Carbonate-Promoted Hydrogenation of Carbon Dioxide to Multicarbon Carboxylates.

CO2 hydrogenation is a potential alternative to conventional petrochemical methods for making commodity chemicals and fuels. Research in this area has focused mostly on transition-metal-based catalysts. Here we show that hydrated alkali carbonates promote CO2 hydrogenation to formate, oxalate, and other C2+ carboxylates at elevated temperature and pressure in the absence of transition-metal catalysts or solvent. The reactions proceed rapidly, reaching up to 56% yield (with respect to CO32-) within minutes. Isotope labeling experiments indicate facile H2 and C-H deprotonations in the alkali cation-rich reaction media and identify probable intermediates for the C-C bond formations leading to the various C2+ products. The carboxylate salts are in equilibrium with volatile carboxylic acids under CO2 hydrogenation conditions, which may enable catalytic carboxylic acid syntheses. Our results provide a foundation for base-promoted and base-catalyzed CO2 hydrogenation processes that could complement existing approaches.

Introduction

Carbon dioxide hydrogenation catalysis is performed industrially for the reverse water–gas shift (RWGS: CO2 + H2 → CO + H2O) and methanation (CO2 + 4H2 → CH4 + 2H2O) reactions to adjust the composition of gases made from fossil fuel steam reforming.[1] Beyond these well-established roles, CO2 hydrogenation is a potential link between renewable energy and the production of carbon-based fuels and chemicals.[2] Renewable H2 can be produced by solar-powered H2O electrolysis with light-to-chemical energy conversion efficiency of up to 15% (>15-fold greater than photosynthesis in most plants), but its high cost relative to CH4 steam reforming and the lack of H2 infrastructure have impeded deployment.[3] CO2 hydrogenation could help address these issues by converting renewable H2 into higher-value, easily stored chemicals. Processes that make carbon–carbon (C–C) bonds are critical for this goal because multicarbon (C2+) compounds generally have higher value and more diverse applications than C1 compounds.[4]

Research in CO2 hydrogenation has been dominated by transition-metal and transition-metal-oxide catalysis. For C2+ products, recent progress has been made in the development of Fe-based catalysts that hydrogenate CO2 to hydrocarbon mixtures with substantial fractions of C5+ alkanes[5−7] or short-chain olefins (e.g., ethylene, propylene).[8] These catalysts are of interest for the production of gasoline and chemical feedstocks, although suppressing CH4 formation, controlling selectivity, and mitigating deactivation by H2O remain challenging. Researchers have also reported heterogeneous and homogeneous catalysts that hydrogenate CO2 to ethanol with high selectivity at moderate CO2 conversion.[9−12] Hydrogenating CO2 to C2+ carboxylic acids has proven to be considerably more difficult, however, with only one report showing low levels of acetic acid production.[13] This gap motivates the investigation of alternative pathways that do not involve transition-metal catalysis. Here we show that alkali carbonates (M2CO3) can promote CO2 hydrogenation to HCO2–, oxalate, and C2+ carboxylates (acetate, propionate, succinate) without the use of catalysts or solvents. Our results reveal facile base-promoted pathways for using H2 to effect C–C bond formation and provide a foundation for the development of new catalytic processes to hydrogenate CO2 to high-value products.

We recently showed that CO32– deprotonates very weakly acidic C–H bonds (pKa > 35 in organic solvents) in solvent-free mixtures of alkali carboxylate (RCO2M) and alkali carbonates (M2CO3) at intermediate temperatures (200–350 °C).[14] The resulting carbanions undergo C–C bond-forming reactions with CO2 to form carboxylates (C–CO2–). This chemistry has been used to prepare furan-2,5-dicarboxylic acid from furoic acid on a 1 mol scale with high yield, highlighting its potential for CO2 utilization.[15] Based on these results, we initially targeted a CO32–-promoted CO2 hydrogenation to oxalate by successive H2 and C–H carboxylation reactions via the following steps: (i) deprotonation of H2 to form hydride (H–); (ii) reaction of H– with CO2 to form formate (HCO2–); (iii) deprotonation of HCO2– to form carbonite (CO22–); and (iv) reaction of CO22– with CO2 to form oxalate (Figure ).

Figure 1

Proposed reaction schemes for CO32–-promoted oxalate synthesis and CO32–-catalyzed reverse water–gas shift (RWGS) reaction. Oxalate is formed by successive CO32–-promoted H2 carboxylation and HCO2– carboxylation (steps 1–4); HCO3– decomposition regenerates 1 equiv of CO32– and CO2 (step 5). RWGS is effected by HCO3–-promoted formate decomposition (steps 1–3a).

