Combining acid-base, redox and substrate binding functionalities to give a complete model for the [FeFe]-hydrogenase

Abstract

Some enzymes function by coupling substrate turnover with electron transfer from a redox cofactor such as ferredoxin. In the [FeFe]-hydrogenases, nature’s fastest catalysts for the production and oxidation of H₂, the one-electron redox by a ferredoxin complements the one-electron redox by the diiron active site. In this Article, we replicate the function of the ferredoxins with the redox-active ligand Cp*Fe(C₅Me₄CH₂PEt₂) (FcP*). FcP* oxidizes at mild potentials, in contrast to most ferrocene-based ligands, which suggests that it might be a useful mimic of ferredoxin cofactors. The specific model is Fe₂[(SCH₂)₂NBn](CO)₃(FcP*)(dppv) (1), which contains the three functional components of the active site: a reactive diiron centre, an amine as a proton relay and, for the first time, a one-electron redox module. By virtue of the synthetic redox cofactor, [1]²⁺ exhibits unique reactivity towards hydrogen and CO. In the presence of excess oxidant and base, H₂ oxidation by [1]²⁺ is catalytic.

Hydrogenase enzymes catalyse the interconversion of H₂ with protons and reducing equivalents. Mimicking the reactivity of these enzymes by means of active-site models is currently of interest because these catalysts rely on inexpensive first-row transition metals and the potential of hydrogen as an energy carrier^1–6. The hydrogenases function by combining cofactors that couple acid–base and redox reactions mediated by the diiron dithiolate core and its cofactors (Fig. 1). The amine cofactor (‘azadithiolate’) relays protons to and from the distal Fe. A 4Fe–4S cluster is attached to the 2Fe–2S core through a single cysteinate residue and provides one redox equivalent to complement the one-electron couple for the diiron dithiolate core. The active site has been characterized in two functional states that differ by one electron: one state, H_red, poised to reduce protons and the other, H_ox, poised to oxidize H₂ (refs 7,8).

Figure 1. Structure of active site for the [FeFe]-hydrogenase and its model.

Both structures contain functionality dedicated to substrate binding as well as the management of redox equivalents and proton equivalents. As shown on the left, the active site consists of a Fe₂(CO)₃(CN)₂ centre bridged by the Brønsted-basic azadithiolate (SCH₂NHCH₂S) cofactor. The redox cofactor, a 4Fe–4S cluster, is attached to a single Fe centre. Substrate binding occurs at the Fe centre distal to the 4Fe–4S cluster and adjacent to the basic amine. The proposed model complex also features an Fe₂(CO)₃ centre bridged by a basic azadithiolate with an alkyl groups R in place of H. A redox-active ligand (Redox-L) simulates the function of the 4Fe–4S cluster, and the phosphine ligands simulate the coordinated cyanides.

Over the last decade, efforts at functional modelling of the H_red state with these systems have identified electrocatalysts for proton reduction that operate via hydrides bound to an apical position on one Fe centre^9–11. The amine cofactor (adt)¹², for which there is increasing direct biophysical evidence^13,14, plays an important role in proton reduction catalysis¹⁵. Several models for the H_ox state have also been reported¹⁶, the defining feature of which is their mixed valency. Models for H_ox, even those containing azadithiolates^17–19, react only slowly with H₂ under forcing conditions¹⁷. Recently, we reported that the addition of mild oxidant to solutions of H_ox models allows for facile oxidation of H₂ (ref. 18). This finding suggests that functional models of the H_ox state require the presence of both an azadithiolate cofactor as well as a suitably tuned redox cofactor. Synthetic redox cofactors^11,20–22 that are intended to mimic the role of the 4Fe–4S cluster have been incorporated into models, but they have exhibited no functional role before this work.

