Switching Site Reactivity in Hydrogenase Model Systems by Introducing a Pendant Amine Ligand.

Hydrogenases are versatile enzymatic catalysts with an unmet hydrogen evolution reactivity (HER) from synthetic bio-inspired systems. The binuclear active site only has one-site reactivity of the distal Fed atom. Here, binuclear complexes [Fe2(CO)5(μ-Mebdt)(P(4-C6H4OCH3)3)] 1 and [Fe2(CO)5(μ-Mebdt)(PPh2Py)] 2 are presented, which show electrocatalytic activity in the presence of weak acids as a proton source for the HER. Despite almost identical structural and spectroscopic properties (bond distances and angles from single-crystal X-ray; IR, UV/vis, and NMR), introduction of a nitrogen base atom in the phosphine ligand in 2 markedly changes site reactivity. The bridging benzenedithiolate ligand Mebdt interacts with the terminal ligand's phenyl aromatic rings and stabilizes the reduced states of the catalysts. Although 1 with monodentate phosphine terminal ligands only shows a distal iron atom HER activity by a sequence of electrochemical and protonation steps, the lone pair of pyridine nitrogen in 2 acts as the primary site of protonation. This swaps the iron atom catalytic activity toward the proximal iron for complex 2. Density-functional theory (DFT) calculations reveal the role of terminal phosphines ligands without/with pendant amines by directing the proton transfer steps. The reactivity of 1 is a thiol-based protonation of a dangling bond in 1- and distal iron hydride mechanism, which may follow either an ECEC or EECC sequence, depending on the choice of acid. The pendant amine in 2 enables a terminal ligand protonation and an ECEC reactivity. The introduction of a terminal nitrogen atom enables the control of site reactivity in a binuclear system.

Introduction

[FeFe] hydrogenase enzymes are highly efficient catalysts with a turnover frequency (TOF) exceeding 6000 s–1 at an electrochemical potential close to the thermodynamic limit for the process of hydrogen generation.[1,2]

Initial crystal structures[3,4] of the enzyme were unable to unambiguously identify the nature of the central atom of the bridging dithiolate ligand. 14N-HYSCORE investigations indicated the possibility of an amine,[5] but a final assignment was only possible from structural [FeFe] model complexes with a well-designed coordination environment.[6,7] The NH group is thought to be favoring the proton reduction process by serving as a proton shuttle to the diiron subsite, thereby assisting in the formation of a key intermediate that involves hydrido–proton interaction.[8] Later, the azadithiolate was also resolved in high-resolution protein X-ray structures.[9,10] While the active site H-cluster is a diiron subsite coordinated by small inorganic CO and CN ligands and is connected to a [4Fe-4S] cubane cluster, only the iron atom distal to the cubane (Fed) is involved in a proton reduction mechanism to release H2 (Figure ).[3,4]

Figure 1

Structure of the [FeFe] hydrogenase active site.

The mechanism of the enzymatic hydrogen evolution reaction (HER) has been thoroughly reviewed in the literature.[2,4,11−13]

The approaches to design bio-inspired model complexes have been extensively studied in several articles.[14−16] In particular, the balance between sites of protonation and hydride reactivity is discussed. A large number of complexes have been investigated with either a pendant amine in the dithiolate-bridged adt complexes, −NH or −NR (R = different groups), or terminal pendant amines to substitute one or more carbonyl ligand(s) (some selected examples are shown in Figure ),[17−41] most of which display alkyl dithiolate bridging ligands. Incorporation of a basic terminal nitrogen-containing ligand can assist proton transfer to the nearest metal ion.[38] When introducing larger bridging aromatic thiolate ligands[42,43] and/or terminal nitrogen-containing ligands,[44,45] a pronounced effect on redox potential and a catalyst stabilization by π–π stacking of the bridging and a terminal aromate can be seen.[46]

Figure 2

Diiron carbonyl complexes with pendant amines.

