	120	n.d.	<0.5
4	120	MeOH/THF (9:1)	118	n.d.	<0.5
5	124	THF	120	n.d.	<0.5
6	123	toluene	121	n.d.	<1
7	120	40 °C	114	n.d.	3
8	122	40 °C	123	n.d.	8
9	133	400–700 nm (0.09 W)	18	n.d.	<0.5
10	130	10 bar	116	n.d.	16
11	132	10 bar	116	n.d.	12

Reaction conditions: CuI/L1/NaBPh4 (1:1:1) was utilized in degassed MeOH under constant CO2 atmosphere, unless indicated by the varied parameter. n.d. = not detected.

Oxalate yields were calculated for signals coinciding with Na2C2O4 added as internal standard after initial measurement and were found to overlap with the signal of iodide.

Prolonged stirring under a constant CO2 atmosphere for 5 days (Table , entries 2 and 3) did not provide detectable quantities of oxalate. Likewise, changing the solvent to a mixture of MeOH and THF (Table , entry 4), THF (Table , entry 5), or toluene (Table , entry 6) as well as illuminating the reaction mixture with visible light (400–700 nm, 0.09 W; Table , entry 9) remained unsuccessful. Interestingly, elevated temperature (40 °C, entries 7 and 8) and elevated CO2 pressure (10 bar, entries 10 and 11) facilitated the appearance of a signal of fitting migration time in the CE analysis (Figures S384–S386, S388, S389), corresponding to 5% and 14% C2O42– yield, respectively (average of two reactions). Again, no oxalate was found by ESI-HRMS analysis, indicating yet again overlapping with the signal of iodide. However, ion suppression during the electrospray ionization process has to be considered and can be controlled by further dilution, adding an internal standard ion, or by coupling with CE.

Various additives, namely, Mg(OTf)2, LiBF4, HCO2Na, and KPF6, showed no improvement of the reactivity when combined with CuI or [Cu(MeCN)4]PF6 and ligand L1 (Table S6). Since iodide exchange with NaBPh4 was found to be incomplete, indicated by the crystallization of [Cu(L1)I] (15) from a THF solution of CuI, NaBPh4, and L1, dehalogenation with AgBPh4 was conducted. However, no improvement of the reaction was detected by CE/NMR analysis after NaOH treatment (2% yield by CE, but integration hampered by partial overlapping; Figure S391), despite an IR signal at 1665 cm–1 (Figure S235). Interestingly, oxalate in low yields was detected for reactions with NaC10H8 in THF, even in the absence of any Cu complex (2–7% yield).

Next, other tridentate N-based ligands were evaluated instead of L1 (Table ). Unfortunately, dipicolylamine (dpa), 1,4,7-benzylated (L2), and 1,4,7-propylated (L3) tacn showed no evidence for CO2 reductive coupling reactivity. In contrast, diethylenetriamine (dien) clearly showcased the limits of CE analysis. A signal that would correspond to a 28% C2O42– yield was observed for a reaction of CO2 with dien, CuI, and NaBPh4 in MeOH over 5 d (Table , entry 4). No intense C–O vibration was observed above 1600 cm–1 (Figures S255 and S256), and doubling the reaction scale caused a noticeable decrease in the potential C2O42– yield to 10% (Table , entry 5). Moreover, performing the reaction under an argon atmosphere resulted in the same CE signal (Figure S414), meaning that the CE signal is most likely corresponding to I– rather than a CO2 reduction product.

Table 3

Treatment of CuI, NaBPh4, and Different Ligands with CO2 in MeOHa

entry	CuI [μmol]	ligand	t [h]	13C NMR (C2O42–)	CE (C2O42– (?)) [%]b
1	123	L2	120	n.d.	<0.5
2	124	L3	120	n.d.	<0.5
3	151	dpa	120	n.d.	2
4	184	dien	119	n.d.	28
5	409	dien	119	n.d.	10
6c	180	dien	120	n.d.	26
7	53	L4	114	n.d.	5
8d	63	L4	118	n.d.	29
9d	187	L4	119	n.d.	19
10d,e	187	L4	119	n.d.	28
11d,f	94	L4	118	n.d.	32
12c,d	63	L4	124	n.d.	3
13c,d	94	L4	118	n.d.	<0.5

Reaction conditions: CuI/ligand/NaBPh4 (1:1:1) was utilized in degassed MeOH under constant CO2 atmosphere, unless stated otherwise. n.d. = not detected.