In designing a model featuring both acid–base and redox functionality, we sought a synthetic cofactor with the following biologically inspired properties: (i) mild redox couple closer to the H₂/H⁺ couple, in the range −0.3 to −1 V versus the ferrocene/ferrocinium couple (Fc^+/0); (ii) chemical inertness, so that reactions would be localized at the Fe₂ core; and (iii) a ligand that would bind tightly to a single Fe centre. Ferrocene-derived ligands seemed an obvious choice due to their chemical inertness and their one-electron redox chemistry²³. Although many ferrocenyl phosphine ligands are known, in almost all cases—most famously dppf—the ferrocene serves simply as an inert scaffold and does not engage in redox. Ferrocenyl phosphines do undergo redox, but only at highly positive potentials, usually >0 V versus Fc^+/0 (640 mV versus normal hydrogen electrode, NHE)²⁴. Such high potentials are incompatible with hydrogenase mimics because they would oxidize the H_ox state, inducing binding of the amine cofactor to Fe. Decamethylferrocene (Fc*), E_1/2 = −550 mV versus Fc^+/0 (187 mV versus NHE), represents an attractive alternative to ferrocene, although it has not been incorporated into monodentate redox-active ligands²⁵. A potential problem with Fc* is its steric bulk, which may discourage coordination. We therefore targeted Cp*Fe(C₅Me₄CH₂PEt₂) (FcP*, Cp* = C₅Me₅), where the highly basic trialkylphosphine substituent is separated from the redox agent by a methylene spacer. As we describe in the following, incorporating FcP* into a model for the H_ox state of [FeFe]-hydrogenase enables the binding of substrates in a manner not previously observed.

Results and discussion

Preparation of FcP* and the reduced active site model

The redox-active ligand FcP* was accessed readily from a procedure involving the concomitant formation of the C–P bond and the generation of an anionic tetraalkylcyclopentadienyl ligand. Commercially available C₅Me₅H was thus converted in one pot to LiC₅Me₄CH₂PEt₂ (ref. 26). This salt was treated with [Cp*FeCl]₂ generated in situ to give FcP* in ~55% isolated yield as a bright yellow solid (Fig. 2)²⁷. In CH₂Cl₂ solution (0.1 M ${\text{NBu}}_4{\text{BAr}}_4^{\mathrm{F}}$, Ar^F = 3,5-C₆H₃(CF₃)₂), FcP* reversibly oxidizes with E_1/2 = −591 mV. Given the mildness of its redox potential, FcP* is well suited to explore redox-induced reactions that occur near the H⁺/H₂ couple²⁸.

Figure 2. Synthesis of FcP*.

The route starts with the formation of the C–P bond and generation of LiC₅Me₄CH₂PEt₂. Combining this organolithium reagent with “(C₅Me₅)FeCl” gives FcP*.

FcP* was installed on a diiron azadithiolato platform by reaction with Fe₂[(SCH₂)₂NBn](CO)₄(dppv), which is known to undergo monosubstitution by basic ligands (dppv = cis-C₂H₂(PPh₂)₂, Bn = CH₂Ph)²⁹. The resulting complex Fe₂[(SCH₂)₂NBn](CO)₃(FcP*)(dppv) (1) was isolated in analytical purity and exhibits spectroscopic properties similar to those of the simpler model Fe₂[(SCH₂)₂NBn] (CO)₃(PMe₃)(dppv) (2), which features a diphosphine and monophosphine ligand on separate Fe centres. The oxidation states of the Fe centres in 1 are Fe(ii)Fe(i)Fe(i). Treatment of 1 with $\mathrm{H}{({\text{OEt}}_2)}_2{\text{BAr}}_4^{\mathrm{F}}$ gave the hydride [1H]⁺ (Fig. 3). The Fe oxidation states within [1H]⁺ are Fe(ii)Fe(ii)Fe(ii). Overall, the spectroscopic properties of 1 and [1H]⁺ are very similar to those for 2 and [2H]⁺, respectively.

Figure 3. Summary of reactions observed for [1]²⁺ with CO and H₂.