Herein, we present two complexes [Fe2(CO)5(μ-Mebdt)(P(4-C6H4OCH3)3)] 1 and [Fe2(CO)5(μ-Mebdt)(PPh2Py)] 2, (P(4-C6H4OCH3)3 = tris(4-methoxyphenyl)phosphine, PPh2Py = diphenyl-2-pyridylphosphine) with identical primary coordination environments consisting of a (CO)3Fe(I)Fe(I)(PX3) core with a 3,4-toluenedithiolate (μ-Mebdt) bridging aromatic dithiolate ligand and terminal phosphine ligands (PX3). Complex 2 only differs in the secondary coordination environment in that the methoxyphenyl groups are substituted by two aromatic phenyl and one pyridine ligand.

Complexes 1 and 2 are investigated for their respective ability to catalyze the proton-assisted hydrogen evolution reaction. The electrochemical behavior of complexes 1 and 2 on treatment with acids (CH3CO2H, CF3CO2H, and HClO4) and the effect of the presence of phosphine ligands without and with a pendant amine are discussed. Although both complexes 1 and 2 are very similar spectroscopically and electrochemically, the quantum chemically suggested reaction mechanisms are strikingly different. Upon one-electron reduction, in the tris(4-methoxyphenyl) terminal phosphine-substituted diiron complex 1, an elongation of one Fe–S bond occurs, which enables thiolate protonation and hydrogen evolution involving the distal iron atom. Both ECEC and EECC mechanisms appear feasible and may be controlled by the strength (pKa) of a given acid. Strong acids have been reported to promote an ECEC pathway in comparison to an EECC pathway promoted by weak acids. In the catalytic cycle, the subsequent E or C step that follows the first E step is determined based on the pKa values of the first reduced intermediate formed and that of the acid used as a proton source.

In the case of complex 2, with a pyridyl group in the phosphine ligand, the one-electron reduction is partially ligand-based and enables protonation of the lone pair of the nitrogen atom as the first step. A proton transfer to the proximal atom affords a reactive terminal hydride and is accompanied by a concerted ligand rotation. By changing the second coordination sphere ligands and introducing a peripheral site of protonation, one can thus control the site reactivity of a binuclear iron catalyst.

Results and Discussion

Preparation and Crystallographic Characterization

The precursor complex [Fe2(CO)6(μ-Mebdt)] was synthesized as reported in the literature.[47] The target complexes [Fe2(CO)5(μ-Mebdt)(P(4-C6H4OCH3)3)] 1 and [Fe2(CO)5(μ-Mebdt)(PPh2Py)] 2 were then prepared by reaction of the precursor complex with the monodentate phosphine ligands P(4-C6H4OCH3)3 and PPh2Py using standard Schlenck line techniques with toluene as the solvent (Scheme S1, see Supporting Information, SI). The complexes were then purified by column chromatography and recrystallized from n-hexane–dichloromethane solutions at low temperature. Blood-red and orange-red plate-like single crystals were obtained for complexes 1 and 2, respectively. The crystallographic parameters and the X-ray structures for complexes 1 and 2 are shown in Table S1 (see SI) and Figure , respectively. The lattice and structural parameters are similar to those reported in the literature[48] but free from any disorder of the N1 and C7 atoms in 2. The expected distorted square pyramidal geometry is found for 1 and 2 around the iron (I) centers.[35,49] The P(4-C6H4OCH3)3 and PPh2Py ligands occupy the apical position, which is similar to that seen for other reported monodentate phosphine-substituted complexes, such as [Fe2(CO)5(μ-pdt)(PPh2Py)], [Fe2(CO)5(μ-Mebdt)(PPh3)], [Fe2(CO)5(μ-bdt)(PPh3)], [Fe2(CO)4(μ-bdt)(PPh3)2], [Fe2(CO)5(μ-bdt)(PPh2Me)], and [Fe2(CO)4(μ-bdt)(PPh2Me)2].[32,45,49] A full list of bond lengths and bond angles for complexes 1 and 2 is given in Table S2 (see SI). The Fe–Fe distances in 1 and 2 are 2.4901(9) and 2.4769(9) Å, respectively, which are close to that reported for the [FeFe] hydrogenase enzyme active site (2.6 Å).[4]

Figure 3

Crystal structures of (a) [Fe2(CO)5(μ-Mebdt)(P(4-C6H4OCH3)3)] 1 and (b) [Fe2(CO)5(μ-Mebdt)(PPh2Py)] 2. Hydrogen atoms have been omitted for clarity.