Aqueous M2CO3 solutions are known to catalyze the water–gas shift (WGS) reaction under high CO pressures at hydrothermal conditions (≥300 °C, >70 bar CO, >100 bar total pressure).[16,17] At 300 °C, the CO32– solution is converted into a HCO2– solution during this catalysis,[18] which implies that CO32– would also be converted into HCO2– by reaction with CO2 and H2 under hydrothermal conditions. These conditions cannot be used to synthesize oxalate, however, because oxalate readily decomposes in aqueous solutions at ≥300 °C.[18] Moreover, the very high pressures, corrosiveness of high-temperature alkali solutions, and large amount of water that must be removed make hydrothermal HCO2– synthesis impractical. Beyond hydrothermal conditions, K2CO3-impregnated Al2O3 has been used to catalyze gas-phase WGS at 400–500 °C.[19] IR spectroscopy has identified surface-adsorbed HCO2– under WGS conditions, which suggests that a K2CO3 or a K+-activated Al2O3 surface can promote CO2 hydrogenation to HCO2–.[20] However, HCO2– cannot accumulate nor be channeled down a pathway to produce something other than CO under these conditions. Researchers have developed many transition-metal complexes that catalyze base-promoted CO2 hydrogenation to HCO2–, some of which operate in H2O with HCO3–/CO32– as the base.[21−25] To our knowledge, however, the conversion of H2, CO2, and CO32– into isolable quantities of HCO2– under nonhydrothermal conditions in the absence of a catalyst has not previously been reported.

Methods to synthesize oxalate starting from HCO2– have been known for a long time.[26] Heating sodium or potassium formate (NaHCO2 or KHCO2) under N2 at 350–410 °C in the presence of a base results in nearly quantitative yield of oxalate and H2 as a byproduct.[27,28] Early thermoanalytical studies,[29] as well as recent DFT calculations,[30] have supported a mechanism in which C–C bond formation takes place via nucleophilic attack of carbonite on HCO2–. In contrast to NaHCO2 and KHCO2, pyrolysis of RbHCO2 or CsHCO2 under N2 results in CO and CO32– with little or no oxalate formation.[31] Cesium oxalate has previously been synthesized in high yield by heating Cs2CO3 under a mixture of CO2 and CO at 380 °C at >200 bar.[32] Conditions to produce oxalate from H2, CO2, and CO32– have not previously been described.

Result and Discussion

Given this backdrop, we hypothesized that if CO32–-promoted CO2 hydrogenation to HCO2– could be performed under nonhydrothermal conditions at 300–400 °C, then HCO2– would undergo subsequent C–C bond-forming reactions. The crystallinity and high melting points (≥800 °C)[33] of M2CO3 compounds normally restrict reactions with gaseous substrates to their surfaces. Instead of using hydrothermal conditions, we hypothesized that the addition of small amounts of H2O vapor to a CO2 hydrogenation reaction would lead to formation of M2CO3 hydrates (M2CO3·nH2O) that would increase ion mobility, enabling subsurface CO32– to react and carboxylate products to accumulate. Hydrates of inorganic alkali salts often have melting points that are several hundred degrees lower than their anhydrous counterparts.[34]

To test this hypothesis, we initially evaluated the effect of adding H2O to CO2 hydrogenations at relatively low temperatures using Cs2CO3, which is the most hygroscopic M2CO3. Reactions were performed in a stainless-steel Parr reactor using 10 mmol of Cs2CO3, and the products were quantified by gas chromatography analysis of the headspace and 1H and 13C NMR analysis of the solids. When dry Cs2CO3 powder was heated for 4 h under a 1:1 CO2:H2 mixture at 230 °C and 25 bar (all pressures are given at the reaction temperature), 830 μmol of CO and 79 μmol of HCO2– were formed (Table ). The latter corresponds to 0.4% yield with respect to CO32– (100% yield is 2 equiv of carboxylate per CO32– reactant; all carboxylate yields are calculated this way). A control experiment in the absence of Cs2CO3 resulted in 290 μmol of CO from RWGS catalyzed by the stainless-steel reactor walls, indicating that ∼500 μmol of CO was produced by Cs2CO3-catalyzed RWGS. These results are consistent with the accumulation of a layer of HCO2– on the surfaces of the Cs2CO3 particles during RWGS catalysis. When the experiment was repeated with the addition of 6 bar H2O, the morphology of the Cs2CO3 was transformed from a powder into a clear waxy solid, which is indicative of hydrate formation during the reaction. The amount of HCO2– formed increased to 605 μmol (3% yield) while only 70 μmol of CO was produced. The suppression of CO production occurs because H2O inhibits RWGS catalysis for both the reactor walls and the Cs2CO3. The amount of CO is far less than the calculated RWGS equilibrium amount in both cases. The formation of much more HCO2– than CO indicates that the HCO2– arises primarily from H– reacting with CO2 instead of HO– reacting with CO under these conditions.

