Ligand “Noninnocence” in Coordination Complexes Vs. Kinetic, Mechanistic, and Selectivity Issues in Electrochemical Catalysis

Ligand “noninnocence” in coordination complexes vs. kinetic, mechanistic, and selectivity issues in electrochemical catalysis

Cyrille Costentina,1,2, Jean-Michel Savéanta,1, and Cédric Tardb,1

aUniversité Paris Diderot, Sorbonne Paris Cité, Laboratoire d’Electrochimie Moléculaire, Unité Mixte de Recherche Université, CNRS 7591, 75205 Paris Cedex 13, France; and bLCM, CNRS, Ecole Polytechnique, Université Paris-Saclay, 91128 Palaiseau Cedex, France

Contributed by Jean-Michel Savéant, July 13, 2018 (sent for review June 21, 2018; reviewed by Marc Fontecave and Clifford P. Kubiak) The world of coordination complexes is currently stimulated by as to where the charge is located but is merely taking care of the the quest for efficient catalysts for the electrochemical reactions bookkeeping of the number of electrons globally exchanged. In underlying modern energy and environmental challenges. Even in the framework of molecular catalysis of electrochemical reac- the case of a multielectron−multistep process, catalysis starts with tions, designing selective efficient catalysts and benchmarking a uptake or removal of one electron from the resting state of the catalyst requires kinetic and mechanistic insights. Relying on catalyst. If this first step is an outer-sphere electron transfer (trig- electronic distribution in the catalyst resting state to predict re- gering a “redox catalysis” process), the electron distribution over activity involving the ligand or the metal may be misleading, as the metal and the ligand is of minor importance. This is no longer shown later on. the case with “chemical catalysis,” in which the active catalyst There is actually no obvious relationship between ligand non- reacts with the substrate in an inner-sphere manner, often involv- innocence and catalytic efficiency that could be used, per se, to ing the transient formation of a catalyst−substrate adduct. The help benchmark catalysts and design more efficient ones. Pursuing fact that, in most cases, the ligand is “noninnocent,” in the sense these goals requires gathering kinetic or transient information that the electron density and charge gained (or removed) from the through application of molecular electrochemistry concepts and resting state of the catalyst are shared between the metal and the techniques. Strategies in this area are enlightened by delineation ligand, has become common-place knowledge over the last half- of two types of catalyzed electrochemical processes (Fig. 1) (4). “ ” century. Insistent focus on a large degree of noninnocence of the In redox catalysis, the active form of the catalyst acts as a mere outer-sphere single-electron transfer. In the other—“chemical ligand in the resting state of the catalyst, even robustly validated ”— by spectroscopic techniques, may lead to undermining the essen- catalysis a catalytic intermediate is formed by an inner-sphere tial role of the metal when such essential issues as kinetics, mech- reaction between the electron transfer-generated active form of the catalyst and the substrate, opening a reaction sequence that anisms, and product selectivity are dealt with. These points are regenerates the starting form of the catalyst and is kinetically more general in scope, but their discussion is eased by adequately docu- favorable than a mere outer-sphere electron transfer. This Sabatier mented examples. This is the case for reactions involving metal- intermediate (5) may be simply an adduct formed by addition of the loporphyrins as well as vitamin B12 derivatives and similar cobalt substrate to the active form of the catalyst (QA in Fig. 1) or another complexes for which a wealth of experimental data is available. type of catalytic intermediate formed with a lower barrier than that of an outer-sphere electron transfer. The identification and the electrochemical reactions | catalysis | contemporary energy challenges | structural characterization of trapped intermediates are ways of ligand/metal noninnocence | cyclic voltammetry unraveling mechanisms, with the possible drawback that the mechanism and kinetics of the catalytic reaction under normal atalysis of electrochemical reactions by low-valent (for re- Cductions; high-valent for oxidations) coordination metal Significance complexes is a topic that has recently garnered boosted atten- tion triggered by modern energy and environmental chal- Coordination complexes are constantly sought as catalysts in lenges. Considering, for example, the case of reductive processes, the transformation of small molecules involved in contempo- the question may arise of whether the electron introduced in the rary energy and environmental challenges. Insistent emphasis starting complex to produce the catalytic species will “sit” on the on the well-established notion of ligand “noninnocence” may metal or on the ligand (or, symmetrically, where is the hole lo- blur the essential role of the metal center when kinetic, cated in case of oxidations), therefore hopefully orienting the mechanistic, and selectivity issues are targeted rather than the follow-up chemistry toward catalysis. This makes us inescapably detailed electronic structure of the starting catalyst molecule. enter the world of “noninnocent ligands,” a terminology that has With the help of modern concepts and techniques of molecular flourished after it was realized that, in redox reactions, both the electrochemistry, several examples are described in which ex- metal and the ligand can host the incoming electron in reduc- tremely efficient catalysis takes place at the metal center de- tions (or holes in oxidations). An excellent retrospective (1) spite an a priori discouraging large delocalization of charge and traces back the origin of this picturesque expression of the rather electron density on the noninnocent ligand. straightforward notion that the metal and ligand orbitals mix, forming global molecular orbitals. The repetitive recent use of Author contributions: C.C., J.-M.S., and C.T. designed research, performed research, and this terminology (2, 3) is thus surprising, noting that the weight of wrote the paper. the ligand contribution in these global molecular orbitals has Reviewers: M.F., Collège de France; and C.P.K., University of California, San Diego. indeed no reason to be negligible, the more so if the ligand has The authors declare no conflict of interest. intrinsic redox properties or, simply, electronic withdrawing/do- Published under the PNAS license. nating ability that render it noninnocent. The same is re- 1To whom correspondence may be addressed. Email: cyrille.costentin@univ-paris-diderot. spectively true for the metal. Conventional designation of the fr, [email protected], or [email protected]. oxidation state of a metallic complex by the formal oxidation 2Present address: Department of Chemistry and Chemical Biology, Harvard University, state of metal, the ligand being viewed as perfectly “innocent” Cambridge, MA 02138. and the metal perfectly “guilty,” has obviously no realistic scope Published online August 24, 2018.