Oxalate yields were calculated for signals coinciding with Na2C2O4 added as internal standard after initial measurement and were found to overlap with the signal of iodide.

Reaction conducted under Ar.

A total of 0.5 equiv of L4 was utilized.

Reaction conducted with 13CO2.

Reaction conducted without NaBPh4.

To evaluate if dinuclear Cu complexes might offer an intrinsic advantage for the reductive coupling of two CO2 molecules, we envisaged ligand L4 as a prime candidate. Thus, we targeted the allyl-derivative L4 which was prepared by reaction of 1,4,7-triazatricyclo[5.2.1.04,10]decane (16)[79] with 1,8-bis(bromomethyl)naphthalene (17)[80] and subsequent allylation (Scheme ). Isolation of [Cu2(L4)I2] (19) further showcased the ability of L4 to coordinate to two Cu centers.

Scheme 2

Synthesis of Ligand L4 and Its Derived CuI Complex 19 and (Left) Molecular Structure of 19

Thermal ellipsoids are displayed at 30% probability, and H atoms are omitted for clarity.

Reactions of L4, CuI, and NaBPh4 toward CO2 in MeOH resulted in the appearance of a CE signal corresponding to 29% C2O42– yield (Table , entry 8; 32% in the absence of NaBPh4, entry 11) which decreased to 19% upon increase of the reaction scale (Table , entry 9). Contrary to previous experiments with dien, blank experiments under argon did not result in the appearance of this CE peak (≤3% yield; Table , entries 12 and 13; Figures S421 and S422). However, no oxalate was detected by ESI-HRMS analysis (Figure S312). Moreover, an experiment with 13CO2 (formed from Na213CO3 and H2SO4), yielding a signal suggesting a 28% C2O42– yield by CE, displayed no observable oxalate signal in the 13C NMR spectrum (Figure S127).

Hence, we exclude the formation of substantial quantities of oxalate with novel ligand L4 as well as L1, L2, L3, dien, and dpa under the studied reaction conditions.

Reactions of α-Ketocarboxylates and Derived CuTp Complexes toward CO2

To evaluate the third literature example, we prepared complex 9 and its 13C5-labeled congener 20 (84%) following the same procedure[62] but utilizing sodium 13C5-3-methyl-2-oxobutyrate (21). 13C-labeling resulted in the expected shift of the C–O stretching frequencies from 1666/1687 cm–1 (9) to 1625/1644 cm–1 (20) (Figure ).

Figure 5

FTIR spectra from reaction of complexes 9 and 20 with air and CO2/O2 in DCM and toluene, respectively (FTIR spectra of 9 and 20 are displayed for comparison).

Exposure of 9 and 20 to air (Scheme ) caused an alteration of the IR spectra resulting in new bands at 1655 cm–1 and a broad band at 1602 cm–1 for 9, the former being in accordance with the initial report.[62] For 20, signals at 1609 and 1644 cm–1 were observed in addition to a broadened band at 1562 cm–1 (Figure ). The former is in agreement with the isotopic shift observed by Takisawa et al. while the latter two remained inconclusive.

Scheme 3

Reaction of Complexes 9 or 20 with Air or CO2/O2 and Subsequent Treatment with NaOH in H2O/Toluene

Therefore, the obtained products were extracted with NaOH and analyzed by NMR spectroscopy and CE. Indeed, the formation of oxalate could be verified (Figures S133–S134 and S423–S425); however, the observed yields (9: 7% [18 h], 54% [48 h]; 20: 23% [18 h]) were significantly decreased in comparison to the literature (89%).[62] It is noteworthy that these yields are determined by CE and calculated assuming a bimolecular mechanism with one oxalate being formed from two α-ketocarboxylate complexes. 13C NMR spectroscopy for the reaction of 20 allowed for the identification of two major byproducts, namely, 2-hydroxy-2-methylpropanoic acid (confirmed by ESI-HRMS; Figure S317) and 2-methylpropanoic acid (Figure ).