Compound numbering and relevant oxidation states are shown, using italics for the ferrocenyl iron centre. The starting compound 1 is in a fully reduced state. Initial oxidation of 1 is localized on the diiron core as shown by FT-IR and EPR measurements. The second oxidation (to [1]²⁺) converts the FcP* centre from ferrous to ferric. Reactions of [1]²⁺ with H₂ and with CO involve substrate binding coupled to intramolecular electron-transfer (ET) from the diiron subunit to the oxidized FcP* ligand.

Redox states of [Fe₂[(SCH₂)₂NBn](CO)₃(FcP*)(dppv)]ⁿ⁺

The influence of the FcP* ligand on the properties of the diiron dithiolate was revealed through a combination of electrochemical and spectroscopic measurements. Because H₂ is a weakly basic ligand, these experiments used the weakly basic aryl borate anion ${\text{BAr}}_4^{\mathrm{F}-}$, which does not compete with potential substrates. In dichloromethane solution, the cyclic voltammetry of 1 exhibits two reversible one-electron redox events (electrolyte: 0.1 M ${\text{NBu}}_4{\text{BAr}}_4^{\mathrm{F}}$), the first at −700 mV, comparable to the [2]^0/+ couple at −643 mV. For [1]⁺, but not for [2]⁺, a second oxidation was observed at −393 mV versus Fc^0/+. The localization of the oxidation changes in [1]ⁿ⁺ was probed by monitoring the titration of 1 with the oxidant ${\text{FcBAr}}_4^{\mathrm{F}}$ using Fourier-transform infrared (FT-IR) spectroscopy (Fig. 4). The addition of 1 equiv. of ${\text{FcBAr}}_4^{\mathrm{F}}$ to a CH₂Cl₂ solution of 1 resulted in a ~60 cm⁻¹ shift in the two ν_CO bands to higher energy. The IR spectrum of [1]⁺ closely matches that for [2]⁺ (ref. 30). Thus, one-electron oxidation results in the formation of a normal H_ox-like mixed valence state^16,19,31,32. Perhaps because of the steric bulk of the FcP* ligand, the salt [1]${\text{BAr}}_4^{\mathrm{F}}$ is noticeably more stable than [2]${\text{BAr}}_4^{\mathrm{F}}$ (ref. 17). For example, solid [1]${\text{BAr}}_4^{\mathrm{F}}$ can be precipitated and handled at room temperature. Addition of a second equivalent of ${\text{FcBAr}}_4^{\mathrm{F}}$ to a solution of [1]${\text{BAr}}_4^{\mathrm{F}}$ shifted υ_CO by only ~4 cm⁻¹, consistent with oxidation away from the diiron centre and hence localized at FcP* (Fig. 4). Thus, the oxidation states in [1]²⁺ are Fe(iii)Fe(ii)Fe(i). The [FcP*]^+/0 couple shifts by 200 mV upon coordination, reflecting both the inductive influence of the appended diiron unit and ion pairing effects^33,34. The electronic interaction between the diiron subunit and the appended ferrocenyl ligand is reminiscent of the coupling between the diiron subunit and the appended 4Fe–4S cluster in the H cluster^35,36.

Figure 4. Infrared spectra probing the localization of the two one-electron oxidations of the reduced model [1].

a–d, Titration of ${\text{FcBAr}}_4^{\mathrm{F}}$ into a CH₂Cl₂ solution of 1 monitored by solution IR spectroscopy: [1] (a), [1] + 1 equiv. ${\text{FcBAr}}_4^{\mathrm{F}}$ (b), [1] + 1.5 equiv. ${\text{FcBAr}}_4^{\mathrm{F}}$ (c), [1] + 2 equiv. ${\text{FcBAr}}_4^{\mathrm{F}}$ (d). On conversion of [1]⁺ to [1]²⁺, an isosbestic point is observed at 1,969 cm⁻¹.