The Fe–Fe bond distance for 1 is, however, slightly elongated compared to that of 2 because of the stronger σ-donor properties of P(4-C6H4OCH3)3 ligand in comparison to the PPh2Py ligand. The methoxy groups in the para-position of the phenyl rings make the phosphorus atom more electron-donating. Moreover, the Fe–Fe bond distance in 1 and 2 is greater than the all carbonyl [Fe2(CO)6(μ-Mebdt)] complex (2.4754 Å)[49] but similar to that reported in related compounds containing apical monodentate phosphine ligands.[45] Characteristic structural parameters for 1 and 2 from calculations are compared with those from the X-ray structural analyses in Table . The agreement between experimental and calculated structural parameters is within ∼0.05 Å for bond distances and 4° for bond angles.

Table 1

Comparison of Relevant Experimental Structural Data for Complexes 1 and 2 with Calculated Data (BP86D3)a

distance/Å	1 (exp)	1–	2 (exp)	2–
Fe–Fe	2.55 (2.49)	2.61	2.53 (2.48)	2.66
Fe–P	2.25 (2.25)	2.24	2.23 (2.23)	2.24
Fep–S	2.32, 2.31 (2.28, 2.27)	2.36, 2.30	2.31, 2.30 (2.28, 2.28)	2.31, 2.34
Fed–S	2.31, 2.30 (2.28, 2.29)	2.76, 2.29	2.31, 2.32 (2.28, 2.28)	2.32, 2.40
(Mebdt)com—p(MeOPh)/Phb	3.64 (3.91)	3.59	3.62 (3.62)	3.68

Additional calculated values are given for the mono-reduced forms to discuss structural changes. The values in parentheses are from X-ray crystallography.

Center-of-mass-to-center-of-mass distance between bridging Mebdt aromatic ring and terminal phosphine aromatic ligand in apical position.

Spectroscopic Characterization

The spectroscopic data (FTIR and NMR) for complexes 1 and 2 match well with those reported for similar complexes (Figures S1–S5, see SI).[48] The FTIR spectra of the mono-substituted complexes 1 and 2 show strong absorption bands in the regions 2042–1916 and 2046–1934 cm–1 in acetonitrile and dichloromethane, respectively, that are assigned to vibrations of terminal carbonyl groups (Table S3, see SI).

The calculated IR spectra of complexes 1 and 2 are in good agreement with the experiment. The spectra of 1 and 2 are almost indistinguishable (Figure S5, see SI). Only after the first one-electron reduction, small differences in the calculated IR spectra of 1/2 become apparent (Figure ). In 1, the coordination environment of the carbonyl ligands becomes slightly unsymmetrical, while in 2, the ligand environment is still symmetric (Figure , right).

Figure 4

Computational IR spectra. Left: for complexes 1/1/1; right: comparison of calculated spectra of 1/1 and 2/2.

Electrochemistry

The redox behavior of the diiron complexes 1 and 2 was investigated in acetonitrile by cyclic voltammetry. The reduction potentials are given in Table , and all potentials are referenced to Fc/Fc+ unless otherwise noted.

Table 2

Electrochemical Data of Complexes 1 and 2

complex	Epc/V	Epa/V	Ecat/V	overpotential	Ecat/V	overpotential
1	–1.68a	0.45b	–2.15c	0.69c	–1.63d	0.76d
2	–1.66a	0.48b	–2.14c	0.68c	–1.60d	0.73d
	–2.92a

Epa (anodic peak potential) for irreversible process.

Epc (cathodic peak potential) for irreversible process.

In the presence of acetic acid (AA).

In the presence of trifluoroacetic acid (TFA).