Table 1

Cs2CO3-Promoted CO2 Hydrogenation at 230 °C with or without the Addition of Water

entry	Cs2CO3	pH2O	ptotal	formate	CO
1	10 mmol		25 bar	79 μmol	830 μmol
2	10 mmol	6 bar	31 bar	605 μmol	120 μmol
3			25 bar		290 μmol
4		6 bar	33 bar		50 μmol

Having established that H2O vapor aids CO32–-promoted CO2 hydrogenation to HCO2–, higher temperatures and pressures were examined to increase the yield and access C2+ products. A series of experiments were performed at 320 °C under an initially 1:1 CO2:H2 mixture at 60 bar. Under these conditions, H2O formed by RWGS was sufficient to form M2CO3·nH2O in situ, obviating the need to add H2O at the start. In the absence of M2CO3, the reactor walls catalyzed formation of CO, CH4, and much smaller amounts of ethane, ethylene, and oxygenates (Figure S7). When 2 mmol of Cs2CO3 was reacted under these conditions for 5 min, carboxylate products were formed in 56% yield, consisting of HCO2– (1.4 mmol, 35%), oxalate (413 μmol, 21%), and a very small amount of acetate (9 μmol) (Table ). As seen at lower temperature, Cs2CO3 catalyzed the RWGS reaction. Notably, Cs2CO3 did not catalyze CH4 formation that was detectable beyond the background amount produced by catalysis on the reactor walls. Increasing the reaction time resulted in a modest increase in HCO2–, decreased oxalate, and continued accumulation of other C2+ oxygenates: after 8 h, acetate, propanoate, and succinate were formed in a combined 4% yield. Control experiments in which Cs2CO3 was heated under only H2 or only CO2 atmosphere yielded no reaction (see the Supporting Information).

Table 2

CO32– -Promoted CO2 Hydrogenation at 320 °C

entry	M+	(M+)2 oxalate	timea	formate	oxalate	acetate	propionate	succinate	% yieldb	ethylenec
1	Cs+		5 min	1.4 mmol	413 μmol	9 μmol	1 μmol	1 μmol	56%	28 μmol
2	Cs+		2 h	1.8 mmol	300 μmol	23 μmol	7 μmol	9 μmol	61%	21 μmol
3	Cs+		8 h	1.9 mmol	371 μmol	50 μmol	34 μmol	31 μmol	70%	22 μmol
4	Cs+	75 μmol	8 h	2.2 mmol	278 μmol	274 μmol	125 μmol	85 μmol	79%	247 μmol
5	Cs+	500 μmol	8 h	2.4 mmol	438 μmol	42 μmol	38 μmol	31 μmol	60%	111 μmol
6	Rb+		8 h	447 μmol	120 μmol	39 μmol	7 μmol	14 μmol	19%	40 μmol
7	K+		8 h	28 μmol		34 μmol	4 μmol		2%	116 μmol

Time at 320 °C, see the SI for temperature ramp.

Total carboxylate yield with respect to CO32–.

Up to 30 μmol of C2H4 produced by reactor walls.

When Rb2CO3 was used instead of Cs2CO3, the same carboxylate products were observed, but the overall yield after 8 h was reduced to 19% (Table ). With K2CO3, the yield was reduced further to only 2%, and oxalate was no longer observed. Interestingly, ∼100 μmol of ethylene was also produced in the presence of K2CO3, indicating that it is possible to close a catalytic cycle and produce a hydrocarbon product. The lower yield of carboxylates with Rb2CO3 and K2CO3 reflects the lower hygroscopicity of these salts compared to Cs2CO3, which suggests that they are less hydrated under the reaction conditions, and consequently less of the CO32– is available to react.

