Report

A Local Proton Source Enhances CO2 Electroreduction to CO by a Molecular Fe Catalyst

See allHide authors and affiliations

Science  05 Oct 2012:
Vol. 338, Issue 6103, pp. 90-94
DOI: 10.1126/science.1224581

Lending a Hand to CO2 Reduction

Although plants and microbes have been reducing CO2 with ease for millennia, people still find it extremely challenging. A cost-effective synthetic scheme for transforming CO2 into fuels and commodity chemicals would be a double boon, lowering atmospheric concentration of the greenhouse gas while supplementing (and perhaps ultimately replacing) dwindling petroleum feedstocks. Toward this end, Costentin et al. (p. 90) show that an iron catalyst for the electrochemical reduction of CO2 to CO gets an efficiency boost from phenol substituents appended to the ligand framework.

Abstract

Electrochemical conversion of carbon dioxide (CO2) to carbon monoxide (CO) is a potentially useful step in the desirable transformation of the greenhouse gas to fuels and commodity chemicals. We have found that modification of iron tetraphenylporphyrin through the introduction of phenolic groups in all ortho and ortho′ positions of the phenyl groups considerably speeds up catalysis of this reaction by the electrogenerated iron(0) complex. The catalyst, which uses one of the most earth-abundant metals, manifests a CO faradaic yield above 90% through 50 million turnovers over 4 hours of electrolysis at low overpotential (0.465 volt), with no observed degradation. The basis for the enhanced activity appears to be the high local concentration of protons associated with the phenolic hydroxyl substituents.

The catalytic reductive transformation of carbon dioxide (CO2) to fuels and commodity chemicals is one of the most important contemporary energy and environmental challenges. A highly negative potential is required to inject an electron into CO2 (1, 2). Reaction pathways that would go through the intermediacy of the resulting anion radical are therefore quite unreasonable in terms of energy and activation. Focusing on the conversion to carbon monoxide (CO), a number of catalysts—mostly coordination complexes of low oxidation state transition metals—have been described (35). Among them, iron(0) porphyrins, electrochemically generated from the iron(II) porphyrin by two successive electron uptakes at a mercury or a glassy carbon electrode, are efficient, CO-selective, and durable catalysts provided they are coupled, in the framework of an electron-push-pull process, with Lewis acids (6) or weak Brönsted acids (7, 8). On the basis of this observed favorable role of proton donors, we reasoned that acid groups attached to the catalyst molecule should have a strong accelerating effect in view of the large local concentration of acid thus present, which would be impossible to introduce in such amounts in solution in the context of bimolecular reaction.

We have indeed found that modification of tetraphenylporphyrin (TPP) through the introduction of phenolic groups in all ortho and ortho′ positions of the TPP phenyl groups, as shown in Fig. 1, leads to a considerable increase of catalytic activity. This is shown in Fig. 2, in which we plot the log of the turnover frequency (turnover number per unit of time), TOF, against the overpotential, η (difference between the standard potential of the CO2/CO couple and the operating electrode potential). The variation of the TOF with the overpotential obtained from cyclic voltammetry of FeTDHPP in N,N′-dimethylformamide (DMF) + 2M H2O in the presence of a saturating concentration of CO2 (0.23 M) is shown as a thick gray segment.

Fig. 1

Investigated iron porphyrins.

Fig. 2

Correlation between turnover frequency and overpotential for the series of CO2-to-CO electroreduction catalysts listed in Table 1. Thick gray segments indicate TOF values derived from “foot-of-the-wave analysis” of the cyclic voltammetric catalytic responses of FeI/0TDHPP and FeI/0TDMPP in the presence of 2 M H2O. Dashed lines indicate Tafel plots for Fe0TDHPP (top) and FeI/0TDMPP (bottom). Also shown are TOF and η values from preparative-scale experiments: The star indicates Fe0TDHPP (this work), and circled numbers indicate the published references for other catalysts specified in Table 1.

