Hydrogenase Enzymes and Their Synthetic Models: The Role of Metal Hydrides

Abstract

Hydrogenase enzymes efficiently process H₂ and protons at organometallic FeFe, NiFe, or Fe active sites. Synthetic modeling of the many H₂ase states has provided insight into H₂ase structure and mechanism, as well as afforded catalysts for the H₂ energy vector. Particularly important are hydride-bearing states, with synthetic hydride analogues now known for each hydrogenase class. These hydrides are typically prepared by protonation of low-valent cores. Examples of FeFe and NiFe hydrides derived from H₂ have also been prepared. Such chemistry is more developed than mimicry of the redox-inactive monoFe enzyme, although functional models of the latter are now emerging. Advances in physical and theoretical characterization of H₂ase enzymes and synthetic models have proven key to the study of hydrides in particular, and will guide modeling efforts toward more robust and active species optimized for practical applications.

Graphical abstract

1. INTRODUCTION

Certain prokaryotes live off energy liberated from the metabolism of simple oxidants and dihydrogen. While this was known even in 1892,¹ it would be another 40 years before bacterial dehydrogenase enzymes were implicated in such processes. In what must have been a startling observation, Bacterium aliphaticum liquefaciens, grown under H₂, O₂, and CO₂, was found to reduce methylene blue (E = 11 mV vs NHE, the normal hydrogen electrode) under a stream of H₂.² This early work was followed by a landmark paper by Stephenson and Stickland, who described bacterial cultures from river mud that reduced NO₃⁻, O₂, and fumarate.^3,4 They wrote that “hydrogen is in some way activated, and this activation can be conveniently expressed H₂ ⇌ 2H without implying anything about the nature of the reaction”, the catalyst for which was termed “hydrogenase”. These H₂-processing metalloenzymes are the basis for many anaerobic bacteria, protozoa, fungi, and algae⁵ collectively producing and consuming an estimated 0.3 Gt of H₂ each year.⁶

Three major classes of hydrogenase (H₂ase) have been identified, members of each featuring either an FeFe, a NiFe, or an Fe core.⁷ The [FeFe]- and [NiFe]-H₂ases have been shown to mediate the interconversion of dihydrogen (H₂) with protons (H⁺) and electrons (e⁻) at high rates⁸ and at potentials very close to those bounded by thermodynamics (eq 1).

$${\mathrm{H}}_2\rightleftharpoons 2{\mathrm{H}}^{+}+2{\mathrm{e}}^{-}$$ (1)

In contrast to these two redox enzymes, the [Fe]-H₂ases (also known as H₂-forming methylenetetrahydromethanopterin dehydrogenases, Hmd) cleave H₂ to then deliver H⁻ to the organic substrate methenyltetrahydromethanopterin, with H⁺ passing to the bulk solvent (eq 2).⁹

$${\mathrm{H}}_2+\text{methenyl}-{\mathrm{H}}_4{\text{MPT}}^{+}\rightleftharpoons {\mathrm{H}}^{+}+\text{methylene}-{\mathrm{H}}_4\text{MPT}$$ (2)

For each reaction, as was hypothesized in 1931, the catalysts are often bidirectional³ and deactivated by CO. Unbeknownst to Stephenson and Stickland, the H₂ases each catalyze the isotope exchange between H₂O and D₂ (eq 3).^{10, 11}

$${\mathrm{H}}_2+{\mathrm{D}}_2\mathrm{O}\rightleftharpoons \text{HD}+\text{HDO}\rightleftharpoons {\mathrm{D}}_2+{\mathrm{H}}_2\mathrm{O}$$ (3)

A further hallmark of H₂ase behavior is their interconversion of ortho- and para-H₂ (eq 4),¹² the two nuclear spin isomers whose differing thermal properties have implications for H₂ storage.¹³

$$\text{ortho}-{\mathrm{H}}_2\rightleftharpoons \text{para}-{\mathrm{H}}_2$$ (4)

Interest in renewable, carbon-neutral energy technologies has motivated considerable biological research into the expression (including biosynthesis),¹⁴ structure, and function of H₂ases. Each of the three enzyme cores and catalytic cycles is described in individual sections below (sections 3, 4, and 5). Before considering details, we must emphasize that the H₂ase families share several mechanistic themes—principal among them H₂ heterolysis—that are in no small part borne of the structural similarities revealed by even a cursory glance at their active sites (Figure 1).

Figure 1.

Schematic representations of the active sites in [FeFe]-H₂ase (left), [NiFe]-H₂ase (center), and [Fe]-H₂ase (right). The presence of H⁻ and H₂ ligands in [FeFe]-H₂ase and [Fe]-H₂ase, respectively, has yet to be confirmed.

Central to the H₂ases is Fe(II),¹⁵ whose ligation to thiolate and CO sees it adopt low-spin configurations most suited to the binding of H⁻ and H₂. Despite the three enzyme classes being phylogenically unrelated,¹⁶ “convergent evolution” causes expression of these common features that are apparently expedient for H₂ processing.¹⁴ Inorganic chemists are motivated to prepare structural, spectroscopic, and/or functional mimics to elucidate the mechanisms of the very enzymes that inspired their synthesis.¹⁷ Given that H₂ processing is championed by the Pt group metals, more cost-effective alternatives would involve the use of purely organic catalysts¹⁸ or those based on earth-abundant metals, like the H₂ases.^19–27 Better yet, their synthetic models may be more robust and mass-producible, with smaller spatial footprints that allow densely decorated, highly active electrodes to be prepared and incorporated into devices such as fuel cells.^28–30

Having identified the importance of Fe–CO–thiolato ensembles in H₂-processing enzymes, one can also reasonably expect complexes of H⁻ and H₂ to be somehow involved in Nature’s plans. These substrate-bearing intermediates, and the synthetic models proposed to mimic them most closely, are the focus of this review. Particular emphasis is placed on work in the past decade, which has seen great progress not only in biochemical and chemical synthesis, but also in the development of experimental and computational methods for the characterization of metal dihydrogen and hydride complexes. Before turning to these themes, some background into the chemistry relevant to these natural and synthetic catalysts is presented.

2. FORMATION AND CHARACTERIZATION OF METAL HYDRIDES

2.1. Formation of Metal Hydrides from H⁺ or H₂

Nature’s handling of H⁺ and H₂ with Fe and Ni follows patterns familiar in synthetic chemistry and catalysis.³¹ For example, electron-rich, low-valent metal sites are susceptible to protonation (oxidative addition of H⁺, Figure 2). In the case of a single metal center, the resulting hydride may vary greatly in its polarity and acid–base properties. When two metals are involved, consideration must also be paid to regiochemistry: will the product feature a terminal hydrido (t-H⁻) or a bridging hydrido (μ-H⁻) ligand?

Figure 2.

Formation of mono- and dinuclear metal hydrides from low valent metals and H⁺. In the latter case, both bridging and terminal hydrido products are possible.

When low- or mid-valent metal centers are exposed to H₂ instead of H⁺, σ-complexes of the type M(η²-H₂) may form, in which intact H₂ binds side-on to M (Figure 3).³² This bonding motif has been described by analogy with the Dewar–Chatt–Duncanson model for metal complexes of olefins.³³ Since η²-H₂ acts as a σ-base and π-acid, the bond between H atoms (bond dissocation energy (BDE) = 436 kJ mol⁻¹ for free H₂)³⁴ is weakened upon coordination, and in the case of electron-rich metals, can be entirely broken to afford M(H)₂ dihydrides through a formal oxidative addition (homolysis).³⁵

Figure 3.

Interactions of low-valent metal sites with H₂. The η²-H₂ ligand can cleave through oxidative addition to afford dihydride complexes.

The homolytic route for hydride formation, while common for group 10 metals and previously proposed based on computations of the [NiFe]-H₂ases,³⁶ is no longer thought to be relevant to the H₂ases. Indeed, the kinetic barriers associated with homolytic H₂ase pathways are prohibitive, and the enzymes instead operate through heterolysis (Figure 4). Heterolysis typically involves higher-valent metals, suitably unsaturated examples of which can ligate H₂. The resulting M(η²-H₂) complexes, when in the presence of a Brønsted base, may experience H₂ fission, but not in the homolytic sense. Rather, the base deprotonates acidic η²-H₂ to afford a metal hydride and a conjugate acid. Such a tug-of-war over the H₂ substrate, which does not result in any change in metal oxidation state,³⁷ is typical of that played out between frustrated Lewis acid and base pairs (FLPs).^38,39 Cleavage is enabled by the remarkable influence electrophilic metal centers have on H₂, whose acidity when bound is highly variable, as is evident from the contrasting behaviors of trans-[Fe(dppe)₂(η²-H₂)H]⁺ (pK_a(THF) = 12; THF = tetrahydrofuran; dppe =1,2-bis(diphenylphosphino)ethane)^40,41 and trans-[Fe(dppe)₂(η²-H₂)CO]²⁺ (pK_a(THF) ≈ pK_a (CH₂Cl₂) = −5).^42,43 Such pK_a values depart greatly from that of the free gas (pK_a(THF) ≈ 50,⁴⁴ pK_a(MeCN) ≈ 50),⁴⁰ underscoring Nature’s need for metals to activate this otherwise very strong bond.

Figure 4.

Interactions of high-valent metal sites with H₂. The η²-H₂ ligand in high-valent complexes can be acidic and is cleaved through deprotonation (heterolysis).

Pentacoordinate, low-spin 16e⁻ centers represent archetypal metal fragments for H₂ binding, and the H₂ adducts of three such motifs are presented in Figure 5. The Fe(II) electrophiles [Fe(dppe)₂CN]⁺ ([1]⁺ and [Fe(dppe)₂CO]²⁺ ([2]²⁺) both form stable η²-H₂ complexes, despite the contrasting donor/acceptor properties of CN⁻ and CO. This is further astonishing in that it is the ligand trans to the H₂ binding site that has the most influence on whether or not a η²-H₂ complex can persist. It turns out that η²-H₂ complexes are stabilized by CN⁻ and CO ligands, although for different reasons in each case. According to Pauling’s principle of electroneutrality, the strong σ-basicity of CN⁻ reduces the electron density Fe requires from the H–H σ-bond, such that the latter remains largely intact.⁴⁵ It may be somewhat counterintuitive that a strong donor such as CN⁻ does not, in this case, promote H₂ oxidative addition. But since σ-bonding between Fe and H₂ is weak, the Fe(dπ)–H₂(σ*) π-backdonation is insufficient to induce splitting. When instead a strongly π-acidic CO ligand³⁵ is ligated to Fe, the system once more attracts the H–H σ-bond. Yet since CO competes for backdonation, the η²-H₂ moiety persists as Fe(dπ)–H₂(σ*) interactions are weak.³³

Figure 5.

Fe(η²-H₂) complexes in the laboratory (top left and right) and in Nature (bottom). Note the presence of five strong-field ligands that serve to support the low-spin Fe(II) electrophile.

A continuum exists between labile M(η²-H₂) and bona fide M(H)₂ complexes, between which are metal complexes where H₂ is activated to the point that heterolysis, but not homolysis, can occur. Residing in the middle of the first row of transition metals, Fe seems to be in such a “sweet spot”. With the judiciously chosen coligands CO and CN⁻, Fe(η²-H₂) species not only form readily, but, as demonstrated by the three H₂ase families, often do so in preference to binding ubiquitous substrates such as H₂O or N₂. Once formed, η²-H₂ complexes may undergo heterolysis to afford hydrido species, general aspects of which are discussed in section 2.2.

2.2. Properties of Hydride Complexes

The strong donor properties of H⁻ are very familiar to inorganic chemists. In general, the donicity of a given ligand L can be quantified by several parameters, including the so-called P_L value.⁴⁶ Defined in terms of ligand effects on the Cr^+/0 couple (P_L = E_1/2[Cr(I/0)(CO)₅L] – E_1/2[Cr(I/0)(CO)₆]), P_L ≡ 0 V for CO, with NO⁺ (1.4 V) and OH⁻ (−1.55 V) having the expected opposite effects on redox. The value for H⁻ (−1.22 V) is less negative than the latter extreme, but more negative than the value for CN⁻ (−1.00 V), consistent with the strong donicity and anionic nature of H⁻. Closely related to the P_L value is the Lever parameter (E_L), which describes redox effects of ligands on Ru(III/II) couples.⁴⁷ The E_L parameter estimated for H⁻ is −0.5 V, in the −0.63 to 0 V range characteristic of strongly donating ligands and π-bases including OH⁻ and RS⁻. While the two scales are certainly influenced by ligand charge, they do provide useful measures of donicity, although they do not account for any steric interactions and synergistic effects between ligands.

Other ligands can be quantitatively compared to H⁻, but how can different complexes of H⁻ be compared? Nominally an anionic ligand, hydride can exist in anything from “hydridic” M–H^δ− moieties to “protic” M–H^δ+ groups. A key parameter in describing hydrides is pK_a, which, for an arbitrary species MH⁺ (eq 5), is found by treating it with a comparable amount of a suitable base (denoted B, eq 6) with a known pK_a (BH⁺) in the solvent being used. If the Brønsted acidities of MH⁺ and BH⁺ differ greatly, then only an upper or lower bound for pK_a(MH⁺) is measurable in practice. If the acidities are comparable, then K_eq (eq 7) and pK_a(MH⁺) may be easily determined.

$${\text{MH}}^{+}\rightleftharpoons \mathrm{M}+{\mathrm{H}}^{+};K_{\mathrm{a}}({\text{MH}}^{+})=[\mathrm{M}][{\mathrm{H}}^{+}]/[{\text{MH}}^{+}]$$ (5)

$${\text{BH}}^{+}\rightleftharpoons \mathrm{B}+{\mathrm{H}}^{+};K_{\mathrm{a}}({\text{BH}}^{+})=[\mathrm{B}][{\mathrm{H}}^{+}]/[{\text{BH}}^{+}]$$ (6)

$${\text{MH}}^{+}+\mathrm{B}\rightleftharpoons \mathrm{M}+{\text{BH}}^{+};K_{\text{eq}}=[\mathrm{M}][{\text{BH}}^{+}]/[{\text{MH}}^{+}][\mathrm{B}]$$ (7)

Matters are more complicated when a base forms an H-bonded adduct with its conjugate acid (eq 8), a process referred to as homoconjugation or self-association.⁴⁸

$$\mathrm{B}+{\text{BH}}^{+}\rightleftharpoons {\mathrm{B}}_2{\mathrm{H}}^{+};K_{\mathrm{f}}=[{\mathrm{B}}_2{\mathrm{H}}^{+}]/[\mathrm{B}][{\text{BH}}^{+}]$$ (8)

For certain weak acid/base pairs in organic solvents (e.g., RCO₂⁻/RCO₂H in MeCN), such effects must be taken into account when the concentration of added base exceeds 1/K_f.⁴⁹ Homoconjugation constants K_f for several acids are known,⁵⁰ and pK_a(MH⁺) can still be computed in a straightforward manner (see the Appendix).⁴⁸ Homoconjugation can be largely avoided by employing aniliniums/anilines, which are expected to exhibit low values of K_f.⁵¹ Values for pK_a(MH⁺) span a wide range,⁴⁴ and rough predictions based on metal and ligand parametrization are possible.⁴¹ The values are typically lower for electron-poor metal complexes and higher for their electron-rich counterparts, with a M–H^δ− description perhaps being more relevant in the latter case (eq 9).

$${\text{MH}}^{+}\rightleftharpoons {\mathrm{M}}^{2+}+{\mathrm{H}}^{-}$$ (9)

The reaction in eq 9 is clearly related to that in eq 5, with the metal hydride now releasing H⁻ rather than H⁺. The Gibbs free energy change associated with eq 9, known as the hydricity (ΔG°(H⁻)),⁵² is readily computed if MH⁺ can transfer H⁻ to acceptors of known hydricity, such a competitive binding method being analogous to determining pK_a by adding a base. Alternatively, if pK_a(MH⁺) and the half-wave potentials (E_1/2) of the 1e⁻ couples [M]^+/0 and [M]^2+/+ (or the 2e⁻ couple [M]^2+/0) are known, hydricity (expressed in J mol⁻¹) can also be calculated according to eq 10 or 11, derivations of which are in the Appendix.^52,53

$$\begin{array}{l}\mathrm{\Delta }G°({\mathrm{H}}^{-})=(RT/{\text{log}}_{10}\mathrm{e})\mathrm{p}{K}_{\mathrm{a}}+F{E}_{1/2}({[\mathrm{M}]}^{+/0})\hfill \\ +F{E}_{1/2}({[\mathrm{M}]}^{2+/+})-2F{E}_{1/2}({[\mathrm{H}]}^{+/-})\hfill \end{array}$$ (10)

$$\begin{array}{l}\mathrm{\Delta }G°({\mathrm{H}}^{-})=(RT/{\text{log}}_{10}\mathrm{e})\mathrm{p}{K}_{\mathrm{a}}+2F{E}_{1/2}({[\mathrm{M}]}^{2+/0})\hfill \\ -2F{E}_{1/2}({[\mathrm{H}]}^{+/-})\hfill \end{array}$$ (11)

where F = 96 485 C mol⁻¹, R = 8.314 J K⁻¹ mol⁻¹ and T is reported in kelvin. It is noted that very few H₂ase models have two consecutive reversible 1e⁻ couples (or, alternatively, a reversible 2e⁻ couple) necessary for this calculation.

The hydricity of a compound is a measure of its propensity to donate H⁻, with the lowest values corresponding to the most hydridic species. As with the pK_a’s of metal complexes, hydricities are also strongly influenced by solvation,⁵⁴ with a narrower range of values expected for more polar solvents. For example, H₂, {Ni[(1,2-bis(dihydroxylmethylphosphino)ethane]₂H}⁺, and HCO₂⁻ have vastly different hydricities in MeCN (76.6, 57.4, and 44.0 kcal mol⁻¹, respectively), the solvent most commonly used for such determinations. The values measured in the stronger dielectric H₂O are much closer (34.2, 30.0, and 24.1 kcal mol⁻¹).⁵⁴ Hydricity is intuitively also related to the electron density at the metal in question, with ΔG°(H⁻) for the relatively electron-rich and -poor complexes [Ni(dmpe)₂H]⁺ (dmpe =1,2-bis(dimethylphosphino)ethane) and [Ni(dppe)₂H]⁺ determined as 48.9 and 62.7 kcal mol⁻¹, respectively.⁵³ Lastly, hydricity is also sensitive to coordination geometry, an aspect that has been explored with [M(diphosphine)₂H]⁺ complexes, and one that is not surprising given that bite angles of chelating ligands can favor certain coordination numbers.^55,56 In particular, ligands that span large angles will stabilize a hydride-free complex, leading to lower hydricity values. Overall, the many factors contributing to hydricity mean that it is certainly possible for a species to have a high pK_a and modest hydricity, as the influence of redox may swing the result either way.

The interrelation between hydride redox, pK_a, and hydricity is summarized in Figure 6. Simple thermodynamic cycles enable determination of free energies, such as hydricity ΔG°(H⁻, II), from experimentally determined pK_a(II), E°(II/I), and E°(I/0) (or E°(II/0)) values as in eq 10. In this context, pK_a(n), ΔG°(H^•,n), and ΔG°(H⁻,n) are respectively associated with equilibria for H⁺, H^•, and H⁻ donation from [MⁿH]⁽ⁿ⁻¹⁾⁺.

Figure 6.

Thermodynamic parameters for redox and acid–base processes for a metal complex M. Colored arrows denote transfer of an electron (blue), proton (red), H^• (black), and H⁻ (green). Adapted from ref 57. Copyright 2012 American Chemical Society.

Until now, metal hydrides have been viewed here as potential sources of H⁺ or H⁻. Yet another important scenario is homolysis of an M–H bond,^58,59 which in practice does not involve free H^• but rather its delivery to a substrate.⁶⁰ The facility of the homolysis reaction can be calculated analogously to hydricity: if pK_a(II) and E°(I/0) (written in eq 13 as pK_a and E_1/2([M]^+/0), respectively) are known, one can compute the H^• donor strength (ΔG°(H^•), eq 12, 13). Once more, details of the cycles used to derive eqs 12 and 13 are presented in the Appendix.

$${\text{MH}}^{+}\rightleftharpoons {\mathrm{M}}^{+}+{\mathrm{H}}^{•}$$ (12)

$$\begin{array}{c}\mathrm{\Delta }{G}^{°}({\mathrm{H}}^{•})=(RT/{\text{log}}_{10}\mathrm{e})\mathrm{p}{K}_{\mathrm{a}}+F{E}_{1/2}({[\mathrm{M}]}^{+/0})\\ -F{E}_{1/2}({[\mathrm{H}]}^{+/0})\end{array}$$ (13)

Our discussion now returns to considering pK_a’s of metal hydrides and the significant and intuitive perturbations redox has on these.^61,62 Indeed, the oxidation of metal hydrides causes large increases in acidity, as the reductive elimination of H⁺ is certainly more favorable when the metal site is more electron poor. One family of species for which the thermodynamics are well studied are the bis(diphosphine) complexes of Ni.⁵⁷ Consider tetrahedral [Ni(dppe)₂], which readily protonates to give the respective 5-coordinate divalent hydride [Ni(dppe)₂H]⁺ (eq 14, pK_a(II) = 16.7). The same cannot be said for oxidized [Ni(dppe)₂]⁺, whose protonation product [Ni(dppe)₂H]²⁺ is a strong acid in MeCN (eq 15, pK_a(III) = −8.6).⁵⁷

$$\begin{array}{l}{[\text{Ni}{(\text{dppe})}_2\mathrm{H}]}^{+}\rightleftharpoons [\text{Ni}{(\text{dppe})}_2]+{\mathrm{H}}^{+};\hfill \\ K_{\mathrm{a}}(\text{II})=[\text{Ni}{(\text{dppe})}_2][{\mathrm{H}}^{+}]/[\text{Ni}{(\text{dppe})}_2{\mathrm{H}}^{+}]\hfill \end{array}$$ (14)

$$\begin{array}{l}{[\text{Ni}{(\text{dppe})}_2\mathrm{H}]}^{2+}\rightleftharpoons {[\text{Ni}{(\text{dppe})}_2]}^{+}+{\mathrm{H}}^{+};\hfill \\ K_{\mathrm{a}}(\text{III})=[\text{Ni}{(\text{dppe})}_2^{+}][{\mathrm{H}}^{+}]/[\text{Ni}{(\text{dppe})}_2{\mathrm{H}}^{2+}]\hfill \end{array}$$ (15)

The marked drop in pK_a upon oxidation of [Ni(dppe)₂H]⁺ is a typical consequence of redox on a mononuclear metal hydride, with this acidity increase often leading to hydride-related redox events being irreversible. One reason the [FeFe]- and [NiFe]-H₂ases employ bimetallic active sites may be to disperse the strong effects of redox over two metal sites.⁶³ This dampens the impact felt by the H⁻ ligand, enhances the reversibility of redox couples, and perhaps lowers activation barriers to catalysis. The [Fe]-H₂ase active site, which is a Lewis acid not required to perform redox, features only a single Fe center, with Nature apparently dispensing with the need for a second metal. A detailed understanding of redox and acid–base properties of H₂ases and functional models requires a number of experimental techniques. Some key methods are discussed in section 2.3, particularly in the context of characterizing hydride-containing species.

2.3. Physical Characterization of Hydrogenase Enzymes and Models

2.3.1. Electrochemistry

In addition to the two hydrogenic substrates H⁺ and H₂, the [FeFe]- and [NiFe]-H₂ases also handle electrons to drive eq 1 in a forward or reverse sense. According to the Nernst equation, at pH 7, p(H₂) = 1 bar, and T = 298 K, the 2H⁺/H₂ couple is at −414 mV vs NHE. Yet hydrogenases only have access to H₂ partial pressures of 1–10 Pa (10⁻⁵–10⁻⁴ bar), such that they must operate between −266 and −296 mV, the narrow range mandated by thermodynamics.⁶ In vivo, electrons travel via Fe–S clusters positioned close enough together (<14 Å is optimal)⁶⁴ to allow tunneling to and from an electron transport protein as the source/sink. In place of the transport protein, in vitro studies of H₂ases instead often employ suitable small molecule electron donors (e.g., methylviologen radical, E_1/2 = −446 mV vs NHE) and acceptors (e.g., benzylviologen, E_1/2 = −358 mV), whose potentials lie on either side of the 2H⁺/H₂ couple at p(H₂) = 1 bar.⁶⁵ Such redox reagents are also used when studying model compounds, which, with notable exceptions,^66,67 almost always lack redox-active Fe–S clusters or other electron shuttling moieties. Organic solvents may be necessary, as many models feature lipophilic ligands (e.g., phosphines) to stabilize their low-valent states. These studies have benefited from a growing body of work on electrochemistry in polar aprotic solvents,⁶⁸ and although pK_a scales are still not uniform, certain reference couples have become standard.⁶⁹ While redox potentials of the enzymes in aqueous solutions are typically reported relative to NHE, potentials of model complexes are most often reported against Fc^+/0 (Fc = FeCp₂; Cp⁻ = cyclopentadienide), a couple which lies at 0.4, 0.63, and 0.69 V relative to NHE in H₂O,⁶⁸ MeCN,⁶⁹ and CH₂Cl₂, respectively.⁶⁸ In terms of electrolyte salts, model studies make extensive use of noncoordinating anions,⁷⁰ with the borates BF₄⁻, B(C₆F₅)₄⁻, and BAr^F₄⁻(3,5-(CF₃)₂C₆H₃)₄B⁻),⁷¹ as well as PF₆⁻, being typical counteranions for both voltammetry⁷² and synthesis. Such anions possess diffuse charge and allow for coordinatively unsaturated cations to be studied “in isolation” and to be given every possible chance to interact with weak ligands like H₂.

An important alternative to sacrificial electron donors and acceptors is the use of an electrode as the electron source/sink, with voltammetric methods⁷³ (as well as associated spectroelectrochemical techniques)⁷⁴ proving extremely useful for characterizing analytes when immobilized or in solution,⁷⁵ and for determining catalytic rates^76,77 and overpotentials (vide infra).^49,78,79 Cyclic voltammetry (CV) has become popular in the study of synthetic hydrides, and many hydride-bearing mimics of [FeFe]- and [NiFe]-H₂ase exhibit electrochemistry that has informed their HER mechanisms, including uncovering the existence of paramagnetic (S = 1/2) hydrides. In contrast to conventional metal hydride chemistry, focused for example on hydrogenation reactions, H₂ases characteristically operate by one-electron (1e⁻) pathways. As described for [Ni-(dppe)₂H]^2+/+, redox imparts large changes in the acidity of hydride species, leading to the reductions and oxidations often being rather irreversible. This reflects the high reactivity of certain hydrides and may hint at desirable catalytic properties.