The assignments of oxidation state from the changes in the IR spectrum are supported by electron paramagnetic resonance (EPR) spectroscopic measurements. The room-temperature EPR spectrum of [1]⁺ is similar to that reported for the related [2]⁺ (ref. 37): the axial spectrum exhibits large hyperfine coupling to a pair of ³¹P centres with A_z(³¹P) = 79.9 MHz. This pattern indicates that the unpaired electron resides mostly on the Fe atom distal FcP* and suggests a formal assignment of [1]⁺ as (FcP*)(CO)₂Fe^II (µ-SR)₂Fe^I (dppv)(CO). The doubly oxidized species [1]²⁺ is, however, EPR-silent from room temperature to 110 K (in 50/50 toluene/CH₂Cl₂ glass). A solution of a 1:1 mixture of [2]⁺ and [Fc*]⁺ exhibits a normal H_ox EPR spectrum. Variable-temperature magnetic susceptibility measurements indicate that [1]⁺ has a single unpaired electron and [1]²⁺ has two unpaired electrons. The absence of an EPR signal for [1]²⁺ may result from the two spins of [1]²⁺ undergoing fast relaxation at 110 K. Indeed, at very low temperatures (4.5 K), a very broad signal is observed at g ≈ 1.65, consistent with the presence of a fast-relaxing triplet state.

Reactions of Fe₂[(SCH₂)₂NBn](CO)₃(FcP*)(dppv)]ⁿ⁺ with CO and H₂

Having established that the model can be poised in three different oxidation states, 1, [1]⁺ and [1]²⁺, we investigated how the oxidized states interact with CO and H₂, two well studied substrates for hydrogenases^38–41. As shown below, the FcP* ligand strongly influences the reactivity of the diiron centre. Although [1]⁺ does not bind CO at room temperature, solutions of [1]²⁺ visibly bind CO at room temperature. Thus, in the presence of CO, purple solutions of [1]²⁺ become orange, indicative of intramolecular redox to yield the all ferrous complex, that is, Fe(iii)Fe(ii)Fe(i) → Fe(ii)Fe(ii)Fe(ii)CO. The reversible binding of CO to [1]²⁺ can be monitored by IR as well as ³¹P NMR spectroscopies. With 1 atm of CO, solutions of [1]²⁺ in dichloromethane show the formation of ~50% of the adduct [1CO]²⁺ (Fig. 5). At lower temperatures the binding of CO is quantitative, leading to diamagnetic solutions of [1(CO)]²⁺ that exhibit well-resolved ³¹P NMR spectra consistent with a single, unsymmetrical isomer. One likely explanation for this interesting reactivity is that CO binding induces electron transfer to the pendant ferrocenium, resulting in a switch from Fe^IIIFe^IIFe^I to Fe^IIFe^IIFe^II (Fig. 3). Studies on the [FeFe]-hydrogenase from Clostridium pasteurianum and Desulfovibrio desulfuricans indicate that CO binding induces partial oxidation of the distal Fe centre^42,43. In the case of [1]²⁺, the effect is sufficiently strong that CO binding triggers full reduction of the ferrocenium centre.

Figure 5. Spectroscopic evidence confirming the reactions of [1]²⁺ with known hydrogenase substrates H₂ and CO.

a, IR spectra of dichloromethane solutions of [1]²⁺ before and after treatment with CO (1 atm) and H₂ (1 atm) at 25 °C. In each case, the formation of a new carbonyl-containing species is indicated by the appearance of new ν_CO bands on the addition of the indicated substrate. b, ³¹P NMR spectrum (202 MHz, CD₂Cl₂) of [1(CO)]²⁺ at −60 °C showing the conversion of the ³¹P NMR-silent paramagnetic [1]²⁺ to the ³¹P NMR-active diamagnetic adduct [1(CO)]²⁺. c, ¹H NMR (500 MHz, CD₂Cl₂) hydride resonance of [1H]⁺ obtained from treatment of [1]²⁺ with H₂ and P(o-tol)₃.