For such molecular catalytic reactions, in which the catalyst is a well-defined molecule, with a well-defined standard potential, turnover frequency and overpotential were traditionally viewed as independent parameters, not allowing a precise comparison of catalyst performances. The obvious assertion that a good catalyst is characterized by a high TOF and a small η (and vice versa) is indeed not very helpful in this purpose. It has recently been shown (9) that turnover frequency and overpotential are in fact linked, for a given catalyst, by a relationship that may be formulated as in Eq. 1 (10) (f = F/RT, where F is Faraday’s constant and R the gas constant):Embedded Image(1)Embedded Image is the rate constant of the catalytic reaction, Embedded Image the standard potential of the catalyst, kS its standard rate constant, and D its diffusion coefficient. An alternative and equivalent formulation in Eq. 2 introduces the overpotential, η, defined earlier, and the turnover frequency at zero overpotential, TOF0 (Eq. 3):Embedded Image(2)Embedded Image (3)(Embedded Image is the standard potential of the global reaction being catalyzed.)

This formulation characterizes the catalyst with three parameters: Embedded Image, the standard potential of the catalyst redox couple, Embedded Image, a measure of intrinsic catalytic activity from a chemical standpoint, and Embedded Image, a measure of the conversion efficiency of the oxidized precatalyst to the reduced active catalyst, relative to mass transport. A graphical representation of Eq. 2 is given in Fig. 2 (dashed line), taking as an illustrative example the FeTDHPP catalyst. There are three successive overpotential domains. At large η, when the electrode potential is set well above the catalyst standard potential, the TOF is governed solely by the catalytic rate constant, regardless of the overpotential. In the opposing situation (Ε>>Embedded Image), logTOF is a linearly increasing function of the overpotential with slope f ʹ = f / ln10 (1/59.3 mV at 25°C). In the transition between these two regimes, the system is controlled partly by the electron transfer and transport term, Embedded Image, giving rise to a linearly increasing function of the overpotential, with slope half the value in the preceding domain. With fast electron transfer catalysts, this intermediary zone tends to vanish.

One consequence of the existence of such a relationship is that one may select the electrolysis electrode potential with no necessity that it should be close to or more negative than the catalyst standard potential. As exemplified later on, it may indeed be more advantageous to operate at low overpotential, even though the resulting TOF decreases with overpotential.

The logTOF versus η correlation diagram in Fig. 2 provides the basis for a rational comparison of the performances of the various molecular catalysts reported so far for the electroreduction of CO2 to CO. Construction of the diagram requires an estimation of the standard potential of the CO2/CO couple, Embedded Image, in order to assign a value to the overpotential in each case. Embedded Image is known as a function of the pH when water is the solvent (11). Adaptation to DMF and acetonitrile, taking into account the stronger acid present, leads to the values given in the first column of Table 1, resulting from estimation that takes into account the electron-, acid-, and CO2-stoichiometry. Without going into the full mechanistic details, the catalytic reduction of CO2 to CO encompasses the stoichiometric reactions depicted in Fig. 3. There are two cases: If the acid introduced the solution is stronger than (CO2 + H2O), Embedded Image depends on the pKa (where Ka is the acid dissociation constant) of the acid present. In the opposite case, (CO2 + H2O) is the strongest acid in the medium. It thus neutralizes the conjugate base of the acid present, giving rise to a stoichiometry in which three molecules of CO2 are involved in the two-electron reaction (Fig. 3). The way in which the values of Embedded Image reported for each case in Table 1 were obtained on these bases is detailed in the supplementary materials.

Table 1

Catalysis of CO2 reduction to CO, showing correlation between turnover frequency and overpotential for the series of catalysts listed.

View this table:
Fig. 3

Simplified reaction scheme for CO2 reduction by iron(0) porphyrins.

The star in Fig. 2 indicates the results of a preparative scale CO2 electrolysis by using electrochemically generated Fe0TDHPP as the catalyst (fig. S1 and accompanying text). After 2 hours at a potential of –1.16 V versus normal hydrogen electrode (NHE), gas chromatographic analysis showed a 94% faradaic yield of CO, with 6% competing hydrogen formation; the averaged current density was 0.31 mA/cm2 (fig. S5). This corresponds to logTOF = 3.5 at 0.466 V overpotential. The catalyst is also remarkably stable: Fifty million turnovers were achieved after 4 hours of electrolysis at this potential, with no degradation of the iron complex. The reason for selecting an electrolysis potential more positive than the standard potential of the catalyst couple derives from the subsequent considerations.