Redox waves of hydrides are often perturbed when H₂ or H⁺ substrate is present. Consider the latter case, for which CV is a typical assay for the hydrogen evolution reaction (HER, the reverse of eq 1) catalytic activity. Assume, for instance, a metal hydride gives rise to a 1e⁻ reduction mechanistically relevant to the HER. The potential of this couple in the absence of H⁺, denoted here E_redox, will certainly depend on the relative electron density at the metal(s) and thus the donor set. When an acid substrate HA is titrated into the hydride/electrolyte mixture, the growth of a wave implicates a situation in which the hydride complex is continually regenerated, and thus participates in a catalytic reaction involving protons. The resultant “catalytic wave” is often described by its potentials at full height (E_cat) or half-height (E_cat/2),⁸⁰ values that may differ from E_redox. The current associated with the catalytic wave, i_c, can be several times i_p, the current at E_redox in the absence of HA. The value i_c will be clear for an ideal (sigmoidal) wave, such that the potentials E_cat and E_cat/2 are precisely determined.⁷⁹ More complicated examples can introduce variance in i_c and, correspondingly, the potentials E_cat and E_cat/2. The latter value is least affected by nonidealities, and is thus the more reliable. If the quotient i_c/i_p is proportional to [HA], then the process is second order in [H⁺], consistent with the HER. At a certain point, one reaches the so-called “acid-independent regime”, in which the addition of HA no longer leads to an increase in i_c, and pseudo-first-order conditions with respect to the catalyst are reached. The maximum obtainable i_c/i_p is proportional to the square root of the turnover frequency (TOF) for H₂ evolution. The latter is an important metric, yet some have questioned its general applicability and instead advocate use of other values including catalytic efficiency (CE = i_c[catalyst]/i_p[HA]).⁸¹

Catalytic currents and/or TOF values are key descriptors of reactivity, yet these kinetic parameters should not be quoted without their associated potentials, on which they are dependent. The potentials E_cat and E_cat/2 ideally will be more positive than the HA reduction in the absence of hydride catalyst, a wave associated with direct reaction at the electrode (e.g., glassy carbon).⁸² Such a process may compete with catalyst-mediated HER at strongly reducing potentials, and one must be aware that currents arising from direct reductions have often been erroneously ascribed to catalysis from model complexes. In any case, E_cat/2 will be more negative than E(H₂/HA), the thermodynamic potential for the H₂/HA couple, although E_cat/2 for an ideal catalyst will be very close to this upper bound. When comparing catalysts, the values E_cat/2, E_cat, and E_redox are to be used in preference to the oft-quoted but rather subjective term “onset potential” (E_onset), the voltage near the base of the catalytic wave at the “onset” of catalysis.^49,73

An important electrochemical figure of merit is overpotential (η = E(H₂/HA) – E), defined as the additional voltage, beyond that required thermodynamically, used to drive a reaction at a specific rate.⁸³ The reaction is driven at a potential E, which may be taken as E_cat/2, E_cat, or another potential. Whatever potential (and thus overpotential) one selects, it is essential that this thermodynamic value is reported if and only if a kinetic value derived from the current is reported.^80,84 Given that η is proportional to the excess electrical energy required, its minimization, while maintaining an acceptable rate of reaction, is highly desirable. Although electron-rich metal catalysts give rise to more negative E_cat/2 values than their electron-poor counterparts, the former can still have reasonably small overpotentials if they can catalyze H₂ evolution from weaker acids, which have more negative thermodynamic potentials E(H₂/HA). In general, one aims to find the weakest acid that will quantitatively protonate the reduced form of the catalyst, whence η and TOF will be optimized. In addition to these catalytic data, stoichiometric conversions can allow one to develop a picture of the HER mechanism(s) at play. Two photons and two electrons must be transferred during the catalytic cycle; these reactions are often denoted as C (chemical) and E (electrochemical) steps. As will become clear, the sequence of these steps (e.g., ECEC vs CEEC) varies from catalyst to catalyst.

Cyclic voltammograms associated with catalytic reactions may be complicated by certain processes including acid homoconjugation, whence H₂ evolves not only following reduction of HA, but also from other sources such as the homoconjugated species H⁺[AHA]⁻.⁴⁹ Digital simulation of voltammograms and supplementary data from chronoamperometry are often beneficial in elucidating mechanisms.⁸⁵ Overall, information from CV contributes in no small part to a detailed picture of processes involved in catalytic H₂ evolution or oxidation. Some case studies involving CV in catalysis are presented in the [FeFe]- and [NiFe]-H₂ase sections 3 and 4. Readers unfamiliar with electrocatalysis and its figures of merit are referred to some useful introductions.^73,80,86,87

2.3.2. Crystallography

Possessing only a single electron, H atoms are often challenging to locate by X-ray crystallography, as any bond they form automatically means a significant amount of electron density is in this bond rather than about the H nucleus. Furthermore, metal hydrides can be difficult to characterize in that Fourier truncation ripples surrounding heavy atoms can overwhelm smaller electron density peaks. These two effects result in H atoms being difficult to locate and their M–H bond lengths typically being underestimated.⁸⁸ Despite these problems, laboratory X-ray analysis of high-quality small molecule metal hydride crystals routinely allows identification and refinement of H⁻ ligand positions. In contrast, synchrotron X-ray diffraction of protein crystals can only afford H atom positions under very favorable circumstances. The sample must diffract to <1 Å (high resolution by protein standards), and only then can careful modeling and refinement reveal hydrides and other H nuclei.^89,90 The resulting crystallographic models are best compared to those from neutron diffraction,^91,92 a technique that exploits the larger scattering cross section of D vs H, and is suitable to the study of macromolecules.⁹³ For example, one might characterize a H₂ase or its hydride-bearing model by X-ray diffraction and compare the metrics to neutron data of the relevant small molecule deuteride (crystals of suitable size for neutron diffraction are realistically more accessible for small molecules than for H₂ases). Such deuterides are typically simple to prepare owing to the H/D exchange properties of many metal hydrides.

2.3.3. Magnetic Resonance Spectroscopy

When considering metal hydrides in particular, certain spectroscopic methods are more informative than others. Among the most useful is nuclear magnetic resonance (NMR) spectroscopy, with spin–orbit coupling causing even diamagnetic Fe–¹H moieties to resonate over a wide range, although typically in the region upfield of Si(CH₃)₄ in which other ligand signals are rarely observed. The chemical shift, as will become clear, is not only dependent on spin–orbit interactions⁹⁴ and the electron-rich or -poor character of the metal(s) to which H⁻ is bound, but also on the terminal or bridging nature of its coordination. NMR has not been successfully applied to the [FeFe]- and [NiFe]-H₂ase enzymes themselves, even in cases where H⁻ is ligated at an S = 0 core. In cases where H⁻ is bound to open-shell metal centers (S > 0), electron paramagnetic resonance (EPR) spectroscopy, as well as its pulsed variants electron nuclear double resonance (ENDOR) and electron spin echo envelope modulation (ESEEM),⁹⁵ have proven extremely powerful. Indeed, the first direct evidence for a metal hydride in Nature was obtained by ENDOR spectroscopic measurement of a [NiFe]-H₂ase (vide infra).⁹⁶ Overall, the latter two techniques are particularly useful for resolution of anisotropic coupling between unpaired spins and ¹H nuclei diagnostic of H⁻ ligation.

2.3.4. Vibrational Spectroscopy

Vibrational spectroscopy provided early breakthroughs in H₂ase structure elucidation, particularly the identification of the CO and CN⁻ cofactors present at the active sites. These cofactors represent key spectroscopic handles because ν_CO and ν_CN bands appear in spectral regions devoid of protein bands. The number and intensities of ν_CO and ν_CN bands not only report on the number of such moieties but also the C–M–C angles between them. In turn, the vibrational frequencies report on metal electron density and the extent to which π-backbonding occurs. But what of M–H and M–H–M chromophores—how easily detected and assigned are their vibrations? Although their intensities are often rather low, ν_MH modes can be identified using infrared (IR) and Raman spectroscopies,⁹⁷ in the latter case taking due consideration of the photolability of many CO- and H⁻-containing complexes. Terminal hydride ν_MH bands often appear in IR spectral regions populated by ν_CO stretches (particularly those of bridging CO ligands).⁹⁸ Bridging hydrides give rise to weaker, lower energy vibrations best observed with resonance Raman, although vibrational mixing with other bands can convolute matters. The challenges discussed here can partly be overcome by studying M–D and M–¹³CO isotopologues, in which case the bands are shifted to lower energy (ν_MH ≈ 2^1/2ν_MD and ν_12CO ≈ (91/87)^1/2ν_13CO, assuming uncoupled harmonic vibrations and m_M ≫ m_D).⁹⁷ Nevertheless, unambiguous assignment of hydride bands is not always possible, and such data are unknown for many hydrides.

The difficulties in identifying metal hydride bands are well met by using a technique known as either nuclear resonance vibrational spectroscopy (NRVS) or nuclear inelastic scattering (NIS),^99–101 which now represents a valuable complement to the two more established vibrational methods. While Raman provides vibrational information following electronic excitation, NRVS does the same with nuclear excitation, requiring inelastic scattering of highly monochromatic (meV-resolved) synchrotron γ- or X-rays¹⁰² by Mössbauer-active nuclides (e.g., ⁵⁷Fe, ⁶¹Ni). The resulting spectra feature bands corresponding to only (and all) vibrations in which the metal nuclide moves. The unparalleled selectivity for probing the inner coordination sphere (even when surrounded by many kilodaltons of proteic mass) has already proven invaluable, as examples in this review will highlight. The intensities of NRVS bands are related to the displacement of the Mössbauer-active nuclide along the beam path. Many NRVS studies focus on low energy vibrations, such as bending and wagging modes, as they often involve large amplitude motion of metals and thus have relatively large NRVS intensities. The disadvantage in using these modes is that their coupled nature necessitates detailed density functional theory (DFT, vide infra) analyses for confident assignments. On the other hand, high energy vibrations, such as M–H/D stretches, see the metal move only very slightly, as the mass of M is large relative to that of H or D. Pure ν_MH modes are thus difficult—although not impossible—to observe.¹⁰³

The inherently high frequency and low intensity of M–H/D modes leads to such NRVS measurements being far from routine, but this will inevitably change given the ever-increasing flux and resolution of synchrotron light sources. At the time of this review, NRVS-equipped facilities include beamlines at the Advanced Photon Source (Chicago, IL, USA), European Synchrotron Radiation Facility (Grenoble, France), and SPring-8 (Hyogo, Japan). Alternatively, seeded free-electron lasers can also provide the necessary monochromatic radiation, although such facilities are currently of limited accessibility and their applications to NRVS are yet to be well developed.¹⁰⁴

When a complex exists for only a short time, its characterization calls for stopped-flow techniques and transient (time-resolved) spectroscopies. For example, the pulsed nature of free-electron lasers, while resulting in lower average fluxes, would enable NRVS measurements on the femtosecond scale. The most widespread transient methods use more traditional laser sources to obtain vibrational¹⁰⁵ and optical spectra.¹⁰⁶ These have been applied to H⁺ transfer dynamics in H₂ases¹⁰⁷ as well as the detection of synthetic mixed-valent hydride species,¹⁰⁸ and are well complemented by computational investigations into such fleeting species. Described in section 2.4 are some of the theoretical approaches used to understand H₂ases and their models, as well as the information that is afforded by such work.

2.4. Theoretical Characterization of Hydrogenase Enzymes and Models

2.4.1. Approaches

The physical characterization of synthetic H₂ase models is typically much simpler than analogous studies on the proteins themselves. The same can be said for computational investigations, and thus while many synthetic species are amenable to in silico characterization, the enzyme structures are typically truncated to contain only the active site and perhaps the second coordination sphere for fully quantum mechanical calculations. Alternatively, mixed quantum mechanical/molecular mechanical (QM/MM) methods, in which only the active site is treated quantum mechanically while the remainder of the enzyme and solvent are treated with a MM force field,^109–113 have been used to study the larger H₂ase enzyme systems.^114–116 The computationally expensive nature of metallic cores, as well as the intractability associated with high levels of theory such as coupled cluster methods, have led to DFT^117,118 being widely used in the study of H₂ases and synthetic models.^119–121 As used with other molecular transition metal electrocatalysts,^122–124 popular functionals include B3LYP,^125,126 BP86,^127,128 B3P86,^126,127 PBE0,^129–131 TPSSh,¹³² M06-L,¹³³ and ωB97XD.¹³⁴ These functionals are typically used with double- or triple-ζ basis sets, as well as polarization for specific atoms including moving H⁺ or H⁻ groups,¹³⁵ and diffuse basis functions for anionic systems.¹³⁶ Additionally, effective core potentials and their corresponding basis sets, such as those of Hay and Wadt (LANL)¹³⁷ or Stuttgart–Dresden (SDD),¹³⁸ may be used for the transition metals to decrease computational cost.¹³⁹ The application of DFT to such complexes has been reviewed in detail.¹⁴⁰

2.4.2. Predicting Observables

Molecular geometries of H₂ase models optimized using the above functionals have been shown to agree quite well with X-ray crystal structures. Geometries can be optimized either in the gas phase, subsequently including solvation free energies using a Born–Haber thermodynamic cycle, or in the solution phase directly.¹⁴¹ Solvation effects are typically incorporated with a polarizable continuum model, in which the solvent is represented as a homogeneous dielectric. These models, such as PCM,^142–144 C-PCM,¹⁴⁵ SMD,¹⁴⁶ and COSMO,¹⁴⁷ are advantageous for treating solute–solvent interactions without requiring conformational sampling, although they do so by inherently neglecting specific solute–solvent interactions such as hydrogen bonding.

With optimized structures in hand, DFT calculations are invaluable for the prediction of further observables such as redox potentials and pK_a values. Accurate predictions of these parameters are necessary given their importance in the H⁺ reduction reaction. The accuracy of calculated reduction potentials and pK_a values has been assessed for many molecular electrocatalysts and provides validation for the computational methods employed.^57,141 Several methods can be implemented to account for the reference electrode in the calculations to allow comparison with experimental measurements. One option is to subtract an experimental or theoretical value for the absolute reduction potential of the reference electrode.^148–152 An alternative is to subtract the value of the reduction potential of the reference electrode calculated using the same level of theory and basis set.^153–156 Yet another option is to calculate the reduction potential of a related half-cell reaction, which has been experimentally measured under similar experimental conditions and with respect to the same reference electrode.^{52,57,141,157,158} Application of a thermodynamic cycle, also called an isodesmic reaction, results in the reference electrode potentials canceling, as has been demonstrated for the calculation of quantitatively accurate redox potentials for synthetic H₂ase models.^159–161 This approach may also be applied to calculate pK_a values, although experimental pK_a measurements are not performed as often. One particular advantage of this approach is the cancellation of systematic errors associated with the level of theory and specified functional and basis set. Furthermore, application of a thermodynamic cycle also avoids the calculation of the free energies of H⁺ and e⁻, as well as the potential of the reference electrode. The application of this thermodynamic cycle is equivalent to the calculation of reduction potentials and pK_a’s relative to related systems with experimentally measured values. Such an approach appears to be the most reliable at the current time.

Along with redox and acid–base behavior, spectroscopic observables are also available through DFT calculations. In particular, the diagnostic IR signatures of the CN⁻ and CO ligands are readily observed and can also be predicted by DFT calculations. However, since the harmonic approximation is often invoked to compute vibrational frequencies, calculated absolute frequencies are generally shifted systematically. For this reason, empirical scaling factors, dependent on the functional and basis set used, are often employed.^162,163 While the absolute vibrational frequencies calculated with DFT are often not quantitatively accurate, shifts in frequencies due to protonations and/or reductions are generally reliable. Nevertheless, absolute vibrational frequencies calculated using the BP86 functional tend to agree remarkably well with the experimental data without the use of scaling factors, which has been attributed to a fortuitous cancellation of errors.¹⁶⁴

Although EPR spectroscopy is widely used to characterize paramagnetic states of the enzyme and models,⁹⁵ the accurate calculation of magnetic spectroscopic parameters is typically a challenging computational task. When calculated within the framework of DFT, the accuracy of parameters such as g-values, hyperfine couplings, zero-field splittings, and quadrupole couplings are very sensitive to the particular system being studied, especially the metal center(s). For example, while DFT calculated values are in good agreement with measurements for organic radicals and biradicals,^165–167 the calculated values for transition metal complexes typically deviate from those observed. In particular, g-shifts are notably underestimated by many functionals for transition metal complexes.^168,169 Hyperfine couplings for transition metal complexes present an additional challenge due to the significant contributions from spin–orbit coupling and spin polarization,¹⁷⁰ with the latter commonly underestimated by DFT.^171,172 However, predicted magnetic spectroscopic parameters are not always unreliable,¹⁷³ with calculated Mössbauer isomer shifts for ⁵⁷Fe being especially promising.^174–176 It has also been suggested that hybrid functionals, which contain a portion of Hartree–Fock exchange, tend to provide better agreement with experiment,^169,177 although at the risk of spin contamination.¹⁷⁸ Nonetheless, the application of these approaches to H₂ase models has seen successes over the years.^179–185

The value of theoretical calculations is not limited to rationalizing experimental observations. The rapid nature of many H⁺ and e⁻ transfers ensures that many species may evade conclusive identification even by transient spectroscopies. Indeed, calculations come to the fore when practical experimental solutions are not to be found. Computational studies are not only useful for the characterization of known H₂ase models, but also have guided the design of new synthetic targets.^186–189 Experimental and theoretical studies in the field are thus driven by one another to a deeper understanding of H₂ases and their models. Our general discussion on characterization ends here, and we now present the state-of-the-art in the characterization and synthetic modeling of the [FeFe]-H₂ases (section 3), [NiFe]-H₂ases (section 4), and [Fe]-H₂ases (section 5).

3. [FeFe]-H₂ASES

3.1. Enzyme Structure and Function

Expressed in bacteria and lower eukaryotes,¹⁹⁰ the [FeFe]-H₂ases are arguably the fastest and most evolved of the H₂ases. While the [FeFe]-H₂ases vary in size (45–130 kDa), each features an active site ensemble known as the H-cluster. With six Fe centers in total, the H-cluster comprises a [4Fe–4S] metallocubane linked to an apical Cys-S⁻ ligand in a [2Fe] moiety of formula {(Cys-S)(NC)(OC)Fe[(SCH₂)₂NH]Fe-(CO)₂(CN)}^2−/3− (Figure 7).

Figure 7.

X-ray structure of the [FeFe]-H₂ase active site from Clostridium pasteurianum (PDB code 3C8Y).¹⁹¹ H atoms are omitted for clarity.

Perhaps the more striking feature of the H-cluster structure is the [4Fe–4S] cluster, a common redox cofactor positioned in an uncommon location—covalently bound at the active site—that emphasizes the importance of e⁻ transfer in catalysis. The Fe centers in the [2Fe] unit are labeled Fe^p (“proximal” to 4Fe–4S) and Fe^d (“distal” to 4Fe–4S), with the latter featuring a vacant site for substrate binding. The low-spin Fe^p and Fe^d centers are linked through S atoms of the azadithiolate (adt²⁻ = ⁻SCH₂N-(H)CH₂S⁻) cofactor. While the 4Fe–4S unit and three or four “auxiliary” Fe–S clusters define an e⁻ transport chain,¹⁹² azadithiolate plays a key role in shuttling H⁺ to or from the active site. Poised over the apical site of Fe^d, the secondary amine influences interconversion of Fe, Fe–H, and Fe(η²-H₂) species by virtue of its acid–base properties. Beyond this, the amine further participates in N⋯H–S-Cys hydrogen-bonding interactions, beginning a pathway for H⁺ transfer between the active site and the protein surface. The arrangement of the amine and Fe^d is constrained such that no bond can form between the two, with this Lewis base and acid constituting a FLP.³⁹ However, this system is unlike most synthetic FLPs in that the [2Fe] site is also redox-active, and is thus well-suited to perform electrocatalysis.

Completing the H-cluster are five “organometallic ligands” (3 × CO, 2 × CN⁻), whose presence at the active site initially came as a surprise, although foreshadowed by the crystallographic analysis of [NiFe]-H₂ases. These chromophores absorb strongly in an uncluttered IR region,^191,193 serving as crucial reporters on the purity of protein samples, which is greatly decreased by any exposure of [FeFe]-H₂ases to O₂. The diatomic ligands are key spectroscopic handles, and while ν_CN modes are less sensitive to H-cluster redox (CN⁻ is a weak π-acceptor), the ν_CO energies can vary over a 200 cm⁻¹ range.⁷ Consequently, IR analyses of [FeFe]-H₂ases are particularly informative, although interpretation can be complicated by the highly coupled nature of these modes.

After accounting for all the ligands, one can now consider the geometry of the [2Fe] unit. If the μ-CO ligand were trans to its current location, at a terminal Fe^d site, one might consider the [2Fe] unit as a pseudosymmetric union of two pyramidal Fe units, much like the vast majority of synthetic [L₃Fe(SR)₂FeL₃] complexes (see section 3.2.1). But the presence of the CO bridge stabilizes an unsymmetrical “rotated” structure, wherein a coordination site on the Fe^d trans to this π-acid is free for substrate binding. The “rotated” structure is further rigidified by a strong Fe^dCN⋯⁺H₃N-Lys358 interaction involving a conserved residue.¹⁹⁴ In addition to the structural role played by the CN⁻ ligands, cyanides also enhance the basicity and lower the redox potentials of the [2Fe] site and, along with the CO ligands, enforce low-spin configurations ideal for H₂ binding.³⁵

On the other hand, have all the ligands been accounted for? Structural analyses suffer from a major complication: protein X-ray crystallography is typically unable to resolve the presence—much less the location—of hydrogenic substrates. Consequently, the protonation state of the amine is not clarified by crystallography. The presence of substrate at the Fe^d binding site is also an open question, and even if an Fe^d–H terminal hydride were a long-lived intermediate, the ligand would typically not be crystallographically locatable. Furthermore, Fe–H and Fe(η²-H₂) moieties are very weak IR and Raman chromophores (see section 2.3.4), and are usually EPR-inactive, such that their identification based on hyperfine interactions is unlikely. Despite the lack of direct evidence for hydride ligation,¹⁹⁵ Fe^d–H intermediates are assumed to exist at least transiently in the catalytic cycle.^24,196 It is altogether conceivable that hydrides might not exist as stabile entities on a nearly flat potential energy landscape associated with an efficient catalytic cycle. While often slow,¹⁹⁷ the protonation and deprotonation of metal centers might be accelerated by proton-coupled electron transfer (PCET), which would exploit the juxtaposition of [4Fe–4S] and adt²⁻ cofactors to keep Fe^d hydride-free. Information on Fe–H species, albeit indirect, has been obtained by studying a mutated [FeFe]-H₂ase from Chlamydomonas reinhardtii. Such work builds on the hypothesis that Fe–H intermediates are destabilized by the PCET machinery. Indeed, introduction of Ser in place of Cys in the H⁺ transfer pathway results in a significant drop in activity, and IR and EPR analyses of the reduced mutant indicated a new form with a H-cluster best described as [4Fe–4S]⁺/Fe(II)Fe(II). Such an assignment is consistent with an Fe–H species, and the presence of the hydride is further indicated by the shift in the bridging CO band upon changing the solvent from H₂O to D₂O.¹⁹⁸

Taking in the active site as a whole, the H-cluster is conformationally rigid, and its overall structure persists throughout the catalytic cycle. Despite this, there is some flexibility within certain components. First, substrate turnover necessitates interaction at the apical site on Fe^d, a metal that must switch between octahedral and square-pyramidal geometries. Second, to function as a H⁺ relay, the secondary amine in the adt²⁻ cofactor is required to undergo rapid inversion (this aspect is illustrated in Appendix A.1, Figure 58). Lastly, the CO bridging Fe^p and Fe^d serves as a “shock absorber”, with its linearity dependent on redox and the chemistry occurring at Fe^d. This reactive site is the location for H⁻/H₂ binding, and its coordination to exogenous CO leads to deactivation of the enzyme. In contrast, H₂O does not poison the active site, consistent with a low-spin π-donor description for Fe^d.

Figure 58.

Proton transport to and from the [2Fe] site in [FeFe]-H₂ase (formal charges omitted). This involves H⁺ movement, denoted by red arrows, between Fe^d, the adt²⁻ cofactor, and the protein (via Cys299), and likely requires pyramidal inversion of the amine.

Particularly important structure–function relationships have been elucidated by reconstitution of the semi-apoenzyme derived from algal H₂ases. The semi-apoenzyme, which contains the Fe₄S₄ cofactor but not the [2Fe] subunit, is readily reconstituted to a fully active enzyme upon treatment with the synthetic diiron complex [NC(OC)₂Fe(adt)Fe(CO)₂CN]²⁻([3]²⁻, Figure 8). The reconstitution is accompanied by loss of CO, attachment of the [4Fe–4S] cluster, repositioning of CN⁻ ligands to transoid, basal positions, and adoption of a “rotated” structure.^199,200

Figure 8.

Reconstitution of apo-[FeFe]-H₂ase from Chlamydomonas reinhardtii with [NC(CO)₂Fe(adt)Fe(CO)₂CN]²⁻.^199,200

The artificial maturations have also been performed using isostructurual, although subtly different, diiron dithiolates including {Fe₂[(SCH₂)₂–X](CN)_x(CO)_6−x}^−/2−, where X = CH₂, O, S, NMe and x = 1, 2.^201,202 While these abiological analogues are readily accepted by the protein, catalytic activity assays indicated that the adt²⁻ cofactor is crucial for H⁺ transfer, and the only semisynthetic [FeFe]-H₂ase that exhibits significant activity is that obtained by reconstitution with [(OC)₃Fe(adt)-Fe(CO)₂CN]⁻. With the functional importance of each H-cluster component now clear, our discussion moves to the various enzyme states and their roles in the catalytic cycle.

The understanding of enzyme mechanisms rests on the identification of characterized states,²⁰³ which number three in the case of [FeFe]-H₂ase. Several crystallographic and spectroscopic studies confirm that the structure of the H-cluster is virtually invariant in all redox forms. In considering the proposed catalytic cycle (Figure 9), focus is consequently placed on the oxidation and protonation states of the Fe centers and adt²⁻ cofactor. Spectroscopic parameters for the three known [FeFe]-H₂ase states are presented in Table 1, and the properties of each are now individually discussed.

Figure 9.

Catalytic cycle proposed for [FeFe]-H₂ase (right)²⁰³ and electron counting for the [2Fe] unit in the three characterized states (left). The totals include Fe 3d electrons and bonding electron pairs from donor atoms. The substrate H atoms are in red for emphasis, although they cannot be distinguished from the amine H atoms.

Table 1.

Oxidation States and Spectroscopic Parameters for [FeFe]-H₂ase HydA1 from Chlamydomonas reinhardtii^a

state	[Fe4S4]n+	[2Fe]	νCO/cm−1	νCN/cm−1	g
Hox	[Fe4S4]2+	Fep(II)Fed(I) or Fep(I)Fed(II)	1964, 1940, 1800	2088, 2072	2.10, 2.04, 2.00
Hred	[Fe4S4]2+	Fep(I)Fed(I)	1935, 1891, 1793	2088, 2072	–
Hsred	[Fe4S4]+	Fep(I)Fed(I)	1954, 1919, 1882	2070, 2026	2.08, 1.94, 1.87

^aProteins from other strains afford similar data for H_ox and H_red.^7,202.

The most common state for biophysical studies is H_ox, so denoted as it is “oxidized”, although not oxygenated (an important distinction for such O₂-sensitive organometallic enzymes). Early Mössbauer studies on H_ox indicated strong coupling between the proximal [4Fe–4S]²⁺ cofactor and the [2Fe] center, with the latter initially described as a low-spin (S = 1/2) Fe(II)Fe(III) tandem.²⁰⁴ Yet, the inadequacy of Mössbauer spectroscopy in distinguishing oxidation states of low-spin Fe led some astray, particularly when low oxidation states such as Fe(I) (“subferrous”) were not considered to be biologically relevant.