Solutions of [1]²⁺, generated by treatment of 1 with 2 equiv. of ${\text{FcBAr}}_4^{\mathrm{F}}$, visibly react on mixing with H₂ (1 atm) over the course of 1 h at 25 °C, a striking result. With previously described models, the reaction occurs only sluggishly and under harsh conditions in the absence of external oxidant³⁰. The ¹H NMR spectrum of the resulting solution revealed signals at high field relative to tetramethylsilane (TMS), a region diagnostic of hydride products. Activation of H₂ is proposed to produce an ammonium hydride product. When the oxidation of H₂ by [1]²⁺ was conducted in the presence of the base P(o-tolyl)₃ (ref. 44), we observed clean formation of equimolar amounts of [1H]⁺ (see above) and phosphonium salt [HP(o-tolyl)₃]⁺ (Figs 3 and 5). Using D₂, we obtained exclusively the corresponding deuterated products [1D]⁺ and [DP(o-tolyl)₃]⁺, as established by ²H NMR spectroscopy. The reaction of [1]²⁺ and H₂ in the presence of P(o-tolyl)₃ proceeds about four times faster than the previously reported reaction of [2]⁺ with H₂ in the presence of Cp*₂Fe⁺ and P(o-tolyl)₃ under identical conditions¹⁸.

Whereas the cation [2]⁺ can only stoichiometrically activate H₂, the dication [1]²⁺ acts as a catalyst for the oxidation of H₂ in the presence of excess oxidant and excess base. We monitored the formation of [HP(o-tolyl)₃]⁺ on introducing an atmosphere of H₂ to a solution containing [1]²⁺, 6 equiv. of P(o-tolyl)₃ and excess oxidant (4 equiv. of ${\text{FcBAr}}_4^{\mathrm{F}}$). After 5 h, ³¹P NMR analysis confirmed the complete conversion of P(o-tol)₃ to [HP(o-tolyl)₃]⁺:

$${\mathrm{H}}_2+2\mathrm{P}{(o-\text{tolyl})}_3+2{\text{Fc}}^{+}\stackrel{\mathbf{1}}{\to }2{\left[\text{HP}{(o-\text{tolyl})}_3\right]}^{+}+2\text{Fc}$$

When the analogous experiment was attempted using [2]⁺, with or without excess oxidant, the catalytic conversion of P(o-tolyl)₃ to [HP(o-tolyl)₃]⁺ was not observed. Catalytic turnover requires the deprotonation of the product hydride [1H]⁺ or, possibly, [1H]²⁺. Although the rate of hydrogen oxidation by [1]²⁺ is only 0.4 turnover/h, far slower than that for the enzyme itself, these results establish additional enabling properties conferred by the FcP* functionality. Not only is the multifunctionalized diiron dithiolate capable of activating H₂, but its hydride also exhibits enhanced acidity upon oxidation, enabling catalytic turnover.

Conclusions

The experiments described above support the concept that the activation of H₂ by mixed valence diiron models benefits from the presence of a mild intramolecular oxidant⁴⁵, a function provided by an appended 4Fe–4S cluster in the [FeFe]-hydrogenases. Our experiments demonstrate that suitably modified ferrocenes can replicate the behaviour of 4Fe–4S clusters without their complications. Our approach shows that redox-complemented diiron systems display a number of properties not seen for simpler models such as redox-induced binding of CO, electronic coupling between the two S = 1/2 centres, and catalytic oxidation of H₂.

Given the importance of redox and especially proton-coupled electron transfer (PCET) in energy conversion schemes⁴⁶, the low potential redox properties of FcP* are expected to be applicable to other themes in biomimetic catalysis.

Methods

All procedures were carried out using standard Schlenk techniques. Tetrahydrofuran (THF) was purified by distillation from Na/benzophenone. Other solvents were dried and degassed by passage through activated alumina and sparging with argon. IR spectra were recorded on a Perkin Elmer Spectrum 100. NMR spectra were recorded on a Varian Unity 500 MHz NMR spectrometer. Additional experimental procedures and details can be found in the Supplementary Information.

