Efficient and selective molecular catalyst for the CO₂-to-CO electrochemical conversion in water

Significance

CO₂-to-CO electrochemical conversion is a key step in the production of liquid fuels through dihydrogen-reductive Fischer–Tropsch chemistry. Among molecular catalysts, iron porphyrins reduced electrochemically to the Fe(0) state are particularly efficient and led to a deeper understanding of mechanisms involving coupled bond-breaking proton–electron transfer processes. The replacement of nonaqueous solvents by water should make the CO₂-to-CO half-cell reaction much more attractive for applications, particularly because it would allow association with a water-oxidation anode through a proton-exchange membrane. Here it is demonstrated that electrochemical CO production catalyzed by a water-soluble iron porphyrin can occur with high catalytic efficiency. Manipulation of pH and buffering then allows conversions from those involving complete CO selectivity to ones with prescribed CO–H₂ mixtures.

Abstract

Substitution of the four paraphenyl hydrogens of iron tetraphenylporphyrin by trimethylammonio groups provides a water-soluble molecule able to catalyze the electrochemical conversion of carbon dioxide into carbon monoxide. The reaction, performed in pH-neutral water, forms quasi-exclusively carbon monoxide with very little production of hydrogen, despite partial equilibration of CO₂ with carbonic acid—a low pK_a acid. This selective molecular catalyst is endowed with a good stability and a high turnover frequency. On this basis, prescribed composition of CO–H₂ mixtures can be obtained by adjusting the pH of the solution, optionally adding an electroinactive buffer. The development of these strategies will be greatly facilitated by the fact that one operates in water. The same applies for the association of the cathode compartment with a proton-producing anode by means of a suitable separator.

One of the most important issues of contemporary energy and environmental challenges consists of reducing carbon dioxide into fuels by means of sunlight (1–3). One route toward this ultimate goal is to first convert solar energy into electricity, which will then be used to reduce CO₂ electrochemically. Direct electrochemical injection of an electron into the CO₂ molecule, forming the corresponding anion radical CO₂^.− requires a very high energy [the standard potential of the CO₂/ CO₂^.− couple is indeed −1.97 V vs. normal hydrogen electrode (NHE) in N,N′dimethylformamide (DMF)] (4, 5). Electrochemical conversion of CO₂ to any reaction product thus requires catalytic schemes that preferably avoid this intermediate. Carbon monoxide may be an interesting step en route to the desired fuels because it can be used as feedstock for the synthesis of alkanes through the classic Fischer–Tropsch process. A number of molecular catalysts for the homogeneous electrochemical CO₂-to-CO conversion have been proposed. They mainly derive from transition metal complexes by electrochemical generation of an appropriately reduced state, which is restored by the catalytic reaction. So far, nonaqueous aprotic solvents (mostly DMF and acetonitrile) have been used for this purpose (5–16). Brönsted acids have been shown to boost catalysis. However, they should not be too strong, at the risk of leading to H₂ formation at the expense of the CO. Trifluoroethanol and water (possibly in large amounts) have typically played the role of a weak acid in the purpose of boosting catalysis while avoiding hydrogen evolution.

One of the most thoroughly investigated families of transition-metal complex catalysts of CO₂-to-CO conversion is that of iron porphyrins brought electrochemically to the oxidation degree 0. The importance of coupling electron transfer and introduction of CO₂ into the coordination sphere of iron with proton transfers required by the formation of CO, CO₂ + 2e⁻ + 2AH ↔ CO + H₂O + 2A⁻, appeared from the very beginning of these studies. Sustained formation of CO was indeed only achieved upon addition of weak Brönsted (17–19) and Lewis acids (20, 21). Such addition of Brönsted acids, however, opens the undesired possibility that the same catalyst that converts CO₂ into CO may also catalyze the reduction of the acid to dihydrogen. This is indeed what happens with Et₃NH⁺ as the acid, limiting the set of acids used for the CO₂-to-CO conversion to that of very weak acids such as propanol, water, and, the strongest one, trifluorethanol (22). Since then, the range of acidity has been extended to phenols, which proved compatible with CO faradic yields close to 100% (23, 24). Installation of phenol functionalities in the porphyrin molecules (Scheme 1) even allowed a considerable improvement of catalysis in terms of catalytic Tafel plots (turnover frequency vs. overpotential) with no degradation of the CO (vs. H₂) faradaic yield (25–27). The problem should, however, resurface upon going to stronger acids. That competition with hydrogen evolution is a general issue for molecular catalysis of the CO₂-to-CO conversion is confirmed by recent findings concerning catalysts derived from terpyridine complexes of first-row transition metals in (90:10, vol:vol) DMF/H₂O mixtures (28).