Figure 6

13C NMR spectroscopic analysis of the NaOH extract from the reaction of 20 with air.

When reactions of 9 and 20 were conducted in toluene under an O2/CO2 atmosphere, analogous IR spectra were obtained (Figure ). Again, C2O42– yields quantified by CE were significantly reduced (9: 7%; 20: 2%), and 2-hydroxy-2-methylpropanoic acid as well as 2-methylpropanoic acid were detected as the primary products by 13C NMR spectroscopy after NaOH extraction (Figure S149–S151). Interestingly, addition of minor quantities of D2O to the reaction of 20 with CO2/O2 in toluene-d8 and analysis by 13C NMR spectroscopy revealed the formation of 13C3-acetone (203.7 ppm [t], 30.3 ppm [d], Figure S131). Subsequent extraction with NaOH and analysis by NMR spectroscopy suggested higher selectivities for C2O42– formation (compare Figure S151 and Figures S153 and S155), while yields assessed by CE were poorly reproducible (25% and 3% for two reactions). Nonetheless, the higher selectivity for C2O42– in the presence of D2O appears to be in accordance with the literature.[62] In addition, ESI-HRMS analysis for reactions of 20 in air and under CO2/O2 atmosphere showed no evidence for the presence of 13C12CO42–, while 13C2O42– was detected for the former and the reaction with CO2/O2 in toluene-d8/D2O (Section S9.2).

Next, in situ formation of 20 starting from CuCl2, sodium 13C5-3-methyl-2-oxobutyrate (21), and KTp in DCM/heptane and exposure to air was attempted and resulted in an oxalate yield of 38% (CE) after 27 h.

To gain further insight into the oxalate formation and evaluate the necessity of defined complexes 9/20, analogous reactions were conducted in the absence of the Tp ligand (Scheme ). For this purpose, a Cu precursor and sodium 3-methyl-2-oxobutyrate (21a) were suspended in DCM containing 10% MeOH and left in air (layered with heptane). Subsequent NaOH treatment revealed the formation of oxalate as the major product with 1 equiv of Cu(BF4)2·6H2O (38–59% by CE, Table S8) or CuCl2 (55–82%). Note that oxalate yields from here on are monomolecular yields assuming one C2O42– per α-ketocarboxylate. In addition, substoichiometric quantities of Cu(BF4)2·6H2O (0.2 equiv) were sufficient to provide 48% C2O42–. In the absence of Cu, 4% C2O42– was detected by CE (Figure S446), while only the starting material was observed by IR spectroscopy (Figure S292).

Scheme 4

Reactions of Different Sodium α-Ketocarboxylates in Combination with a Cu2+ Source under Oxidative Conditions and Subsequent NaOH Treatment for Analysis via NMR and CE

In addition to these aerial oxidations, Cu(BF4)2·6H2O and 21 were stirred under a CO2/O2 atmosphere in CD2Cl2/CD3OD (10:1, v:v) and C6D6/CD3OD (10:1, v:v) for 4 d. In these cases, 13C NMR spectroscopic analysis confirmed the formation of 13C3-acetone (Figure S131). Subsequent NaOH treatment provided oxalate in 19% (CD2Cl2) and 39% (C6D6) yield, respectively. The former was lower due to a noticeable amount of 2-methylpropionic acid resulting from decarboxylation (Figures S156 and S158). The reason for the favored decarboxylation in CD2Cl2 is not clear and might be a result of DCl traces. It should be noted that in no case mixed isotopic constitutions of C2O42– were detected by ESI-HRMS analysis (Section S9.2), substantiating the lack of CO2 incorporation starting from both, defined complexes 9/20, or a CuII salt combined with the α-ketocarboxylate.