(Diethylphosphinomethyl)nonamethylferrocene (FcP*)

Li[C₅Me₄CH₂PEt₂] was prepared by slow addition of 22 ml of 2,3,4,5-tetramethylfulvene²⁶ solution (0.21 M in toluene) to 0.49 g (4.56 mmol) of LiPEt₂ in 100 ml of THF at −78 °C. The reaction mixture was allowed to warm to room temperature overnight, resulting in a pale yellow solution of Li[C₅Me₄CH₂PEt₂]. Finely ground FeCl₂ (0.578 g, 4.5 mmol) was added and stirred with 100 ml of THF for 2 h. The flask was then cooled to −78 °C and a pre-cooled solution of 0.648 g (7.1 mmol) of LiCp* in 60 ml THF was added by cannula over 10 min. The mixture was stirred at −30 °C until it turned olive green (about 1 h). The flask was then cooled to −78 °C, and the solution of Li[C₅Me₄CH₂PEt₂] was added by cannula over the course of 1 h. The mixture was allowed to warm to room temperature and stirred overnight, during which time the solution became dark. Solvent was removed under vacuum. The resulting dark solid was extracted into 90 ml of pentane, which was filtered through Celite. The resulting yellow/orange filtrate was evaporated under vacuum and purified by chromatography on silica gel. Elution with 30/70 toluene/pentane yielded a fast-moving yellow band of decamethylferrocene; subsequent elution with 50/50 toluene/pentane yielded a second yellow band of desired product followed by a third yellow band of (C₅Me₄CH₂PEt₂)₂Fe using 100% toluene. Yield: 1.05 g (55%). ¹H NMR (500 MHz, C₆D₆, 20 °C): δ 2.38 (s, 2H), 1.82 (s, 6H), 1.67 (s, 15H), 1.66 (s, 6H), 1.28 (m, 4H) 1.00 (dt, J_H-H = 7 Hz, J_P-H = 14.5 Hz, 6H). ³¹P{¹H} NMR (202 MHz, C₆D₆, 20 °C): δ −18.4 (s). Anal. Calcd for C₂₉H₃₄FeP (found): C, 69.56 (69.97); H, 9.49 (9.62).

Fe₂(adt^Bn)(CO)₃(FcP*)(dppv)

A 250 ml Schlenk flask was charged with a solution of 72 mg (0.17 mmol) of FcP* and 142 mg (0.17 mmol) of Fe₂(adt^Bn)(CO)₄(dppv) (ref. 17) in 100 ml toluene. The solution was photolysed in a Pyrex Schlenk tube at 450 nm using an light-emitting diode array. After ~12 h, the reaction was judged complete by IR spectroscopy, and solvent was removed under vacuum to give 182 mg (87%) product. ¹H NMR (500 MHz, C₆D₆, 20 °C): δ 8.07–7.93 (m, 6H), 7.48–7.39 (m, 6H), 7.30–7.09 (m, 13H), 6.74 (d, J_P-H = 5 Hz, 2H), 3.05 (d, J_H-H = 5 Hz, 2H), 3.00 (s, 1H), 2.77 (d, J_H-H = 10 Hz), 1.98 (d, J_H-H = 5 Hz), 1.72 (s, 6H), 1.72 (s, 6H), 1.69 (m, overlapping, ~4H), 1.66 (s, 15H), 1.04 (dt, J_P-H = 15 Hz, J_H-H = 7 Hz, 6H). ³¹P{¹H} NMR (162 MHz, C₆D₆, 20 °C): δ 93.9 (s, 2P), 59.9 (s, 1P). IR (CH₂Cl₂): ν_CO = 2,022 (w), 1,958 (s), 1,945 (m), 1,907 (s), 1,897 (s), 1,881 (m) cm⁻¹. Anal. Calcd for C₆₂H₇₃Fe₃NO₃P₃S₂ (found): C, 62.34 (62.86); H, 6.13 (6.21); N, 1.01(1.16).