The yield of CO2 hydrogenation can be improved by seeding M2CO3 with small amounts of oxalate. When 2 mmol of Cs2CO3 was combined with 75 μmol of cesium oxalate (Cs2C2O4) and subjected to the optimal CO2 hydrogenation conditions, the yields of HCO2– and oxalate were similar, but 500 μmol of acetate, propanoate, and succinate was produced, corresponding to 14% yield (Table ). This effect was only observed when small amounts of Cs2C2O4 relative to Cs2CO3 were added: when 500 μmol of Cs2C2O4 was added, the yields were very similar to reactions using Cs2CO3 alone (Table S1). Increased yields of CO2 hydrogenation were also seen when seeding Rb2CO3 with rubidium oxalate, although the main effect in this case was to increase the yield of formate and oxalate (Table S1).

The formation of oxalate, C2+ carboxylates, and ethylene demonstrates the ability of CO32– to couple CO2 hydrogenation to C–C bond formation. To probe the mechanistic pathways to these products, a series of experiments was performed using isotopically labeled substrates. We first assessed the acid–base properties of H2 and formate via H/D exchange experiments. When 0.5 mmol of deuterated cesium formate (DCO2Cs) and 0.25 mmol of Cs2CO3 were heated to 250 °C under 2.5 bar H2 (34 equiv) for 1 h, 22% incorporation of H into formate was observed with no detectable decomposition (Figure a). When the experiment was repeated in the absence of Cs2CO3, 5% H incorporation was observed. A control experiment performed with DCO2Cs and Cs2CO3 under N2 showed less than 1% H incorporation, which confirms that H2 is the source of proton rather than adventitious moisture. These results are consistent with reversible deprotonation of H2 to H– and HCO2– to carbonite (CO22–) in a HCO2Cs medium at elevated temperature, providing independent evidence for the acid–base reactions hypothesized for CO32–-promoted CO2 hydrogenation to oxalate (Figure ). Notably, cesium and potassium carbonite have been prepared and spectroscopically characterized in Ar matrices,[35] and a crystal structure has recently been obtained for carbonite bound to Ti(IV) centers.[36]

Figure 2

Evidence for reversible H2 and C–H deprotonations by CO32– in alkali salt. H/D isotope exchange between (a) DCO2– and H2; (b) CD3CO2– and H2; and (c) CH3CH2CO2– and CD3CO2–. The amount of exchange was quantified by 1H NMR.

H/D exchange was more facile with cesium acetate as the reaction medium. When 0.5 mmol of deuterated cesium acetate (CD3CO2Cs) and 0.25 mmol of Cs2CO3 were heated to 230 °C for 1 h under 2.5 bar H2, >95% H incorporation was observed (Figure b). In the absence of Cs2CO3, 81% H incorporation was observed. When CD3CO2Cs and Cs2CO3 were heated under N2 instead of H2, 20% of H/D scrambling was observed, which is attributed to exchange with adventitious H2O in the Parr reactor or salt mixture. These results provide a second example of H2 deprotonation in alkali carboxylate and indicate that acetate can be deprotonated or tautomerized to form ethene-1,1-bis(olate) (CH2CO22–) or 1-hydroxyethen-1-olate (CH2C(OH)O–) (see Figure below). The high degree of exchange in the absence of CO32– indicates that acetate itself serves as a base for these reactions. Finally, heating CD3CO2Cs, cesium propanoate (CH3CH2CO2Cs), and Cs2CO3 under N2 at 320 °C resulted in complete H/D exchange at the 2 position of propanoate but no exchange at the 3 position (Figure c).

Figure 4

Proposed pathways to C2+ products. (a) Overview of proposed mechanistic pathway for CO2 hydrogenation. (b–e) Individual steps leading to acetate, propanoate, succinate, and ethylene. The steps for RWGS catalysis and oxalate formation are shown in Figure .

We next used 13C labeling to identify possible intermediates for the formation of the higher-order C2+ carboxylate products. Unfortunately, when Cs2CO3 was seeded with Cs+ formate–13C or Cs+ oxalate–13C2, the label was rapidly lost to 12C/13C exchange with CO2 and/or decomposition during the temperature ramp to 320 °C, making it impossible to track the label into C2+ products. To avoid very rapid loss of the 13C label, it was necessary to switch to Rb+ and use modified conditions: A mixture of 0.5 mmol of Rb+ formate–13C and 1 mmol of Rb2CO3 was heated at 295 °C under 30 bar of CO2 and 3 bar of H2O for 1 h. The addition of water prevented rapid formate decomposition, and the omission of H2 prevented the formation of large amounts of nonenriched formate. The reaction resulted in 171 μmol of formate, which showed no 13C enrichment, and 8 μmol of acetate, which showed 12% 13C at the C2 position and no enrichment at C1 (Figure a and Figure S4). This result demonstrates that a small amount of formate–13C was converted into acetate before 13C/12C exchange between formate and CO2 was complete. The absence of labeling at C1 is the result of 13C/12C exchange between the C1 of acetate and CO2, which was observed in an independent experiment starting with acetate-13C2 (Figure S6).