A suite of advanced EPR techniques has greatly informed the current picture of H_ox. While its EPR signals at g = 2.10, 2.04, and 2.00 shift slightly depending on the organism, these invariably are consistent with an Fe(II)Fe(I) description for the [2Fe] core. Although biophysical reports indicate Fe^p to be monovalent, analyses of synthetic models consistently point to the reverse assignment, supported on general grounds by considering the favorability of a +I oxidation state for pentacoordinate sites such as Fe^d in H_ox. In any case, the electronic coupling of the [4Fe–4S]²⁺ cofactor and [2Fe] core allows mixing of excited states and enhanced hyperfine coupling (isotropic A(⁵⁷Fe) ≈11–12 MHz for each moiety). The value of paramagnetic centers as spectroscopic handles also comes to the fore in the identification of ¹⁴N hyperfine, which represented the first evidence for the identity of the adt²⁻ cofactor.²⁰⁵

Poised to bind H₂, the H_ox state also strongly binds CO at the apical Fe^d site adjacent to the secondary amine.²⁰⁶ The binding of ¹³CO is stereospecific,²⁰⁷ which adds weight to our picture of the relatively rigid active site. When exogenous CO is taken up by H_ox, its spin becomes more delocalized across the H-cluster, although EPR analysis indicates that the cubane remains in a [4Fe–4S]²⁺ state. An Fe(1.5)Fe(1.5) description for [2Fe] is thus logical, given that both Fe^p and Fe^d are now octahedrally coordinated with similar ligand sets. The deactivated, CO-bound state features strong coupling between the [4Fe–4S] and [2Fe] units, which is significant mechanistically in that binding of H₂ (and especially of H⁻) might also enhance interaction between these two subsites and thus be involved in PCET.²⁰⁸

When [FeFe]-H₂ase exists at potentials more negative than −395 mV, it is not H_ox, but rather its product H_red, that predominates. The latter state is EPR-silent and is thought to feature a [4Fe–4S]²⁺ cluster (also EPR-silent) and an S = 0 [2Fe] site. It was formerly assumed that H_red features a H⁻ ligand as part of an Fe(II)Fe(II), H⁻ core. However, X-ray spectroscopy does not support the presence of an Fe–H moiety in H_red, which is instead considered to be an isoelectronic Fe(I)Fe(I) form. This description is consistent with IR data indicating that the H_red core is more electron-rich than that of H_ox, with the terminal ν_CO bands undergoing, on average, a bathochromic shift of 25 cm⁻¹ on forming H_red. The two states also differ in the geometry of the semibridging CO ligand. In H_ox, the CO bridges to Fe^p, with the Fe^d–C–O unit being highly bent,²⁰⁹ perhaps in order to enable some spin delocalization. In the diamagnetic H_red state, the Fe^d–C–O is almost linear.

A third active state has been uncovered through studies on an [FeFe]-H₂ase that, unlike most such proteins, lacks Fe–S clusters outside of the H-cluster. This recently identified H_sred (“super reduced”) form in Chlamydomonas reinhardtii is observed at or below −540 mV, about 150 mV more negative than the H_ox/red couple. Both the [4Fe–4S]⁺ cubane and the Fe(I)Fe(I) [2Fe] site adopt their lowest oxidation states, a situation that would not be long-lived were auxiliary Fe–S clusters present to accept electrons. Nevertheless, H_sred is assumed to be a transient intermediate in all [FeFe]-H₂ases. Its IR spectrum features ν_CO bands more similar to H_ox than H_red, but one ν_CN band shifts by 45 cm⁻¹ to lower energy, perhaps indicating a structural change as such stretches are insensitive to metal electron density.²¹⁰ Described now are current proposals regarding the enzyme catalytic cycle and mechanism, subjects that are still topics of debate.

In considering the stoichiometry of the catalytic cycle, one notes that the two redox equivalents demanded by the H₂/2H⁺ couple cannot be provided by the H_ox/red couple alone, a problem conveniently addressed by invoking H_sred, which is 2e⁻ more reduced than H_ox (Figure 9). Consistent with present biophysical information, it is suggested that oxidation of H₂ involves its binding at the Fe^d site in H_ox to give H_ox (H₂).²⁰³ The Fe(η²-H₂) fragment in the latter species is deprotonated by the amine of the adt²⁻ cofactor, resulting in the ammonium hydride heterolysis product H_ox(H⁺, H⁻). A PCET step is now proposed, in which deprotonation is associated with e⁻ transfer from [2Fe] to the [4Fe–4S]²⁺ cluster. The resulting [4Fe–4S]⁺,Fe(II)Fe(II),H⁻ species may tautomerize to H_sred, whose rapid oxidation (e⁻ transfer) affords H_red and then H_ox, thereby completing the cycle. The reverse processes apply to H⁺ reduction, which involves H⁺ transfer from adt²⁻ to a reduced Fe^d center in concert with e⁻ transfer from [4Fe–4S]⁺ to give the same intermediate invoked for H₂ oxidation.

The catalytic mechanism can be further appreciated from the perspective of e⁻ counting. Summing 3d electrons and electron pairs for each Fe–ligand bond in the [2Fe] subsite, one notes that H_ox is a 33e⁻ dimer, and H_red is a 34e⁻ dimer (Figure 9), with 34e⁻ being consistent with a bimetallic complex featuring a metal–metal bond. In order to maintain this electron count, binding of exogenous CO to Fe^d in H_red is accompanied by dissociation of the Cys-S–Fe^p bond.²¹¹ If the [2Fe] site carried a H⁻ ligand, as seems likely in at least a transient form, the 18e⁻ rule is still obeyed, since the H⁻ complex would remain 34e⁻ regardless of its description as Fe(II)Fe(II),H⁻ or Fe(I)Fe(I),H⁺. To reiterate, as the H_ox/H_red couple differs by only one electron, it is clear that [4Fe–4S] must supply the extra electron or hole required for the H₂ ⇌ 2H⁺ + 2e⁻ reaction.

While characterization data for each catalytic intermediate are not available, what we do know is that the [FeFe]-H₂ases are some of the fastest enzymes.¹⁴ Rates vary by 2 orders of magnitude for H₂ evolution but only by a factor of 2 or 3 for H₂ oxidation. Compared to the rates for H₂ evolution, H₂ oxidation is always faster, sometimes by as much as 3 orders of magnitude.²¹² Perhaps the best characterized in terms of structure and rates is the H₂ase I (note: organisms often have two or more different H₂ases) from Clostridium pasteurianum. In this case, the turnover frequencies (TOFs) at 30 °C for H₂ oxidation and evolution are 25 000 and 5700 s⁻¹, respectively.²¹²

The remarkably high catalytic rates of [FeFe]-H₂ases are competitive with those exhibited by Pt metal.²¹³ Yet, as stated in the Introduction, one motivation for the study of [FeFe]-H₂ases is the preparation of base metal catalysts with the hope that reproducing the native structure will afford the native function. While long-lived hydrides are unlikely to be involved in the [FeFe]-H₂ase mechanism, it is assumed that complexes of the form Fe(II/I)Fe(II),H⁻ do have a transient presence (Figure 9). Presented in section 3.2 are synthetic hydrides relevant to the FeFe active site. The high fidelity hydride models are typically terminal, but, for historical purposes, some bridging species will be described in discussion of early work. Moreover, catalytically important bridging hydrides will also be considered.

3.2. [FeFe]-H₂ase Synthetic Modeling

3.2.1. Early Diiron Dithiolates and Their Hydrido Complexes

The crystallographic analysis of [FeFe]-H₂ase attracted particular interest from organometallic chemists, for whom the active site immediately brought to mind “classic” diiron dithiolates that predated the determination of the enzyme structure by 70 years.²¹⁴ Renewed attention was paid to Fe(I)Fe(I) carbonyls such as [(OC)₃Fe(SEt)₂Fe(CO)₃]²³ and, later, [(OC)₃Fe(pdt)Fe(CO)₃] ([4], pdt²⁻ = 1,3-propanedithiolate; Figure 10),²¹⁵ owing to the presence of Fe, RS⁻, and CO in these complexes as well as the diiron (and, in fact, all) H₂ase active sites.

Figure 10.

Archetypal diiron thiolato hexacarbonyl [4] can be converted to its hydride [4(μ-H)]⁺ only with very strong acids.⁴⁴ An alternative structural representation of [4(μ-H)]⁺, using full arrow and half-arrow notation, is provided below.

There exists an uncanny and promising resemblance between [FeFe]-H₂ase active sites and the archetypal synthetic low-valent hexacarbonyls, with their 34e⁻, stable Fe–Fe bonded motif.²¹⁶ Yet, in reality the latter are too stable, to the point that their oxidations are strongly anodic (e.g., E_1/2([4]^+/0) = 0.65 V vs Fc^+/0)²¹⁷ and their Brønsted basicities are low. As discussed above, a requirement of the [FeFe]- and [NiFe]-H₂ases is their participation in both acid–base and redox at mild pH and potentials. Although [4] is inert to HBF₄·Et₂O, it is protonated by HOTf (trifluoromethanesulfonic acid) such that lower and upper bounds for the acidity of [4(μ-H)]⁺ in C₆D₅F are known: −9 <pK_a([4(μ-H)]⁺) < 0.⁴⁴ However, protonation with HOTf does not proceed smoothly, and isolation of an Fe(II)(μ-H)Fe(II) derivative of [4] was found to necessitate action of the superacid generated from [SiEt₃]B(C₆F₅)₄ and HCl.²¹⁸ In the preparation of [4(μ-H)]⁺ from [4], while the 2e⁻ Fe–Fe bond in the latter formally reduces H⁺ to afford a H⁻ ligand, the internuclear separation in the resulting hydride is still typically comparable to twice the covalent radius²¹⁹ of low-spin Fe (2 × 1.32 Å = 2.64 Å). Thus, although the 3d⁶ Fe(II) sites are not expected to interact strongly with each other, a dashed bond between metals is nevertheless an oft-adopted notation. While not used in this review, a more rigorous notation exists for the representation of Fe(μ-H)Fe systems and three-center two-electron bonds in general.²²⁰ This convention avoids depicting direct Fe–Fe bonds (Figure 10, bottom), and employs the typical arrows for L-type (charge-neutral, 2e⁻) donors, and a half-arrow for the bridging hydride. Such a method is useful when counting electrons, with the half-arrow making clear that Fe(μ-H)Fe units involve three-center two-electron rather than three-center four-electron bonds, as the simple line drawing might imply.

The acid–base and redox reactivity of the diiron(I) dithiolato hexacarbonyls is greatly enhanced upon replacement of π-acidic CO ligands with stronger donors—a strategy that has proven very generalizable. For example, the installation of one or two CN⁻ ligands occurs under mild conditions to afford, in the latter case, a complex bearing all the diatomic ligands present in the [FeFe]-H₂ase active site.²²¹ While acids only protonate the N atom in [(OC)₃Fe(pdt)Fe(CO)₂CN]⁻ (the Fe–Fe bond being insufficiently basic), the dicyanide [3]²⁻ protonates to give hydrides.^222–224 In situ analysis revealed an N-protonated intermediate prior to the formation of [NC(OC)₂Fe(pdt)(μ-H)Fe(CO)₂CN]⁻ ([3(μ-H)]⁻).

The asymmetric derivative [Me₃P(OC)₂Fe(pdt)Fe-(CO)₂CN]⁻([5]⁻) could be converted to its stable conjugate acid [Me₃P(OC)₂Fe(pdt)(μ-H)Fe(CO)₂CN] ([5(μ-H)], Figure 11),²³ which exhibits a ¹H NMR resonance at −17 ppm typical of a μ-H⁻ ligand. The charge-neutral hydride [5(μ-H)] undergoes electrochemical reduction at −1.57 V vs Fc^+/0, the pseudoreversibility of which indicates that a mixed-valent complex is accessible (vide infra). Further protonation of [5(μ-H)] affords [Me₃P(CO)₂Fe(pdt)(μ-H)Fe(CO)₂(CNH)]⁺ ([5(μ-H)H]⁺), whose irreversible reduction at E_pc = −1.47 V is accompanied by H₂ evolution. In what is a common indirect indicator of catalytic activity, the reductive current increases upon titration with strong acids, including HCl, H₂SO₄, and toluenesulfonic acid (HOTs), with a TOF of 0.0067 s⁻¹ at −1.2 V vs Ag/AgCl in the latter case.²²⁵ While modest, the activity of species such as [5(μ-H)] contrasts the inactivity of the related bis(phosphine) hydride [Me₃P(CO)₂Fe(SMe)₂(μ-H)Fe-(CO)₂PMe₃]⁺ (δ(¹H) −15.6 ppm),²²⁶ which lacks an effective H⁺ relay. This role is apparently well-served by CN⁻, despite its arrangement perhaps not being ideal for H⁺ transfer to the metal sites. Overall, these findings highlight an important design feature for functional [FeFe]-H₂ase (and [NiFe]-H₂ase) models: the basic/oxidizable metal(s) must be proximal to a basic moiety.

Figure 11.

Catalytic cycle for electrocatalytic proton reduction mediated by [5(μ-H)].²³

The ambidentate nature of CN⁻ was recognized as a complicating factor that saw it largely replaced in models by simpler ligands of (ideally) comparable σ-donicity. While donors such as carbenes,^227–231 isonitriles,^232–235 and even nitrosyls^236,237 have been used, tertiary phosphines have proven the most useful.²³

3.2.2. Mixed-Valent Hydrides

The 34e⁻ Fe(II)(μ-H)Fe-(II) species discussed above represent models for a putative [FeFe]-H₂ase form tautomeric to the H_sred state, an unobserved species which likely bears a terminal hydride (Figure 9, bottom right). The catalytic cycle for H₂ evolution necessitates the reduction of diferrous hydrides to mixed-valent derivatives, both for synthetic catalysts and in the native catalytic cycle (the state H_ox (H⁺, H⁻)). While there is reason to believe that synthetic versions of these hydrides may be unstable (e.g., toward bimolecular decomposition and H₂ evolution), the pseudoreversibility of Fe(II)(μ-H)Fe(II/I) couples in certain complexes suggested that a mixed-valent species might be persistent. Such odd-electron dinuclear species are often classified according to the system of Robin and Day.^238,239 A Robin–Day class I complex features structural asymmetry associated with localization of the singly occupied orbital on one site. A class III complex features structurally indistinguishable metal sites, each with a “genuinely nonintegral valence”.²⁴⁰ Between these two extremes lie the class II complexes, whose metal sites are distinguishable, but not very different.

At this point it is important to acknowledge a large, but quite distinct body of work concerning hydrogen evolution reaction (HER) electrocatalysis mediated by hexacarbonyls [(OC)₃Fe-(dithiolate)Fe(CO)₃] and pentacarbonyls [(OC)₃Fe-(dithiolate)Fe(CO)₂L]. Such ligand sets, in contrast to that of the CN⁻-containing active site and models, result in higher oxidation potentials but make accessible very reduced (but abiological) states such as Fe(0)Fe(I),²⁴¹ particularly when electron-poor ligands, such as 3,6-dichloro-1,2-benzeneditholate,²⁴² are used. The protonation of these (often monoanionic) Fe(0)Fe(I) complexes would afford mixed-valent Fe(II)Fe(I)/Fe(1.5)Fe(1.5) hydrides, species that have been invoked during the HER as mediated by the [FeFe]-H₂ases. However, the necessary formation of Fe(0)Fe(I) species leads to hexa- and pentacarbonyls displaying high overpotentials,⁶⁷ and these parent compounds can be considered to have less fidelity to [FeFe]-H₂ases than their more substituted analogues, with which this subsection is now concerned.

An early study on mixed-valent diiron hydrides reported the reduction of [Me₃P(OC)₂Fe(pdt)(μ-H)Fe(CO)₂PMe₃]⁺ ([6(μ-H)]⁺) with acenaphthylene anion radical to generate the neutral complex [Me₃P(OC)₂Fe(pdt)(μ-H)Fe(CO)₂PMe₃] ([6(μ-H)], Figure 12).¹⁰⁸ The product features an EPR resonance at g = 2.0066 (near that of the free electron g_e = 2.0023) split by ¹H (A_iso = −41.7 MHz) and two equivalent ³¹P nuclei (A_iso = −75.8 MHz). A Robin–Day class III delocalized Fe(1.5)(μ-H)Fe(1.5) description for the complex was supported by DFT studies, which suggested that 70% of the unpaired spin density resides on the two Fe centers, with approximately 35% on each site.¹⁰⁸

Figure 12.

The 1e⁻ reduction of a diamagnetic Fe(II)(μ-H)Fe(II) hydride affords a Fe(1.5)(μ-H)Fe(1.5) mixed-valent hydride.¹⁰⁸

The radical hydride [6(μ-H)] is rather labile, as evidenced by its poorly reversible oxidation and an estimated t_1/2 ~ 1 s at 25°C. Relatives of the present system include Fe(edt)(μ-H)Fe (edt²⁻ = 1,2-ethanedithiolate),¹⁰⁸ Fe(bdt)(μ-H)Fe (bdt²⁻ = 1,2-benzenedithiolato),²⁴³ and Fe(SH)₂(μ-H)Fe derivaties,²⁴⁴ all of which are similarly fragile. However, protonation of [dppv-(OC)Fe(pdt)Fe(CO)dppv] ([7], dppv =1,2-bis-(diphenylphosphino)ethene) affords a more sterically encumbered diferrous hydride [dppv(OC)Fe(pdt)(μ-H)Fe(CO)-dppv]⁺ ([7(μ-H)]⁺, Figure 13), which sustains reduction to afford the isolable neutral hydride [dppv(OC)Fe(pdt)(μ-H)Fe(CO)dppv] ([7(μ-H)]).²⁴⁵ X-ray crystallography revealed the product to feature an asymmetric Fe(II)(μ-H)Fe(I) core. The structural and EPR data are consistent with DFT calculations indicating that the Fe(I), which bears two-thirds of the spin density, is more distant from the hydride than is Fe(II) (1.82 vs 1.61 Å), on which most of the remaining spin resides. In solution, a small fraction of this asymmetric Robin–Day class II species converts to a C₂-symmetric complex with the expected Fe(1.5)(μ-H)Fe(1.5) class III description. The latter product features dppv ligands with so-called “apical–basal” stereochemistry, in which they each bind an apical site that is trans to μ-H, as well as a basal site trans to pdt²⁻.

Figure 13.

Formation and isomerism of the mixed-valent hydride [7(μ-H)]. The μ-H⁻ ligand is unaffected by acid.²⁴⁵

Treatment of mixed-valent bis(dppv) hydride [7(μ-H)] with D⁺ yields the diferrous hydride and 1/2D₂ rather than [dppv(OC)Fe(pdt)Fe(CO)dppv]⁺ ([7]⁺) and HD. This surprising observation implicates a spectator role for μ-H⁻ in this complex during HER, probably reflecting the influence of the two bulky diphosphines and the robustness of Fe(II)(μ-H)Fe(II) motifs in general.

The relatively high-valent paramagnetic hydride [(Cp*)Fe-(bdt)(μ-H)Fe(Cp*)] (Cp*⁻ = pentamethylcyclopentadienide) is generated by reduction of the unusual diferric species [(Cp*)Fe(bdt)(μ-H)Fe(Cp*)]⁺ (E_1/2 = −0.91 V vs Fc^+/0). Protonation of the Fe(II)(μ-H)Fe(III) complex returns the Fe(III)(μ-H)Fe(III) complex and induces evolution of H₂ (0.5 equiv).²⁴⁶ These bridging hydride ligands, even on reduced diiron cores, can often be so inert they resist protonation even with strong acids. Indeed, the HER reactions described here can be considered to proceed by an outer-sphere mechanism, throughout which the coordination sphere of the catalyst is unchanged. If instead HER occurs with protons contacting the metal site(s), then it corresponds to an inner sphere mechanism, involving the intermediacy of a mixed-valent dihydride complex. Diiron dihydrides are described in section 3.2.3.

3.2.3. Dihydrides

The privileged bis(dppv) motif is well-suited to stabilizing many Fe oxidation and protonation states, among which are FeFe complexes in which two hydride ligands are present. Displacement of CO from [7(μ-H)]⁺ in MeCN affords the activated species [dppv(OC)Fe(pdt)(μ-H)Fe-(MeCN)dppv]⁺ ([8(μ-H)(MeCN)]⁺), and subsequent treatment with BH₄⁻ cleanly gives [dppv(OC)Fe(pdt)(μ-H)Fe(t-H)dppv] ([8(μ-H)(t-H)], Figure 14).²⁴⁷ The displacement of a labile MeCN ligand by a hydride source was inspired by the related synthesis of [(Me₃P)₂(CO)Fe(edt)Fe(CO)(PMe₃)₂(t-H)]⁺ discussed below. The product features not only a bridging hydride but also a terminal hydride (denoted t-H⁻); these cis ligands give rise to ¹H NMR resonances at δ −18.9 and −12.2 ppm, respectively (intramolecular exchange rate ≈1 s⁻¹ at −25°C).

Figure 14.

Preparation of the dihydride [8(μ-H)(t-H)] from sources of H⁺ and H⁻.²⁴⁷

While sequential addition of H⁺ and H⁻ to an Fe(I)Fe(I) precursor represents one route to dihydrides, perhaps a more obvious (and generally applied) method is oxidative addition of H₂. In the case of [dppv(OC)Fe(edt)Fe(CO)₃] ([9], Figure 15), photolytic decarbonylation affords a transient 32e⁻ species that binds and cleaves H₂ to afford [dppv(OC)Fe(edt)(μ-H)Fe(t-H)(CO)₂] ([10(μ-H)(t-H)], δ −12.8 (t-H), −14.9 ppm (μ-H)).²⁴⁸ DFT calculations indicate a product of C_s symmetry, with the trans nature of the hydride ligands contrasting the cis arrangement in [8(μ-H)(t-H)].²⁴⁸

Figure 15.

Preparation of a dihydride through oxidative addition.²⁴⁸

A closely related process of CO dissociation and H₂ activation occurs for diruthenium analogues. For example, irradiation of [Cy₃P(OC)₂Ru(pdt)Ru(CO)₂PCy₃] under an H₂ atmosphere affords the trans dihydride [Cy₃P(OC)₂Ru(pdt)(μ-H)Ru(t-H)(CO)PCy₃],²⁴⁹ which yields [Cy₃P(CO)₂Ru(pdt)(μ-H)Ru-(η²-H₂)(CO)PCy₃]⁺ on treatment with [H(OEt₂)₂]BAr^F₄ in CH₂Cl₂. When [D(OEt₂)₂]BAr^F₄ is instead used, D incorporation is only observed as η²-HD and not as a D⁻ ligand, a result that highlights the inherent stability of μ-H⁻ over t-H⁻ ligands.

The discussion on synthetic models has, until now, focused on bridging hydride complexes. Yet, both [8(μ-H)(t-H)] and [10(μ-H)(t-H)] feature t-H⁻ ligands that are a key motif in the [FeFe]-H₂ase mechanism. The preparation of terminal hydride complexes has been the subject of intense research activity, much of which is summarized in section 3.2.4.

3.2.4. Terminal Hydrides

In considering the structure of dihydride [8(μ-H)(t-H)], one might say that the presence of the strongly donating μ-H⁻ ligand directs the second H⁻ to adopt the important terminal position. But what of the situation when only a single H⁻ ligand is present? Can a terminal monohydride, similar to that proposed in the enzyme mechanism, be observed in a model complex?

The large number of bridging hydrides reported is in part due to such species typically being thermodynamic products of [L₃Fe(dithiolate)FeL₃] protonation. Yet it just so happens that terminal hydrides, of varying kinetic stability, feature commonly (but not always)²⁵⁰ as intermediates in this reaction. Their presence was initially inferred from electrochemical studies on the HER activity of the related diphosphide [(OC)₃Fe(μ-PPh₂)₂Fe(CO)₃].²⁵¹ The first direct observation of terminal hydrides arising from protonation was found by studying the action of HBF₄·Et₂O on [OC(dppe)Fe(pdt)Fe(CO)₃].²⁵² At 298 K in CH₂Cl₂ solution, the sole product was bridging hydride [(dppe)(CO)Fe(pdt)(μ-H)Fe(CO)₃]⁺, in which a dibasal phosphine (i.e., C_s symmetry) was indicated by a characteristic high-field ¹H NMR triplet at −14.1 ppm (²J_PH = 21 Hz). When protonation was instead monitored at 203 K, a singlet at −4.33 ppm could be observed, such a shift being significantly downfield of resonances expected for bridging hydrides. The lack of coupling to any ³¹P nuclei indicated that initial protonation, perhaps counterintuitively, occurred at the relatively electron-poor Fe(CO)₃ fragment to afford [(dppe)(OC)Fe(pdt)Fe-(CO)₃(t-H)]⁺. This terminal hydride, upon warming to 243 K, converts to a mixture of apical–basal and dibasal forms of the bridging hydride, with the latter being the only FeFe species at 298 K. As many further examples will demonstrate, low-temperature protonation studies of this type are a powerful tool in studying the formation of terminal hydrides for a variety of FeFe systems.

Despite efforts in monitoring protonations of Fe(I)Fe(I) species, the first diiron terminal hydride complex to be fully characterized arose from a rather different route involving hydride transfer to an Fe(II)Fe(II) precursor. Treatment of [(Me₃P)₂(CO)Fe(edt)Fe(CO)(PMe₃)₂MeCN]²⁺([11(MeCN)]²⁺) with either AlH₄⁻ or BH₄⁻ at −25°C induced the formation of [(Me₃P)₂(CO)Fe(edt)Fe(CO)(PMe₃)₂(t-H)]⁺ ([11(t-H)]⁺) by displacement of MeCN (Figure 16). Isolated as a strikingly green species,²⁵³ [11(t-H)]⁺ exhibits a ¹H NMR resonance at −4.6 ppm coupled to just two ³¹P nuclei. The product features an IR-active ν_FeH band at 1844 cm⁻¹, consistent with the terminal nature of the hydride ligand. While M–H vibrations are often not easily assigned,²⁵⁴ the frequency is comparable to that calculated using DFT (1908 cm⁻¹), with two lower frequency bands (1352 and 1151 cm⁻¹) predicted for the analogous bridging isomer.²⁵³

Figure 16.

Preparation (top) and X-ray structures of [11(t-H)]⁺ and ([11(μ-H)]⁺ (bottom). Non-hydride H atoms are omitted for clarity.²⁵³

Although terminal hydride [11(t-H)]⁺ could be crystallized (r_FeFe = 2.565 Å, r_FeFe= = 1.498 Å), it is of limited thermal stability; in solution at room temperature it isomerizes to the red, C₂-symmetric bridging hydride ([11(μ-H)]⁺ (−20.6 ppm, r_FeFe = 2.610 Å, rFeFe= = 1.656, 1.602 Å) by a first-order process (k = 2 × 10⁻⁴ s⁻¹ at 294 K). Both t-H and μ-H isomers are unreactive toward H₂O, but the terminal form does liberate H₂ upon treatment with the strong acids HOTf or [H(OEt₂)₂]BAr^F₄ in the presence of MeCN to give back [11(MeCN)]^2+.255 This finding highlights the considerably more hydridic nature of t-H⁻ ligands.²⁴⁹ Such work further emphasizes the need to suppress the isomerization of terminal to bridging species. This process is typically irreversible, although theoretical work has suggested that excitation of Fe(pdt)(μ-H)Fe species to a low-lying triplet state can afford the terminal isomer.²⁵⁶

The t-H → μ-H isomerization has been examined using DFT calculations on [(dppv)(CO)Fe(edt)Fe(CO)₃(t-H)]⁺ ([12(t-H)]⁺), [(dppv)(CO)Fe(edt)Fe(PMe₃)(CO)₂(t-H)]⁺, and [(Me₃P)₂(CO)Fe(edt)Fe(PMe₃)₂CO(t-H)]⁺ ([11(t-H)]⁺). These isomerizations may conceivably occur through Bailar (trigonal) or Ray–Dutt (rhombic) twists, with the pathway dependent on the ligand set.²⁵⁷ For these complexes, the free energy barriers to Ray–Dutt twisting are lower by 4.4–7.7 kcal/mol. When bulky ligands are present, certain isomerizations are unlikely to occur by Ray–Dutt twists since these would involve high-energy intermediates such as bridging phosphine complexes (Figure 17). Such motifs need not be invoked if isomerization proceeds through Bailar twists. For example, [12(t-H)]⁺ was predicted to convert to its bridging tautomer through a multistep pathway involving only Ray–Dutt twists, while [11(t-H)]⁺ was calculated to isomerize by both Ray–Dutt and Bailar twists.