Complex 1 with the P(4-C6H4OCH3)3 ligand displays one irreversible reduction wave at −1.68 V. On the other hand, complex 2 with PPh2Py moiety (with a pendant amine in the monodentate phosphine ligand) displays two irreversible reduction peaks at −1.66 and −2.92 V (Figure ). The irreversibility of the reduction peaks was also confirmed from the CVs at different scan rates (Figure S6, see SI). The one-electron reduction was assigned based on the bulk electrolysis experiments (Figure S7, see SI). The peaks for oxidation of complexes 1 and 2 appear at +0.45 and +0.48 V, respectively (Figure S8, see SI). The more negative potentials for 1 and 2 in comparison to [Fe2(CO)6(μ-Mebdt)] are due to substitution of CO with phosphine ligands.[49] An additional reduction wave was also observed at −2.02 and −1.90 V for 1 and 2, respectively. The appearance of reduction peaks for complex 1 at a more negative potential than 2 is in agreement with FTIR and NMR data and shows that the P(4-C6H4OCH3)3 ligand is slightly more electron-donating than PPh2Py. The observation of anodic peaks between −1.29 and −1.31 V for 1 and 2 suggests slow exchange of a CO ligand for a weakly bound solvent molecule after reduction. A similar ligand exchange has been previously reported for [2Fe2S] complexes containing monodentate phosphine ligands.[50−52]

Figure 5

Cyclic voltammograms (CVs) for reduction of complexes 1 (1.40 mM) (—) and 2 (1.46 mM) (- - -) in acetonitrile (vs Fc/Fc+) at a scan rate of 0.1 V s–1. Inset: Full-range CVs of complexes 1 (—) and 2 (- - -) in acetonitrile at a scan rate of 0.1 V s–1.

The calculated structures of the mono-reduced species 1 and 2 are given in Figure together with the calculated unpaired spin density distribution. For complex 2, the unpaired spin density is evenly distributed between the metal ions with spin densities of 0.31 and 0.28 and changes in structural parameters are only minor (Table ). In 1, however, one of the distal Fe–S bonds is broken and the distance increases from 2.31 Å in 1 to 2.76 Å in 1 which opens up accessible site for protonation (Table ). Here, the reduction is occurring at the distal iron atom. The unpaired spin density is 0.74 at the distal iron and close to zero at the proximal iron. The bridging sulfur atom has an unpaired spin of 0.11. The calculated changes in Gibbs free energies for the first reduction step of 1/1 (−72 kcal/mol for BP86; −70 kcal/mol with B3LYP) and 2/2 (−69 kcal/mol for BP86; −70 kcal/mol with B3LYP) are almost indistinguishable and again confirm the very similar electronic properties of the compounds. The calculated redox potential of −1.62 V agrees well with the experimental value.

Figure 6

Structures and unpaired spin density distributions at 0.005 au of mono-reduced Fe0FeI species: (A) 1 and (B) 2.

The electrocatalytic properties of complexes 1 and 2 were investigated in the presence of acids of varying strengths. Upon addition of acetic acid (AA), new reduction peaks (cathodically shifted) were observed at potentials of −2.15 and −2.14 V for 1 and 2, respectively. The current height for these new reduction peaks increased with acid concentration initially and leveled off after addition of ∼200 mM acid (Figure ).

Figure 7

Cyclic voltammograms of (a) complex 1 (1.1 mM) and (b) complex 2 (1.1 mM) in acetonitrile without acid (- - -) and with increasing amounts (6–197 mM) of acid (—) (CH3COOH) at a scan rate of 0.1 V s–1.