9104–9109 | PNAS | September 11, 2018 | vol. 115 | no. 37 www.pnas.org/cgi/doi/10.1073/pnas.1810255115 Downloaded by guest on October 1, 2021 electrode solution direct Catalytic and Noncatalytic Reactions of Electrogenerated - A e electrochemical Low-Valent Metalloporphyrins, Vitamin B12s, and Similar reaction potential redox catalysis energy products (no selectivity) Cobalt Complexes B “I”/“0” products (no selectivity) The Fe porphyrin couple is involved in several catalytic e- P B redox catalyzed two-step chemical catalysis processes, notably catalysis of the CO2-to-CO electrochemical electrochemical QA reaction conversion, where the systematic analysis of ligand substitution A Q Q + A has led to the design of the most efficient catalysts of this re- B Product (selectivity) P + B action today (7–9). As detailed later on, the Fe0 resonance form e- P chemically catalyzed QA (or CI) is considered, among the mesomeric forms, as the best repre- electrochemical reaction QA no catalysis sentation of the active catalyst at the start of the catalytic loop, Q Â Ã Â Ã Â Ã 0,2− I,− II A reaction coordinate Fe ðPÞ ↔ Fe ðP · −Þ ↔ Fe ðP :; 2 − Þ .

Fig. 1. Contrasting redox and chemical catalysis of an electrochemical reaction (in the case of a reduction). A, substrate; B, products; P/Q, the catalyst From the very first times the catalytic or stoichiometric reactivity couple; QA, the Sabatier adduct or intermediate. of doubly reduced FeII porphyrin were investigated, until very recently, various spectroscopic techniques have indicated that the FeII resonance form appears as largely predominant over the operating conditions would not be the simple extension of what they 0 Fe form (10–21). Taking as an example the reduction of CO2 to are under intermediate trapping conditions. A particularly effective CO, the starting molecular reductant would be the FeII reso- way of investigating and operating electron transfer-triggered ca- nance form of the doubly reduced FeII porphyrin and the final talysis is offered by molecular electrochemistry. Control of the product FeIICO. From this perspective, the metal oxidation electrode potential provides a continuous variation of the concen- number seems to be invariant over the course of the reaction, tration of the active catalyst at the electrode, while the current af- and hence the ligand is not only noninnocent but also “redox fords straightforward access to the catalytic reaction kinetics. An active.” However, in this example as well as others, reasons for additional advantage is that these experiments can be carried out in 0 a microelectrolytic context, where the recording of the current nevertheless insisting on an Fe -based depiction of the mech- (kinetic) vs. electrode potential (thermodynamic) can be carried anism of these reactions are founded on kinetic and selectivity out with negligible consumption of the substrate by means of a experimental investigations. nondestructive technique such as cyclic voltammetry (CV), thus Pursuing the application of the Sabatier principle (Fig. 1), avoiding cumbersome preparative-scale electrolyses. electrophiles more potent than CO2 are predicted to give rise to In redox catalysis, the distribution of the incoming electron in noncatalytic reactions in which the Sabatier intermediate is so the lowest unoccupied molecular orbit of the active catalyst (for stable that it is itself the reaction product (blue curve in Fig. 1). reductions, or of the departing electron in the highest occupied This is indeed the case with n-BuBr as the electrophile (22), “0” molecular orbital for oxidation) is not of much importance, since where its reaction with Fe porphyrins directly produces a II the catalyst works globally as an outer-sphere electron donor (or perfectly characterized n-butyl-Fe complex, with no detectable acceptor, respectively). This is not the case with chemical ca- ligand alkylation. This seems difficult to explain by means of a talysis where the formation of an intermediate involving the active catalyst and the substrate is an essential step in agreement with the general Sabatier principle (5) as illustrated in Fig. 1. The 6 figure also shows that the Sabatier intermediate should not be 4 Fe“I”OEP too stable at the risk of its decomposition becoming the slow step 2 Fe“0”F20PP of the catalytic process, eventually hampering completely catalysis. “I” 0 Fe (C12)2TPP In most cases, molecular catalysis is a multielectron process, “I” -2 Fe TPP Fe“0” (diC3Ph) TPP hence the idea that the uptake of a second electron would be 2 eased by the “redox