Figure 17.

Mechanisms for the isomerization of terminal to bridging hydride complexes.

Since the first preparation of a diiron terminal hydride, several more such complexes of varying stability have been prepared. Their syntheses typically follow a biomimetic route involving (low-temperature) protonation of Fe(I)Fe(I) species. In the case of the bis(dppv) complex [dppv(OC)Fe(pdt)Fe(CO)dppv] ([7]) protonation with strong acids such as HBF₄·Et₂O (weak acids will not suffice) at −25 °C affords the terminal hydride [dppv(OC)Fe(pdt)Fe(CO)(t-H)dppv]⁺ ([7(t-H)]⁺, δ(¹H) −3.5 ppm) as an isolable kinetic product. Solutions of this species are stable for minutes at 0 °C,²⁵⁸ an effect of the bulky, strongly electron-donating ligands that hinder isomerization.²⁵⁹ At room temperature, the hydride converts to asymmetric and symmetric isomers of the bridging species [dppv(OC)Fe(pdt)-(μ-H)Fe(CO)dppv]⁺ ([7(μ-H)]⁺, δ(¹H) −14.5 and −15.6 ppm, respectively) described above. While stereodynamics and HER mechanisms are discussed in section 3.2.6, both the t- and μ-H propanedithiolates catalyze the HER, although with modest TOFs of 5 and 3 s⁻¹, respectively.

Protonation at a terminal site in preference to the Fe–Fe bond is puzzling²⁶⁰ because most synthetic Fe(I)Fe(I) species do not have an empty apical coordination site. Due consideration must also be paid to the thiolate S atoms, whose nonbonding electron pairs allow for thiol intermediates in the protonation reaction. The S atoms in synthetic [FeFe]-H₂ase models have indeed been shown to be competent Brønsted bases. For example, protonation of [(OC)₂(Me₃P)Fe(edt)(μ-H)Fe(CO)₂(PMe₃)] affords, according to IR spectroscopy, the expected hydride [(OC)₂(Me₃P)Fe(edt)(μ-H)Fe(CO)₂(PMe₃)]⁺, as well as the thiol [(OC)₂(Me₃P)Fe(Hedt)Fe(CO)₂(PMe₃)]⁺.²⁶¹ Similarly, treatment of [(OC)₃Fe(3,4-dichlorobenzene-1,2-dithiolate)Fe-(dppp)CO] with HBF₄·OEt₂ (1 equiv) results in the protonation of one S atom.²⁶² The involvement of heteroatom-protonated species can greatly influence the regiochemistry of hydride formation because heteroatom proton relays are known to play key roles in reducing barriers in what would otherwise be intermolecular protonations of metal sites. Indeed, heteroatom bases of comparable thermodynamic basicity to low-valent metal centers have much greater kinetic basicities.²⁶³ These effects are borne out in the contrasting protonations of dithiolate [dppv(OC)Fe(pdt)Fe(CO)₃]²⁶⁴ ([13], Figure 18) and a related diphosphide [dppv(OC)Fe(pdpp)Fe(CO)₃] ([14], pdpp²⁻ = 1,3-propanedi(phenylphosphide)).²⁶⁵ Treatment of [13] with [H(OEt₂)₂]BAr^F₄ at −90 °C rapidly affords [dppv(OC)Fe-(pdt)Fe(t-H)(CO)₃]⁺ ([13(t-H)]⁺), which undergoes slow isomerization to the bridging hydride [13(μ-H)]⁺. Under identical conditions, the diphosphide transforms slowly and exclusively to the bridging species [dppv(OC)Fe(pdpp)(μ-H)Fe(CO)₃]⁺ ([14(μ-H)]⁺) with no intermediates observed. With lone electron pairs available on ligated pdt²⁻ but not pdpp²⁻, it is apparent that heteroatom participation not only accelerates metal protonation, but also directs it to terminal sites.

Figure 18.

Bridging secondary phosphido ligands do not represent sites for protonation, and only bridging hydrides are observed.²⁶⁵ In contrast, bridging thiolato ligands can be protonated, and serve as relays to afford terminal and bridging hydride complexes.²⁶⁴

Assuming the dppv(CO)Fe^I fragment is more basic than Fe^I(CO)₃, it may seem strange that the t-H⁻ ligands in complexes such as [12(t-H)]⁺ are located at the tricarbonyl sites. This stereochemistry is rationalized in terms of the steric bulk of the diphosphine, as well as by considering a “rotated” form of the neutral complex in which a bridging CO ligand is trans to a vacant terminal Fe site (Figure 19). In the extreme case, one CO ligand moves to a bridging position to afford [dppv(CO)Fe(edt)(μ-CO)Fe(CO)₂]. This isomer may be considered a 2e⁻ mixed-valent species in which the pseudooctahedral site is Fe(II), with the square-pyramidal basic site being Fe(0). According to DFT calculations, the two isomers of this “rotated” species are both transition states, of which an asymmetric form, in which dppv occupies an apical and basal site (apical–basal), is only slightly more stable.²⁶⁵

Figure 19.

Rotated and unrotated isomers of [12], an Fe(I)Fe(I) model for H_red.²⁶⁵

The role of S atoms in hydride formation was also implicated in the protonation and isomerization of the electron-rich tetraphosphine [(Me₃P)₂(OC)Fe(pdt)Fe(CO)(PMe₃)₂] ([15], Figure 20).²⁵⁰ Its treatment with [H(OEt₂)₂]BAr^F₄ at −90 °C gave a 2:1 mixture of symmetrical bridging hydride and a terminal hydride (δ(¹H) −18.8 and −2.2 ppm, respectively). Also formed was a nonhydride (putatively thiol) species that converts to both bridging and terminal hydrides at −60 °C, with the t-H → μ-H conversion being, encouragingly, rather slow (t_1/2 = 2.5 h at 20 °C) in the absence of acid catalyst.

Figure 20.

Protonation of [15] affords a thiol intermediate en route to a terminal hydride. Formation of the bridging hydride thermodynamic product is acid-catalyzed.²⁵⁰

While models discussed in this section share much in common with the [FeFe]-H₂ase active site, the synthetic terminal hydrides lack the conformational rigidity of the enzyme H-cluster. In the latter case, the distal Fe adopts a reactive “rotated” state in which a terminal site is available for binding H⁻ or H₂ substrate.²⁶⁶ This rotated structure is perhaps stabilized by hydrogen bonding to the protein, but electronic factors may also be important. In model systems, this geometry can only be reproduced in unsymmetrically substituted complexes, including the Fe(I)Fe-(I) compounds [dppv(OC)Fe(2,2-diethyl-1,3-propanedithiolate)(μ-CO)Fe(CO)₂]²⁶⁶ and [dmpe(OC)Fe-(adt^Bn)Fe(CO)₃].²⁶⁷

3.2.5. Comparison of Bridging and Terminal Hydrides

While the thermal stability of bridging hydrides was discussed above, these isomers are also typically more resistant to electrochemical reduction than their terminal counterparts. In view of the strong σ-donicity of the H⁻ ligand,⁴⁷ 1e⁻ reduction of Fe(II)Fe(II)(t-H) occurs at the ferrous center not bound to the hydride (analogous to Fe^p in [FeFe]-H₂ase). This situation is not possible in the case of Fe(II)(μ-H)Fe(II), as H⁻ exerts its cathodic effect on the couples of both metals.²⁵⁹ A corollary is that H⁻ binds Fe(II) considerably more tightly than Fe(I), as is illustrated by the disparate Fe–H distances in the mixed-valent complex [dppv(OC)Fe(pdt)(μ-H)Fe(CO)dppv] ([7(μ-H)]). Theoretical calculations also indicate that reductions of terminal hydrides occur at more positive potentials than those of bridging hydrides,¹⁸⁶ although some special cases have been identified in which this trend is reversed. These situations arise when bulky ligands are present, with reduction of the μ-H complex alleviating strain by allowing the metals to distance themselves from one another.¹⁸⁶

Discussion of μ-H isomers can be illustrative and guide the synthesis of more stable terminal hydrides, but one must keep in mind that μ-H species do not come into consideration in the [FeFe]-H₂ase mechanism. For the enzyme catalytic cycle (Figure 9), DFT studies indicate that t-H to μ-H isomerization comes with a prohibitively large electronic energy barrier (29 kcal mol⁻¹),²⁶⁸ in no small part due to the H-bonding of CN⁻ ligands to nearby residues preventing turnstile rotation. Moreover, direct protonation of the Fe–Fe bond—bypassing any t-H species—is even more unlikely, and the endothermic H⁺ transfers from Lys358 to H_ox and H_red are accompanied by kinetic barriers of 52 and 39 kcal mol⁻¹, respectively. These large barriers arise because direct H⁺ transfer would require movement of Lys358 and cleavage of its H-bond to Glu361. In contrast to such processes, terminal hydride formation is exothermic and requires an electronic activation energy of only 6.9 kcal mol⁻¹, with the barrier to H₂ formation being 4.1 kcal mol⁻¹.²⁶⁸ The redox properties of diiron dithiolates allows one to estimate rates of protonation of the Fe–Fe bond (the highest occupied molecular orbital (HOMO)).²¹⁷ In the case of an [FeFe]-H₂ase protein operating at −414 mV vs NHE at pH 7, the estimated rate is an order of magnitude lower than the observed rates for the enzyme. This once more underscores that the enzyme operates through terminal hydrides, and it does so because of distinct kinetic advantages. Of course, the key component facilitating terminal protonation is the azadithiolate cofactor. Synthetic models featuring adt²⁻ (or its derivatives) are discussed in section 3.2.6, along with their mechanisms for H₂ evolution.

3.2.6. Azadithiolato Hydrides

The synthesis of the archetypal [(OC)₃Fe(pdt)Fe(CO)₃] simply involves treatment of an Fe(0) carbonyl with the dithiol H₂pdt. In contrast, the instability of free H₂adt or H₂adt^R ((adt^R)²⁻ = ⁻SCH₂N(R)-CH₂S⁻); R = alkyl, etc.) species or simple salts thereof²⁶⁹ necessitates indirect synthetic routes for [(OC)₃Fe(adt)Fe-(CO)₃] derivatives. Among the methods used, the Mannich-type condensation of CH₂O, amines, and [(OC)₃Fe(SH)₂Fe(CO)₃] has proven useful in affording [(OC)₃Fe(adt)Fe(CO)₃],²⁷⁰ N-substituted species [(OC)₃Fe(adt^R)Fe(CO)₃], as well as oxadithiolate [(OC)₃Fe(odt)Fe(CO)₃] (odt²⁻ = (⁻SCH₂)₂O).²⁷¹ Conveniently, the ligand substitution chemistry of these heteroatom-bridged hexacarbonyls mirrors that of the pdt²⁻ parent, and several tertiary phosphine- and CN⁻-substituted derivatives are known. These models thus feature an ideal arrangement with an Fe(I)Fe(I) core of high thermodynamic basicity proximal to an amine proton relay with high kinetic basicity.

A major contributor to the kinetic favorability of terminal hydrides is the azadithiolate cofactor, which plays a key role in the catalytic activity of models and the enzyme. The azadithiolate in the enzyme was confirmed by an elegant series of experiments in which apo-HydA (the [FeFe]-H₂ase from Chlamydomonas reinhardtii lacking the [2Fe] subsite) was reconstituted (“maturated”) with the synthetic diiron dicyanides [NC-(CO)₂Fe(adt)Fe(CO)₂CN]²⁻ ([3]²⁻), [NC(CO)₂Fe(odt)Fe-(CO)₂CN]²⁻, and [NC(CO)₂Fe(pdt)Fe(CO)₂CN]²⁻.^199,200,272 Each maturation product exhibited ν_CO and ν_CN bands matching those of native HydA, indicating that the diiron complexes had been accepted by the apoprotein. Yet only the apo-HydA + [NC(CO)₂Fe(adt)Fe(CO)₂CN]²⁻ product (believed to be the holoenzyme) could reproduce the activity of authentic HydA. Because the activity of the enzyme is so high, the protein maturated with the flawed cofactor pdt²⁻ is still catalytically competent, albeit with <1% activity of HydA.²⁰¹ Such work confirms not only that the dithiolate cofactor contains an amine, but further that this H⁺ relay is essential to the efficiency of [FeFe]-H₂ases.

The divergent reactivity of propanedithiolates and azadithiolates, apparent also in purely synthetic species, is of great importance. While [(Me₃P)₂(OC)Fe(pdt)Fe(CO)(PMe₃)₂] ([15]) converts primarily to the bridging hydride upon treatment with [H(OEt₂)₂]BAr^F₄ (vide supra), azadithiolate [(Me₃P)₂(OC)Fe(adt)Fe(CO)(PMe₃)₂] ([16]) converts rapidly and exclusively to the terminal hydride [(Me₃P)₂(OC)Fe-(adt)Fe(t-H)(CO)(PMe₃)₂]⁺ ([16(t-H)]⁺, δ(¹H) −2.29 ppm).²⁵⁰ Considering the similar electronic properties of the low-valent species (ν_CO = 1860, 1839 and 1857, 1836 cm⁻¹ for pdt²⁻ and adt²⁻ complexes, respectively), this regiochemical selectivity is ascribed to the influence of the secondary amine, which provides a kinetic protonation pathway via an ammonium intermediate. The effect is so marked that even weak acids (e.g., NH₄PF₆) can induce rapid t-H formation in the azadithiolate complex, with no reactivity observed for the propanedithiolate complex. While S-protonation was invoked when describing the reactivity of [15], such thiol intermediates are no longer relevant when a basic amine is present, especially when it is poised above the terminal site. Such an arrangement has been crystallographically confirmed in the diiron azadithiolato terminal hydrides [(Me₃P)₂(OC)Fe(adt)Fe(t-H)(CO)-(PMe₃)₂]⁺([16(t-H)]⁺) and [dppv(CO)Fe(adtH)Fe(t-H)-(dppv)(CO)]²⁺ ([17(t-H)H]²⁺, Figure 21).^250,258

Figure 21.

X-ray structures of [16(t-H)]⁺ and [17(t-H)H]²⁺. Nonionizable H atoms are omitted for clarity.^250,258

The structure of [16(t-H)]⁺ features not only the synergistic trans arrangement of the t-H⁻ donor and μ-CO acceptor, but also a positioning of the amine such that N–H⋯H–Fe dihydrogen bonding (2.042 Å) is possible, as also indicated by one-dimensional nuclear Overhauser effect NMR experiments. It does not take much imagination to see that the Fe–H (1.487 Å) moiety could arise from tautomerization of an ammonium intermediate, followed by inversion at N such that this interaction (as well as the anomeric-type interactions featuring the lone pair) can be established. The related tetraphosphine [17(t-H)H]²⁺, perhaps owing to the lower donicity of dppv versus bis(PMe₃) ligation, has a slightly shorter Fe–H bond (1.438 Å). Given that the dppv complex was crystallized with the dithiolate in the ammonium state, the difference could also be ascribed to the stronger dihydrogen bonding (1.879 Å) of this now more acidic moiety to the H⁻ ligand. DFT studies have shown that the N–H⋯H–Fe interaction is strongly affected by the presence of H-bond acceptors such as BF₄⁻, and that in the absence of the counteranions the H⋯H distance contracts to 1.40 Å.¹⁶¹ The participation of an ammonium hydride state for the enzyme has been proposed, with such a species (Figure 9, middle right) arising from the protonation of H_sred.

The thermodynamic influence of dihydrogen bonding has been the subject of many experimental and theoretical studies. In particular, vibrational frequency shifts induced by such bonding, as well as the final H⋯H distances, have been used in combination with DFT calculations to correlate observables with the strengths of these interactions.²⁷³ In the present case of [16(t-H)]⁺ and [17(t-H)H]²⁺, the stabilization afforded by dihydrogen bonding amounts to approximately 2 and 4 kcal mol⁻¹, respectively, based on the H⋯H distances. In all, such interactions play a small but not insignificant role in reactivity, and they have also been implicated in the [Fe]-H₂ase catalytic cycle (section 5.1).²⁷⁴

Although the amine group plays a key role in second coordination sphere interactions and directing terminal protonation, it is perhaps unfortunate that the t-H → μ-H isomerization in synthetic models occurs regardless of which dithiolate is present.²⁶⁴ Thus, both complexes in Figure 21 irreversibly convert to their respective bridging hydride thermodynamic products. In the dppv-containing system, the conformational dynamics have been extensively studied both in situ and in silico (Figure 22). DFT calculations indicate the kinetic protonation product of [17] to be the ammonium [17H]⁺.¹⁶¹ The proton is rapidly relayed to a terminal coordination site at the nearest Fe center, which is 1.3 pK_a units (on an MeCN scale) more basic than the amine. Experimentally, low-temperature protonation of [17] readily affords terminal hydride [17(t-H)]⁺ (δ(¹H) −4.2 ppm), which eventually isomerizes to bridging hydrides of pseudo C₂ and then C₁ symmetry. The final hydride complex [17(μ-H)]⁺ (−14.3 ppm; Figure 22, center left) is much less acidic (pK_a > 18.6) than [17(t-H)]⁺ (pK_a = 16), reflecting the higher thermodynamic basicity of the Fe–Fe bond relative to the terminal sites in Fe(I)Fe(I) species. Consistent with NMR data, the pK_a values suggest the conversion should be almost quantitative, as [[17(μ-H)]⁺]/[[17(t-H)]⁺] > 10^2.6.²⁵⁸ A similar t-H → μ-H isomerization process is at play for the doubly protonated derivatives. The ammonium terminal hydride [17(t-H)H]²⁺ is 9.5 pK_a units more acidic than [17(t-H)]⁺, in line with the former only being relevant when strong acids are employed. It turns out that [17(t-H)H]²⁺ plays an important role in catalysis. Having addressed the stoichiometric protonations of diiron species, the discussion now moves to their catalytic reactions. As will become clear, such complexes—most notable of which is the dication [17(t-H)H]²⁺—can exhibit a broad range of HER activity that can be readily appreciated by considering their structures.

Figure 22.

Conformational dynamics accompanying the protonation of [17] (inversion of ammonium/amine centers not depicted).¹⁶¹

3.2.7. Proton Reduction Catalysis

Understanding the protonation of [FeFe]-H₂ase models is a prerequisite for the study of HER electrocatalysis, in which acid–base chemistry plays a central role. Hydrides of the form FeFe(t-H) or Fe(μ-H) Fe are important intermediates, but few HER investigations on diiron models include conclusive characterization of hydride-bearing states. Several surveys on the HER mediated by FeFe species are available,^81,275–277 and presented here is rather a description of the salient components of diiron dithiolate catalysts, such that one can identify the often subtle structural and electronic differences between systems and rationalize their influence on catalytic H⁺ reduction. For instance, the donicity of ligands bound to the FeFe cores will govern which oxidation states are accessible during the catalytic cycle. The electron density at the metal further influences which protonation states predominate, an aspect that is also affected by protic groups in the second coordination sphere and the nature of the acid employed. With these aspects in mind, the design challenges associated with [FeFe]-H₂ase modeling, as well as some potential solutions, will become more apparent. In terms of evaluating electrocatalysis, it is once more emphasized that a description should include both thermodynamic information (e.g., potential) and a corresponding kinetic parameter (e.g., TOF).⁷³ In all, catalysts acting at low overpotentials and high turnover frequencies are the most desirable.

Before turning to specific examples, possible pathways for inner-sphere HER mediated by an arbitrary reduced metal complex L_nM are presented in Figure 23. With respect to FeFe complexes, catalysis typically involves conversion of an Fe(I)-Fe(I) species (an H_red model denoted L_nM) to the doubly protonated and doubly reduced state L_nMH₂. Poised to release H₂, this latter state may take the form of a dihydrogen complex L_nM(η-H₂), a dihydride L_nM(H)₂, or another tautomer. The exact nature of the relevant hydride-bearing intermediates for a given catalyst depends on the mechanism at play, which is influenced by metal coordination environment, Brønsted acid strength, and catalytic conditions. As described in section 3.2.2, a scarce few diiron hydrides have been characterized aside from the 34e⁻ diferrous systems.¹⁹⁶ Furthermore, it is rare for even the latter well-studied compounds to be directly observed in catalysis.

Figure 23.

Many proton and electron transfers that can be involved in the HER.

Discussed now are examples of [(OC)₃Fe(dithiolate)(CO)₃] catalysts, the “first-generation” [FeFe]-H₂ase models on which the majority of HER studies have focused,²⁷⁶ in part due to the straightforward preparation and robust nature of such hexacarbonyl compounds. With respect to the archetypal [(OC)₃Fe(pdt)Fe(CO)₃] ([4]), the use of electrochemical and spectroelectrochemical measurements, digital simulations,^278,279 and DFT calculations has revealed much about its HER mechanism (Figure 24).²⁸⁰ The electron-poor Fe(I)Fe(I) core in [4] is not readily protonated (section 3.2.1), and catalytic reduction of HOTs instead begins with reduction of [4] (E_1/2 = −1.55 V vs Fc^+/0). The monoanion is now sufficiently basic to undergo protonation by HOTs, and the resulting μ-H species is more easily reduced (E > −1.55 V) than [4]⁻ such that an additional reduction occurs. The formally Fe(I)Fe(I) anion [(OC)₃Fe(pdt)(μ-H)Fe(CO)₃]⁻ ([4(μ-H)]⁻) can now undergo protonation at either of its Fe or S atoms. Protonation at a metal site affords an Fe(II)Fe(II) dihydride poised to release H₂ and regenerate [4] (albeit slowly) in an overall ECEC mechanistic cycle. Another possible form of the doubly-protonated species is a Fe(I)Fe(I) hydride in which one thiol binds only one Fe site. The latter species readily undergoes reduction (−1.75 V) and quickly releases H₂ as part of a CECE cycle. A common reduced hydride intermediate can thus perform catalysis by two separate mechanisms occurring at distinct potentials and rates.

Figure 24.

HER mechanism associated with [4].²⁸⁰

Electrocatalytic proton reduction by the 1,2-benzenedithiolate congener [(OC)₃Fe(bdt)Fe(CO)₃] ([18]) proceeds by a mechanism different from those described above. While the exact pathway is dependent on the acid pK_a, experimental and DFT studies indicate that this very electron-poor complex must receive 2e⁻ (−1.31, −1.33 V vs Fc^+/0 in MeCN) to give the formally Fe(0)Fe(0) dianion [18]²⁻, protonation of which yields the Fe(I)(μ-H)Fe(I) species [18(μ-H)]⁻ (Figure 25).⁸⁵ This hydride product is susceptible to further reaction with acids of pK_a < 23 (e.g., HOTs), resulting in slow release of H₂ that regenerates [18] and completes the EECC catalytic cycle. Additional reduction of [18(μ-H)]⁻ is necessary when weaker acids (pK_a > 23, e.g. HOAc) are used. This is followed by protonation and rapid H₂ release to complete an overall ECEC mechanism.

Figure 25.

HER mechanism associated with [18].⁸⁵

The reader may recall that it is an Fe(I)Fe(I) form of [FeFe]-H₂ase, namely H_sred, that is proposed to undergo tautomerization/protonation to afford a terminal hydride (section 3.1). Contrasting the HER mechanism of the enzyme with those of its hexacarbonyl models allows one to tease out the influences of ligand sets and acid–base cofactors: the first and second coordination spheres. With respect to ligand effects, electron-poor models (i.e., those with six CO ligands) require reduction below an Fe(I)Fe(I) state before their protonation can occur. Even then, it is a μ-H rather than t-H species that invariably forms. In terms of the second coordination sphere, the presence of Brønsted basic sites within a model complex can have a strong influence on the electrocatalytic mechanism and potential at which catalysis occurs.²⁸¹ Basic residues are protonated in acidic media, resulting in the catalyst bearing less negative charge and giving rise to milder reductions relative to those of more anionic species. For example, the mechanism proposed for HOTs reduction²⁸² catalyzed by [(OC)₃Fe(adt^R)Fe(CO)₃] ([19], R = ⁱPr, CH₂CH₂OCH₃) begins with protonation of the amino group to form [19H]⁺ (Figure 26). The resulting Fe(I)Fe(I) cation can be reduced at −1.2 V vs Fc^+/0, nearly 450 mV more positive than the wave for the neutral complex [(OC)₃Fe(adt^R)Fe(CO)₃]. The reduced ammonium [19H] is then protonated to generate a mixed-valence ammonium hydride [19(μ-H)H]⁺, which immediately reduces at potentials below −1.2 V and slowly releases H₂. While the structure of the [19(H)₂] product is not obvious, what is clear is that an acid–base cofactor enables a mechanism involving mixed-valence hydrides in place of abiological, highly reduced Fe(I)(μ-H)Fe(I) intermediates. In the presence of the strong acid HBF₄·Et₂O, [19(H)₂] can not only undergo slow H₂ loss, but instead may also be protonated to afford [19(H)₃]⁺, whose reduction (−1.4 V) causes fast release of H₂. Of the two CECE mechanisms, as was illustrated in the previous two examples, the reduction occurring at the more negative potential results in faster H₂ production.

Figure 26.

HER mechanism associated with [19].²⁸²

The low basicity of hexacarbonyl derivatives means that their conversion to hydride-bearing species necessitates either the use of superacids or electrochemical reduction prior to protonation. The second process requires highly negative potentials and, thus, can incur significant overpotentials, even when an azadithiolate (adt^R)²⁻ is present. At no point does ammonium [19H]⁺ convert to an amine hydride form, reflecting a poorly basic metal core. The observation that another Fe(I)Fe(I) species, namely H_red, readily protonates to form a hydride indicates deficiencies in the inner coordination sphere of these “first-generation” models. As described earlier in this section, these are addressed by substitution of CO ligands for strong σ-donors such as CN⁻, phosphines, and N-heterocyclic carbenes. But how many substitutions must be made? Replacing two carbonyls is apparently insufficient, as exemplified in the reduction of HOTs catalyzed by [(dppe)(OC)Fe(pdt)Fe(CO)₃]. In this case, HER requires reduction of either a bridging hydride [(dppe)(OC)Fe(pdt)(μ-H)Fe(CO)₃]⁺ (−1.3V vs Fc^+/0) or the parent complex (−1.5 V).²⁸³ Of course, Nature eschews bridging hydrides and does not operate via Fe(I)Fe(0) intermediates, allowing for much milder HER potentials. The first model to exhibit H⁺ reduction electrocatalysis through a terminal hydride was [(OC)(dppv)Fe(pdt)Fe(dppv)(CO)] ([7]), whose treatment with HBF₄·Et₂O affords the Fe(II)Fe(II)(t-H) complex [7(t-H)]⁺, a species that persists for minutes at 20 °C. The hydride reduces at −1.64 V vs Fc^+/0, and catalytic reduction of HBF₄·Et₂O at this potential proceeds with a TOF of 5 s⁻¹. In contrast, the Fe(II)(μ-H)Fe(II) isomer [7(μ-H)]⁺ reduces at −1.84 V with an associated TOF of 3 s⁻¹ (Table 2).