However, the peak currents for the initial reduction peaks of 1 and 2 increase only slightly on addition of acetic acid. The cyclic voltammograms also displayed cathodically shifted second reduction peaks at −2.47 V for 1 and −2.44 V for 2 in the presence of increasing amounts of AA. Acetic acid reduction was observed at −2.56 V vs Fc/Fc+ in the absence of a catalyst, while the catalytic reductions were observed at potentials less than −2.56 V (Figure S9, see SI). The peak catalytic currents were comparable for both 1 and 2. The plot of current versus acid concentration (AA) for the complexes is shown in Figure S10 (see SI), wherein the current increases and then levels off with a maximum of 200 mM acid in the solution, indicating that catalysis becomes independent of acid concentration. The coulometry data also support the electrocatalytic activity of the complexes (Figure S11, see SI). There was not much difference in the overpotentials (0.69 V for 1 and 0.68 V for 2) associated with the catalytic reduction of acetic acid (EAcOH0 = −1.46 V)[53] by both the complexes (Table ). Similar types of model complexes have been reported previously with large catalytic currents and overpotentials.[54,55] The first reduction potentials 1/1 and 2/2 are not sufficient to reduce protons from acetic acid. This is indicated by the first reduction peaks of the complexes for which the current does not increase on addition of acid. However, there are examples of complexes that give a notable increase in peak current at the first reduction potential in the presence of acetic acid.[32,55] Henceforth, a mechanism of electrochemical reduction (E), followed by a first protonation (C) appears more plausible. The monoprotonated species 1H and 2H are then reduced at −2.15 and −2.14 V, respectively.

The electrocatalytic properties of 1 and 2 were also investigated in the presence of perchloric acid and trifluoroacetic acid (TFA). With HClO4, catalysis was observed for both the complexes (addition of up to 15 mM of acid) with an overpotential of 1 V (EHClO0 = −0.26 V).[53,54] Since low currents were observed, detailed investigations were not carried out. The slow reaction with perchloric acid could be related to catalyst instability and loss of reactivity. With TFA, new reduction peaks were observed at −1.63 and −1.60 V for 1 and 2, respectively, that shifted cathodically on further addition of acid.[53,54] A second set of peaks at −2.10 and −1.93 V for 1 and 2, respectively, was also observed (Figure ). The appearance of multiple peaks with TFA as a proton source could be attributed to homoconjugation of the acid.[56,57] The background currents without catalyst in the presence of TFA due to direct reduction of protons at the glassy carbon electrode were found to be negligible for potentials in the range of −1.2 to −1.9 V and hence, the acid-induced currents in the presence of catalysts can be attributed to catalytic turnover (Figure S12, see SI). The plot of current versus acid concentration (TFA) for the complexes is shown in Figure S13 (see SI).

Figure 8

Cyclic voltammograms of (a) complex 1 (1.1 mM) and (b) complex 2 (1.1 mM) in acetonitrile without acid (- - -) and with increasing amounts (1–55 mM for 1 and 1–78 mM for 2) of acid (—) (CF3COOH) at a scan rate of 0.1 V s–1.

Reaction Mechanism of Proton Reduction

As mentioned above, the mono-reduced state 1 (E) possesses an accessible site of protonation at one of the bridging Mebdt thiolate sulfur atoms. The Gibbs free energies for sulfur protonation of −16 kcal/mol (BP86) and −27 kcal/mol (B3LYP) show the feasibility for the first step of protonation (C). Only the application of implicit solvation leads to a positive Gibbs free energy change of +18 or +1 kcal/mol, respectively. This is an indication of the problem of treating charged compounds reliably in an implicit solvation model. It is this difference in solvation energies between 1 and 1SH of 29 kcal/mol that renders the calculated protonation energies in a solvent to become positive. Upon the second one-electron reduction, a spontaneous intramolecular proton transfer from 1SH to the distal iron atom to give 1FeH occurs (−95 kcal/mol with BP86 and −80 kcal/mol with B3LYP). This recovers the site of protonating at the Mebdt thiolate. The second step of protonation (ECEC) to give 1FeHSH again is clearly possible with calculated protonation Gibbs free energies of −11 kcal/mol (BP86) and −35 kcal/mol (B3LYP). Again, the large difference in solvation energies between 1FeH and 1FeHSH of 31 kcal/mol gives positive Gibbs free energies for protonation in solvent. The release of molecular hydrogen from 1FeHSH is thermodynamically driven by −27 and −26 kcal/mol and recovers complex 1 (Scheme ).