Rather than deriving the logTOF versus η relationship characterizing the Fe0TDHPP catalyst from a series of preparative scale electrolyses at various potentials, it can be obtained in a more rapid and detailed manner by treating the catalytic cyclic voltammetric responses of FeTDHPP as follows. In the absence of CO2, FeIIITDHPP shows three waves, in DMF, corresponding successively to the FeIII/FeII/FeI/Fe0 redox couples (Fig. 4A). As with FeTPP (68), catalysis takes place at the most negative wave, meaning that the catalyst is the iron(0) complex. In the absence of CO2, the FeI/Fe0 wave is not quite reversible at the slow scan rate, 0.1 V/s, at which the catalytic experiments are run. Raising the scan rate allows the determination of the FeI/Fe0 standard potential, Embedded Image= –1.333 V versus NHE (fig. S3). Introduction of CO2 results in a 60-fold increase of the current at the level of the FeI/Fe0 wave, (Fig. 4B) indicating a fast catalytic reaction. As discussed in detail elsewhere (9), the S-shaped current potential curve expected in these conditions is hampered by the occurrence of side phenomena such as substrate consumption, self-inhibition, deactivation of the catalyst, and possibly others that interfere more and more as the current increases. These undesirable phenomena are thus minimized at the foot of the current potential curve. Shown in Fig. 4, C and D, is how inspection of the foot of the catalytic current potential response can therefore be used to obtain the TOF as a function of the electrode potential. The foot-of-the-wave analysis involves a portion of the current-potential curve that stands at potentials substantially more positive than Embedded Image. It consists in plotting a fitting function FIT (supplementary text, section 6) against Embedded Image. The value of Embedded Image is then derived from the slope of the linear portion of the FIT diagram, as shown in Fig. 4, D, F, and H. Using this value to obtain the logTOF–η relationship (Eqs. 1 to 3) closes the derivation of the catalyst characteristics from easy-to-measure cyclic voltammetric data as opposed to the more tedious preparative-scale tests. The consistency of the two approaches is confirmed by the fact that the point representing the result of preparative-scale catalysis (Fig. 2, star) stands on the logTOF versus η curve derived from cyclic voltammetry. This also explains why the electrode potential for preparative scale electrolysis was selected to be more positive (by 170 mV) relative to the catalyst standard potential: In this potential region, the abovementioned side-phenomena, which would have perturbed the preparative-scale electrolysis as they perturbed the cyclic voltammetric responses, are minimized. It is in this low overpotential region that the catalyst is used at the best of its catalytic capabilities: A current density of 0.31 mA/cm2 at a 0.466 V overpotential is obtained even though the operating potential, –1.16 V versus NHE, is then well below the catalyst standard potential, –1.33 V versus NHE.

Fig. 4

Cyclic voltammetry in DMF + 0.1 M n-Bu4NPF6 electrolyte at 0.1 V/s of 1 mM of the three iron porphyrins (Fig. 1) after normalization to the FeII/FeI peak current, Embedded Image. (A) FeTDHPP +2 M H2O. (B) FeTDHPP +2 M H2O in the presence (upper trace) and absence (lower trace) of 0.23 M CO2. (C) FeTDHPP +2 M H2O in the presence of 0.23 M CO2. (E) FeTDMPP +2 M H2O in the presence of 0.23 M CO2. (G) FeTPP + 3 M PhOH in the presence of 0.23 M CO2. (D, F, and H) Foot-of the-wave analyses of the voltammograms in (C), (E), and (G), respectively.