Interestingly, NaOH treatment under an Ar atmosphere decreased the oxalate yield to 17% (from 49% in air, Tables S8 and S10), and significant quantities of starting material were observed by 13C NMR spectroscopy (Figures S163 and S164).

Finally, evaluation of different α-ketocarboxylates was performed to complement the results of our investigation. The reaction of sodium phenylpyruvate with Cu(BF4)2·6H2O in DCM/MeOH in air over 46 h resulted in 29% oxalate (CE). In contrast, sodium pyruvate gave only 6% by CE after 16 h, close to the blank experiment. However, we observed oxidative decarboxylation to acetate by 13C NMR spectroscopy (Figure S165). Moreover, 2-hydroxy-3-methylbutyric acid did not undergo oxidative degradation to oxalate.

While the formation of oxalate was confirmed for defined complexes 9 and 20 as well as for reactions in the absence of Tp, we found no evidence for the incorporation of CO2 into the product oxalate and, thus, cannot confirm the reported conversion of CO2 into oxalate.

Discussion

Reduction of [FeII(tmtaa)] with Na in THF, yielding [FeI(tmtaa)Na(THF)3] (7), and subsequent reaction with CO2 was reported to result in two distinct reactions depending on the employed solvent.[63] Formation of [FeII(tmtaa)(CO)(THF)] (8) and Na2CO3 via reductive disproportionation of CO2 was observed in THF. In contrast, oxalate was the claimed product from the reaction in toluene assessed by titration with KMnO4.

Based on the previous work, we investigated the reaction starting from 1 and Na, NaC10H8, and K followed by treatment with CO2 in toluene or THF (Scheme ). For the reactions with K, we were able to confirm the formation of CO, in the form of complex 8, and CO23– resulting from reductive disproportionation of CO2. Yet, we did not observe the reported solvent effect and identified the distinct CO band of complex 8 (1911–1912 cm–1) for reactions in both toluene and THF. Interestingly, we did not detect any oxalate by NMR spectroscopy or CE analysis (<0.5% based on 1) after aqueous extraction of the resulting reaction products regardless of the utilized reductant, thus ruling out insufficient solubility preventing CO2 reductive coupling.

Scheme 5

Revised Reaction of [Fe(tmtaa)] with Strong Reductants Followed by CO2 Treatment Yielding “CO” and M2CO3 but Not M2C2O4

This lack of oxalate formation highlights two major hurdles associated with studying the reduction of CO2 into oxalate, namely, irreproducibility and ambiguous analysis. Standardization of KMnO4 with the help of oxalate constitutes an established procedure.[81] However, KMnO4 is likewise able to oxidize the metal catalyst (in this case Fe2+) and most organic compounds.[82,83] Therefore, oxidation of residual Fe2+ or an organic impurity could have resulted in a false positive for the KMnO4 titration. Thus, our results emphasized the importance of combining different analytical methods to confirm the presence of C2O42–. Furthermore, the reactions with [Fe(tmtaa)] showcased that strong reductants capable of facilitating CO2•– formation do not guarantee reductive coupling to oxalate.

Our investigations on the second selected literature example further corroborate that both of these pitfalls are not limited to the single example of [Fe(tmtaa)]. Peacock and co-workers reported a CuI complex bearing an allylated tridentate ligand (L1) that was converted into the dinuclear complex [(L1)Cu(μ-C2O4)Cu(L1)](BPh4)2 (2) upon reaction with exhaled air, CO2 (21%), or CsHCO3 (53%) in MeOH.[61] Besides a characteristic IR signal at 1660 cm–1, a crystal structure of the oxalate complex and its magnetic moment were reported to prove the reductive coupling of carbon dioxide. A major advantage of this approach seems to be the absence of additional reductant. Surprisingly, despite the interest in the chemistry of tacn-derived Cu complexes,[80,84−91] no further developments of this system were reported, possibly highlighting the unique properties of allyl-substituted L1.