Table 2.

Redox and Catalytic Properties of FeFe Dithiolato Hydrides

hydride complex	abbreviation	δ(1H)/ppm	E1/2a	Ecata	acid used	TOF/s−1	η/V
[(OC)2(Me3P)Fe(pdt)(μ-H)Fe(CO)2(CN)]	[5(μ-H)]	−17.08	−1.57	−1.47	HOTs	–	0.75
[(OC)2(PMe3)Fe(pdt)(μ-H)Fe(CO)2(PMe3)]+	[6(μ-H)]+	−15.3	−1.39b	−1.4	HOTs	–	0.75
[dppv(OC)Fe(adt)Fe(t-H)(dppv)(CO)]+	[17(t-H)]+	−4.2	−1.64b	−1.49	ClCH2CO2H	5000	0.71
[dppv(OC)Fe(adtH)Fe(t-H)(dppv)(CO)]2+	[17(t-H)H]2+	−4.95	−1.4b	−1.11	CF3CO2H	58000	0.51
[dppv(OC)Fe(adt)(μ-H)Fe (dppv)(CO)]+	[17(μ-H)]+	−14.8,c −13.7d	−1.86	−1.72	ClCH2CO2H	20	0.90
[dppv(OC)Fe(adtH)(μ-H)Fe(dppv)(CO)]2+	[17(μ-H)H]2+	−15.5,c −14.3d	−1.77
[dppv(OC)Fe(pdt)Fe(t-H)(dppv)(CO)]+	[7(t-H)] +	−3.5	−1.67e	−1.49e	HBF4·Et2O	5	1.32
[dppv(OC)Fe(pdt)(μ-H)Fe(dppv)(CO)]+	[7(μ-H)] +	−15.6,c −3.5d	−1.8e	−1.78e	ClCH2CO2H	3	0.95

^aRelative to Fc^+/0 in CH₂Cl₂/NBu₄BAr^F₄.

^bIrreversible.

^cSymmetric isomer.

^dUnsymmetric isomer.

^eIn CH₂Cl₂/NBu₄PF₆.

The above example illustrates that, as described in section 3.2.5, terminal hydrides are privileged in that their reductions are both mild (due to the availability of a hydride-free Fe(II) site as an oxidant) and can give rise to higher TOFs (due to t-H⁻ ligands being more reactive than μ-H⁻ ligands). The 1e⁻ reduction of [7(t-H)]⁺ is likely followed by direct protonation of its t-H⁻ ligand to evolve H₂, a process in which no doubly protonated and singly reduced derivative of [7] (analogous to H_ox (H⁺, H⁻)) is formed. A much faster HER would result if a heteroatom base, proximal to H⁻, could serve as a “landing site” for H⁺. The important combination of an electron-rich bimetallic core and a pendant base is present in [dppv(OC)Fe(adt)Fe(dppv)(CO)] ([17], Figure 27), whose terminal hydride derivative [dppv-(OC)Fe(adt)Fe(t-H)(dppv)(CO)]⁺ ([17(t-H)]⁺) is relatively long-lived and catalyzes the mild and rapid HER (−1.49 V vs Fc^+/0, TOF = 5000 s⁻¹) from ClCH₂CO₂H in CH₂Cl₂. The bridging hydride tautomer [17(μ-H)]⁺ is an inferior catalyst: E_cat = −1.72 V, TOF = 20 s⁻¹. Furthermore, a remarkable effect is observed if [17(t-H)]⁺ acts on the stronger acid CF₃CO₂H, whence catalysis occurs at −1.11 V with a TOF of 58 000 s⁻¹. The improvement in both catalytic potential and TOF was attributed to the high activity of the oxidizing, doubly protonated dication [dppv(OC)Fe(adtH)Fe(t-H)(dppv)(CO)]²⁺ ([17(t-H)H]²⁺).

Figure 27.

Highly active HER catalyst [17] operates by different mechanisms that depend on acid strength.²⁵⁸

The attractive catalytic properties of [17] have motivated computational investigations into its activity.¹⁶⁰ Unlike other synthetic models described in this review so far, the two protonations of [17] precede any reduction events, a situation that parallels [FeFe]-H₂ase. Reduction of [17(t-H)H]²⁺ induces a contraction of the N–H⋯H–Fe interaction, such that a spin-delocalized intermediate Fe(1.5)Fe(1.5)(η²-H₂) forms.^95,208 Release of H₂ from this labile species affords a substrate-free Fe(II)Fe(I) species, whose rapid reduction affords the starting Fe(I)Fe(I) complex. This study underscores once more the synergistic roles that terminal hydrides and proximal protic moieties play in efficient HER catalysis.

The FeFe HER catalysts discussed here make use of an electrode to mimic the native [4Fe–4S] electron relay cluster. An electron relay can be incorporated into synthetic models in the form of a redox-active moiety that may serve as an electron source/sink. Pentacarbonyl [20] features such a ligand—a phosphole—and the mechanism of its HER has been studied in CH₂Cl₂ solution.²⁸⁴ Precatalyst [20] enters the HER cycle upon double protonation and 2e⁻ reduction (−1.44 V vs Fc^+/0, Figure 28). An ECCE mechanism affords the active species [20(μ-H)H], which takes the form of a mixed-valence bridging hydride featuring a singly reduced, protonated ligand. Catalytic reduction of [Et₃NH]BF₄ occurs at −2.0 V, at which [20(μ-H)H] undergoes a second 2e⁻ reduction and protonation in an ECE process, resulting in an Fe(I)(μ-H)Fe(I) species of formula [20(μ-H)(H)₂]⁻. This species is indicated by spectroelectrochemical and DFT studies to adopt a structure in which an Fe–S bond is cleaved and a doubly protonated, doubly reduced ligand is present. Intermediate [20(μ-H)(H)₂]⁻ is the catalyst resting state; its protonation and release of H₂ is rate-determining (k_cat = 10⁵ M⁻¹ s⁻¹). Overall, the phosphole ligand serves as a reservoir of reducing equivalents, enhancing PCET in a system that is both O₂-tolerant and very fast. This system also catalyzes reduction of H₂SO₄ at −0.66 V at TOF = 70 000 s⁻¹ (k_cat = 3.5 × 10⁴ M⁻¹ s⁻¹ in acid-independent regime). The value of the redox-active ligand is emphasized in that catalysis by [20(μ-H)(H)₂]⁻ is so fast even though its mechanism is “handicapped” by operation via bridging hydrides.

Figure 28.

HER catalytic cycle proposed for [20], which bears a redox-active phosphole ligand.²⁸⁴

3.2.8. Hydrides from H₂

Mimicking the catalytic activity of [FeFe]-H₂ases requires a synthetic FeFe complex to mediate not only the evolution of H₂, but also its oxidation. Such a conversion is of interest for many reasons, a principle one being the need for cost-effective alternatives to Pt electrodes in fuel cells. As noted in section 2.1, the heterolysis of H₂ is challenging in that free H₂ is only weakly acidic (pK_a(THF)⁴⁴ ≈ pK_a(MeCN) ≈ 50),⁴⁰ a hurdle that can be overcome by its binding to a cationic metal site such as Fe(II). While successes in catalytic H₂ oxidation are less numerous than those in the HER, several synthetic Fe catalysts can cleave H₂. Some of these bear little structural resemblance to the [2Fe] subsite; these “bioinspired” models have been reviewed elsewhere.^24,26,281 The focus here is on “biomimetic” systems—diiron thiolates in particular—that either catalytically oxidize H₂ or convert to a hydride form under H₂. The ideal biomimetic system would mimic H_ox, the Fe(I)Fe(II) enzyme state that binds and cleaves H₂, using it as a source of H⁻.

Early work on H₂ activation by [FeFe]-H₂ase models did not result in oxidation of H₂ or its conversion to hydrides. Rather, it was shown that [(Me₃P)(CO)₂ Fe(pdt)(μ-H)Fe-(CO)₂(PMe₃)]⁺ ([6(μ-H)]⁺), a Fe(II)(μ-H)Fe(II) complex, mediates H/D scrambling under D₂ and ambient light (Figure 29).²⁸⁵ The process is rather slow, with exchange of the hydride in CH₂Cl₂ solution under 7–8 bar of D₂ requiring 4 weeks. In a related experiment, a H₂/D₂ mixture, exposed to direct sunlight and [6(μ-H)]⁺ in CH₂Cl₂, afforded HD. H/D exchange was markedly slower in the absence of light, and did not occur at all when the coordinating solvent CH₃CN was used. It was thus proposed that light opens a coordination site by inducing either reversible decarbonylation or isomerization of the bridging hydride to a terminal hydride (the latter possibility was addressed in section 3.2.4).²⁵⁶ Further studies of H/D exchange with [6(μ-H)]⁺ or [(Me₃P)(CO)₂Fe(pdt)(μ-SMe)₂Fe(CO)₂(PMe₃)]⁺ indicated that both complexes catalyzed D₂/H₂O exchange, a characteristic reaction of H₂ases. The basic requirements for H/D exchange catalysis could thus be formulated: one needs a pentacoordinate Fe(II) Lewis acid to bind (H/D)₂ and an accessible Brønsted base to deprotonate coordinated (H/D)₂. These roles are possibly fulfilled by the μ-H complex and water, respectively.

Figure 29.

H/D scrambling mediated by [6(μ-H)]⁺. This complex may undergo either decarbonylation or μ-H → t-H isomerization such that a vacant site for D₂ binding is accessible.²⁸⁵

While the [2Fe] subsite extracts a H⁻ ligand from the H₂ substrate, early models could not replicate this activity. Photolytic decarbonylation of [(dppv)(CO)Fe(edt)Fe(CO)₃] ([9]) affords an all-important free Fe site, yet the unsaturated intermediate, instead of heterolyzing H₂, oxidatively adds it through a very slow process.²⁴⁸ As mentioned above, oxidative addition of H₂ is not observed for H₂ases, which operate exclusively by heterolysis. It is also conceivable that metals extract H⁺ (rather than H⁻) from H₂, as in the complete conversion of [6] to its hydride [6(μ-H)]⁺ under H₂ (1 atm) in the presence of the H⁻ acceptor B(C₆F₅)₃. This FLP can be viewed as the reverse of the enzymatic system, in which the amine accepts H⁺ and Fe^d receives H⁻.

Initial attempts at activating and oxidizing H₂ with synthetic [FeFe]-H₂ase models utilized compounds in their Fe(I)Fe(I) or Fe(II)Fe(II) forms, neither of which is relevant to the biological mechanism, which proceeds through an Fe(I)Fe(II) state. As discussed in section 3.1, the mixed-valent core of H_ox exhibits a rotated Fe^d coordination geometry exposing a site on an electrophilic Fe center to bind H₂ (or the inhibitor CO). More recent synthetic studies have thus involved the preparation and reactivity studies of H_ox models. While the first such complexes, obtained by oxidation of Fe(I)Fe(I) H_red models, were only observed transiently in spectroelectrochemical and stopped-flow studies,²⁸⁶ a series of stable H_ox models appeared following reports of [(CO)₂(PMe₃)Fe(pdt)Fe(CO)₂(IMes)]⁺(IMes =1,3-bis(2-mesityl)imidazolylidine) and [Me₃P(CO)₂Fe(edt)-Fe(CO)(dppv)]⁺.^230,287 The products, readily prepared by oxidation of Fe(I)Fe(I) parent complexes with Fc⁺, reproduced many features of H_ox. For example, the [(CO)₂(Me₃P)Fe] fragment in [(CO)₂(Me₃P)Fe(pdt)Fe(CO)₂(IMes)]⁺ adopts a rotated structure with a semibridging CO ligand. Meanwhile, [Me₃P(CO)₂Fe(edt)Fe(CO)(dppv)]⁺ mimics H_ox behavior in that it binds CO.

A more complete H_ox model also contains an azadithiolate ligand, [Me₃P(CO)₂Fe(adt^Bn)Fe(CO)(dppv)]⁺ ([21]⁺). In considering the structure of [21]⁺, one notices that it features not only the mixed-valent core but also a H⁺ relay poised for H₂ heterolysis.^200,288,289 This prototype was the first H_ox model to activate H₂ and afford a hydride product, albeit in a bridging form and at very high H₂ pressure (Figure 30).²⁹⁰ The reaction mechanism is not obvious a priori, but could be proposed based on the redox potentials of the relevant species. A likely scenario involves [21]⁺ binding and heterolyzing H₂ to afford the mixed-valence ammonium hydride [21(t-H)H]⁺. An additional [21]⁺ complex may then oxidize and deprotonate (or abstract H^• from) [21(t-H)H]⁺. The slow nature of the H₂ activation is likely due to a combination of weak H₂ binding, as well as the necessity for two complexes to act on a single H₂ molecule, one as an H₂ binder and the other as an oxidant.

Figure 30.

Stoichometric activation of H₂ with [21]⁺ in the presence²⁹¹ and absence²⁹⁰ of an external oxidant.

The above result prompted investigation of a potentially catalytic and biomimetic route, in which external base and oxidant were employed.²⁹¹ In fact, [21]⁺ effects H₂ heterolysis in the presence of the mild oxidant [FeCp*₂]⁺ (Fc*⁺) as well as the (optional) weak base P(o-tolyl)₃. The reaction proceeds over the course of hours under only 2 atm H₂ at 0 °C to quantitatively afford [21(μ-H)]⁺ and [HP(o-tolyl)₃]⁺ (k_obs = 2.2 × 10⁻⁵ s⁻¹ at 0°C). Given that the rate-determining step is likely H₂ binding, it was thought that the stronger electrophile [(OC)₃Fe(adt^Bn)Fe-(CO)(dppn)]⁺ ([22]⁺, dppn = 1,8-bis(diphenylphosphino)-naphthalene) might more effectively cleave H₂. Under identical conditions (although using the slightly stronger oxidant Fc⁺) [22]⁺ heterolyzes H₂ 20 times faster (k_obs = 4.8 × 10⁻⁴ s⁻¹ at 0 °C) than [21]⁺ and ~10⁴ times faster than [21]⁺ in the absence of oxidant at room temperature. The kinetic isotope effect measured for H₂/D₂ heterolysis by [22]⁺ and Fc⁺ was inverse (KIE = 0.8).

The above results bring to light the three requirements for H₂ activation. Not only does the system require a Lewis acidic Fe site and a Brønsted basic amine, but it also requires an electron sink. In Nature, these roles are served by Fe^d, adt²⁻, and the proximal [4Fe–4S] cluster, respectively. The first synthetic model addressing these three design principles was an analogue of [21]⁺ in which the PMe₃ ligand was replaced by the redox-active phosphine Cp*Fe(C₅Me₄CH₂PEt₂) (abbreviated FcP*).⁶⁶ In this way, the complex [(FcP*)(CO)₂Fe(adt^Bn)Fe(CO)(dppv)] ([23], Figure 31) preserves the electronic properties of [21], as the redox auxiliary is insulated by a CH₂ spacer group. The triiron complex reversibly oxidizes at its Fe(I)Fe(I) core (−700 mV vs Fc^+/0) and at the FcP* metalloligand (−393 mV) to afford the unique H_ox model [23]²⁺ featuring Fe(I), Fe(II), and Fe(III) centers. Encouragingly, [23]²⁺ exhibited increased affinity toward CO, such that at low temperature the diamagnetic triferrous complex [23(CO)]²⁺ forms quantitatively. While the inclusion of redox cofactor Fc*P resulted in only a 2-fold increase in H₂ activation rate relative to the [21]⁺/Fc*⁺ system, [23]²⁺ catalytically oxidizes H₂ to protons and electrons. Exposure of [23]²⁺ to H₂ in the presence of excess Fc⁺ and P(o-tolyl)₃ resulted in complete conversion to Fc and [HP(o-tolyl)₃]⁺, although at an extremely slow rate (TOF = 0.4 h⁻¹). However, the triiron system is bidirectional in that the H_red model [23] is a catalyst for the HER,⁹⁸ such that its combination with [H(OEt₂)₂]BAr^F₄ (10 equiv) and Fc* (5 equiv) resulted in three turnovers during a 3 h experiment. The functional [FeFe]-H₂ase model [23] represented a significant step forward, although its activity does not approach that of the native system. The markedly lower activity of [23]²⁺ is thought to result from its low affinity for H₂. DFT calculations on [23]²⁺ indicate that H₂ binding requires the relatively bulky benzyl group to shift to an energetically unfavorable conformation to allow H₂ access to an open Fe site.¹⁸⁸

Figure 31.

Catalytic cycle proposed for [23]-mediated H₂ evolution and oxidation (clockwise and counterclockwise, respectively).^66,292

Catalytic H₂ oxidation and formation mediated by the triiron system are plagued by the formation of μ-H complexes, which are more inert than their active t-H tautomers. The bridging species [23(μ-H)]⁺ is not a HER catalyst, although it maintains some H₂ oxidation activity. This deactivation is apparent as rates for catalytic reactions are much lower than those for stoichiometric reactions, implying sluggish deprotonation of the μ-hydride might restrict catalytic turnover.

The oxidation of H₂ by H_ox models does not require the acid–base cofactor to take the form of an azadithiolate, nor does it require the redox cofactor to be attached to the FeFe core. This is demonstrated with [(OC)₃Fe(pdt)(μ-dppf)Fe(CO)] (dppf =1,1′-bis(diphenylphosphino)ferrocene), which catalyzes H₂ oxidation at the surface of an electrode in the presence of pyridine as the base.²⁹³ A separate body of work on azadiphosphines²⁹⁴ culminated in the preparation of [(OC)₃Fe-(pdt)Fe(PNP)(CO)] ([24], Figure 32), whose H⁺ relay Ph₂PCH₂N(ⁿPr)CH₂PPh₂ (PNP)²⁹⁵ causes bridging hydride [24(μ-H)]⁺ to undergo rapid H/D exchange with CH₃CO₂D. Importantly, these hydrides are almost instantaneously deprotonated by relatively weak bases such as aniline and water, suggesting μ-H complexes might undergo turnover during H₂ oxidation.

Figure 32.

Catalytic cycle proposed for H₂ oxidation mediated by PNP complex [24(μ-H)]⁺.²⁹⁵

It was found that [24]⁺ catalytically oxidizes H₂ in the presence of excess Fc⁺ and P(o-tolyl)₃, illustrating that the redox auxiliary need not be directly appended to the FeFe core to achieve catalytic turnover.²⁹⁶ However, the proximity of the H⁺ relay is absolutely necessary, as replacement of PNP’s ⁿPrN group with a nonbasic CH₂ fragment affords a catalytically inactive species. Oxidation of H₂ by the [24]⁺/Fc⁺ system (TOF = 0.6 h⁻¹ at 25 °C) is only marginally greater than the triiron system (0.4 h⁻¹), suggesting that the extremely low rates of catalytic H₂ oxidation by model complexes are primarily a result of their inherently poor affinity for H₂.

3.3. Concluding Remarks and Future Challenges

Considerable efforts have gone into synthetic modeling of the [FeFe]-H₂ases, no doubt targeted at reproducing the activity of these remarkably efficient enzymes. The ease with which “first generation” models (i.e., hexacarbonyls) can be prepared has led the majority of synthetic and catalytic studies to focus on these systems. Yet their structural and functional resemblance to the active site is inferior to that of more substituted, electron-rich systems. The latter can operate biomimetically through terminal hydride intermediates and at oxidation states matching those proposed for [FeFe]-H₂ase. Synthetic studies concerned with fine-tuning the first and second coordination spheres by CO substitution and acid–base incorporation, respectively, have been rewarded with HER activities comparable to the enzymatic catalysts, albeit at overpotentials greater than 0.5 V. The quest for fully functional bidirectional models has led to the development of sophisticated models featuring not only a pendant base proximal to an electron-rich FeFe core, but also a redox cofactor.²⁹⁷ These complete models for the H-cluster mediate H₂ oxidation, yet they do so at rates several orders of magnitude less than the native system.

Given the successes and shortcomings of functional models, what avenues seem promising from the vantage? Certainly the [FeFe]-H₂ase active site is subject to an array of weak interactions that are difficult to replicate in small molecule models. Further, the highly evolved mechanisms allowing for H₂ and H⁺ transport to and from the Fe^d site cannot be found in purely synthetic systems. Whereas chemists are jubilant in replicating the role of the first and even the second coordination spheres, mutation studies make it clear that the third coordination sphere is also crucial to high catalytic rates.²⁹⁸

Despite the challenges inherent to synthetically approximating the protein environment around the active site, there are some specific shortcomings in current models that restrict their activities. In the case of H_ox models, these are poised to activate H₂ yet bind it rather weakly. This relative unreactivity of H_ox models is also reflected in their weak binding of CO, a strong inhibitor of [FeFe]-H₂ase. Until synthetic models show higher affinity for CO, they are unlikely to exhibit high rates for H₂ oxidation.²⁹⁷ Remarkably little emphasis has been placed on a rather important design aspect—matching redox potentials of synthetic models to those of [FeFe]-H₂ases, which must operate between −266 and −296 mV vs NHE. Most desirable are diiron dithiolates with reversible electrochemistry around −900 mV vs Fc^+/0 mimicking the Fe(I)Fe(II/I) H_ox/H_red couple. This criterion is satisfied by only a handful of models, including bis(dppv) complexes. While H₂ases operate at low H₂ pressures, the most sought after synthetic catalysts are those that can handle higher pressures by exhibiting more cathodic redox waves. Such reducing models are yet to be reported, and the cationic nature of most systems contrasts the tetra- and trianionic H_red and H_ox states, respectively (assuming adt²⁻ is not doubly N-protonated). Finally, one obvious problem with models is that reduced states almost invariably adopt nonbiomimetic structures, not unlike the age-old [(OC)₃Fe(SR)₂Fe(CO)₃] species. In the enzyme, the [2Fe] subsite adopts a preorganized rotated structure necessary for catalysis. Reproducing this is naturally a difficult task as CO and CN⁻ are free to move and are not functionalizable. But by employing bulky dithiolates and unsymmetrical substitution, the rotated structure can be stabilized in select cases.^266,267,299 It will be necessary to target more conformationally rigid models to enforce this geometry and prevent bridging hydride formation.

It is anticipated that efforts in further tuning biomimetic FeFe models will bring us closer to the enzyme activity. This goal may also be approached from an entirely different angle, and much success has been reported in research concerning bioinspired systems, whose structures are considerably different from that of the native catalyst. Among these is a remarkably active monoiron PNP (azadiphosphine) complex,³⁰⁰ which merges the pendant base motif of [24] into an otherwise rather different structure that nevertheless serves as an efficient bidirectional catalyst. The discussion concerning [FeFe]-H₂ases and their models ends here, and we now consider the [NiFe]-H₂ases in section 4. As will become apparent, despite being biologically unrelated, these two classes of H₂ase share similar mechanistic themes that make their study even more instructive.

4. [NiFe]-H₂ASES

4.1. Enzyme Structure and Function

The [NiFe]-H₂ases are older³⁰¹ and more abundant and O₂-tolerant³⁰² than the [FeFe]-H₂ases. The former, more robust family of catalysts is expressed in bacteria, but unlike [FeFe]-H₂ases is also found in archaea and cyanobacteria (though not in lower eukaryotes).¹⁹⁰ The [NiFe]-H₂ases are heterodimeric (63 + 29 kDa),^303–305 with the larger subunit housing an active site featuring a redox-active Ni(S-Cys)₄ center, of which two ⁻S-Cys ligands bridge to an organometallic, redox-inactive Fe-(CN)₂(CO) fragment (Figure 33).

Figure 33.

X-ray structure of the [NiFe]-H₂ase active site from Desulfovibrio vulgaris Miyazaki F (PDB code 4U9H).⁸⁹ Nonionizable H atoms are omitted for clarity.

A striking aspect of the active site is the unusual arrangement of S donor atoms around the Ni center. While (Cys546-S)Ni(S-Cys84) defines an almost straight line (176.3°), the (Cys549-S)Ni(S-Cys81) angle is considerably smaller (107.9°) such that Ni adopts a seesaw geometry approaching that of SF₄. As for the (Cys-S)₂Fe(CN)₂(CO) center, it is best described as a distorted square pyramid in which CO occupies the apical site. The structural parameter τ, which ranges from 0 to 1 for the square-pyramidal and trigonal bipyramidal extremes, is in this case equal to (173.7° − 162.7°)/60° = 0.18.³⁰⁶ The metals are separated by a distance (2.57 Å) comparable to the sum of the Ni and Fe covalent radii (1.24 Å + 1.32 Å = 2.56 Å).²¹⁹ The (Cys-S)₂Ni(μ-S-Cys)₂Fe(CN)₂(CO) core is common to all enzyme states, and while the metrics described here are for the Ni-R form (vide infra), they are very similar to those determined experimentally or computationally for other states. One aspect that does change between redox states is the nature of a third coordination site between the metals. When occupied, this site completes the distorted square-pyramidal and octahedral ligand arrangements about Ni and Fe, respectively. The reduced, active forms of [NiFe]-H₂ase have the site either vacant or home to a bridging hydrido ligand. These catalytically relevant states are presented in Figure 34.

Figure 34.

Catalytic cycles proposed for [NiFe]-H₂ase. The electronic spin (S) and electron count of the NiFe cores (the sum of 3d electrons and electrons in metal–ligand bonds) are presented below each structure.³⁰⁷ A H₂O ligand may be weakly bound to Ni in Ni-SI_a.

The name “Ni-R” is given to the most reduced enzyme state. In a biophysical context, the “most reduced” species has the greatest number of electrons and prevails at low redox potentials (“reduced” does not necessarily refer to low metal oxidation states, as might be understood by synthetic chemists). Although (t-H)Ni(II)(μ-H)Fe(II) dihydrides^181,308 and Ni(II)Fe(II)(η-H₂) dihydrogen-bound formulations have been proposed based on computations,³⁰⁹ the current picture for Ni-R is that of a 34e⁻ Ni(II)(μ-H)Fe(II) core. This is further believed to be low-spin,¹⁰³ despite some earlier computational predictions to the contrary.^310,311 While the identities of the active site heavy atoms have been agreed upon for some time, only recently has X-ray crystallographic data been of sufficient resolution to allow unambiguous location of the μ-H⁻ ligand.⁸⁹ The hydrido binds Ni (1.58 Å) somewhat tighter than Fe (1.78 Å), and its presence has since been verified by ⁵⁷Fe NRVS, with the observation of a characteristic δ_NiHFe bending vibration at 675 cm⁻¹.^103,312

IR spectroscopy indicates the existence of three isoelectronic Ni-R subspecies, whose pH-dependent speciation indicates that they differ in their protonation states. More basic than bridging thiolates, the terminal thiolates are the most obvious H⁺ acceptors. The highest-resolution X-ray diffraction data for Ni-R have been modeled with an active site featuring protonated Cys546 (Figure 33).⁸⁹ However, as the difference map has electron density on either side of the Cys546 S atom, an argument can be made for S to be unprotonated and undergo anisotropic motion perpendicular to the Ni–S bond. Indeed, aside from the terminal thiolates, an Arg509 residue proximal to Ni represents another viable H⁺ relay (vide infra).³¹³ In any case, the situation is much clearer when it comes to electron transport, with a chain of 4Fe–4S, 3Fe–4S, and 4Fe–4S clusters clearly defining a tunneling path to and from the active site.