Scheme 1

Suggested Reaction Mechanisms for Acid-Assisted Proton Reduction of Complex 1

Left: Gibbs free energy difference in an implicit solvent model (COSMO); Right: Gibbs free energy difference in the absence of a solvent. All values in kcal/mol.

Interestingly, the HER is centered at the distal iron atom only: protonation and one-electron reduction steps involve one of the bridging thiolates and the distal iron only (Scheme ). The proximal iron atom is stabilized by π–π stacking interaction with one of the terminal 4-C6H4OCH3 groups. This interaction is persistent throughout the entire catalytic cycle and obstructs large structural rearrangements of the iron atoms and the bridging μ-Mebdt ligand (Figure S14, see SI). Actually, during the HER of 1, the Fe–Fe bond distance only changes from 2.55 Å in 1 to 2.73 Å in 1FeHSH. Likewise, the center-to-center distance between the bridging aromatic Mebdt ligand and aromatic rings of tris(4-methoxyphenyl)phosphine ligand varies between 3.43 Å in 1FeH (strong stabilizing effect, one Fe–S bond broken, iron atom protonated) and 3.74 Å in 1FeHSH (weak stabilization enabling release of H2) (Figure S14, see SI). A similar stabilizing effect by a bridging aromatic thiolate and a terminal aromatic ring of the phosphine substituent was also observed in an earlier report by our group.[46] An EECC mechanism with two consecutive one-electron steps to give 1 (with Gibbs free energy changes of −140 kcal/mol for BP86 and −123 kcal/mol for B3LYP) and two sequential steps of protonation also appears feasible since protonation will be facile and swift then (Scheme S2, see SI). This is also evident from the reduction peaks at −2.47 and −2.44 V, which can be associated with the reduction of 1–/2– to 12–/22–. However, when the acetic acid concentration is increased, the hydride-containing species 1H/2H are generated and are then reduced at −2.15 and −2.14 V.[23] Both mechanisms may operate in parallel and be controlled by the pKa of the acid vs closely spaced redox potential.[58]

The overall mechanistic picture of HER of complex 2 appears to be similar to that of 1 at first glance only (Scheme ). The first one-electron reduction to afford 2 occurs with a change in Gibbs free energies of −69 and −70 kcal/mol for BP86 and B3LYP, respectively, and thus almost identical to the reduction 1/1. Unlike in 1, this first electrochemical reduction step does not induce the opening of an iron–sulfur bond and the thiolate to become available as a site for protonation. Neither the Fe–Fe bond distances (2.53 Å in 2 and 2.66 Å in 2) nor the Fe–S bond distances (2.30–2.32 Å in 2 and 2.31–2.40 Å in 2) change significantly. Also, the center-to-center distance between the aromatic ring of the bridging Mebdt ligand and a p-methoxyphenyl group of 3.62 Å in 2 and 3.68 Å in 2 remains almost unchanged. Although the reduction in 1 is Fed (0.74)- and thiolate (0.11)-based, in 2, the unpaired spin is delocalized over the two iron atoms (0.39 at the Fed and 0.28 at the proximal Fep) and, to a lesser degree, over one of the sulfur atoms (0.05) and the nitrogen atom of the pyridine (0.05). This demonstrates that (i) the simultaneous stability and plasticity of complex are able to easily accommodate an extra electron without large structural changes, and (ii) modulation of the terminal ligand leads to a change from reduction of only Fed in 1 to a delocalized Fed/Fep reduction in 2.

Scheme 2

ECEC Mechanism for Acid-Assisted Proton Reduction of Complex 2

This partial reduction of the pyridine ring increases its electron richness, making the pyridine nitrogen basic enough to become protonated (to give 2H) in the subsequent step. Since all other covalent bonds are intact, this is the only possible site of protonation in 2. The Gibbs energies of protonation in the gas phase of −64 (BP86) and −79 kcal/mol (B3LYP) become +4 and −8 kcal/mol, respectively, when using a COSMO model due to the difficulty with differently charged complexes.