noninnocent” ligand, acting as an “electron -4 ” -6 reservoir for the metal. In the absence of a coupled chemical 6 Fe“0”ETIOP step, the uptake of a second electron is generally more difficult 4 than that of the first (4). The molecular factors that govern the Fe“0”TPP separation of the two ensuing standard potentials are, anyway, 2 Fe“I”OEP − 0 properties of the common metal ligand molecular orbitals, Fe“0”F20PP rather than properties depending separately on the metal and -2 -4 the ligand, making irrelevant the “electron reservoir” notion. It CHEMISTRY should be noted that, in practice, molecular catalysis is not only -6 6 multielectron processes but also multistep processes, in which Fe“0”ETIOP 4 electron transfers (E) are coupled with chemical steps (C) along Fe“0”TPP “EEC”- or, more frequently, “ECE”-type reaction sequences 2 Fe“0”F20PP (4, 6). For a typical example, see Fig. 7. The mesomeric metal− 0 ligand relation may well continue to be at work in the successive -2 intermediates involved in the global reaction. -4 Metalloporphyrins offer a remarkable illustration of these is- -6 sues. This will be the object of Catalytic and Noncatalytic Reac- 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1.2 -1.4 tions of Electrogenerated Low-Valent Metalloporphyrins, Vitamin ΔG0(eV) B12s, and Similar Cobalt Complexes, where catalytic and non- Δ 0 — Fig. 2. (Left) Rate constant (k) vs. reaction standard free energy ( G )plotfor catalytic reactions will be analyzed. Vitamin B12s the famous the reaction of the three butyl bromides (n-, s-, t-, from top to bottom) with the Co(I) “supernucleophile”—raises similar problems, albeit that a iron porphyrins shown in DMF + 0.1 M Bu4NBF4 (solid diamonds). The solid circles recent report concerning a quite similar cobalt complex introduces a represent a series of aromatic anion radicals playing the role of outer-sphere very different mechanism invoking “ligand noninnocence” (see Fig. one-electron donors. The full lines represent the fitting of these data points by 5). Ligand Noninnocence and Ligand Substitution Effects: Through- the quadratic law for the dissociative electron transfer of the three butyl bro- Structure and Through-Space Effects explores the relation between mides according the Morse curve model of dissociative electron transfer in ref. 4. ligand noninnocence and ligand substitution effects. (Right) Structures of the porphyrins.

Costentin et al. PNAS | September 11, 2018 | vol. 115 | no. 37 | 9105 Downloaded by guest on October 1, 2021 “ ” “ ” potential than the Fe I / 0 couple which still gives rise to a reversible wave (Fig. 3A). Upon raising the acid/catalyst concentration ratio, the catalytic wave increases in height and shifts in the negative direction, thus merging with the 2/2’ wave, while the reversibility disappears. This behavior is typical of a “total” catalysis situation where the catalytic reaction is so fast that the current is controlled by the diffusion of the substrate to the electrode surface (4). Very similar results were obtained with CHF2CO2H as the acid. The preceding comparison of the different behaviors of the reduced state of the iron and copper complexes is a good illustration of a more general trend. Accumulation of electron density and negative charge on the ligand indeed quite often leads to its saturation (hydrogenation, carboxylation, alkylation, etc., eventually followed by demetallation) and thus to the de- Fig. 3. CV of TPPFeIIICl (A, 0.96 mM; B, 0.65 mM) in DMF + 0.1 M Et NClO at activation of the catalyst. This is current practice, even if these 4 4 events—being considered as failures in strategy—are not often a mercury electrode in the presence of Et3NHCl (A, 1.6 mM; B, 7.1 mM). Scan rate is 0.1 V/s. Temperature is 25 °C. Reprinted with permission from ref. 24. reported and analyzed in detail (25). Copyright 1996 American Chemical Society. What happens when the metal is varied while keeping the same ligand can further be illustrated with catalysis by metalloporphyrins of the electrochemical reduction of 1,2-dibromo- reaction scheme involving an FeII complex with no change of the cyclohexane into the corresponding olefin (Fig. 4) (26). Fe oxidation state along the reaction path, It clearly appears (Fig. 4) that the Zn and Cu porphyrins behave as outer-sphere electron donors, as does the free base, Fe0,2−ðPÞ + n-BuBr → n-BuFeII,−ðPÞ + Br− whereas the Co, Fe, and Ni data points stand well above the outer- vs. sphere line. Additional electrochemical and stereochemical FeIIðP:,2− Þ + n-BuBr → n-BuFeII,−ðPÞ + Br−.