Scheme 1.

Iron-porphyrin catalysts for CO₂-to-CO electrochemical conversion.

The results thus obtained in nonaqueous or partially aqueous media enabled the discovery of remarkably efficient and selective catalysts of the CO₂-to-CO conversion. They were also the occasion of notable advances in terms of mechanisms and theory of concerted bond-breaking proton–electron transfer (29).

It must, however, be recognized that, from the point of view of practical applications, the use of nonaqueous solvents is not the most exciting aspect of these results. One would rather like to use water as the solvent, which would render more viable the CO₂-to-CO half-cell reaction as well as its association with a water-oxidation anode through a proton-exchange membrane. Several approaches are conceivable in this respect. One of these consists of coating the electrode with a film, possibly making use of the water insolubility of the catalyst (see the three last entries of table 3 in ref. 5, refs. 30 and 31, and ref. 32 and references therein). One may also attempt to chemically attach catalyst molecules onto the electrode having in mind their use, or possible use, of the resulting films in water (33, 34).

Our approach consisted of devising a catalyst fully soluble in water and able to convert CO₂ to CO selectively as well as efficiently with regard to overpotential and turnover frequency. Full solubility in water, indeed, allows an easy manipulation of pH and buffering of the system. It will also help the design of a full cell associating the cathode compartment with a proton-producing anode by means of a suitable separator (35). The challenges ahead in this endeavor were as follows. CO₂ is poorly soluble in water ([CO₂] = 0.0383 M) (36). Even more seriously, it is partially converted (K_hydration = 1.7 × 10⁻³) into carbonic acid, CO₃H₂, which has a first ionization pK_a of 3.6, that is, an apparent pK_a of 6.4 (36). These features make it a priori possible that the CO₂-to-CO conversion might be seriously challenged by H₂ evolution from reduction of carbonic acid and/or hydrated protons (37, 38). Indeed, a previous attempt showed poor CO₂/H₂ selectivity and practically no catalytic effect in cyclic voltammetry (37). In the same vein, Ni cyclam seems to be a selective and efficient catalyst of the CO₂-to-CO conversion in water under the condition of being operated with a mercury electrode, pointing to favorable specific interactions of the catalytic species with the mercury surface (38, 39), whereas the use of a carbon electrode leads to much less efficiency in terms of rate (maximal turnover frequency of 90 s⁻¹) and turnover (four cycles) (15).

We have found that the iron tetraphenylporphyrin in which the four paraphenyl hydrogens have been substituted by trimethylammonio groups, hereafter designated as WSCAT (Scheme 1), fulfills these stringent requirements at neutral pHs.

As seen in Fig. 1A, a very high catalytic current is observed in cyclic voltammetry of half a millimolar solution of WSCAT saturated with CO₂ at pH 6.7. In the absence of CO₂ (Fig. 1B) three successive waves are observed when starting from the Fe^III complex with Cl⁻ as counter ions and presumably as axial ligand. As expected, the shape of the first, Fe^III/II, wave reflects the strong axial ligation by Cl⁻ (40). The second wave is a standard reversible Fe^II/I wave. The third wave is the wave of interest for catalysis. It is irreversible, as opposed to what is observed in DMF (Fig. 1C), where all three waves are one-electron reversible waves, including the third Fe^I/0 wave, as with the simple FeTPP (Scheme 1) (see, e.g., refs. 23 and 24). The irreversibility and somewhat increased current observed here in water presumably reflects some catalysis of H₂ evolution from the reduction of water. The considerable increase in current observed in the Fe^I/0 potential region when CO₂ is introduced into the solution is a clear indication that catalysis is taking place. It is roughly similar to what has been observed in DMF under 1 atm CO₂ in the presence of a weak acid such as phenol (Fig. 1D and refs. 23 and 24).

Fig. 1.