[HFe₂(adt^Bn)(CO)₃(FcP*)(dppv)]${\text{BAr}}_4^{\mathrm{F}}$

Via 1 and H(OEt₂)₂${\text{BAr}}_4^{\mathrm{F}}$

In a Schlenk flask, a solution of 15 mg (12.5 mmol) of 1 in 2 ml CH₂Cl₂ was treated with a solution of 12.66 mg (12.5 µmol) H(OEt₂)₂${\text{BAr}}_4^{\mathrm{F}}$ in 5 ml CH₂Cl₂. The resulting solution was stirred for 30 min, during which time the solution colour changed from brown to red-orange. The product was precipitated by adding 25 ml hexanes. The resulting red-brown coloured solid was washed with small portions of hexane and dried under vacuum. Yield: 22 mg (85% yield) of [1H]${\text{BAr}}_4^{\mathrm{F}}$. ³¹P NMR (202 MHz, CD₂Cl₂, 20 °C): δ 93.2 (dd, J_P-P = 5 Hz, 4 Hz, 1P), 89.1 (d, J_P-P = 4 Hz, 1P), 54.3 (d, J_P-P = 5 Hz, 1P); ¹H NMR (500 MHz, CD₂Cl₂, 20 °C): δ −15.33 (ddd, J_P-H = 22 Hz, 19.3 Hz, 3.8 Hz). IR (CH₂Cl₂): ν_CO = 2,023 (m) and 1,973 (s) cm⁻¹. Anal. Calcd for C₉₄H₈₅BF₂₄Fe₃NO₃P₃S₂ (found): C, 54.59 (54.92); H, 4.14 (4.09); N 0.68 (0.83).

Via H₂ and [1]²⁺

In a Schlenk flask, 15 mg (12.5 µmol) of 1, 3.6 mg (12.5 µmol) of P(o-tol)₃ and 26.2 mg (25 µmol) of ${\text{FcBAr}}_4^{\mathrm{F}}$ were dissolved in 5 ml CH₂Cl₂, resulting in a deep purple solution. Hydrogen was then bubbled through the solution, which changed from deep purple to red-orange. IR spectroscopy showed that conversion to [1H]⁺ was complete within 1 h. The IR, ³¹P NMR, and ¹H NMR spectra of the product matched those for [1H]${\text{BAr}}_4^{\mathrm{F}}$ prepared from treatment of 1 with H(OEt₂)₂${\text{BAr}}_4^{\mathrm{F}}$ as above.

[Fe₂(adt^Bn)(CO)₄(FcP*)(dppv)][${\text{BAr}}_4^{\mathrm{F}}$]₂, [1(CO)][${\text{BAr}}_4^{\mathrm{F}}$]₂

In a Schlenk flask, a solution was prepared of 15 mg (12.5 µmol) of 1 and 26.2 mg (25 µmol) of ${\text{FcBAr}}_4^{\mathrm{F}}$ in 5 ml CH₂Cl₂. CO was then bubbled through the solution, which changed colour from deep purple to orange within seconds. Solution IR spectroscopy showed new ν_CO bands at 2,077, 2,051 and 2,018 cm⁻¹. On purging with dry Ar, these bands disappeared, restoring the original colour and IR spectrum of [1]²⁺. To probe the reaction by ³¹P NMR spectroscopy, a J. Young tube was charged with 4.2 mg (3.5 µmol) of 1, 7.3 mg (7 µmol) ${\text{FcBAr}}_4^{\mathrm{F}}$ and 0.75 ml CD₂Cl₂. The ³¹P NMR spectrum showed no resonances. The tube was pressurized with CO (2 atm) and cooled to −50 °C, resulting in a colour change from purple through orange to lime-green. ³¹P NMR (202 MHz, CD₂Cl₂, −50 °C): δ 68.6 (dd, J_P-P = 10 Hz, 6 Hz, 1P), 68.1 (d, J_P-P = 6 Hz, 1P), 39.1 (broad, 1P).