The stepwise oxidation and deprotonation¹⁰⁷ (−436 mV vs NHE, pH 7.4)³¹⁴ of EPR-silent Ni-R affords the 33e⁻ Ni-C state (Ni(III)(μ-H)Fe(II), “catalytic intermediate”), perhaps the most well-characterized of the active forms. This EPR-active (g = 2.21, 2.15, 2.01) low-spin (S = 1/2) species has been the subject of several X-ray structural determinations, although none with sufficient resolution to unambiguously locate the μ-H⁻ ligand. Rather, this moiety has been confirmed by the pulsed EPR methods HYSCORE and ENDOR, with signals for the paramagnetic Ni(III) site featuring anisotropic spin coupling to a ¹H nucleus.^96,315 The five anionic ligands surrounding the Ni center undoubtedly play a role in taming the otherwise oxidizing +III state, whose polarizing nature results in terminal cysteinato ligands almost certainly being deprotonated,³¹⁶ in contrast to the situation with Ni-R.

Oxidation and deprotonation (−375 mV vs NHE, pH 7.4)³¹⁴ of Ni-C gives Ni-SI_a (“silent active”), a low-spin hydride-free 32e⁻ Ni(II)Fe(II) state. This electron count does not take into account a weakly Ni-bound H₂O ligand,³¹⁶ upon release of which an electrophilic core is exposed to heterolyze H₂. Yet before this transformation is described, the potentially important 33e⁻ Ni-L (Ni(I)Fe(II), “light”) form, intermediary between Ni-C and Ni-SI_a, should also be acknowledged. Originally observed by low-temperature irradiation of Ni-C, the isoelectronic Ni-L is a tautomer in which reductive elimination of the proton allows formation of a Ni–Fe bond,¹⁷⁹ although with little change in the internuclear separation.⁸⁵ ENDOR and HYSCORE point to the proton residing on a terminal Cys546,³¹⁷ with prolonged irradiation leading to H⁺ departure from the active site.³¹⁸ Should interconversion between Ni-C and Ni-SI_a involve concerted PCET, then Ni-L need not be invoked in mechanisms. Computational studies had proposed involvement of Ni-L in catalysis,³⁶ but this was contraindicated by the apparent need for light irradiation that would not necessarily be available in vivo. However, Ni-L formation has been demonstrated to occur in the dark³¹⁹ as well as during catalysis, although its presence is fleeting.^107,307,320 The divergent observations regarding the stability of Ni-L have been attributed to the redox state of the proximal [4Fe–4S]^2+/+ cluster.³²¹ When this cluster is oxidized, Ni-C converts to Ni-SI_a, either directly through PCET, or by tautomerization to Ni-L and donation of an e⁻ to the cluster. If instead the cluster is in its reduced state, the intermediate Ni-L may persist. Analogous to these redox effects, pH also has an influence on the Ni-C to Ni-SI_a conversion. At low pH, this reaction is slowed, presumably due to the protonation of the residue that must accept a proton from the Ni(III)HFe(II) core of Ni-C.³²² While the participation of Ni-L in turnover is still a point of contention, selected data for this state, as well as the three confirmed catalytic states Ni-R, Ni-C, and Ni-SI_a, are given in Table 3.

Table 3.

Structural and Spectroscopic Parameters of Active [NiFe]-H₂ase States

state	core	rNi–Fe/Å	rNi–H/Å	rFe–H/Å	νCO/cm−1	νCN/cm−1	g
Ni-R	Ni(II)(μ-H)Fe(II)	2.57a	1.58a	1.78a	1944	2074, 2061	–
Ni-C	Ni(III)(μ-H)Fe(II)	2.57b	1.59b	1.70b	1961	2085, 2074	2.21, 2.15, 2.02
Ni-L	Ni(I)Fe(II)	2.56c	–	–	1911	2062, 2048	2.28, 2.11, 2.05
Ni-SIa	Ni(II)Fe(II)	2.61d	–	–	1943	2086, 2075	–

^aDistances from X-ray structure.⁸⁹

^bDFT-calculated values.³¹⁶

^cDFT-calculated values.¹⁷⁹

^dDFT-calculated value from a structure bearing a H₂O ligand and protonated Cys ligand. IR bands from Desulfovibrio vulgaris Miyazaki F.³¹⁴ EPR g-values from Allochromatium vinosum.³²³

With the active states described, one can now complete a catalytic cycle by activation of H₂ with Ni-SI_a. Upon dissociation of any aquo ligand present, Ni-SI_a features seesaw Ni(II) and square-pyramidal Fe(II), both of which are electrophilic 16e⁻ metal centers that may conceivably bind and cleave H₂. On general grounds, one might consider this to be more likely to occur at Fe(II), a middle transition metal with just enough nuclear charge to impart electrophilicity while still retaining π-basicity for synergistic H₂ binding. Yet the consensus is that while both metals may participate, Ni does the bulk of the work, not only because it is optimally positioned near the end of a hydrophobic H₂ transport channel,³²⁴ but also because its unusual geometry is particularly reactive, as indicated by DFT calculations.^310,325 Recent work has suggested that the orientation of Cys ligands has a strong influence on spin densities at Ni and S.^326,327 Furthermore, several computational studies on truncated active sites indicate that rotation of terminal Cys modulates the relative stability of low- and high-spin Ni,^310,328,329 although the protein environment may also have nontrivial effects on Ni electronic structure.^325,330 In the case of a Ni(II)Fe(II) core such as that in Ni-SI_a, one would expect square-planar and tetrahedral Ni to give rise to S = 0 and 1 states, respectively. The seesaw Ni(S-Cys)₄ coordination present in all states is rigidly poised between these two typical geometries, and while Ni-SI_a has no EPR signal, its singlet–triplet gap is not large.³²⁹

The conversion of Ni-SI_a to Ni-R involves the reactive Ni(II)Fe(II) core being “tamed” upon receiving the μ-H⁻ ligand, with the fate of the other heterolysis product, a proton, still unclear. A substantial body of computational work³³¹ as well as the high-resolution Ni-R X-ray structure⁸⁹ implicates Cys546, although questions have arisen as to whether the terminal thiolate is sufficiently basic²⁴ to deprotonate the as-yet-unobserved Ni(II)(η²-H₂)Fe(II) intermediate. In an alternative mechanism, a more basic Arg509 serves as the H⁺ acceptor, much as the azadithiolate cofactor does for [FeFe]-H₂ase (Figure 35). Notably, Arg509 is highly conserved, and its basic guanidine group (pK_a ≈ 13.8)³³² is only 4.5 Å away from Ni.³¹³

Figure 35.

Alternative mechanism for H₂ activation by [NiFe]-H₂ase, in which a proximal Arg residue (rather than a terminal Cys ligand) relays H⁺ to and from the active site.³¹³

While there is debate as to exactly how [NiFe]-H₂ases process H₂, there is little doubt that they do it incredibly quickly. A graphite-supported film of [NiFe]-H₂ase isolated from Allochromatium vinosum exhibited diffusion-limited TOF values for H₂ oxidation (>6000 s⁻¹, pH 7, 45 °C, 0.2 V vs SHE). The corresponding current density was similar to that of a Pt electrode, although for the latter case this could be achieved at lower potentials of −0.4 V.³³³ Given the large footprint of [NiFe]-H₂ases, their activity must be superior to that of the individual Pt sites, although the accessibility of the active site to the substrate is cited as a source of overpotential.³³⁴ Alternatively, the H₂ oxidation can be driven using benzylviologen as a sacrificial electron acceptor, as demonstrated with [NiFe]-H₂ase from Desulfovibrio gigas (2200 s⁻¹, 30 °C, −358 mV).³³⁵ Many examples of [NiFe]-H₂ase are bidirectional, and the same Desulfovibrio gigas enzyme also catalyzes H⁺ reduction (640 s⁻¹, 30 °C, methylviologen at −446 mV).³³⁵

The activities of [NiFe]-H₂ases can vary greatly. These differences may arise from several factors, an important one being the strength with which the protein binds H₂. Stronger H₂ binding is exhibited by enzymes expressed for H₂ uptake, such as the Ralstonia [NiFe]-H₂ases. Their H₂ production rates are modest (70 s⁻¹, pH 5.5, 40 °C, −0.45 V vs SHE), yet they can operate in the presence of O₂. Indeed, the strong binding of H₂ makes it an even more potent inhibitor than O₂ and CO.³³⁶ Notably, O₂ and CO irreversibly deactivate [FeFe]-H₂ase and Pt catalysts, respectively.

Inhibition of the HER by H₂ can be quantified by considering the Michaelis constant K_M,^337,338 which is related to how much the activity is attenuated by a given H₂ concentration.³³⁹ In the case of [NiFe]-H₂ases, K_M values are typically ~10⁻⁷−10⁻⁴ M, whereas those for [FeFe]-H₂ase are greater and fall in the range ~10⁻⁴−10⁻³ M.³³⁹ Thus, while trace H₂ can inhibit the HER in many [NiFe]-H₂ases, the [FeFe]-H₂ases bind H₂ more weakly and are less prone to product inhibition. Such measurements are consistent with the typical biasing of [FeFe]-H₂ases toward H₂ formation, with [NiFe]-H₂ases usually being expressed to oxidize H₂. Interestingly, under high pressures of H₂, certain [FeFe]-H₂ases can still mediate H₂ oxidation faster than [NiFe]-H₂ases.

As has been described above, hydrides are central to the catalytic activity of [NiFe]-H₂ases, the only H₂ases for which hydride-bearing species have been confirmed. Further comparison of the H₂ases points to the contrasting bridging hydride cores of Ni-R and Ni-C with respect to the terminal hydride proposed for [FeFe]-H₂ase. Lacking azadithiolate, the [NiFe]-H₂ases shuttle protons around entirely with amino acid residues (and a H-bonded H₂O network) and feature a dinuclear core more strongly anchored to the protein scaffold. Relative to the diiron enzymes, one might say that the [NiFe]-H₂ases make up for in robustness what they lose in activity. The heteronuclear active sites represent attractive targets for synthetic modeling studies, which, through reproducing [NiFe]-H₂ase structure and spectroscopy (Table 3), aim to reproduce enzyme function. Several reports have described NiFe complexes strongly resembling the [NiFe]-H₂ase active site. These efforts are summarized in section 4.2, with particular emphasis placed on complexes bearing hydride ligands.

4.2. [NiFe]-H₂ase Synthetic Modeling

4.2.1. Hydrides from H⁺

Synthetic modeling of the [NiFe]-H₂ases, relative to that of the [FeFe]-H₂ases, is a much younger area of study.^340,341 The low-spin Fe(CN)₂(CO) fragment in the former is reminiscent of the distal Fe(CN)(CO)₂ site in the latter, and has similarly caught the eye of synthetic chemists. Certain difficulties associated with preparing heterobimetallics led early [NiFe]-H₂ase modeling to be directed toward mononuclear Ni or Fe models of their respective subsites. With respect to metal hydrides, particular attention was paid to Fe–CN–CO systems, including [Fe(CN)₂(CO)₃]²⁻. This dianion undergoes protonation, not at the N atoms, but rather at Fe(0) to give three isomers of the octahedral Fe(II) product [HFe(CN)₂(CO)₃]⁻, one of which is shown in Figure 36.³⁴² With four of the six ligands in the Ni-R state of the enzyme, this hydride remains an attractive synthetic module, and its potential aqueous solubility may lend itself to reconstitution studies akin to those of [FeFe]-H₂ases.

Figure 36.

Initial work on hydride-containing [NiFe]-H₂ase models focused on the Fe subsite, whose Fe(CO)(CN)₂H fragment is replicated by [25]⁻.³⁴²

Two CO ligands in [HFe(CN)₂(CO)₃]⁻ are liberated during its conversion to [Fe(dppv)(CO)(CN)₂H]⁻ ([25]⁻), a product that replicates the enzyme structure well, with the diphosphine dppv representing a prosthesis for the bridging Cys thiolates. IR spectroscopy (ν_CO = 1936 cm⁻¹; ν_CN = 2087, 2080 cm⁻¹) indicates the monocarbonyl product to be electronically similar to the enzyme subsite, although the ν_CN values are lower than those for Ni-R, whose CN⁻ ligands participate in H-bonding.

Hydrides containing both Fe and Ni are rare, and at one point examples of these were limited to clusters such as the μ³-H⁻-containing [NiFe₃H(CO)₁₂]⁻ metallotetrahedron³⁴³ or the μ-H⁻-bridged [NiFe₅HN(CO)₁₄]²⁻ and [Ni₂Fe₄HN(CO)₁₄]²⁻ octahedra bearing interstitial N³⁻.³⁴⁴ Of much greater relevance to the [NiFe]-H₂ase active site was the report of the first NiFe thiolato hydride, [(dppe)Ni(pdt)(μ-H)Fe(CO)₃]⁺ ([26(μ-H)]⁺, Figure 37).³⁴⁵

Figure 37.

Synthesis of a Ni(I)Fe(I) thiolate [26],⁶³ which can accept H⁺ either from acid or from H₂ heterolysis to afford the hydride derivative [26(μ-H)]⁺.³⁴⁵

Yet before discussing hydride [26(μ-H)]⁺ and its derivatives, one must consider the low-valent conjugate bases from which they are prepared. The Ni(I)Fe(I) species [(dppe)Ni(pdt)Fe-(CO)₃] ([26]),³⁴⁶ diamagnetic on account of its metal–metal bond (2.467 Å),³⁴⁷ is a 34e⁻ complex similar to the homobimetallic [dppe(OC)Fe(pdt)Fe(CO)₃]. Comproportionation of the Ni(II) complex [(dppe)Ni(pdt)] with an Fe(0) source such as [Fe(benzylideneacetone)(CO)₃] or [Fe₂(CO)₉] affords [26], which was initially purported to be unstable,³⁴⁶ but is in fact rather robust. A more generalizable synthesis for [26] was later developed in which [(diphosphine)Ni(dithiolate)] and cis-[FeI₂(CO)₄] are combined, with 2e⁻ reduction of the resulting iodide [(diphosphine)Ni(dithiolate)(μ-I)Fe(CO)₃]⁺ affording the neutral species [(diphosphine)Ni(dithiolate)Fe-(CO)₃].³⁴⁸ Such complexes reproduce the Ni(SR)₂Fe “butterfly” and apical CO ligand at the [NiFe]-H₂ase active site, although differences in the coordination spheres of the synthetic and native systems are not insignificant. Much like the modeling of [FeFe]-H₂ases, in which phosphines are extensively used as prostheses for CN⁻ groups and the Cys-S–[4Fe–4S] metalloligand, the diphosphine in [(diphosphine)Ni(dithiolate)Fe-(CO)₃] is a substitute for terminal Cys-S⁻ ligands. Another significant difference can be found in the set of diatomic ligands, with the three CO ligands in these models contrasting Nature’s choice of one CO and two CN⁻ groups. The stereoelectronic consequences of these changes will be discussed in the context of the first generation of synthetic NiFe hydrides.

The finding that the archetypal Ni(I)Fe(I) complex [26] sustains protonation to afford a well-defined hydride product represented a significant advance in the field of [NiFe]-H₂ase modeling. This protonation reaction was found to proceed via a 2e⁻ mixed-valent conformer, with dppe rotation resulting in a square-planar environment around Ni (see section 4.2.3).¹⁶⁰ Oxidized (divalent) Ni is now strongly stabilized, with concomitant cleavage of the NiFe bond resulting in Fe being in a reduced state. Overall, this redox isomerization closely parallels the “rotation” of Fe(I)Fe(I) to afford a reactive Fe(II)Fe(0) species. In the present case, DFT calculations indicate that the latent Ni(II)Fe(0) isomer is ~10⁸ times more basic than the Ni(I)Fe(I) form and is geometrically preorganized to form hydride [26(μ-H)]⁺.¹⁶⁰ With its Ni(II)(μ-H)Fe(II) core, the robust, air-stable product [26(μ-H)]⁺ represented the first model for the Ni-R enzyme state. However, Ni-R does not form by protonation of a low-valent species—indeed no [NiFe]-H₂ase states feature a Ni(I)Fe(I) or Ni(II)Fe(0) core. The 26 to [26(μ-H)]⁺ reaction is rather a low-valent analogue of the Ni-L to Ni-C conversion, in which oxidative addition of a H⁺ transforms Ni(I)Fe(II) to Ni(III)(μ-H)Fe(II). Yet [26(μ-H)]⁺ is nevertheless an important model with instructive characterization. Curiously, the H⁻ ligand gives rise to a high-field ¹H NMR resonance (−3.53 ppm) that appears in a chemical shift region typical of FeFe(t-H) terminal hydrides rather than Fe(μ-H)Fe species. The ligand was found by X-ray crystallography (Figure 38) to be somewhat closer to Fe (1.460 Å) than to Ni (1.637 Å).³⁴⁵ Relative to [26], the metals in [26(μ-H)]⁺ are further apart (2.613 Å) such that they are out of bonding distance, although a Ni⋯Fe interaction is sometimes drawn. The H⁻ ligand has a rigidifying influence on the structure, resulting in a single sharp ³¹P NMR resonance for the dppe ligand and two ¹³C resonances for the CO ligands.

Figure 38.

X-ray structure of [26(μ-H)]⁺. Non-hydride H atoms are omitted for clarity.³⁴⁵

The structure of [26(μ-H)]⁺ is supported by vibrational spectroscopy in the form of Raman³⁴⁹ and ⁵⁷Fe NRVS measurements.¹⁰³ The latter method, applied to ⁵⁷Fe-labeled isotopologue [(dppe)Ni(pdt)(μ-H)⁵⁷Fe(CO)₃]⁺ ([26′(μ-H)]⁺), indicated that H⁻ does indeed bond more strongly to Fe (ν_HFe = 1479 cm⁻¹) than to Ni (ν_NiH = 1022 cm⁻¹). The vibrational frequencies of the corresponding deuteride agree with the ν_MH ≈ 2^1/2ν_MD approximation for M–H/D harmonic oscillators in which M is much heavier than H/D. Relative to stretches, bending modes associated with μ-H⁻ ligands are much lower in energy, with the frequency of the δ_NiHFe bending vibration for [(dppe)Ni(pdt)(μ-H)⁵⁷Fe(CO)₃]⁺ (758 cm⁻¹) approaching that for the enzyme (675 cm⁻¹). The asymmetric hydride binding in [26(μ-H)]⁺ has also been probed by valence-to-core X-ray emission spectroscopy.³⁵⁰ In this case, the Fe spectra exhibit strong Fe–H signatures, while the Ni spectrum shows no evidence for its interaction with a light bridging atom. The method further allows estimation of M–H bond lengths, and although these come with moderately high uncertainties (~0.15 Å), such information is a useful complement to other analyses.

The tightness with which hydride binds Fe is rationalized in terms of the high electrophilicity of the Fe(CO)₃ center in [26(μ-H)]⁺, whose weighted average ν_CO frequency (2043 cm⁻¹) is considerably higher than that of the Fe(CO)(CN)₂ fragment in Ni-R (1944 cm⁻¹). The electronic influences described here result in [26(μ-H)]⁺ being somewhat acidic (pK_a = 10.7, Table 4), with the μ-H group being rather protic (e.g., it readily exchanges with D₂O) and insensitive to excess acid. The latter fact simplifies the synthesis of this and related NiFe hydrides, and is in contrast to the preparation of FeFe hydrides, which often involve strictly stoichiometric quantities of acid to avoid H₂ formation.

Table 4.

Spectroscopic and Structural Parameters for NiFe Thiolato Hydrides

hydride complexa		δ(1H)/ppm	νCO/cm−1	pKad	rNi–Fe/Å	rNi–H/Å	rFe–H/Å
[(dppe)Ni(pdt)(μ-H)Fe(CO)3]+	[26(μ-H)]+	−3.53	2082, 2024	10.7	2.613	1.637	1.460
[(dcpe)Ni(pdt)(μ-H)Fe(CO)3]+	[27(μ-H)]+	−3.00	2078, 2017	13.6	2.684	1.905	1.535
[(dppe)Ni(edt)(μ-H)Fe(CO)3]+	[28(μ-H)]+	−5.7	2084, 2025	11.3	2.596	1.843	1.578
[(dppe)Ni(pdt)(μ-H)Fe(CO)2PPh3]+	[29(μ-H)]+	−3.08b	2016, 1964	14.9	2.643	1.890b	1.487b
[(dppe)Ni(pdt)(μ-H)Fe(CO)2PPh2(2-py)]+	[30(μ-H)]+	−3.19b	2022, 1971
[(dppe)Ni(pdt)(μ-H)Fe(CO)2P(OPh)3]+	[31(μ-H)]+	−3.45b	2031, 1981
[(dcpe)Ni(edt)(μ-H)Fe(CO)3]+	[32(μ-H)]+	−5.3	2080, 2019	13.6
[(dppe)Ni(edt)(μ-H)Fe(CO)2PPh3]+	[33(μ-H)]+	−8.15,b −5.13c	2068, 1912	14.0
[(dppe)Ni(pdt)(μ-H)Fe(dppe)CO]+	[34(μ-H)]+	−6.34,b −3.01c	1954, 1938	2.656c	1.847c	1.555c
[(dppe)Ni(pdt)(μ-H)Fe(dppbz)CO]+	[35(μ-H)]+	−6.01,b −3.14c	1949, 1935	2.671a	1.788a	1.798a
[(dppv)Ni(pdt)(μ-H)Fe(dppv)CO]+	[36(μ-H)]+	−5.79b	1962, 1952	16.6	2.646b	1.79b	1.56b

^aIsolated as BF₄⁻ salts.

^bAsymmetric isomer.

^cSymmetric isomer.

^dpK_a is in PhCN.

Several derivatives of [26(μ-H)]⁺ have been prepared in order to bring the structure and spectroscopy of models more in line with Ni-R. One component in such compounds that is not often varied is the dithiolate, with pdt²⁻ and edt²⁻ giving rise to the most stable species owing to their high basicity and favorable chelate ring sizes. Indeed, while FeFe complexes of many mono-and di-, alkyl-, and arylthiolates are known, the NiFe systems are less modular. Although the variety of diphosphines at Ni is not large (dppe, dcpe, and dppv have been reported), the Fe(CO)₃ center allows more substitutional scope. The hydrides [(diphosphine)Ni(pdt)(μ-H)Fe(CO)₃]⁺ undergo reaction with mono-⁶³ and diphosphines³⁵¹ to afford [(diphosphine)Ni-(pdt)(μ-H)Fe(PR₃)(CO)₂]⁺ and [(diphosphine)Ni(pdt)(μ-H)Fe(diphosphine)(CO)]⁺ complexes, respectively (Figure 39). One limitation of this method is its potential incompatibility with highly basic phosphines such as PCy₃, which deprotonate [26(μ-H)]⁺ rather than participate in ligand substitution at Fe.

Figure 39.

Synthesis of phosphine-substituted Ni(pdt)(μ-H)Fe derivatives is best performed from hydrido tricarbonyl [26(μ-H)]⁺.⁶³

Complexes of the type [(diphosphine)Ni(pdt)(μ-H)Fe(PR₃)-(CO)₂]⁺ feature basal PR₃ ligands and are thus of lower symmetry than their C_s-symmetric tricarbonyl parents. When edt²⁻ is present, appreciable quantities of C_s-symmetric apical PR₃ isomers coexist with the C₁-symmetric species. In either case, substitution of one CO ligand for PPh₃ causes negligible change in ΔG°(H⁻), although the weighted average ν_CO decreases by ~50 cm⁻¹ and the pK_a increases by 3–4 units.⁶³ Even greater effects are observed when diphosphines are employed, and the first such complex [(dppe)Ni(pdt)(μ-H)Fe(dppe)(CO)]⁺ ([34(μ-H)]⁺) was initially prepared by decarbonylation of [(dppe)Ni(pdt)Fe(dppe)(CO)₂]²⁺ with BH₄⁻. A cleaner and more general route involves photolysis of a solution containing tricarbonyl hydride [(diphosphine)Ni(pdt)(μ-H)Fe(CO)₃]⁺ and diphosphine. The complexes [(diphosphine)Ni(pdt)(μ-H)Fe(diphosphine)(CO)]⁺ typically exist as two isomers: one with the Fe-ligated diphosphine being apical–basal and another in which it is dibasal. The structure of the latter isomer corresponds well to that of Ni-R, whose two basal CN⁻ ligands are mimicked by the diphosphine in the present species. Such bis(diphosphine) species also approach Ni-R in terms of their electron density at Fe, with the two isomers of [26(μ-H)]⁺ exhibiting low ν_CO frequencies (1954 and 1938 cm⁻¹). The effect of substitution on the weighted average ν_CO is approximately additive, with these values being ~100 cm⁻¹ lower than that for the tricarbonyl. As expected, changes in Fe ligation have large effects on ν_CO and pK_a, and the latter is consistent with the μ-H⁻ being closely associated with the Fe site. Changes at Ni (e.g., substitution of dppe with dcpe) are less influential on these parameters but do affect redox (section 4.2.3).

The present complexes feature Ni(μ-H)Fe cores whose geometries certainly vary with the identity of the coligand(s). While the bond length difference r_Ni–H − r_Fe–H is 0.177 Å in the case of [26(μ-H)]⁺, this doubles to 0.370 Å for the Ni(dcpe) analogue [27(μ-H)]⁺.³⁴⁸ This latter complex features the most electron-rich Ni site and the most electron-poor Fe site, and one might argue that an increase in donicity of the Ni ligand set means that it need not bind H⁻ so tightly. However, a conclusive trend in H⁻ binding is difficult to tease out as ligand substitution often has steric consequences. Indeed, interactions between the phosphine substituents in the PPh₃ complex [(dppe)Ni(pdt)(μ-H)Fe(PPh₃)(CO)₂]⁺ ([29(μ-H)]⁺)⁶³ lead to a bond length difference of 0.403 Å that is greater than that in [26(μ-H)]⁺, a counterintuitive result if just considering electronic effects. Such steric interactions are also at play in the bis(diphosphine) complexes, and while there is structural variation, it is significant that the r_Ni–H − r_Fe–H value for [(dppe)Ni(pdt)(μ-H)Fe-(dppbz)(CO)]⁺ ([35 (μ-H)]⁺, dppbz = 1,2-bis-(diphenylphosphino)benzene)³⁵¹ of −0.010 Å indicates that increasing electron density at Fe can shift H⁻ toward Ni such that the structures approach that of [NiFe]-H₂ase.

4.2.2. Related Dithiolato Hydrides

As might be expected, the properties of heterobimetallic dithiolato hydride complexes are sensitive to not only variations in the ligand sets, but also changes in the metals used. The ease of preparation and high stability of the Ni(pdt)(μ-H)Fe models above have inspired the synthesis of congeners in which Ni or Fe is replaced by other metals, typically heavier elements from their respective groups. While such work departs somewhat from replicating the [NiFe]-H₂ase active site, it may provide information about the factors governing reactivity of NiFe systems. It is also possible that these “less biomimetic” systems perform even better catalytically. The ligand set chosen for Fe (or its replacement) often varies, with that of the group 10 metal typically bearing two neutral donors in addition to two thiolates. For example, square-planar building blocks of the form [ML₂(SR)₂] feature prominently, and although the following examples have phosphines as the neutral ligand (e.g., L₂ = dppe), the formation of bimetallics can in principle occur with a range of building blocks.