FeII porphyrins are indeed known for their appetite for Lewis 8 bases as axial ligands and even more for double-bond-forming 6 Co Fe ligands such as O, CO, carbenes, etc., but not for electrophiles. Ni H2 This is confirmed by the SN2 character of the iron alkylation process as a result of a systematic investigation of the reaction Cu “ ” “ ” 4 kinetics of a series of Fe I and Fe 0 porphyrins with n-,s-,t-butyl bromides, compared with that of aromatic anion radicals acting Zn 2 as outer-sphere electron donors (Fig. 2) (23). Some contribution of the Fe0 resonant form ought thus to be present in the initial complex to initiate the dynamic pro- 0 cess that leads to its alkylation in which the ligand–metal re- 00 EEcat- targ (V) dox swap is likely to be influenced by the approach of the -2 electrophile reactant to the iron center. It should be noted, in -0.5 -0.7 -0.9 -1.1 -1.3 -1.5 this connection, that the contribution of the FeII(P:,2‒)form is likely to be less important in the electrochemical conditions [N,N-dimethylformamide (DMF) + 0.1 M Et4NClO4 solu- tions] than in the spectroscopic studies where the negative charges borne by the porphyrin ring are stabilized by ion pairing with the sodium ions deriving from the reductive reagent used in much less polar solvents (ethers) than DMF. An even more striking example is provided by the catalysis of H2 evolution from the reduction of + “I”/“0” Et3NH by the Fe tetraphenylporphyrin (TPP) couple in DMF (24). The faradaic efficiency of dihydrogen formation is 100%, with no degradation of the porphyrin catalyst after more than 1 h of electrolysis. By contrast, the complex resulting from the addition of two electrons to TPPCuII does not catalyze hydrogen evolution under the same conditions despite the fact that the standard po- Fig. 4. Catalysis of the electrochemical reduction of 1,2-dibromocyclohex- ‒ ‒ tential of the TPPCuII /II2 couple (−1.63 V vs. SCE) is almost the ane (Top Right) into the corresponding olefin (Top Left) in DMF + 0.1 M “I”/“0” same as the standard potential of the TPPFe couple (−1.60 V Bu4NBF4 by anion radicals of aromatic hydrocarbons (solid circles) and by the − + vs. SCE) (24). Instead, addition of the acid leads to a 3e + 3H redox couples obtained by one-electron reversible reduction of ETIOP porphyrins (see Fig. 2 for the definition of the ETIOP ring) of CoII,FeII,NiII,ZnII, hydrogenation of the ring. II In the absence of the acid, the iron porphyrin exhibits three and Cu and the free base (H2) (solid squares), from the data in ref. 26. The successive reversible one-electron CV waves corresponding to full line represents the fitting of these data points by the quadratic law for “ ”− “ ” − the successive formation of the FeII,FeI , and Fe 0 2 com- the dissociative electron transfer by outer-sphere one-electron donors according to the Morse curve model in ref. 4. (Bottom) Catalysis of the elec- plexes. Upon addition of the acid, the first two waves remain trochemical reduction of 1,2-dibromocyclohexane (OlX ). For redox catalysis, unchanged (this is shown in Fig. 3 for wave 1/1’, representing the 2 “II”/“I” “0”2− ’ the key step is the dissociative electron transfer step in which the aromatic Fe couple). Generation of Fe at wave 2/2 triggers the anion radicals, but, also, the one-electron reduced ZnII and CuII porphyrins and •‒ appearance of a catalytic irreversible wave, noted as 2C in Fig. 3. free base play the role of outer-sphere electron donors, D . The two metals Noteworthy also is the presence, at more negative potentials, are perfectly innocent, and the ligand is entirely guilty. For chemical catalysis, of a small but distinct reversible wave (3/3′), which features the the key step is the halonium abstraction by the noninnocent metal at the +I II−/I2− reversible hydride FeH couple. At low acid/catalyst con- oxidation degree as for Co, Fe, and Ni, with ‒XM“III”+L as the Sabatier catalytic centration ratios, the catalytic wave occurs at a more positive intermediate.