We may then compare, using the logTOF-vesus-η representation of Fig. 2, the performances of the present Fe0TDHPP catalyst with those of the other molecular catalysts reported in the literature in DMF and CH3CN as solvents. We excluded from the comparison a few systems for the following reasons. First, instable systems decompose after a few turnovers, as in (12) and (13). Then, the attractive system described in (14) likely involves heterogeneous interaction of a nickel cyclam complex with the mercury electrode surface [a comparison with a glassy carbon electrode is provided in (15)]. It is thus not possible to estimate the amount of catalyst participating in the reaction and hence to make a reliable comparison with the other catalysts in Table 1. Likewise, recent studies in which imidazolium (16) or pyridinium cation (17) are added to the aqueous solution and the electrode material (silver in the first case, platinum in the second) obviously plays a crucial, but still mysterious, role are clearly beyond the scope of molecular catalysis. The way in which the representative point of each catalyst was derived from the corresponding literature data are detailed in the supplementary text and in Table 1, which gives the values of TOF0 in each case. It appears that the Fe0TDHPP catalyst is slightly more efficient in terms of TOF (by a factor of ~10) than the most efficient catalyst previously reported. The latter is a rhenium complex, whereas the FeTDHPP catalyst uses a much cheaper metal.

Coming back to cyclic voltammetry, the catalytic properties of Fe0TDMPP were compared with those of Fe0TDHPP in order to highlight the essential role of the OH protons in the remarkable efficiency of the latter catalyst. The catalytic Fe0TDMPP wave (fig. S2 provides more cyclic voltammetry of FeIIITDMPP in the absence and presence of CO2) and the associated foot-of-the-wave analysis, which underlies the lower dashed line in Fig. 2, are shown in Fig. 4, E and F. In the potential domain examined, this catalyst gives rise to rather high TOFs. However, this activity comes at the cost of large overpotentials. The comparison made at the level of intrinsic properties, as captured by TOF0, shows that Fe0TDMPP is a considerably poorer catalyst than Fe0TDHPP by a factor of ~1 billion, which is not even worth assaying at a preparative scale.

The comparison highlights the crucial role of the phenolic protons in this venture. This is confirmed by the observation that the TPPFe(I)/Fe(0)CO2 catalytic wave increases with the addition of phenol in the solution (9). The exact mechanism of the interference of phenol is not known at present but is likely to be of the same push-pull (electron-proton) type as previously proposed for other acids (8). Whatever the details of the mechanism, the enhanced FeTDHPP catalytic activity is thus related to the very high local concentration of phenolic protons. In this context, it is interesting to compare quantitatively the catalytic reactivities of FeTDHPP and of FeTPP in the presence of a high concentration of phenol. The catalytic response of FeTPP in the presence of 3 M phenol is shown in Fig. 4G, together with the corresponding foot-of-the-wave analysis in Fig. 4H. The corresponding characteristics of this catalyst are Embedded Image= –1.41 V versus NHE and Embedded Image= 3.2 × 104 s−1. The latter figure is to be compared with the rate constant for FeTDHPP, 1.6 × 106 s−1, leading to an estimate that the eight phenolic OH groups in the molecule are equivalent to a 150 M phenol concentration. Similar reasons are likely to explain the high reactivities observed in the presence of a local proton donor in other cases, such as in the cleavage of O-O bonds, concerted with proton and electron transfer (18), in the catalytic conversion of O2 into H2O (19), and in the catalysis of hydrogen production and oxidation (20).

Supplementary Materials

www.sciencemag.org/cgi/content/full/338/6103/90/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S6

References (2540)

References and Notes

  1. The standard potential of the CO2/CO2 couple has been estimated to be –1.97 V versus NHE in N,N′-dimethylformamide + 0.1M Et4NClO4. The standard rate constant was 6 × 10−3 cm/s, with a transfer coefficient of 0.4 (2).
  2. The last term in Eq. 1 is lacking in the derivations of (9) because electron transfer between the electrode and the catalyst was assumed to be unconditionally fast. In the present case, catalysis is so fast that the catalyst electron transfer kinetics cannot be neglected. Full proof of Eq. 1 is given in the supplementary materials, section 5.
  3. Acknowledgments: Partial financial support from the Agence Nationale de la Recherche (ANR 2010 BLAN 0808) is gratefully acknowledged.
View Abstract

Navigate This Article