The second one-electron reduction of 2H to 2H is accompanied with a change in the Gibbs free energy of −78 and −58 kcal/mol for BP86 and B3LYP, respectively. Large structural changes are not observed during this step. Minor changes in iron–iron bond distances by 0.3 Å and Mebdt–terminal ligand distance by 0.14 Å rationalize the stability during the HER of complex 2 (Figure S15, see SI). A proton transfer of the pyridinium proton to the proximal iron atom is exothermic by −6 kcal/mol for BP86 and −24 kcal/mol with B3LYP to give the terminal hydride species 2FeH.

In 2FeH, a Fe–H hydride bond of 1.54 Å is formed and accompanied by a concerted rearrangement of the phosphine ligand: upon proton transfer, the phosphine ligands rotate by 104° around the Fe–P bond so that the π–π stacking phenyl in 2 and 2– is replaced by the pyridine ring. In 2FeH, now the pyridine occupies the apical position and establishes stacking interactions with the benzendithiolate (with a center-to-center distance of 3.60 Å). Simultaneously, the pyridine ring undergoes a rotation around the C2–C5 axis of the five-membered ring such that the nitrogen atom is pointing away from the proximal iron atom (Figure ) and no longer accessible for a second protonation step. Upon generation of an iron hydride at Fep, one of the proximal iron–thiolate bonds breaks (Fe–S distance of 3.92 Å). This monocoordinate thiolate can then be protonated to give 2FeHSH, a formal Fe(0)Fe(0) species which bears an iron hydride and a thiol ligand. Here, the iron–iron distance reduces from 2.89 Å in 2FeH– to 2.72 Å in 2FeHSH and the Mebdt–py distance reduces from 3.60 to 3.56 Å (Figure S15, see SI). A close to thermoneutral transfer of the proton from the thiol to the proximal iron atom gives the intermediate 2FeHH, which corresponds to a weakly bound μ-H2 complex with iron–hydrogen distances of 1.77 and 1.91 Å and a H–H bond distance of 0.85 Å. This corresponds to an activated H2 complex (H–H bond distance of 0.75 Å). From 2FeHH, molecular hydrogen is easily released with a change in Gibbs free energy of −6 (for BP86) and −18 kcal/mol for B3LYP to complete the cycle.

Figure 9

Intramolecular proton transfer of the pyridinium proton from 2H– (left) to the proximal iron atom Fep to yield 2FeH– (right). This transfer is accompanied by a phosphine ligand rotation to bring the pyridine into the stacking position with Mebdt and a pyridine ring rotation around the C2–C5 axis.

Conclusions

Two complexes 1 and 2 with proton-assisted hydrogen evolving catalytic activity were characterized in detail using various spectroscopic techniques and quantum chemical calculations. The complexes only differ in the terminal monodentate phosphine ligands P(4-C6H4OCH3)3 and PPh2Py with very similar structural (X-ray) and electronic properties (νCO values and reduction potentials). The newly synthesized complexes are catalytically active in the presence of various acids (acetic acid, trifluoroacetic acid, and perchloric acid). The catalytic reduction peaks were observed at −2.15 and −2.14 V for 1 and 2, respectively, while in the absence of catalysts, the reduction for acetic acid was observed at −2.56 V vs Fc/Fc+.

The stabilization of a terminal (4-C6H4OCH3)3 phosphine ligand by a stacking interaction with a bridging μ-naphthalene-2-thiolate ligand significantly reduced structural changes during the catalytic cycle, hindered the rotation of the phosphine ligand, and the reaction proceeded involving a bridging hydride and a protonated thiolate species.[46] In complexes 1 and 2, the formation of such a μ-hydride seems to be avoided by a stacking interaction between the bridging benzenedihiolate and a terminal aromatic ring. In the absence of a bridging hydride, the terminal hydrides may react with thiol protons and release H2 in the absence of an energy barrier.