9106 | www.pnas.org/cgi/doi/10.1073/pnas.1810255115 Costentin et al. Downloaded by guest on October 1, 2021 very fact that addition of proton donors draws the reaction toward CO2 catalytic reduction rather than toward irreversible saturation of the porphyrin ring is a further indication that the catalytic chemistry takes place at the iron center despite the likely ‒ modest contribution of the Fe0,2 (P) resonant form at the start of the reaction. On this basis, and with the help of a systematic CV investigation, the mechanism depicted in Fig. 7 could be established (8). The formation of a, Fe‒CO2 adduct as the first step of the catalytic process is confirmed by low-temperature resonance Raman (32). Furthermore, in the same vein, installing the acid functionalities inside the catalyst molecule in positions favoring H-bond stabilization and protonation of the initial Fe‒CO2 adduct considerably improves the catalytic efficiency, according to the mechanism summarized in Fig. 8 (33). Ligand Noninnocence and Ligand Substitution Effects: Through-Structure and Through-Space Effects As seen in the preceding sections, ligand noninnocence amounts to the metal−ligand mesomeric relation involving common mo- I Fig. 5. Alkylation of a Co complex. lecular orbitals. Rather than repetitive invocation of ligand noninnocence, it therefore appears more fruitful to take advantage of this situation to investigate and rationalize the effects of in- investigations of other vicinal 1,2-dihalohalides (27) showed troducing substituents in the ligand. Substituents are expected that two types of mechanisms take place, as summarized in to change the standard potential of the catalyst couple along Fig. 4. with the electron density and charge on the metal, and hence Another interesting example of alkylation, reminiscent of what “ ” we have previously discussed with Fe 0 porphyrins, is provided the catalytic efficiency. These through-structure effects are by the system shown in Fig. 5 (28). The oxidative addition of the simply the result of the combination of the orbitals of the non- alkylation reagent RX on the anionic cobalt complex shown in innocent ligand and of the noninnocent metal to form a common molecular orbital, which passes on to the metal the electronic Fig. 5 follows an SN2 mechanism. According to ref. 28, Co would remain perfectly innocent at the oxidation degree III, donating or withdrawing effect of the substituents. If catalysis bearing a +1 charge and (very unlikely for an CoIII complex) no is of the redox type, such variation of substituents will mostly axial ligand (Fig. 5, Top). This full “innocence” of the metal result in a change of the standard potential of the catalyst outer- would be maintained all along the course of the reaction, which sphere redox couple, with a moderate effect on the associated seems in contradiction to the fact that alkylation takes place intrinsic barrier. solely at the metal and not at any atom of the ligand, as should In the case of chemical catalysis, for a reduction, introduction be the case, at least in part, if the whole electron density and of an electron-withdrawing substituent leads to a favorable de- negative charge were borne by the ligand. crease of the overpotential due to a positive shift of the catalyst This is very similar to the previous discussion of the reaction of standard potential and, at the same time, an unfavorable de- “ ” Fe 0 porphyrin with alkyl halides. For the same reasons, the crease of the turnover frequency (i.e., the apparent pseudo-first- reactant is best formulated by the same type of resonance hybrids order rate constant of homogeneous reduction of the substrate as shown above. by the reduced form of the catalyst, the rate constant kcat), due to The cobalt complex in Fig. 5 and its alkylation reactions are the decrease of the electron density on the metal atom. These reminiscent of vitamin B12s (Fig. 6). This is a potent nucleophile, giving rise to organocobalt derivatives (29, 30), and, for this reason, traditionally represented by its CoI resonance form (Fig. 6). It II III is also a weak Brönsted base, in which, unlike the Co or Co B12, the cobalt is not coordinated to the endogenous benzimidazole axial ligand (gray tint part of Fig. 6) (30). It results from the preceding discussion that, even if the main

reaction of vitamin B12s takes place at the cobalt, there is no CHEMISTRY reason that the corrin ligand should be totally innocent. Indeed, theoretical calculations attempting to reproduce spectrochemical results indicate that there is a substantial, albeit not total, contribution of 67% of the d8 CoI configuration with a 23% contribution of CoII–noninnocent corrin ring (31). However, these figures should be taken with extreme caution, according to the author himself. They are certainly not an obstacle for vitamin B12s to be a vigorous nucleophile at cobalt. The same is true, for the same reasons, with the Fig. 5 cobalt complexes. Coming back to the catalysis of the CO -to-CO electro- “ ” 2 chemical conversion by the TPPFe I/0 couple, it should be recalled that the noninnocence of the porphyrin ligand, with the predominant contribution of the FeII(P:,2‒) resonant form, appeared, at first, as a discouraging obstacle to an efficient catalyst. The turnover number was desperately low, to the benefit of the irreversible saturation of the porphyrin (7). Faradaic effi- ciencies get considerably better when Lewis or Brönsted acids are added as cosubstrates, to the point of reaching 100% (8). The Fig. 6. Vitamin B12s.