Figure 3

13C-labeling experiments to probe C–C bond formation pathways. (a) Heating Rb+ formate–13C and Rb2CO3 under CO2 and H2O results in unlabeled formate and a small amount of acetate that is enriched in 13C at the 2 position, providing evidence that formate is a precursor to acetate. (b) A short RbCO3-promoted CO2 hydrogenation reaction seeded with Rb+ oxalate–13C2 resulted in oxalate and formate with 7% 13C labeling and a small amount of acetate with 6% and 2% 13C labeling at the 1 and 2 positions, respectively. The lower amount of 13C labeling at the 2 position of acetate relative to oxalate is inconsistent with oxalate being a precursor to acetate. (c) A CsCO3-promoted CO2 hydrogenation reaction seeded with acetate-2-13C resulted in 13C-label incorporation into propanoate and succinate, demonstrating that acetate is a precursor to these higher-order carboxylates. In all cases, the products and amount of 13C labeling were quantified by 1H and 13C NMR.

To probe if oxalate is reduced to form acetate, a mixture of 1.5 mmol of Rb2CO3 and 0.5 mmol of Rb+ oxalate–13C2 was heated at 320 °C under 60 bar 1:1 H2:CO2 using a short reaction time of 15 min to prevent complete exchange of the 13C with CO2. This reaction resulted in 448 μmol of oxalate, 261 μmol of formate, 6 μmol of acetate, and trace amounts of other products. Oxalate and formate each showed 7% 13C labeling, whereas acetate was 6% labeled at the C1 position and only 2% at the C2 position (Figure b and Figure S5). This result is inconsistent with acetate formation via oxalate reduction because this pathway would result in greater 13C labeling at C2 of acetate compared to C1 because of exchange at C1 via reversible malonate formation. Moreover, the extent of labeling at C2 of acetate should be greater than or equal to the 13C labeling remaining in oxalate. The results are consistent with an alternative C–C bond-forming step for acetate and other C2+ products (see below). Finally, to probe if acetate is the precursor to other C2+ products, 2 mmol of Cs2CO3 was seeded with 100 μmol of cesium acetate-2-13C and heated at 320 °C under 1:1 CO2:H2 at 60 bar for 8 h. The label was found in propanoate-2-13C (21 μmol), propanoate-3-13C (10 μmol), and succinate-2-13C (40 μmol), indicating that acetate is the precursor to these oxygenates (Figure c and Figure S3). The overall carboxylate yield also increased substantially with acetate seeding, similar to the effect of seeding a reaction with oxalate.

The isotope exchange experiments are consistent with the mechanistic pathways to formate and oxalate shown in Figure . These pathways are reversible, as indicated by the rapid 13C/12C exchange between 13C-labeled formate or oxalate and 12CO2. Because oxalate does not appear to be a precursor to acetate (see above), we propose that the C–C bond of acetate is formed by CO22– attack on CO2-activated HCO2– (HCO2CO2–), resulting in 2-oxoacetate (glyoxylate) and expulsion of CO32– (Figure b). Glyoxylate could be converted into acetate by the following sequence: (i) H– reduction of the aldehyde to form 2-oxidoacetate; (ii) addition to CO2 to form 2-carboxylatooxy acetate; and (iii) H– displacement of CO32–. The rapid 13C/12C exchange between acetate-1-13C and 12CO2 may be explained by reversible CO32–-promoted C–H carboxylation of acetate to form malonate. These pathways are consistent with the oxalate–13C2 experiment. The small amount of 13C enrichment at C2 of acetate in this reaction was likely produced from H13CO2– that came from oxalate–13C2 decomposition. The labeling at C1 most likely arose from 12C/13C exchange between unlabeled acetate and 13CO2 produced by oxalate–13C2 decarboxylation. This interpretation assumes that CO2 produced in the salt from oxalate decarboxylation can react with nucleophiles before escaping to the gas phase.