In our hands, initial experiments with CO2 bubbling indicated an oxidation of the CuI complex formed from CuI, NaBPh4, and L1. However, isolation of a μ-hydroxo CuII complex (10) highlighted that trace O2 might compromise the distinct color change associated with CuI oxidation as an analytical tool. Moreover, IR bands in the 1630 cm–1 region observed for the products from these CO2 bubbling experiments might indicate the formation of a CO2 reduction product, despite the 30 cm–1 difference with respect to desired complex 2. Given the proximity of C–O vibrations of CO2 derived compounds, such as NaHCO3 or HCO2Na in the 1550–1700 cm–1 range (Figure S197), additional analytical data for the identification of CO2 reduction products was desirable. Therefore, we prepared the BF4– analogue of complex 2 and confirmed the possibility of oxalate removal by extraction with aqueous NaOH. This extraction after initial FTIR analysis allowed for detection of oxalate by CE and 13C NMR as well as quantification by CE. With this methodology in hand, the conversion of solid bicarbonates into oxalate with a CuI source and L1 was evaluated but provided no evidence for the reaction of interest. Likewise, evaluation of various reaction parameters, such as solvents, temperature, CO2 pressure, light, or different additives did not yield competing quantities of oxalate. As part of these experiments, a shortcoming of the CE analysis was encountered. Due to similar mobility constants of iodide and oxalate,[78] overlapping of their signals during CE analysis could be expected and was confirmed by control measurements. Moreover, we observed signals coinciding with oxalate especially for reactions in the absence of NaBPh4 or at elevated temperature or pressure. Nonetheless, ESI-HRMS showed no evidence for oxalate formation, substantiating that these signals result from I– rather than C2O42–.

Thus, for a precise determination of oxalate in such and related processes, the use of different analytical methods is essential.

Aside from extensive investigations with L1, we employed our approach with initial FTIR analysis and subsequent NaOH treatment prior to NMR and CE to tridentate N-based ligands, namely, dpa, dien, and two triazacyclononane derivatives. Likewise, these ligands were incapable of promoting the formation of oxalate when combined with CuI and NaBPh4 under the studied conditions. For reactions with dien, a blank experiment in the absence of CO2 was performed to demonstrate that the observed signal in the CE likely corresponded to I–, thereby highlighting another valuable tool for studying CO2 reductive coupling.

Finally, we investigated whether a structurally similar dinuclear Cu complex derived from L4 might be active in the conversion of CO2 into oxalate. Analogous 1,8-bis((1,4,7-triazacyclononanyl)methyl)naphthalene (18) was reported to serve as a valuable ligand for the detection of C2O42– via fluorescence spectroscopy in combination with Eosin-Y due to selective binding to the derived Cu complex.[92] The ability of L4 to coordinate two Cu centers was established by the synthesis of [Cu2(L4)I2] (19). However, L4 did not facilitate the formation of oxalate when combined with CuI or CuI/NaBPh4. Again, ESI-HRMS was employed to assign the CE signal with fitting migration time to I– rather than oxalate. In addition, a reaction with 13CO2 was conducted to further corroborate this assignment by the absence of the 13C NMR signal of C2O42–.

Moreover, GC analysis of the reaction headspace for selected examples combined with 13C NMR spectroscopy provided no evidence for substantial CO2 reduction with the investigated CuI complexes, albeit formate traces were identified for few reactions (also under oxidative conditions in air—see Table S4, entry 4). Finally, to rule out the possibility of an oxidative process yielding μ-C2O4 complex 2, we treated the in situ formed [CuI(L1)X] (X = I, BPh4) with air or O2. Neither of these reactions resulted in oxalate formation.

In conclusion, the results from our extensive investigations with various ligands, additives, and reaction conditions (Scheme ) highlight the problems arising from irreproducibility of CO2 reductive coupling by a Cu complex.[71] While we cannot entirely exclude the formation of small quantities of oxalate sufficient for crystallization of the complex of interest based on the combination of analytical methods employed by us, we cannot confirm the previously reported yields for oxalate complex 2.