Costentin et al. PNAS | September 11, 2018 | vol. 115 | no. 37 | 9107 Downloaded by guest on October 1, 2021 0.4

0.3

0.2

0.1

0 Ecat (V vs. SHE) 0 -0.8 -1 -1.2 -1.4 -1.6

Fig. 7. Catalysis of the electrochemical CO2-to-CO conversion by the “I/0” TPPFe couple in DMF + 0.1 M n-Bu4BF4 in the presence of various weak Brönsted acids. (Top Left) Cyclic voltammograms of FeTPP recorded in the

absence (blue) and presence (red) of CO2 and PhOH substrates. (Top Right) Fig. 9. Through-structure and through-space substituent effects in the ca- “I”/“0” Correlation between the rate constant of the rate-determining step and the talysis of the electrochemical CO2-to-CO conversion by the Fe porphyrin + pKa of the acid. (Bottom) Mechanism. couples in DMF 0.1 M n-Bu4BF4. Shown is linear free energy correlation between the overall catalytic reaction and the standard potential of the catalyst couple as appears for the through-structure substituent effects effects are reversed for electron-donating substituents, and the (FeTPP, FeF5TPP, FeF10TPP, FeF10TPP) (36). Through-space electrostatic ef- case of an oxidation is strictly symmetrical. fects departing from the correlation appear with charged substituents (Fe-o- An ideal situation would be that substitution would both shift TMA, Fe-p-TMA, Fe-p-SULF) (34). 0 Ecat toward positive values and increase the overall catalytic rate constant, kcat (for reductions; the opposite for oxidations), in other words, an optimization in terms of both overpotential and kinetics. In fact, it appears that, as far as through-structure substituent effects are concerned, the two effects play in opposite directions. A good illustration is provided by catalysis of the electro- “I”/“0” chemical CO2-to-CO conversion by Fe porphyrin couples. Fig. 9 shows a linear free energy relationship between the log of the global catalytic rate constant TOF = k = K K k (Fig. 7) −FeICO •− Sabatier intermediate max cat 1 2 3 2 and the standard potential of the catalyst couple taken as thermo- dynamical index of the correlation (black straight line in Fig. 9), with a global correlation coefficient of 2. Dissection of this global effect along each step allows a better understanding that what makes that an advantage in terms of overpotential is compensated by a disadvantage in terms of TOFmax. Most of the effect of an electron-donating substituent, which renders E0 more negative, indeed results from an increase of The pending O-Hs serve two purposes: cat the nucleophilicity of the Fe0 center and of the Brönsted basicity a) to stabilize the CO Sabatier intermediate b) to provide a high local 2 of the oxygens in the initial iron‒CO adduct—two effects that by means of H-bonding concentration of proton donor 2 increase the overall catalytic rate constant. These are reverted for an electron-withdrawing substituent. A more ambitious quest i is to overcome these limitations with substituents that do not exclusively exert their influence through the electronic structure of the metal complex. Such is the case with charged substituents that may stabilize key intermediates through electrostatic inter- actions, while still having a through-structure effect. Typical examples are provided by catalysis of the CO2-to-CO electrochemical conversion by TPPFeI/0 substituted by four (positively charged) trimethylammonio groups in ortho position or para position or by E four (negatively charged) sulfonic groups in para position (Fig. 9), + as results from the relevant electrostatic stabilization and de- Fig. 8. Catalysis of the electrochemical CO2-to-CO conversion in DMF − •− “I/0” 2 0 ↔ I 0.1 M n-Bu4BF4 by the Fe couple of two TPPs bearing phenol substituents stabilization of the negatively charged Fe CO2 Fe CO2 pri- in ortho, ortho’ of their phenyl groups in the presence of PhOH. mary intermediate. This effect is small in both cases because of the