The acidity of metal hydride and dihydrogen complexes (their pKa’s) is relevant for many processes in catalysis and biology, hydrogen storage and generation, and fuel cell electrocatalysis.[59] The sequential events of electrochemical (E, reduction) and chemical (C, protonation) steps require a detailed balance between ligand proton affinities (pKa) and transition-metal hydricities. The fine-tuning of both is critically dependent on the charge of the conjugate base complex. Here, in complex 2, the pKa value of a pyridinium (5.25 in water) is apparently very close to that of the proximal Fep terminal hydride.

By incorporating a pyridine nitrogen atom into a binuclear FeFe complex, the site-selective reactivity can be shifted from the distal (in 1) to the proximal (in 2) iron atom. The intramolecular stabilization by aromatic stacking interactions obstructs large structural rearrangements and complex degradation. The introduction of pendant amines as an initial site of protonation, then proton relay and directing the transfer of protons in mononuclear[60−62] and binuclear[33] complexes has been inspired by the design of the [FeFe] hydrogenase enzymatic system. The growing understanding and control of proton shuffling to the development of molecular electrocatalysts for energy storage[63] and hydrogen production[64] is striking.

Experimental Section

General Procedure

All of the experiments were carried out in an inert atmosphere using Schlenk techniques unless otherwise specified. [Fe2(CO)6(μ-Mebdt)] was prepared according to the procedure reported in literature.[49] All of the starting materials and anhydrous solvents were obtained from Sigma-Aldrich and used without any further purification. The deuterated solvents were also obtained from Sigma-Aldrich. The 1H and 31P NMR spectra were recorded at room temperature in CDCl3 with a JEOL 400 MHz NMR spectrometer. The FTIR spectra were recorded from dichloromethane and acetonitrile solutions of the samples over the range of 400–4000 cm–1 on a PerkinElmer FTIR spectrometer. The UV–vis spectra for the complexes 1 and 2 were recorded on a PerkinElmer Lambda-25 spectrophotometer. The elemental analyses were carried out with a Vario Micro Cube elemental analyzer.

Electrochemical measurements were conducted in acetonitrile with 0.1 M tetrabutylammoniumhexafluorophosphate (Fluka, electrochemical grade) as a supporting electrolyte that was dried in vacuum at 383 K. Cyclic voltammetry was carried out using an Autolab potentiostat with a GPES electrochemical interface (Eco Chemie). The working electrode was a glassy carbon disk (diameter, 3 mm; freshly polished) for cyclic voltammetry. A platinum wire was used as the counter electrode. The reference electrode was a non-aqueous Ag/Ag+ electrode (CH Instruments, 0.010 M AgNO3 in acetonitrile). All of the potentials (text, tables, and figures) are quoted against the ferrocene–ferrocenium couple (Fc/Fc+); ferrocene was added as an internal standard at the end of the experiments. All solutions were prepared from dry acetonitrile (Sigma-Aldrich, spectroscopic grade, dried with MS 3 Å). Controlled potential coulometry (CPC) was performed on the same instrument with the three-electrode setup described earlier. The experiment was carried out with continuous stirring and purging of argon gas at a fixed potential.

X-ray Crystallography

Single crystals of 1 and 2 were grown by slow evaporation of hexane–dichloromethane solutions at low temperature. X-ray data of 1 and 2 were collected on Oxford X-Calibur-S and Oxford SuperNova single-crystal X-ray diffractometers using Mo-Kα radiation, respectively. Significant crystallographic parameters and refinement details are given in Tables S1 and S2 (see SI). The structures were solved and refined by full-matrix least-squares techniques on F2 using the SHELX-97 (SHELXTL program package).[65]

Computational Details

ADF2018.105[66,67] was used with the BP86[68,69] and B3LYP[70,71] exchange-correlation functionals, Grimme’s dispersion correction with Becke–Johnson damping,[72] and a TZP Slater-type basis set for all atoms.[73] The solvent (acetonitrile) was considered in a COSMO solvation model[74,75] using the Klamt atomic radii and surface definition with corrections for outlying charges. Thermodynamic corrections were obtained using statistical thermodynamics from BP86 calculations under standard conditions.[76] Redox potentials were calculated following refs (44, 77) and are given relative to an Fc/Fc+ reference electrode in acetonitrile.