Scheme 6

Irreproducibility in the Formation of Oxalate from CO2 or MHCO3 with [CuI(NNN)]+ Complexes Studied under Various Conditions, with Multiple Ligands and Different Additives

Besides irreproducibility and ambiguous analysis, conflicting reactivity causing the formation of oxalate without incorporation of CO2 can be misleading and constitutes another major pitfall for investigations on the reductive dimerization of CO2.[66,72] This is showcased by the third selected example based on Cu α-ketocarboxylate complexes. In 2014, Fujisawa and co-workers reported the unusual formation of a dinuclear oxalate complex 3 ligated by Tp which was isolated from the reaction of [Cu(Tp)(O2CC(O)CH(CH3)2)] (9) with air, O2, or CO2/O2.[62] The reaction was proposed to proceed via formation of a CO2•– intermediate resulting from decomposition of the α-ketocarboxylate in the presence of O2. Incorporation of CO2 into the product was evaluated with 13CO2/O2 causing an isotopic shift in the C–O vibration of 49 cm–1. However, the demand for oxidative reaction conditions to facilitate a reduction paired with our experience on the oxidative degradation of ascorbate encouraged us to gain additional insight in the underlying reaction mechanism.

To evaluate CO2 incorporation into oxalate, we prepared complex 9 and its 13C labeled analogue 20. It is noteworthy that our initial attempt to utilize [(Tp)CuBr] (22) was impeded by the formation of an undesired byproduct ([(Hpz)2CuBr], 23) indicating the decomposition of the Tp ligand, an infamous property of these scorpionate ligands.[93−99]

Our initial reactions of 9 and 20 with air in CH2Cl2/heptane or CO2/O2 in toluene indeed produced oxalate. However, the obtained yields (by CE) were significantly reduced compared to the literature, and analysis by 13C NMR revealed the formation of 2-hydroxy-2-methylpropanoic acid and 2-methylpropanoic acid. These products likely resulted from oxidative decarboxylation of the α-ketocarboxylate and account for the decreased C2O42– yield. Formation of 2-hydroxy-2-methylpropanoic acid was mentioned in the original report and ascribed to dry air, highlighting the necessity of water.[62] Thus, reaction of 20 with CO2/O2 in toluene-d8 with addition of D2O was conducted and displayed enhanced selectivity for oxalate over decarboxylation products by 13C NMR analysis, albeit the oxalate yield showed poor reproducibility according to CE. Interestingly, 13C3-acetone was detected by 13C NMR for this reaction prior to treatment with NaOH. This hints at a different mechanism for the formation of oxalate, namely, oxidative C2–C3 bond scission rather than the reported C1–C2 cleavage. This is further corroborated by the observation of decarboxylation products resulting from oxidative C1–C2 scission in combination with poor C2O42– yields. Moreover, ESI-HRMS analysis was found to be a valuable tool in combination with isotopic labeling. ESI-HRMS measurements confirmed the presence of 13C2O42– for reactions of 20 with air or CO2/O2 (in toluene-d8/D2O) while no mixed isotopic constitutions of oxalate were observed.

Incorporation of CO2 via the originally proposed mechanism should partially result in mixed isotope oxalate. In contrast, should oxalate formation proceed via oxidative C2–C3 bond cleavage, one would not expect any CO2 incorporation and, thus, no formation of 13C12CO42–. Oxidative decarboxylation of α-ketocarboxylates via C1–C2 bond cleavage is well-known for Tp complexes of Fe.[100,101] Likewise, the oxidative C2–C3 cleavage of phenylpyruvate yielding oxalate has been observed for Fe and Co complexes ligated by a trispyridylamine derivative.[100,102] Furthermore, Fe-based enzymes Dke1 (acetylacetone-cleaving enzyme) and HPPD (4-hydroxyphenylpyruvate dioxygenase) were found to convert 4-hydroxyphenylpyruvate into 4-hydroxybenzaldehyde and oxalate or 2,5-dihydroxyphenylacetic acid and CO2, respectively.[103] These differences in reactivity are influenced by the electronic structure of the respective metal center and the ability of the α-ketocarboxylate to undergo enolization.[100,102,103]