In a similar manner, we propose that propanoate is formed by deprotonation of acetate to form CH2CO22–, attack on HCO2CO2– to form 3-oxopropanoate, and analogous H– additions to reduce the aldehyde (Figure c). This pathway could account for the formation of propanoate-2-13C when seeding the reaction with acetate-2-13C. The formation of propanoate-3-13C from acetate-2-13C suggests an additional pathway wherein CH2CO22– attacks acetate in an SN2-like fashion to expel CO22– as a leaving group. The inability to deprotonate the 3 position of propanoate indicates that it is not a precursor to succinate. We propose that succinate is formed by attack of CH2CO22– on 2-oxoacetate or 2-oxidoacetate, followed by H– reduction (Figure d). This pathway is consistent with the presence of only one 13C label in succinate when a reaction is seeded with acetate-2-13C. A possible pathway to ethylene formation consists of the following: (i) addition of acetate to CO2 to form acyl carbonate; (ii) H– addition to form acetaldehyde and expel CO32–; (iii) H– addition to form ethoxide; (iv) addition to CO2 to form monoethylcarbonate; and (v) E2 elimination (Figure e). This pathway closes the cycle to render CO2 hydrogenation catalytic in CO32–.

Channeling CO2 hydrogenation toward acetate or higher-order products is intrinsically challenging because of the large number of elementary steps required and the multitude of branching points. Carbonate-promoted CO2 hydrogenation has the additional complexities that the reaction medium is heterogeneous, and it changes substantially as CO32– is consumed and carboxylate products accumulate.

The boost in the yield of acetate, propanoate, and succinate by doping the starting M2CO3 salt with small amounts of acetate or oxalate may arise from changes to the hygroscopicity or other physical properties of the salt mixture that affect the initial rates of hydrogenation and C–C bond formation. Further adjustment of the initial salt composition or morphology may provide an avenue to improve the C2+ carboxylate yields.

An alternative to trying to optimize carboxylate yield is to close a catalytic cycle by protonating a carboxylate product and removing a volatile carboxylic acid. To investigate whether carboxylates are in equilibrium with volatile carboxylic acids via protonation by H2O under the reaction conditions, a trapping experiment was performed. The Parr reactor was configured with a mixture of 0.5 mmol of acetate-2-13C and 1 mmol of Rb2CO3 in a glass vial and a separate sample of 2 mmol of Rb2CO3 at the bottom of the glass insert to act as an acetic acid trap. After 1 h at 320 °C under 60 bar 1:1 CO2:H2, NMR analysis of the solids in the vial and in the insert showed a nearly 1:1 distribution of the acetate-2-13C in the two containers with 75% recovery, which indicates that carboxylates are in equilibrium with volatile carboxylic acids. Similar results were obtained with Cs2CO3 (Table S2). This result suggests that it is possible to close the cycle and use CO32– as a catalyst for CO2 hydrogenation to C2+ carboxylic acids, which could be removed and collected using a flow reactor.

The reversibility of formate and oxalate formation enables selective oxalate synthesis in a one-pot, two-step process. In the first step, 1 mmol of Cs2CO3 was heated to 340 °C under 60 bar of 1:1 CO2/H2, which yields a product mixture composed predominantly of formate and oxalate. This material was then subjected to 60 bar of CO2 at 340 °C for 2 h. The final product mixture was composed of 417 μmol of oxalate (42%), unreacted Cs2CO3, and trace amounts of other carboxylates (Figure S9). The procedure is selective for oxalate because some of the formate produced in the first step undergoes C–H carboxylation to oxalate in the second step while the remainder is decomposed back to CO32– and CO2/H2 or CO/H2O. Oxalate is potentially a useful feedstock for the synthesis of other valuable C2 compounds. For example, dimethyl oxalate can be hydrogenated very efficiently and selectively to either ethanol or ethylene glycol using various heterogeneous catalysts.[37,38] Combining this chemistry with the CO32–-promoted CO2 hydrogenation described here requires an efficient way to convert cesium oxalate into dimethyl oxalate.

In summary, we have described conditions to convert alkali carbonate, CO2, and H2 into carboxylates in the absence of catalysts or solvent. The major products are formate and oxalate, but substantial quantities of acetate, propanoate, succinate, and ethylene can also be produced. The reactions occur at intermediate temperatures and pressures and require H2O vapor to hydrate the alkali carbonates. H2 and C–H bonds are readily deprotonated under these conditions, generating H– and C-centered nucleophiles that undergo H–C and C–C bond-forming reactions. This chemistry may provide a foundation for designing base-catalyzed CO2 hydrogenation reactions that generate C2+ products.