Cyclic voltammetry of 0.5 mM WSCAT at 0.1 V/s, temperature 21 °C. (A) In water + 0. 1 M KCl brought to pH 6.7 by addition of KOH under 1 atm CO₂. (B) Same as A but in the absence of CO₂. (C) Same as B but in DMF + 0.1 n-Bu₄NBF₄. (D) Same as C but under 1 atm CO₂, and presence of 3 M phenol.

What is the nature of the catalysis observed in the Fe^I/0 potential region? Does it involve acid reduction (CO₃H₂ and/ or H⁺) or conversion of CO₂ into one of the reduction products? The answer is given by the results of preparative-scale electrolyses.

A first series of preparative-scale electrolyses was performed at −0.97 V vs. NHE over electrolysis times between 1 and 4 h. The pH was brought to 6.7 by addition of KOH. The current density was ca. 0.1 mA/cm² in all cases. CO was found to be largely predominant with formation of only a very small amount of hydrogen. Over five of these experiments the average faradaic yields of detected products were as follows: CO, 90%; H₂, 7%; acetate, 1.4%; formate, 0.7%; and oxalate, 0.5%. The catalyst was quite stable during these periods of time. The variation of pH during electrolysis was small, passing from 6.7 to 6.9. The decrease in peak current registered before and after electrolysis was less than 5%.

A longer-duration electrolysis (Fig. 2) was carried out at a somewhat less negative potential, −0.86 V vs. NHE, leading to a quasi-quantitative formation of CO (faradaic yield between 98% and 100%). After 72 h of electrolysis, the current was decreased by approximately half but CO continued to be the only reaction product.

Fig. 2.

Electrolysis of a 0.5 mM WSCAT solution in water + 0.1 M KCl brought to pH 6.7 by addition of KOH under 1 atm CO₂ at −0.86 V vs. NHE. Temperature 21 °C. The reaction products were analyzed at the end of each day. Charge passed (Top); current density (Bottom).

As an example of pH manipulation, the addition of a 0.1 M formic acid buffer at pH 3.7 resulted in the exclusive formation of hydrogen. With a 0.1 M phosphate buffer adjusted at the same pH, 6.7, as that where the electrolyses with no additional buffer were carried out, a 50–50 CO–H₂ mixture was obtained.

Although deserving a precise kinetic analysis, a likely interpretation of the factors favorable to CO formation against H₂ evolution is summarized in Scheme 2. Despite Fe⁰ porphyrins’ being good catalysts of H₂ evolution (22), the fact that the reaction that converts CO₂ into CO₃H₂ is thermodynamically uphill and kinetically slow favors the CO-formation pathway, making it able to resist the fast H₂ evolution pathway in pH-neutral media. Another favorable factor is the wealth of H-bonding possibilities offered by water, a factor that has been shown to be quite important in stabilizing the primary Fe⁰–CO₂ adduct and therefore in boosting catalysis as transpires from the results obtained with OH-substituted tetraphenylporphyrins CAT and FCAT (Scheme 1) (25–27). Addition of a buffer, such as phosphate in 0.1 M concentration, opens an additional catalytic pathway for H₂ evolution able to compete with the CO-formation pathway even if the pH is kept at the same neutral value. Indeed, along this pathway, the delivery of protons (by PO₄H₂⁻) is not under a stringent kinetic limitation as it is in the pathway represented in Scheme 2. It is nevertheless clear that a future detailed mechanistic and kinetic investigation of the effect of pH changes in buffered and nonbuffered media is clearly warranted, based on cyclic voltammetry and determination of the CO and H₂ faradaic yields.

Scheme 2.

Competition between CO formation and H₂ evolution.

In view of the paucity of data concerning homogeneous molecular catalysis of the CO₂-to-CO conversion in water, benchmarking with other catalysts in terms of overpotential and turnover frequency does not seem possible at the moment. We may just note that a preliminary estimation of the maximal turnover frequency through the foot-of-the-wave analysis (discussed below) leads to the exceptionally high figure of 10⁷ s⁻¹ (i.e., a second-order rate constant of 2.5 × 10⁸ M⁻¹⋅s⁻¹).