9108 | www.pnas.org/cgi/doi/10.1073/pnas.1810255115 Costentin et al. Downloaded by guest on October 1, 2021 large distance between the para position and the charge on the transfers or formation of transient intermediates associating the initial intermediate. It becomes very large when the trimethy- substrate and the active catalyst are operating. Involvement of lammonio groups are placed in the ortho position, giving rise to the metal gives rise to a richer palette of possibilities in terms of the most efficient catalyst of the CO2-to-CO electrochemical efficiency and specificity. The drawback of recent insistence on conversion at present (34). Another type of initial adduct stabili- noninnocence of ligands is that it may miss the really important zation is at work in the catalysis of the CO2-to-CO electrochemical endeavors in the design of efficient catalysts. More-reasonable conversion by the two FeI/0 porphyrins bearing acid functionalities approaches consist in gathering and analyzing more experi- as discussed previously (Fig. 8). mental kinetic data by use of all of the resources of modern molecular electrochemistry. Among these, catalysts’ bench- Conclusion marking by means of catalytic Tafel plots is an essential tool in As realized for a long time, ligands in coordination complexes the quest for more-efficient catalysts. They are profitably ap- are noninnocent in the sense that the metal and the ligand or- plied to investigate through-structure and through-space effects bital mix to form common molecular orbitals. These are able to of systematic variations of the ligand while keeping the same host or expel electrons. That the additional electron or hole thus metal and vice versa (35). Further progress in analysis of such created is delocalized over the metal and the ligand, making mechanistic subtleties is expected to result from the extension further reactions involve one or the other of these parts of the of kinetic studies. This concerns experiments as well as clues molecule, or both competitively, is a truism. However, the from quantum chemical computations. In the latter case, an question is relevant to the contemporary efforts in the field of urgent task should be not only to analyze the ground state of energy and environment, which requires the design of efficient the active form of the catalyst but also to follow the kinetics of catalysts for the activation of small molecules. The relative lo- its reaction with the substrate and watch the changes of the cation of an incoming electron or hole is not a crucial one in the electron distribution along the reaction coordinate. On the case of redox catalysis in which the rate-determining step is an experimental side, many kinetic data can be accessed by drawing outer-sphere electron transfer between the active form of the on existing studies, but more in-depth investigations might be ad- catalyst and the substrate. The situation is quite different with visable, such as kinetic isotope effects and temperature-dependent chemical catalytic processes in which inner-sphere electron experiments.

1. Eisenberg R, Gray HB (2011) Noninnocence in metal complexes: A dithiolene dawn. 20. Römelt C, et al. (2017) Electronic structure of a formal iron(0) porphyrin complex