The most closely related complexes to [26(μ-H)]⁺ are its analogues [(dppe)Pd(pdt)(μ-H)Fe(CO)₃]⁺ ([37(μ-H)]⁺) and [(dppe)Pt(pdt)(μ-H)Fe(CO)₃]⁺ ([38(μ-H)]⁺). These are spectroscopically almost identical to their NiFe relative [26(μ-H)]⁺ and are also isostructurual, including the asymmetry of the μ-H⁻ bridge toward Fe. Additionally, each is prepared in high yield by treatment of a [(dppe)M(pdt)Fe(CO)₃] precursor with HBF₄·Et₂O. A distinguishing feature of the PtFe example is that the low-valent species [(dppe)Pt(pdt)Fe(CO)₃] ([38]) is a bona fide M(II)Fe(0) complex. No Pt–Fe bond exists in this species, which is “frozen” in a reactive conformation rather than in a tetrahedral arrangement observed for the NiFe system.

More often than is the case with the Ni site, the Fe site can also be altered to feature other elements. The most common substitute is Ru, which has rich organometallic chemistry and forms strong, stable bonds to H⁻, H₂, and C-donor ligands, as evidenced by the formation of [Cy₃P(CO)₂Ru(pdt)(μ-H)Ru(t-H)(CO)PCy₃] from H₂ (section 3.2.3).²⁴⁹ Indeed, the synthesis of NiRu complexes [(dxpe)Ni(pdt)Ru(cymene)] ([39], dxpe = dppe; [40], dxpe = dcpe; Figure 40)³⁵² suggests that 14e⁻ FeL₃ and Ru(cymene) fragments are to some extent interchangeable. The tetrahedral Ni coordination might suggest a Ni(I)Ru(I) description for these complexes, but the geometry is also consistent with a Ni(0)Ru(II) assignment, to which DFT calculations point. Combination of either [39] or [40] with [H(OEt₂)₂]BAr^F₄ affords the corresponding hydrides [39(μ-H)]⁺ and [40(μ-H)]⁺. The zerovalent Ni center is thought to be the initial site of protonation, although the X-ray structure of [39(μ-H)]⁺ reveals a familiar motif in which the H⁻ ends up closer to Ru (1.54 Å) than to Ni (1.65 Å).

Figure 40.

Synthesis of NiRu³⁵² and NiW dithiolates³⁵³ and their respective hydrides.

There is no particular reason why the metal to replace Fe must be chosen from group 8. However, it is thought that 14e⁻ fragments are most suitable, and indeed Mo(CO)₄ and W(CO)₄ fragments readily bind the Ni chelates [Ni(R₂PCH₂CH₂S)₂] and [(R₂PCH₂CH₂PR₂)Ni(pdt)] to form Ni(II)M(0) bimetallics.³⁵⁴ While the products do not feature metal–metal bonds and M is octahedrally bound, the propensity of group 6 metals to form 7-coordinate complexes can be exploited to prepare the Ni(II)(μ-H)W(II) species ([41(t-H)]⁺).³⁵³ The high acidity of this hydride (which is deprotonated by excess Et₂O) and low catalytic activity result from the electron-poor nature of the W(II) center (average ν_CO = 2006 cm⁻¹), and perhaps also from its coordination geometry not allowing H⁻ to interact with Ni.

An ever-increasing number of hydrides have been prepared that reproduce many aspects of the [NiFe]-H₂ase active site. In terms of structure and spectroscopy, some are certainly closer than others to Nature’s blueprint, while some do not even include Ni and/or Fe. The key question is, how do the differences between the enzyme and its models affect catalysis? This question is addressed in section 4.2.3 with respect to some complexes already described here, as well as some further examples.

4.2.3. Electrochemistry and HER Catalysis

With active sites that exhibit both acid–base and redox behavior, the [NiFe]-H₂ases are suitably equipped to mediate proton reduction (HER). The acid–base and redox criteria are met by several synthetic [NiFe]-H₂ase models, and their electrochemistry, studied using methods described in section 3.2.7, is now described. Considered first are the family of complexes [(diphosphine)Ni(dithiolate)(μ-H)Fe(CO)_3–nL_n]⁺, not only because they were the first HER-active models to be reported, but also because variation in the ligand sets allows development of structure–function relationships. All but the most electron-rich of these Ni(μ-H)Fe species are not oxidizable. Surprisingly, the oxidation of the bis(dppe) complex [34(μ-H)]⁺ occurs at a mild potential (0.15 V vs Fc^+/0), although with low reversibility. When treated with a chemical oxidant (acetylferrocenium) at −25 °C, the ν_CO bands are shifted by large amounts (68 cm⁻¹), indicative of a Ni(II)(μ-H)Fe(III) product whose oxidation states are reversed relative to the Ni(III)(μ-H)Fe(II) description for Ni-C. In contrast to the oxidations, the reductions of Ni(μ-H)Fe are more well-defined and more relevant to the HER mechanisms of these synthetic catalysts. Some key electrochemical measurements are presented in Table 5.

Table 5.

Redox and Catalytic Properties of NiFe Dithiolato Hydrides

hydride complexa		E(NiHFe+/0)b	Ecat/2b,c	acid used	TOF/s−1	η/V
[(dppe)Ni(pdt)(μ-H)Fe(CO)3]+	[26(μ-H)]+	−1.34	−1.20	CF3CO2H	20	0.50
[(dcpe)Ni(pdt)(μ-H)Fe(CO)3]+	[27(μ-H)]+	−1.56	−1.46	ClCH2CO2H	50	0.59
[(dppe)Ni(edt)(μ-H)Fe(CO)3]+	[28(μ-H)]+	−1.33	−1.23	CF3CO2H	240–310	0.49
[(dppe)Ni(pdt)(μ-H)Fe(CO)2PPh3]+	[29(μ-H)]+	−1.49	−1.30	CF3CO2H	50	0.60
[(dppe)Ni(pdt)(μ-H)Fe(CO)2PPh2(2-py)]+d	[30(μ-H)]+	−1.49	−1.30	CF3CO2H	50	0.60
[(dppe)Ni(pdt)(μ-H)Fe(CO)2P(OPh)3]+	[31(μ-H)]+	−1.44	−1.32	CF3CO2H	50	0.62
[(dcpe)Ni(edt)(μ-H)Fe(CO)3]+	[32(μ-H)]+	−1.47	−1.45	ClCH2CO2H	20	0.59
[(dppe)Ni(edt)(μ-H)Fe(CO)2PPh3]+	[33(μ-H)]+	−1.47	−1.45	ClCH2CO2H	60–120	0.54
[(dppe)Ni(pdt)(μ-H)Fe(dppe)CO]+	[34(μ-H)]+	N/D	−1.61, −1.65	ClCH2CO2H	23	0.85
[(dppe)Ni(pdt)(μ-H)Fe(dppbz)CO]+	[35(μ-H)]+	N/D	−1.61, −1.70	ClCH2CO2H	30	0.85
[(dppv)Ni(pdt)(μ-H)Fe(dppv)CO]+	[36(μ-H)]+	N/D	−1.54	ClCH2CO2H	29	0.75

^aIsolated as BF₄⁻ salts.

^bIn volts, determined in CH₂Cl₂, relative to Fc^+/0.

^cPotential at i_cat/2.

^d2-py =2-pyridyl.

Potentials for the [(diphosphine)Ni(dithiolate)(μ-H)-FeL₃]^+/0 couple are strongly related to the nature and number of phosphine donors at both the Ni and Fe sites, being most negative when strongly donating ligand sets are involved. As is the case with the reduction of Fe(II)(μ-H)Fe(II) species, the initial 1e⁻ reduction product of Ni(II)(μ-H)Fe(II) may take multiple forms, viz., Ni(I)(μ-H)Fe(II), Ni(II)(μ-H)Fe(I), and Ni(1.5)(μ-H)Fe(1.5). The reversibility of this couple varies greatly and correlates with the stability of the radical hydride complex, the formation of which is a key step in catalysis (vide infra). The reduction is irreversible in the case of tri- and dicarbonyls with the exception of [31(μ-H)]^+/0, where the [31(μ-H)]⁰ radical is thought to be stabilized by spin delocalization onto the P(OPh)₃ ligand in a manner similar to that observed for [31]⁺.¹⁸⁰ When two diphosphines are incorporated into Ni(II)(μ-H)Fe(II) species, the reversibility of the 1e⁻ reductions approaches unity (e.g., for [36(μ-H)]⁺, Figure 41, |i_pa/i_pc| = 0.94 at 0.1 V s⁻¹, room temperature). Electron transfer can also be induced chemically, and treatment of [(dppv)Ni(pdt)(μ-H)Fe(dppv)CO]⁺ with CoCp*₂ allows isolation of the stable charge-neutral hydride. Reduction causes only a modest red shift, with Δν_CO = −37 cm⁻¹, consistent with Ni-centered redox and thus a Ni(I)(μ-H)Fe(II) description. This is further corroborated by EPR and DFT studies, the latter indicating that significant spin density resides on Ni.³⁵¹

Figure 41.

X-ray structure of [36(μ-H)]⁺, one of the most active synthetic NiFe HER catalysts. Non-hydride H atoms are omitted for clarity.³⁵¹

The electron-rich di- and monocarbonyl catalysts operate at high rates, although their E_cat/2 values are, unsurprisingly, rather negative. Such electron-rich catalysts give rise to more negative E_cat/2 values, yet can still have reasonably small overpotentials if they can mediate H₂ evolution from weaker acids. As stated in section 2.3.1, one should find the weakest acid that will protonate the reduced form of the catalyst, which is the Ni(I)Fe(I) species in the present case. As is clear from the data in Table 5, the more cathodic E_cat/2 values have yet to be offset by employing weaker acids for the NiFe systems. Overpotentials for the current catalysts remain >0.5 V, and reducing this key metric while maintaining high TOFs is a fundamental challenge in HER catalysis.

The mechanism by which [(diphosphine)Ni(dithiolate)(μ-H)FeL₃]⁺ catalysts operate involves sequential e⁻ and H⁺ transfers. Hydrides of the form Ni(II)(μ-H)Fe(II) (Figure 42, top right) are typically unaffected by acid, and the catalytic cycle thus involves their 1e⁻ reduction to a mixed-valent complex. Examples of these reduced species have been observed to react instantaneously with acid,³⁵¹ perhaps with S atoms serving to relay protons to the metal sites. Following protonation, loss of H₂ affords mixed-valent species Ni(II)Fe(I) complexes (for tri- and dicarbonyls) or Ni(I)Fe(II) complexes (for monocarbonyls), the latter case matching the assignment for the Ni-L state of [NiFe]-H₂ase. Reduction of the S = 1/2 intermediates gives the neutral, hydride-free species. The reduced hydride-free species can exist in a reactive Ni(II)Fe(0) form in which Ni is planar, as well as a Ni(I)Fe(I) resting state with tetrahedral Ni.¹⁶⁰ The tetrahedral isomer can readily convert to the reactive form (with a DFT-calculated free energy barrier of ΔG^‡ ≈ 6.7 kcal mol⁻¹ and an experimentally estimated barrier of 9.5 kcal mol⁻¹ for [26]¹⁶⁰), whose high basicity allows for protonation and completion of the catalytic cycle.

Figure 42.

HER catalytic cycle proposed for the present NiFe dthiolato hydrides.¹⁶⁰

Given that HER catalysts featuring a mono- or diphosphine at Fe operate at the highest rates, it is likely that the basicity conferred by these ligands is important for catalysis. Indeed, one or both of the H⁺ transfers may be rate-determining, and the lack of an effective H⁺ relay apart from the bridging pdt²⁻ may represent a shortcoming of these models, despite the negligible effect replacing PPh₃ with P(2-pyridyl)Ph₂ has on the TOF. This relay aspect has been investigated with the unusual Ni(I)Fe(I) complex Ni(xbsms)Fe(CO)₃ ([43], Figure 43), in which only one of the thiolates is bridging.³⁵⁵ The xbsms²⁻ ligand and its relatives are popular ligands in [NiFe]-H₂ase synthetic modeling, with flexible polydentate donors of this type often being referred to as podands. Treatment with HBF₄·Et₂O affords conjugate acid [43H]⁺, whose average ν_CO is only modestly (43 cm⁻¹) shifted relative to [43] (cf. 66 cm⁻¹ for [26] → [26(μ-H)]⁺). Crystallographic analysis confirms the product as a Ni(I)Fe(I) thiol, although DFT calculations suggest that the Ni(II)(μ-H)Fe(II) hydride tautomer is only slightly higher in free energy (~5 kcal mol⁻¹). Cyclic voltammetry studies indicate [43] mediates electrocatalytic reduction of CF₃CO₂H (E_cat/2 = −1.43 V vs Fc^+/0, η = 540 mV), although bulk electrolysis suggests that the TOF is low in this case (0.0014 s⁻¹ at −1.6 V). The poor basicity of [43], which is not completely protonated by CF₃CO₂H, may well be responsible for this sluggishness. The use of stronger acids and/or substitution of CO ligands for more basic donors is likely to increase the activity of the present system.

Figure 43.

Interconversion between thiol and hydride tautomers of [43H]⁺.³⁵⁵ There is evidence for both thiol and hydride moieties in the structure for Ni-R.

Given the scope of this review, the focus here has largely been on bimetallic HER catalysts that are isolable in one or more hydride-bearing forms. However, these represent a subset of catalytically active [NiFe]-H₂ase models, and many others have not been characterized as their hydride derivatives, although hydrides are assumed as HER intermediates.^356,357 Consider the Ni(I)Fe(II) radical [CpNi(pdt)Fe(dppe)CO] ([44], ν_CO = 1901 cm⁻¹),¹⁵⁹ a very electron-rich model for the Ni-L state. The conjugate acid [CpNi(pdt)(μ-H)Fe(dppe)CO]⁺ ([44(μ-H)]⁺, Figure 44), also a radical, is predicted by DFT studies to have spin density localized mainly on the Ni coordination sphere, with almost twice as much density at Ni than Fe. Such a finding is significant, as the Ni(III)(μ-H)Fe(II) core implicated would match that of Ni-C. The hydride [44(μ-H)]⁺ has so far eluded characterization, as while it is generated by addition of acid (1 equiv) to [44], the hydride is instantly reduced to [44(μ-H)] by surrounding [44]. After reaction of [44(μ-H)] with remaining acid, the overall stoichiometry is [44] + H⁺ → [44]⁺ + 1/2H₂, a reaction that also proceeds electrocatalytically using CF₃CO₂H as the H⁺ source (E_cat/2 = −1.16 V, TOF = 4 s⁻¹). The DFT calculations suggest that the H⁺ reduction mechanism can proceed through one of two pathways, depending on whether the dppe ligand is dibasal or apical–basal. Overall, this example demonstrates well that the challenges in [NiFe]-H₂ase modeling cannot solely be met by preparing more electron-rich NiFe dithiolates. The DFT structure calculated for [44(μ-H)]⁺ features asymmetrically bound μ-H⁻, with the Ni–H bond (1.80 Å) being longer than the Fe–H bond (1.60 Å).¹⁵⁹ Despite the presence of the Cp⁻ ligand, the Ni(III) site remains a strong oxidant, as it does not bind H⁻ as tightly as Fe. The oxidizing power and reducing power of [44(μ-H)]⁺ and [44], respectively, illustrate the large effects protonation can have on redox (and vice versa). Such NiFe models rely on the geometric persistence of the NiCp subunit, and the square planar ⇌ tetrahedral isomerization accompanying the reduction of 4-coordinate Ni^II does not come into play here.

Figure 44.

DFT-calculated structure for the putative species [44(μ-H)]⁺, whose Ni(III)(μ-H)Fe(II) core mimics that in Ni-C. Non-hydride H atoms omitted for clarity.¹⁵⁹

This review has so far highlighted examples from the growing list of thiolate- and hydride-containing heterobimetallic complexes, many of which exhibit electrocatalytic behavior.^356,358 Even rather electron-poor bimetallic species can often be converted to hydrides because the proton chemical potential is readily increased by changing acid concentration and/or employing a superacid. The preparation of Ni(μ-H)Fe species from H₂ is an arguably more challenging matter and is addressed in section 4.2.4.

4.2.4. Hydrides from H₂

As is the case for the modeling of [FeFe]-H₂ases, very few [NiFe]-H₂ase models exist in which a H⁻ ligand is derived from H₂ heterolysis.²⁶ Before these examples are described, two less-biomimetic situations are addressed, in which model complexes instead accept either a H⁺ or a H^• from H₂.

The first case is exemplified by the conversion of the Ni(I)Fe(I) species [(dppe)Ni(pdt)Fe(CO)₃] ([26]) to the Ni(II)Fe(II) hydride [(dppe)Ni(pdt)(μ-H)Fe(CO)₃]⁺ ([26(μ-H)]⁺) not by using the typical acid HBF₄·Et₂O, but rather by using a combination of H₂ (1 atm) and B(C₆F₅)₃.⁶³ This presumably heterolytic pathway makes use of the [26]/B(C₆F₅)₃ FLP, although the roles are reversed relative to the enzyme. Indeed, the Ni-SI_a + H₂ ⇌ Ni-R conversion involves a Lewis acidic metal site and a basic Cys (or Arg) residue. The present situation is perhaps more relevant to the [Fe]-H₂ase mechanism (vide infra), in which the Lewis acidic substrate methenyl-H₄MPT⁺, whose role here is played by B(C₆F₅)₃, accepts H⁻ in a metal-mediated process.

Metal complexes can also use H₂ as a source of H^•, although this does not necessarily implicate a homolytic activation pathway. Consider the low-valent Ni(0)Ru(II) species [39], which is readily oxidized by Fc⁺ to afford the Ni(I)Ru(II) cation [39]⁺ (Figure 40). Under H₂, the latter complex converts to the diamagnetic Ni(II)(μ-H)Ru(II) product ([39(μ-H)]⁺), whose alternative synthesis is from [39] and [H(OEt₂)₂]BAr^F₄. The reaction with H₂ is slow (~12 h) and poorly understood,³⁵⁹ and it is as yet unclear whether the process is first or second order in [39]⁺. The latter case would be less relevant to the H₂ases, whose active sites are deeply nested in their protein shells and cannot work in concert with a second H₂ase.

Biomimetic [NiFe]-H₂ase models that utilize H₂ as a H⁻ source are best designed by mimicking the structure of Ni-SI_a, whose electrophilic Ni(II)Fe(II) core is amenable to the binding and cleavage of H₂. A popular class of building blocks exists in Ni complexes of N₂S₂- (2 × amine + 2 × thiolate) or S₄-donor (2 × thioether + 2 × thiolate) podands.³⁶⁰ As with (diphosphine)-Ni(dithiolate) species, low-spin Ni(II) podands have been incorporated into many bimetallic systems, one example of which is [Ni(podand^Me)Ru(OH₂)(C₆Me₆)]²⁺ ([45(OH₂)]²⁺, Figure 45, top left).³⁶¹

Figure 45.

NiRu dithiolate [45(OH₂)]²⁺ activates H₂ to afford hydride [45(μ-H)(OH₂)]⁺. Catalytic H₂ oxidation, using Cu²⁺ as the electron acceptor, is proposed to involve a dihydride intermediate.³⁶¹

As proposed for Ni-SI_a, the Ni(II)Ru(II) complex [45(OH₂)]²⁺ incorporates a weakly bound OH₂ ligand into its structure, with a 14e⁻ Ru(C₆Me₆) fragment serving as a Fe(CN)(CO)₂ surrogate. Under H₂ (1 atm) at room temperature, diamagnetic [45(OH₂)]²⁺ converts to the paramagnetic hydride [45(μ-H)(OH₂)]⁺ with concomitant release of H⁺ to the (buffered) solution. Octahedral on account of an apical aquo ligand, the Ni(II) site in the product (S = 1) contrasts that in Ni-R (S = 0), wherein no H₂O is bound. Precise atomic coordinates of the synthetic model were obtained by X-ray and neutron diffraction, the latter utilizing deuteride [45(μ-D)(OH₂)]⁺ (prepared from [45(OH₂)]²⁺ and D₂) and affording accurate Ni–D (1.859 Å) and Ru–D (1.676 Å) distances. The two proposed mechanisms for H₂ activation are (i) arene slippage to η⁴-hapticity, with subsequent H₂ ligation and cleavage at Ru, and (ii) H₂ ligation and cleavage at Ni. In both cases it was suggested that the Ru-bound aquo deprotonates η²-H₂, but an alternative route that appears at least as likely involves a free H₂O molecule serving as the base.

Hydride [45(μ-H)(OH₂)]⁺ is a catalyst for H₂ oxidation, mediating 2e⁻ reduction of excess Cu²⁺ under H₂, with the concomitant release of 2H⁺. The proposed mechanism is unusual in that each cycle requires activation of two molecules of H₂ and involves a putative dihydride intermediate.³⁶² The latter species is invoked based on H/D-exchange experiments, as well the observation that [45(μ-H)(OH₂)]⁺, despite its protic nature, requires H₂ to convert to the low-valent Ni(I)Ru(I) complex. The hydride [(H₂O)Ni(podand^Me)(μ-H)Ru(C₆Me₆)]⁺ was further demonstrated to mediate hydrogenation of aldehydes to their corresponding alcohols,³⁶³ as well as to serve as an anodic catalyst in a fuel cell^28,29 in which the cathodic reaction was mediated by [Ni(podand^Me)(η²-O₂)Ru(Cp*)]⁺.³⁶⁴ The processing of H₂ is accompanied here also by O₂ activation, with the more electron-rich Cp*⁻ derivative allowing access to Ni(II)-Ru(IV) peroxos. More recently, the Ni(II)Fe(IV) peroxo [Ni(podand^Me)(η²-O₂)Fe(Cp*)]⁺ could also be prepared from O₂,³⁶⁵ a reaction mimicking the oxidase behavior exhibited by certain O₂-tolerant [NiFe]-H₂ases.³⁶⁶

The structure of hydride [45(μ-H)(OH₂)]⁺ certainly differs from that of Ni-R, not least in the use of Ru(II) for Fe(II). The promising reactivity and aqueous solubility of the synthetic model³⁶² inspired further research into related podand complexes, including those with NiFe cores. A close relative of [45(μ-H)(OH₂)]⁺ is the 34e⁻ complex {Ni(podand^Et)Fe-(MeCN)[P(OEt)₃]₃}²⁺ ([46(MeCN)]²⁺, Figure 46), which preserves many important features while also having the targeted NiFe composition. The organometallic Fe(CN)₂(CO) site at [NiFe]-H₂ase is mimicked here by a Fe[P(OEt)₃]₃ fragment, with the phosphite ligands chosen for their σ-donating and π-accepting properties.³⁶⁷

Figure 46.

Activation of H₂ by Ni(II)Fe(II) species [46(MeCN)]²⁺ affords hydride [46(μ-H)]⁺ (top). X-ray structure of [46(μ-H)]⁺ (bottom; non-hydride H atoms omitted for clarity).³⁶⁷

Demonstrating reactivity akin to its NiRu analogue, [46(MeCN)]²⁺ transforms under H₂ to the hydride [46(μ-H)]⁺, although not without the use of a strong base to accept the H⁺ byproduct. The product here is diamagnetic, allowing for assignment of the H⁻ resonance (δ(¹H) −3.57 ppm), whose shift is on par with values for [(diphosphine)Ni(dithiolate)(μ-H)Fe(CO)_3–nL_n]⁺. The use of D₂ in place of H₂ confirms both the resultant hydride structure and the origin of the new bridging ligand identified according to IR (ν_NiHFe = 1687 cm⁻¹, ν_NiDFe = 1218 cm⁻¹) and ESI-MS data. X-ray crystallography of the hydride afforded bond lengths for Ni–H (2.16 Å) and Fe–H (1.57 Å) slightly smaller than those obtained by neutron diffraction using the deuteride (Ni–D 2.18 Å, Fe–H 1.577 Å). The latter measurements provide a better picture of the true nuclear positions, although in both cases the Fe-biased binding of H⁻ is clear. Nevertheless, the complex [46(μ-H)]⁺ is interesting in that it exhibits hydridic properties, liberating H₂ upon treatment with HBF₄·Et₂O. Similar to Ni-R, it also undergoes oxidation, although only irreversibly and at a rather high potential (0.45 V). A reversible wave at −0.42 V may arise from a [46(μ-H)]^+/0 couple, which indicates that the mixed-valent hydride [46(μ-H)] is potentially isolable.

The report that hydride [46(μ-H)]⁺ could be prepared from H₂ was followed almost immediately by one describing synthetic hydrides of even higher fidelity to Ni-R, both in structure and in function. One model in question was prepared from [(dppe)-Ni(pdt)Fe(CN)₂(CO)₂], one of many complexes that mimic the Ni-SCO state (Ni-SI_a + CO ⇌ Ni-SCO), an inactive enzyme form in which Fe(II) is coordinatively saturated due to its 2CO + 2CN⁻ ligands. The bis(triarylborane) adduct ([47(CO)]) undergoes decarbonylation with Me₃NO to presumably afford a coordinatively unsaturated electrophile that, despite its lack of a positive charge, extracts H⁻ from H₂ in the presence of the Me₃N decarbonylation coproduct (Figure 47).³⁶⁸

Figure 47.

Activation of H₂ by in situ generated coordinatively unsaturated Ni(II)Fe(II) species affords hydride models for Ni-R (top). X-ray crystal structure of anion [47(μ-H)]⁻ (bottom; F and non-hydridic H atoms omitted for clarity).³⁶⁸

The product [47(μ-H)]⁻ and its relatives are the first [NiFe]-H₂ase models containing both CN⁻ and H⁻ groups. Aside from its CO occupying a basal site, the Fe coordination sphere in [47(μ-H)]⁻ completely reproduces that in Ni-R, with the BR₃ groups simulating the H-bonding exhibited by the CN⁻ ligands in Ni-R. Nevertheless, the H⁻ ligand (δ(¹H) −7.05 ppm) is still closer to Fe (1.516 Å) than to Ni (1.710 Å), although the asymmetry is less pronounced than in [46(μ-H)]⁻. A subtlety differentiating [47(μ-H)]⁻ from the many Ni(II)(pdt)(μ-H)Fe(II) species is the distortion in the Ni coordination, with the lesser of the P–Ni–S angles (147.7°) departing from linearity, although not to the extent of the respective S–Ni–S angle in Ni-R (107.9°). The distorted Ni geometry in these systems is unlikely to be reproduced by podand-containing complexes such as those described above, for which square-planar Ni sites are apparently favored.

The hydridic behavior of [47(μ-H)]⁻ is evidenced in its dihydrogen bonding interactions with weak acids (e.g., Fe– H⁻⋯⁺HNMe₃) and its liberation of H₂ when treated with the stronger acid PhNH₃⁺. These interactions are no doubt promoted by the anionic nature of [47(μ-H)]⁻, an aspect that also affects redox. Hydride [47(μ-H)]⁻ and its dcpe analogue [48(μ-H)]⁻ oxidize at −0.08 and −0.10 V vs Fc^+/0, respectively, with the similarity of these values suggesting that oxidation is Fe-centered and generates a Ni(II)(μ-H)Fe(III) species (cf. Ni-C Ni(III)(μ-H)Fe(II)). Significantly, [47(μ-H)]⁻ is an electrocatalyst for H₂ oxidation (E_cat/2 ≈ −0.08 V, TOF = 0.98 s⁻¹) in the presence of 1,8-diazabicycloundec-7-ene (DBU, pK_a = 24.3 in MeCN) as a base. The observation that DBU does not deprotonate [47(μ-H)]⁻ mandates that the anion first be oxidized in the catalytic cycle, with the implicated Ni(II)(μ-H)Fe(III) species serving as a rare example of a synthetic high-valent NiFe hydride. Work aimed at a more accurate synthetic model for Ni-C will certainly address the ⁻CNBR₃ ligands, whose high ν_CN frequencies (2162, 2137 cm⁻¹) and low donicity (2 × ⁻CNBR₃ is less donating than dppe) contrast the ⁻CN ligands in the enzyme. Concomitantly, the installation of stronger donors at Ni will be required to ensure that Ni(II) is more easily oxidized than Fe(II) in complexes of the present type.