We are thus led to make the comparison with the characteristics of other catalysts obtained in an aprotic solvent such as DMF or acetonitrile, starting from the results shown in Fig. 1 C and D. The standard potential of the Fe^I/Fe⁰ couple in DMF (Fig. 1C) is −1.23 V vs. NHE. A systematic analysis of the wave obtained under 1 atm CO₂ and presence of 3 M phenol was carried out according to the foot-of-the wave approach, which aims at minimizing the effects of side phenomena that interfere at large catalytic currents. This technique has been previously described in detail and successfully applied in several instances (23–27). It was applied here, assuming that the reaction mechanism is of the same type as for FeTPP in the presence of phenol (Scheme 3) (24).

Scheme 3.

Likely mechanism of the CO₂-to-CO conversion pathway.

Combination of the foot-of-the wave analysis with increasing scan rates, which both minimize the effect of side phenomena (23–27), allowed the determination of the turnover frequency as a function of the overpotential, leading to the catalytic Tafel plot for the WSCAT catalyst shown as the red curve in Fig. 3. The turnover frequency, TOF, takes into account that the molecules that participate in catalysis are only those contained in the thin reaction-diffusion layer adjacent to the electrode surface in pure kinetic conditions (23, 24). The overpotential, η, is the difference between the standard potential of the reaction to be catalyzed and the electrode potential. Correlations between TOF and η provide catalytic Tafel plots that are able to benchmark the intrinsic properties of the catalyst independently of parameters such as cell configuration and size. Good catalysts appear in the upper left corner and bad catalysts in the bottom right corner of Fig. 3. These plots allow one to trade between the rapidity of the catalytic reaction and the energy required to run it. The other Tafel plots shown in Fig. 3 are simply a repetition of what has been established in detail in ref. 24.

Fig. 3.

Catalytic Tafel plots in DMF or acetonitrile. See ref. 27 for details and references.

It clearly appears that WSCAT is the best catalyst of the set of molecules represented in Fig. 3. It is expected that the electron-withdrawing properties of the para-N-trimethylammonium groups leads to a positive shift of the Fe^I/Fe⁰ couple, being thus a favorable factor in terms of overpotential (positive shifts of 200 mV vs. FeTPP, 100 mV vs. CAT, and 40 mV vs. FCAT). What is more surprising is that this effect, which tends to decrease the electron density on the iron and porphyrin ring at the oxidation state 0, does not slow down the catalytic reaction to a large extent. Systematic analysis of the factors that control the standard potential and the catalytic reactivity as a function of the substituents installed on the porphyrin ring is a clearly warranted future task.

For the moment, we may conclude that substitution of the four parahydrogens of FeTPP by trimethylammonium groups has produced a water-soluble catalyst that is able, for the first time to our knowledge, to catalyze quantitatively the conversion of carbon dioxide into carbon monoxide in pH-neutral water with very little production of hydrogen. This seems a notable result in view of the hydration of CO₂, producing carbonic acid—a low pK_a acid—the catalytic reduction of which, and/or of the hydrated protons it may generate, into hydrogen might have been a serious competing pathway. Although a number of mechanistic details should be worked out, this notable result seemingly derives not only from the relatively small value of the hydration constant but also from the slowness of this reaction. As judged from its performances in DMF, this catalyst moreover seems particularly efficient in terms of catalytic Tafel plots relating the turnover frequency to the overpotential. Manipulation of the pH under unbuffered conditions or by introduction of an additional buffer should open the way to the production of CO–H₂ mixtures in prescribed proportions. The substitution by the four trimethylammonium groups ensured the water solubility of the catalyst. It may also be an adequate way of immobilizing it onto the electrode surface by means of a negatively charged polymer to be associated with a water-oxidation anode through a proton-exchange membrane.

Materials and Methods

Cyclic voltammograms were obtained in a three-electrode cell, the working electrode being a 3-mm-diameter glassy carbon disk. The experiments were carried out either under argon or carbon dioxide atmosphere at 21 °C, with a careful compensation of the ohmic drop. Electrolyses were performed with a carbon crucible as working electrode, the volume of the solution was 5 mL, and the active surface area was ca. 16 cm². In this case the ohmic drop was minimized by immersing the reference electrode directly into the solution as close as possible to the working electrode. Then, the quantification of the products (CO and H₂) was performed by chromatography analyses of the gas evolved in the headspace, with calibration curves for both H₂ and CO being determined separately by injecting known quantities of pure gas. Experimental details concerning the synthesis of WSCAT and full complementary experimental details can be found in Supporting Information.