Inorg Chem 50:9741–9751. relevant to CO2 reduction. Inorg Chem 56:4746–4751. 2. Berben LA, de Bruin B, Heyduk AF (2015) Non-innocent ligands. Chem Commun 21. Römelt C, et al. (2018) Electronic structure and spin multiplicity of iron tetraphe- (Camb) 51:1553–1554. nylporphyrins in their reduced states as determined by a combination of resonance 3. Kaim W (2012) The shrinking world of innocent ligands: Conventional and non- Raman spectroscopy and quantum chemistry. Inorg Chem 57:2141–2148. conventional redox-active ligands. Eur J Inorg Chem 2012:343–348. 22. Lexa D, Savéant J-M, Wang DL (1986) Electroreductive alkylation of iron porphyrins. 4. Savéant J-M (2006) Elements of Molecular and Biomolecular Electrochemistry: An Iron(III), iron(II) and iron(I) alkyl complexes from the reaction of doubly reduced iron(II) Electrochemical Approach to Electron Transfer Chemistry (John Wiley, Hoboken, NJ). porphyrins with alkyl halides. Organometallics 5:1428–1434. 5. Sabatier P (1920) La catalyse en chimie organique (Librairie Polytechnique, Paris). 23. Lexa D, Savéant J-M, Su K-B, Wang DL (1988) Single electron transfer and nucleophilic 6. Costentin C, Savéant J-M (2014) Multielectron, multistep molecular catalysis of elec- substitution. Reaction of alkyl bromides with aromatic anion radicals and low oxi- trochemical reactions: Benchmarking of homogeneous catalysts. ChemElectroChem 1: dation state iron porphyrins. J Am Chem Soc 110:7617–7625. 1226–1236. 24. Bhugun I, Lexa D, Savéant J-M (1996) Homogeneous catalysis of electrochemical hy- – 7. Hammouche M, Lexa D, Savéant J-M, Momenteau M (1988) Catalysis of the electro- drogen evolution by iron(0) porphyrins. J Am Chem Soc 118:3982 3983. chemical reduction of carbon dioxide by iron(0) porphyrins. J Electroanal Chem 249: 25. Costentin C, Passard G, Savéant J-M (2015) Benchmarking of homogeneous electro- 347–351. catalysts: Overpotential, turnover frequency, limiting turnover number. J Am Chem – 8. Costentin C, Robert M, Savéant J-M (2015) Current issues in molecular catalysis illus- Soc 137:5461 5467. 26. Lexa D, Savéant JM, Su KB, Wang DL (1987) Chemical vs. redox catalysis of electro- trated by iron porphyrins as catalysts of the CO2-to-CO electrochemical conversion. Acc Chem Res 48:2996–3006. chemical reactions. Reduction of trans-1,2-dibromocyclohexane by electrogenerated 9. Costentin C, Robert M, Savéant J-M, Tatin A (2015) Efficient and selective molecular aromatic anion radicals and low oxidation state metalloporphyrins. J Am Chem Soc 109:6464–6470. catalyst for the CO2-to-CO electrochemical conversion in water. Proc Natl Acad Sci 27. Lexa D, et al. (1990) Outer-sphere and inner-sphere processes in reductive elimina- USA 112:6882–6886. tion. Direct and indirect electrochemical reduction of vicinal dibromoalkanes. JAm 10. Cohen IA, Ostfeld D, Lichtenstein B (1972) Characterization of a d iron system. Tet- Chem Soc 112:6162–6177. raphenylporphineiron (I) anion. J Am Chem Soc 94:4522–4525. 28. Smith AL, Hardcastle KI, Soper JD (2010) Redox-active ligand-mediated oxidative 11. Lexa D, Momenteau M, Mispelter J (1974) Characterization of the reduction steps of addition and reductive elimination at square planar cobalt(III): Multielectron reac- Fe(III) porphyrins. Biochim Biophys Acta 338:151–163. tions for cross-coupling. J Am Chem Soc 132:14358–14360. 12. Mashiko T, Reed CA, Haller KJ, Scheidt WR (1984) Nature of iron(I) and iron(0) tet- 29. Schrauzer GN, Deutsch E, Windgassen RJ (1968) The nucleophilicity of vitamin B12. raphenylporphyrin complexes. Synthesis and molecular structure of (dibenzo- J Am Chem Soc 90:2441–2442. 18-crown-6)bis(tetrahydrofuran)sodium (meso-tetraphenylporphina to)ferrate and 30. Lexa D, Savéant J-M (1983) The electrochemistry of vitamin B12. Acc Chem Res 16: bis[tris(tetrahydrofuran)sodium] (meso-tetraphenylporphinato) ferrate. Inorg Chem 23: 235–243. – 3192 3196. 31. Jensen KP (2005) Electronic structure of Cob(I)alamin: The story of an unusual nu- 13. Hickman DL, Shirazi A, Goff HM (1985) Deuterium NMR spectroscopic studies of low- cleophile. J Phys Chem B 109:10505–10512. CHEMISTRY – – + valent iron porphyrin species. Inorg Chem 24:563 566. 32. Mondal B, Rana A, Sen P, Dey A (2015) Intermediates involved in the 2e /2H re- 14. Reed CA (1982) In Electrochemical and Spectrochemical Studies of Biological Redox duction of CO2 to CO by iron(0) porphyrin. J Am Chem Soc 137:11214–11217. – Components (American Chemical Society, Washington, DC), Vol 201, pp 333 356. 33. Costentin C, Passard G, Robert M, Savéant J-M (2014) Pendant acid-base groups in 15. Donohoe RJ, Atamian M, Bocian DF (1987) Characterization of singly reduced iron(II) molecular catalysts: H-bond promoters or proton relays? Mechanisms of the conver- porphyrins. J Am Chem Soc 109:5593–5599. sion of CO2 to CO by electrogenerated iron(0)porphyrins bearing prepositioned 16. Yamaguchi K, Morishima I (1992) Low-valent iron porphyrins. NMR evidence for. phenol functionalities. J Am Chem Soc 136:11821–11829. pi.-anion-radical character in two-electron-reduced iron(III) meso- or. beta.-pyrrole- 34. Azcarate I, Costentin C, Robert M, Savéant J-M (2016) Through-space charge in- substituted porphyrins. Inorg Chem 31:3216–3222. teraction substituent effects in molecular catalysis leading to the design of the most

17. Sinyakov GN, Shulga AM (1993) 1H NMR spectra and electronic structure of reduced efficient catalyst of CO2-to-CO electrochemical conversion. J Am Chem Soc 138: iron porphyrins: Fe(II), Fe(I) and Fe(0) porphyrins. J Mol Struct 295:1–14. 16639–16644. 18. Anxolabehere E, Chottard G, Lexa D (1994) Highly reduced iron porphyrins: UV-vis 35. Costentin C, Savéant J-M (2017) Toward an intelligent design of molecular electro- and resonance Raman spectroelectrochemical studies of FeTPP and FeTF5PP. New J catalysts. Nat Rev Chem 1:0087. Chem 18:889–899. 36. Azcarate I, Costentin C, Robert M, Savéant J-M (2016) Dissection of electronic sub-

19. De Silva C, Czarnecki K, Ryan MD (1999) Visible and resonance Raman spectra of low stituent effects in multielectron-multistep molecular catalysis. Electrochemical CO2- valent iron porphyrins. Inorg Chim Acta 287:21–26. to-CO conversion catalyzed by iron porphyrins. J Phys Chem C 120:28951–28960.

Costentin et al. PNAS | September 11, 2018 | vol. 115 | no. 37 | 9109 Downloaded by guest on October 1, 2021