Based on our experimental data from complexes 9 and 20, a C2–C3 bond scission mechanism explains the formation of oxalate. These results contrast with the initially reported CO2 incorporation based on an isotopic shift to 1607 cm–1 upon utilization of 13CO2/O2.[62] In addition, considerable quantities of decarboxylation products and low oxalate yields beg the question of how the high reported yields were obtained. Based on the selectivity enhancement upon addition of D2O and improved C2O42– yield after prolonged exposure to air, we envisaged that decomposition of the complexes and perhaps even Tp, as in our synthesis of [(Tp)CuBr], could occur.[93−99] The resulting Cu2+ might then facilitate the oxidative decomposition of the α-ketocarboxylate in analogy to its activity in the oxidation of ascorbic acid.[104,105]

Indeed, exposing a CuII precursor (CuCl2 or Cu(BF4)2·6H2O) and 21a in CH2Cl2/MeOH/heptane to air and extracting the resulting residue with NaOH gave significantly higher oxalate yields by CE (38–82%; mononuclear yield), in accordance with the higher signal intensity in the 13C NMR. Even substoichiometric quantities of Cu(BF4)2·6H2O produced 48% oxalate (by CE), while blank experiments without Cu2+ confirmed its role in the oxalate formation. NMR experiments with Cu(BF4)2·6H2O and 21 in CD2Cl2/CD3OD or C6D6/CD3OD confirmed the formation of 13C3-acetone, which is in accordance with the proposed oxidative C2–C3 bond cleavage (Figure S131). Again, no 13C12CO42– was detected by ESI-HRMS for reactions with 21, highlighting the lack of CO2 incorporation into oxalate.

For a better understanding of the proposed C2–C3 bond cleavage, the NaOH treatment was conducted under an Ar atmosphere. This resulted in a notable decrease in the oxalate yield from 48% to 17%, and unconverted α-ketocarboxylate was detected by NMR spectroscopic analysis. Hence, the basic pH facilitates the oxidative C2–C3 bond scission in the presence of Cu2+. This indicates that enolization of the α-ketocarboxylate might be essential for the formation of oxalate via C2–C3 bond cleavage. To gather additional evidence for this hypothesis, further experiments were conducted with related α-ketocarboxylates. Replacement of 21a with sodium phenylpyruvate, which would likewise allow for enolization, provided oxalate in 29% yield. In contrast, sodium pyruvate gave only 6% C2O42–, which is close to that of the blank experiment. These results further corroborate that the structure of the α-ketocarboxylate, allowing for enolization or stabilization of an intermediate radical species, is another key parameter determining the selectivity for C1–C2 vs C2–C3 bond cleavage.

Based on our experimental results, we propose the following reaction pathways for 9/20 (Scheme ): Well-defined [Cu(Tp)(O2CC(O)CH(CH3)2)] complexes facilitate primarily oxidative decarboxylation via C1–C2 bond scission. Low quantities of oxalate should be a result of C2–C3 bond cleavage by [Cu(Tp)]+ or by its decomposition products, possibly resulting from hydrolysis in the presence of H2O. In contrast, Cu2+ facilitates the oxidative degradation of suitable α-ketocarboxylates into oxalate. Here, the oxidation under basic pH plays an integral part. Besides these primary reactions, the underlying oxidative degradation processes appear to be more complicated and might involve follow-up condensation reactions during NaOH treatment, as indicted by the 13C NMR spectroscopic analysis (Section S7.4).

Scheme 7

Proposed Reaction Pathway for the Formation of Oxalate and Acetone via Oxidative C2–C3 Bond Cleavage

Isobutyric acid and 2-hydroxyisobutyric acid were identified, presumably resulting from oxidative decarboxylation upon C1–C2 bond scission.

In conclusion, this example highlights that oxidative processes resulting in the formation of oxalate can readily be misinterpreted. Therefore, cautious analysis with a combination of analytical techniques as well as suitable control experiments to evaluate the incorporation of CO2 are required to avoid this pitfall.

Conclusions

The use of renewable energy to produce organic matter has remained a central challenge for many decades. In this context, the reductive dimerization of carbon dioxide to oxalate can be considered as a first step. Hence, this apparently simple reaction continues to be of interest for basic science as well as more recently in the context of alternative energy technologies, highlighted by the “OCEAN” project of the European Union, which explores the conversion of CO2 into oxalic acid and its utilization as a platform chemical.[28]

While investigating the desired CO2 reductive dimerization, we faced significant problems and describe here the potential pitfalls based on our studies. Looking exemplarily at three previously published works, we identified irreproducibility, insufficient analysis, misleading analytical data, and conflicting reactivity as the main hurdles, which arguably hamper the development of novel and enhancement of established protocols for CO2 reduction to oxalate.

Specifically, using [Fe(tmtaa)] and alkali metal-based reductants, we could not verify the formation of oxalate in toluene or THF previously claimed by titration with KMnO4. Yet, the importance of suitable analytical methods became apparent.

Similarly, in the presence of triazacyclonone-ligated Cu complexes, no CO2 reductive coupling was observed. Here, various reaction parameters, additives, and tridentate ligands were used without providing C2O42–. However, the limitations of single analytical methods, namely, IR and NMR spectroscopy as well as CE, were highlighted. Fortunately, the combination of these methods and suitable control experiments offered a reliable approach to investigate CO2 reductive coupling.

Finally, in the reaction of [Cu(Tp)(O2CC(O)CH(CH3)2)] (9) with air or CO2/O2, oxalate results from oxidative cleavage of the substrate with no evidence for CO2 incorporation. Interestingly, formation of oxalate via oxidative decomposition of the α-ketocarboxylate was enhanced in the presence of Cu2+ without any additional ligand. As part of these investigations, we presented the second[66] example for a potentially misleading outcome of an isotopic labeling study with 13CO2 within this field. Ambiguous results can arise from the formation of additional CO2 reduction or hydration products (formate or bicarbonate), oxidative decomposition reactions resulting in the formation of carboxylic acid or carbonyl compounds, and variable solvent contamination with, e.g., DMF.

Based on all these results, we advise the following guidelines to enhance future progress in this cutting-edge field:

Utilization of a combination of orthogonal analytical methods, for instance, FTIR spectroscopy combined with NMR spectroscopy and CE analysis, is indispensable. In addition, electrospray mass spectrometry, especially coupled with CE, should allow for appropriate validation of experimental results. Suitable alternatives to CE represent ion chromatography, high-performance liquid chromatography, or gas chromatography in combination with derivatization or oxidative decomposition of oxalic acid (the latter being less compatible with studying CO2 reduction).[67,106−110] These techniques have proven reliable in the quantification of oxalate in biological samples, foods, or waters from pulp and paper processes.

In general, a cautious strategy for reaction design and implementation of control reactions to exclude oxidative processes should be made. We highly encourage researchers working in this area to examine oxidative conditions or reactions under exclusion of CO2 to validate that a reductive process accounts for the formation of C2O42–.

While isotopic labeling with 13CO2 in combination with IR spectroscopic analysis can provide valuable insight into the mechanism of oxalate formation, we would like to raise awareness of the limited significance of the assignment of a CO2 reductive coupling mechanism based on isotopic labeling with vibrational spectroscopy as the sole analytical tool. Here, NMR spectroscopy and ESI-HRMS offer great compatibility and provide additional information to ensure the validity of the proposed mechanism and correct identification of the reaction product.

Albeit the interest in CO2 reductive coupling seems to suffer from the variety of complications highlighted herein, we believe the area offers great potential for the fundamental understanding of the reactivity of carbon dioxide. One of the apparently simplest C–C bond formations certainly deserves more attention, because a mechanistic understanding allowing for a general design of suitable catalysts enabling this transformation would be highly desirable in view of the need for efficient CO2 utilization. We believe the lessons presented throughout this article will contribute to paving the way for reliable progress within this field and, hopefully, promote the development of guiding principles for the construction of C≥2 compounds from CO2.