4.2.5. Bioinspired Systems

The mechanism by which [NiFe]-H₂ase operates involves redox at Ni, with the Fe center serving as a Lewis acid permanently in the +II oxidation state. This observation has led many chemists to investigate mononuclear Ni catalysts with the requisite redox and acid–base properties. Such species are often termed “bioinspired” as they have less in common with H₂ases than the “biomimetic” examples described here thus far. Although several Ni complexes, particularly those featuring dithiolenes,³⁶⁹ are known to mediate the HER, discussed here are the more privileged examples that exhibit H₂ activation. There are certainly fewer stable examples of Ni(η²-H₂) complexes than of Fe(η²-H₂) analogues, in part a reflection of η²-H₂ complexes almost always being octahedral. This rarity had initially led to doubts about Ni performing H₂ cleavage in [NiFe]-H₂ase, although this is now thought to be the case. Early evidence for Ni interaction with H₂ could be found from [49H], which converts to its D isotopologue [49D] under D₂ (Figure 48).³⁷⁰ The formation of a short-lived D₂ complex is thought to be followed by heterolysis of D₂, with Ni receiving D⁻ and thiolate accepting D⁺.

Figure 48.

H/D scrambling implies a Ni(η²-H₂) intermediate (top).³⁷⁰ Such Ni(η²-H₂) species can exhibit heterolysis with³⁷¹ and without an external base³⁷² (middle and bottom, respectively).

One isolable dihydrogen complex is [50(η²-H₂)]⁺, whose ¹H NMR resonance at −3.21 ppm is characteristic of a η²-H₂ ligand. Illustrating the H₂ heterolysis reaction, this cation reacts with NEt₃ to afford the corresponding hydride [50(H)].³⁷¹ A similar example exists in [51(η²-H₂)]⁺, although this H₂ adduct exhibits H₂ fission without the need for an external base.³⁷² Indeed, the amido group is well-placed to accept a H⁺ while remaining bonded to Ni, resulting in the low thermal stability of [51(η²-H₂)]⁺. In general, it is important to note that observation of H₂ heterolysis does not rule out a homolytic pathway. Indeed, an oxidative addition route involving a Ni(IV) dihydride is proposed in the case of [51(η²-H₂)]⁺. Lastly, the most stable Ni(II)(η²-H₂) complex to date is {[(2-Ph₂PC₆H₄)₃Si]Ni(η²-H₂)}⁺, which features a tetradentate, anionic SiP₃-donor tripod.³⁷³ In a sense, this complex mirrors the 5-coordinate Ni(II) center in the (Cys-S)₂Ni(η²-H₂)(Cys-S)₂Fe(CN)₂(CO) species proposed as an enzyme intermediate. While the complexes [50(η²-H₂)]⁺ and [51(η²-H₂)]⁺ illustrate biorelevant facets of nickel hydrides, their redox chemistry has not been developed.

Molecular Ni complexes with favorable acid–base and redox properties allow mediation of both H₂ oxidation and evolution. Such bidirectional catalysts remain key synthetic targets. One prominent class of these blends the well-understood protonation behavior of bis(diphosphine)nickel complexes⁵⁵ with the H⁺ relay motif in [FeFe]-H₂ases. These complexes feature heterocycles with phosphine donor groups and amine relays (denoted “P₂N₂”, Figure 49).⁸⁶

Figure 49.

Examples of the [Ni(P₂N₂)₂] family of electrocatalysts mediate both H₂ oxidation (counterclockwise) and evolution (clockwise).⁸⁶ This cycle is a simplification that neglects certain side pathways, and the order of the steps depends on the identities of substituents R and R′.

Variation of R and R′ substituents in [Ni(P₂N₂)₂] allows tuning of steric and electronic properties such that certain reactions are favored. For example, when R = Cy and R′ = ^tBu, the Ni derivative is a fast H₂ oxidation electrocatalyst, at least by synthetic standards (TOF = 51 s⁻¹ under H₂ (1 atm) at −0.77 V vs Fc^+/0).^27,374 When instead R = Ph and R′ = CH₂CH₂OMe, a bidirectional catalyst is afforded, albeit one that effects H₂ oxidation and production very slowly. Bidirectional catalysts typically operate at low overpotentials, in this case possibly facilitated by concerted PCET.^375,376 It was later found that the HER activity of the Ni complexes is greatly enhanced by using a “P₂N” ligand, a seven-membered analogue of “P₂N₂” ligands bearing only one NR′ substituent. When R = R′ = Ph, [Ni(P₂N)₂] mediates the HER from protonated N,N-dimethylformamide (pK_a = 6.1) extremely rapidly (TOF > 100 000 s⁻¹ at −1.13 vs Fc^+/0), albeit at a high overpotential (η = 625 mV).³⁷⁷ Whether performing H₂ oxidation or evolution, it is clear that H⁺ relays play a large role in facilitating substrate transfer. The present class of compounds combines Ni with amines reminiscent of adt²⁻, the latter being positioned far enough away such that Ni–N bonding is unfavorable, yet close enough to rapidly relay H⁺ to and from the metal.²⁸¹

4.3. Concluding Remarks and Future Challenges

The synthetic modeling of [NiFe]-H₂ases has afforded an ever-growing collection of heterobimetallic complexes, the structures and reactivities of which are incrementally approaching those of Nature’s blueprint. The high efficiency of H₂ases requires that each catalytically active state be of similar free energy³⁷⁵ and exist in an almost flat free energy landscape in which kinetic barriers to H₂ oxidation or evolution are minimized.¹⁴¹ Reproducing these rates in synthetic systems will necessitate more rigid complexes, and no doubt motivate the synthesis of new polydentate ligands. Indeed, Ni(diphosphine)₂-based catalysts oscillate between tetrahedral Ni(0) and planar Ni(II) during catalysis in contrast to the rigid NiS₄ coordination sphere of the enzyme. This molecular reorganization may well be a contributing factor to the high overpotentials several conformationally flexible models exhibit. While rigidity appears important, not just any ligand will suffice, and ideal designs include those that constrain the complex, particularly the Ni center, in the unique geometry between square planar and tetrahedral. The seesaw coordination, indicated by DFT studies to be important in H₂ scission,³²⁵ has yet to be replicated. Synthetic chemists will need to look beyond classic ligands, such as podands, which strongly bias square-planar coordination. This biasing of one redox state over another may be a reason why no biomimetic catalyst has been characterized in all of the states proposed for its catalytic cycle. On the other hand, ligand flexibility has been shown to be important for PCET in the [Ni(P₂N₂)₂] and related catalysts with proton relays. In such cases, thermal motions enable the pendant amine to move closer to the Ni center with a relatively low energy penalty in order to facilitate proton transfer.³⁷⁶ Although the geometry of the Fe site in models is more easily managed, holding the CO ligand trans to the hydride binding site poses a challenge.³⁷⁸

A key shortcoming with models prepared thus far is that no system can exist in a Ni(III)(μ-H)Fe(II) form analogous to Ni-C. The continuing quest to prepare more electron-rich complexes can be readily tracked by considering redox potentials and ν_CO frequencies. This work will involve shifting the μ-H⁻ ligand toward Ni to stabilize a +III oxidation state.^379,380 Additionally, this state would be favored by the inclusion of stronger anionic ligands including thiolates or possibly even phosphides or amides. Such ligands may also serve as H⁺ relays, although their propensity to bridge metals must be suppressed. The innovations necessary to tackle these and other challenges in the modeling of [NiFe]-H₂ases will not only bring chemists closer to useful synthetic analogues, but also deepen our understanding of the unusual chemistry in Nature’s toolkit.

5. [Fe]-H₂ASES

5.1. Enzyme Structure and Function

Of all the H₂ produced in Nature, approximately half is used in the conversion of CO₂ to CH₄, a process that underpins the metabolism of methanogens. To this end, these archaea express four [NiFe]-H₂ases, as well as, when starved for Ni, one [Fe]-H₂ase³⁸¹ (the “third hydrogenase”, 43 kDa).³⁸² Discovered after³⁸³ the more O₂-sensitive FeFe and NiFe proteins,³⁸⁴ [Fe]-H₂ase is not a redox enzyme and thus does not feature Fe–S clusters. In fact, it was initially purported to be completely free of Fe,³⁸⁵ although it is now known as a metalloenzyme after all,³⁸⁶ whose activity centers around a Cys-ligated Fe-guanylylpyridinol cofactor (Fe-GP). This structure was shown by IR (ν_CO = 2011, 1944 cm⁻¹)³⁸⁷ and NRVS³⁸⁸ analyses to contain a Fe(CO)₂(S-Cys) fragment bound to a unique guanylylpyridinol/pyridonate (GP) through its acyl (ν_CO = 1697 cm⁻¹) and pyridyl donors. The GP ligand can be isolated, although in a form with the acyl group oxidized to the corresponding carboxylate.³⁸⁹ Initial holoenzyme structural studies included this additional O atom, although subsequent mutagenesis and crystallography³⁹⁰ confirmed that the acyl form is indeed at the active site (Figure 50).³⁹¹

Figure 50.

X-ray structure of the [Fe]-H₂ase active site from Methanocaldococcus jannaschii (PDB code 3F4Z).³⁹⁰ H atoms are omitted for clarity.

The Fe(CO)₂(S-Cys)(guanylylpyridinol/pyridonate) unit is square pyramidal (τ = (175.1° – 173.7°)/60° = 0.02),³⁰⁶ with the two CO ligands being cis: one trans to the basic thiolate and the other trans to the pyridine/pyridonate. The acyl ligand exerts its trans effect on a sixth site, which when not passivated by a H₂O ligand, is available to bind H₂ (as well as inhibitors such as CO). When the enzyme adopts an “open” state, H₂O is displaced by methenyl-H₄MPT⁺ (Figure 51). After folding to the closed state, there remains a 4 Å-wide hydrophobic channel accessible to solvent and H₂. Diffusion of the latter to the active site and release of a proton from the pyridinol moiety to a neighboring His residue affords a Fe(η²-H₂)(CO)₂(S-Cys)-(guanylylpyridonate) complex proximal to methenyl-H₄MPT⁺. The H₂ substrate is now primed to undergo heterolysis, with the basic O atom of the pyridonate³⁹² and methenyl-H₄MPT⁺ serving as the H⁺ and H⁻ acceptors, respectively. Release of the product and ligation of H₂O completes the cycle.³⁹³ Overall, the consumption of H₂ is close to isoergic (ΔG = −5.5 kJmol⁻¹), as one would expect given the comparable potentials for the H⁺/H₂ (−414 mV) and methenyl-H₄MPT⁺/methylene-H₄MPT (−390 mV) couples at pH 7, as well as the rates of H₂ splitting (TOF = 215 s⁻¹, pH 7.5) and formation (TOF = 555 s⁻¹, pH 6.5) mediated by [Fe]-H₂ase.³⁸³

Figure 51.

Catalytic cycle proposed for [Fe]-H₂ase.⁹

Inasmuch as pyridonate is similar to CN⁻ in terms of donor/acceptor properties,³⁹⁴ the active site of [Fe]-H₂ase shares certain features with that of [FeFe]-H₂ase.³⁹⁵ Each has a basic cofactor (secondary amine or pyridonate O atom) adjacent to the transiently vacant Fe coordination site. Yet [Fe]-H₂ase does not perform redox and is not highly O₂-sensitive,³⁹⁴ and its Fe site is invariably found to exist in a low-spin divalent state, certainly a good fit for H₂ binding and heterolysis.

The catalytic cycle for [Fe]-H₂ase, like that for [FeFe]- or [NiFe]-H₂ase, is proposed to involve hydride species. One mechanism, based on DFT calculations of a truncated active site model, invokes an elongated dihydrogen bond of the form Fe(II)–H⋯H–O_pyridinol.²⁷⁴ However, calculations in which all amino acids are considered indicate that Fe hydrides need not be considered if H₂ scission is concerted. In such a case, the Fe(η²-H₂)(guanylylpyridonate) group, in which the Fe(η²-H₂) moiety has an estimated pK_a of 9,³⁹⁶ converts to Fe(guanylylpyridinol) without the intermediacy of a hydride, as H⁺ abstraction by pyridonate and H⁻ migration to the substrate occur simultaneously.³⁹⁷ The methylene-H₄MPT product then leaves the active site, but not before the protein undergoes large-scale movement to open its binding pocket. This study employs molecular dynamics (MD) simulations to probe the opening and closing of the binding pocket, and quantum mechanical/molecular mechanical (QM/MM) calculations to investigate H₂ splitting on select MD snapshots.³⁹⁷ In contrast to [Fe]-H₂ase, no large displacements are required to shuttle small substrates to the cores of [FeFe]-H₂ase and [NiFe]-H₂ase.

While not an electrocatalyst (and thus not relevant to fuel cell applications), [Fe]-H₂ase does deliver H⁻ stereoselectively to its substrate. The enzyme could thus serve as a basis for catalysts in organic synthesis, although perhaps due to the many specific supramolecular interactions involved in binding substrate, no molecules other than methenyl-H₄MPT⁺ induce H₂ cleavage.³⁹⁸ Time will tell whether the methenyl-H₄MPT⁺/H₂/Fe-GP ternary complex is absolutely necessary or whether the H⁻ delivery can be generalized. Motivated by this, as well as the prospect of learning more about the catalytic cycle, synthetic chemists have taken to reproducing the [Fe]-H₂ase active site.³⁹⁹

5.2. [Fe]-H₂ase Synthetic Modeling

5.2.1. Models of CO-Inhibited [Fe]-H₂ase

H₂ases are inhibited by CO, which often (but not always, see section 3.1) outcompetes H₂ for metal binding sites, thereby filling their coordination spheres. In the case of [Fe]-H₂ase, the inhibited enzyme features a low-spin facial Fe(II)(CO)₃ fragment giving rise to ν_CO bands at 2074, 2020, and 1981 cm⁻¹ (Table 6). Synthetic modeling of [Fe]-H₂ase began with mimics of the inhibited enzyme, which are described here before highlighting efforts to mimic the coordinatively unsaturated Fe site in the active states.

Table 6.

CO-Stretching Bands for Active and CO-Inhibited Forms of [Fe]-H₂ase, as well as for Selected Synthetic Models

species	νCO/cm−1
[Fe]-H2ase	2011, 1944
[Fe]-H2ase-CO	2074, 2020, 1981
[52]	2075, 2020, 1981
[53]	2032, 1987
[54]	2026, 1961
[55]	2027, 1964
[56]	2022, 1958
[57]	2068, 2050, 1996
[58]	2084, 2018, 2003
[59]	2013, 1950
[60]	2022, 1958
[61HBr]	2030, 1974
[62]	2003, 1940
[63I]	1945
[RuCp(CO)2]−	1887, 1802

The finding in 2009 that existing X-ray data for [Fe]-H₂ase indicated an Fe-bound acyl ligand inspired a flurry of synthetic activity to reproduce this common organometallic fragment. Despite the multitude of octahedral ferrous complexes, the enzyme has a unique ligand set, and work in the area has focused on reproducing the structure (inner coordination sphere) and spectroscopy (IR bands). For example, the first acyl thiolato monoiron complex [Fe(CO)₃(SPh)(2-diphenylphosphinobenzoyl)] ([52], Figure 52) features five of the six donors present at the [Fe]-H₂ase active site and gives rise to ν_CO bands closely matching those of the enzyme.^400,401 Moreover, the synthesis of [52] by oxidative addition of a thioester to Fe(0) is biomimetic in that the guanylylpyridinol/pyridonate (GP) cofactor is also derived from a thioester.⁴⁰² Much like the CO-inhibited enzyme, [52] readily loses CO, although the 5-coordinate product converts to a dinuclear species.

Figure 52.

Early models for CO-inhibited [Fe]-H₂ase feature some of the donor groups present in the enzyme.^400,401,403

Complexes of Fe featuring 2-pyridonate moieties have also been reported, one example of which is [Fe(CO)₂(PPh₃)I(6-methyl-2-pyridonate)] ([53]).⁴⁰³ A dicarbonyl, this complex further contrasts the active site in that the pyridonate group in [53] binds through both N and O atoms.⁴⁰⁴ A key feature of [Fe]-H₂ase is the positioning of the basic O atom near but not bonded to the Fe site. Early efforts toward octahedral analogues of Nature’s 5-ccordinate FLP are now described.

Synthetic investigations into 2-pyridylcarbamoyl^405,406 and pyridylacyl complexes have afforded models closely resembling the CO-inhibited [Fe]-H₂ase active site. Indeed, this motif has been incorporated into complexes of the form [Fe(CO)₂(2-pyridinethiolate)(2-pyridylacetyl)] ([54–56], Figure 53), which feature donors resembling the native ligands, with an additional N-donor.^407–409 While [54–56] feature protected acylpyridinols, Fe complexes of the unmasked ligand have also been prepared, the coordination sphere in tricarbonyl [57] being similar to that in the CO-inhibited enzyme with I⁻ in place of Cys-S⁻. These pyridylacyl complexes are not assembled through a biomimetic thioester oxidative addition, but instead form by more traditional organometallic routes. In the case of [54]–[56], this involves nucleophilic attack of [Fe(CO)₄]²⁻ onto a 2-picolyl tosylate, with subsequent insertion of CO into the Fe–alkyl bond.^408–410 In the synthesis of [58], Fe instead serves as an electrophile, with [Fe(CO)₅] undergoing addition of 2-picolide anion such that a CO ligand is converted to an acyl.⁴⁰⁷

Figure 53.

Acyl- and carbamoylpyridine models for CO-inhibited [Fe]-H₂ase.

The tricarbonyls [57] and [58] were employed in an [Fe]-H₂ase reconstitution study analogous to that described for [FeFe]-H₂ase (vide supra). Incubation of [57] or [58] with apo-[Fe]-H₂ase, which lacks the Fe-GP cofactor,⁴¹¹ afforded products whose IR spectra indicated ligation of Fe(CO)₂(acylpyridinol/acylmethylpyridinol) to Cys176. Although the methylated derivative was unreactive toward H₂/methenyl-H₄+MPT⁺ or H⁺/methylene-H₄MPT, the acylpyridinol-containing protein catalyzed these reactions. The semisynthetic system was 2 orders of magnitude slower than the native one, a result attributed to subtle electronic differences between GP and the acylpyridinol ligand. Nevertheless, this work highlights the importance of pyridinol/pyridonate as a H⁺ relay, and refutes some older DFT studies that, by analogy with [NiFe]-H₂ases, implicated Cys-S⁻ as a relay.^114,274

Given that pentacoordinate Fe is crucial to the activity of H₂ases, section 5.2.2 highlights 16e⁻ Fe complexes that most closely mimic the [Fe]-H₂ase active site in both structure and reactivity. This will serve as a prelude to studies into H₂ activation with [Fe]-H₂ase models.

5.2.2. Models of Active [Fe]-H₂ase States

A significant breakthrough in modeling the [Fe]-H₂ases came with the finding that a combination of thiolate, pyridylacyl, and other π-bonding ligands can stabilize unsaturated Fe(II) sites. Addition of thiolate to [58] replaces not only I⁻ but also CO, such that square-pyramidal complex [59] forms (Figure 54).⁴¹² However, this species does not bind H₂O or H₂ (or O₂), a result attributed to its electron richness. Although the similarity of its ν_CO frequencies to those of the enzyme suggests that other factors may also play a role, DFT calculations indicate that H₂ binding is thermodynamically unfavorable (ΔG = 10.2 kcal mol⁻¹).⁴¹³

Figure 54.

Pentacoordinate acylpyridine models [59] and [60] replicate inner Fe coordination sphere of the active [Fe]-H₂ase states, yet do not bind H₂.^412,414

Building on the chemistry of [59], the higher-fidelity structural model [60] features the hydroxyl group characteristic of [Fe]-H₂ase.⁴¹⁴ The species was identified according to NMR and IR spectroscopies, but was found to be rather labile (although less so than its 2,6-dimethylphenylthiolato and 1-propanethiolato congeners). Although spectroscopically and structurally similar to the active site, it is unreactive toward H₂. This may indicate that secondary interactions involving the [Fe]-H₂ase pocket, the inclusion of methenyl-H₄MPT⁺, and the presence of a more substituted pyridyl ligand are important factors in the activation of H₂.²⁶

Another example of a five-coordinate [Fe]-H₂ase model could be prepared from dehydrobromination of octahedral Fe(II) carbamoyl [61HBr] (Figure 55).⁴¹⁵ Similar in structure to [60], the proposed kinetic product could not be characterized, as it rapidly converted to the cyclometalated species [62]. The latter product can also be prepared from hydride [61(H)H] (δ(¹H) −5.1 ppm), which, in turn, is generated from [61HBr] and HBEt₃⁻. While displacement of Br⁻ with H⁻ is not a biomimetic transformation, the donor atoms around [61(H)H] match those of the active site in a transient hydride-bearing intermediate. Such an enzyme state has yet to be experimentally confirmed, although some theoretical studies invoke it as a short-lived species, very similar to [61(H)H], which is unstable above −40 °C. While the generation of [61(H)H] indicates that the present type of carbamoyl species can support a hydride ligand, the second coordination sphere of [61(H)H] does not feature a basic moiety to engage in H₂ heterolysis.

Figure 55.

Preparation and reactivity of the Fe hydrido carbamoyl [61(H)H].⁴¹⁵

The close resemblance that the Fe coordination spheres of acylpyridine [61] and related species have to that of [Fe]-H₂ase indicates that differences in reactivity may arise from secondary interactions. In search of H₂ heterolysis, such structural models have been modified to incorporate more basic fragments into their architecture, even though this may decrease their resemblance to the active site. For example, combination of tricarbonyl [58] with the bidentate ligand Et₂PCH₂N(Me)-CH₂PEt₂ affords monocarbonyl [63I], from which I⁻ dissociation can occur (Figure 56).⁴¹⁶ The putative 5-coordinate species is well set up to cleave H₂, although no direct evidence of the ammonium hydride has been reported, even under high H₂ pressures. The activation of H₂ by [63]⁺ was evidenced in the scrambling of H₂/D₂ mixtures, with the additional scrambling of H₂ and CH₃CO₂D suggesting a heterolytic mechanism is at play. Bearing some similarity to bioinspired Ni(PNP) and Fe(PNP) electrocatalysts, the present species represents the first functional model for [Fe]-H₂ase and underscores the importance of a proximal base for FLP activity. Even though no hydride complex could be detected, such work importantly demonstrates that an Fe-based [Fe]-H₂ase model can activate H₂ outside of the enzyme in the absence of the native methenyl-H₄MPT⁺ substrate.

Figure 56.

Polar hydrogenation of an aldehyde catalyzed by [63]⁺.⁴¹⁶

The inertness toward H₂ shown by many 5-coordinate Fe complexes described above comes as a disappointment, but is unsurprising given that ferrous dihydrogen complexes are characteristically cationic.³⁵ Furthermore, when observed, either directly with [61(H)H] or indirectly in the case of [63]⁺, the hydrides are unstable. This has led efforts in functional mimicry to also consider structures departing from the Fe(pyridylacyl) paradigm. For example, triarylimidazolium salts of [RuCp-(CO)₂]⁻ represent FLPs that parallel the biological combination of methenyl-H₄MPT⁺ and the Fe-GP cofactor (Figure 57).⁴¹⁷ Indeed, the electrophilic C centers of the triarylimidazolium (δ(¹³C) 169.6 ppm) and methenyl-H₄MPT⁺ (165.34 ppm)⁴¹⁸ would appear to be comparably reactive. When treated with H₂, the synthetic Lewis pair heterolyzes H₂ with the cation receiving H⁻ and the Ru⁰ center accepting a H⁺ by oxidative addition.

Figure 57.

Complex [RuCp(CO)₂]⁻ exhibits [Fe]-H₂ase behavior, effecting H₂ heterolysis and delivery of H⁻ to an imidazolium substrate resembling methenyl-H₄MPT⁺.⁴¹⁷

The heterolysis reaction proceeds fastest when a polymeric Ru–CO–Cp species is used in place of [RuCp(CO)₂]⁻. A slightly more biomimetic choice would of course be [FeCp-(CO)₂]⁻, but it directly reduces the imidazolium cation. Overall, H₂ heterolysis is similar to the H₂ fission mediated by the [(dppe)Ni(pdt)Fe(CO)₃]/B(C₆F₅)₃ FLP (section 4.2.3).⁶³ Since the H⁻ and H⁺ heterolysis products are delivered respectively to the organic substrate and metal center, the process is not biomimetic.

5.3. Concluding Remarks and Future Challenges

Synthetic modeling of [Fe]-H₂ases, particularly since the clarification of the active site structure in 2009, has afforded several high-fidelity structural mimics. This work began with the preparation of octahedral Fe(II) complexes featuring some or all of the thiolate, carbonyl, acyl, and pyridonate/pyridinol ligands present in the native protein. More relevant to the active enzyme are 5-coordinate species, in which a site is free for H₂ or H⁻ binding. Such 16e⁻ species have been reported, but typically do not exhibit [Fe]-H₂ase reactivity despite closely mimicking the structure and spectroscopy. This points to the need for tuning the second coordination sphere, the importance of which is demonstrated by artificial reconstitution studies of [Fe]- and [FeFe]-H₂ase. Indeed, the high H–H bond strength means that H₂ heterolysis can only be facile when both Lewis acidic and basic components are present and well arranged. To this end, incorporation of an artificial amine H⁺ relay in [63]⁺ afforded a catalyst that activates and transfers H₂. No η² –H₂ or H⁻ complex intermediates have been identified in this case, yet the same is true for [Fe]-H₂ases, with the possible role of Fe–H moieties still a subject of debate. Future efforts in this area may address these questions and also aim to synthetically replicate the cooperative binding of methenyl-H₄MPT⁺ and H₂ to the enzyme active site.

6. CONCLUSIONS

The study of H₂ases reminds us that organometallic chemistry is literally alive and well, providing the machinery that allows Nature to process the most abundant element in the universe. It is worth mentioning here that examples of metal hydrides in biology are not limited to H₂ases. For example, ENDOR studies on nitrogenase (N₂ase) suggest the presence of two hydrides attached to its Fe₇MoS₉C cofactor during H⁺ reduction.⁴¹⁹ Each of the chemically equivalent hydrides is thought to bridge two Fe centers, a situation rather reminiscent of Fe(pdt)(μ-H)Fe model complexes. It is hardly surprising that hydrides play a role in N₂ases, as these exhibit H₂ase behavior in that they can process protons and electrons independently of N₂ reduction to NH₃.⁴²⁰ An inner sphere mechanism for this HER would necessitate metal hydride participation. Such chemistry is also at play in the related enzyme nickel–iron carbon monoxide dehydrogenase ([NiFe]-CODHase), which utilizes protons and electrons in the interconversion of CO₂ with CO and H₂O.⁴²¹ In this case, the Fe₄NiS₄ cofactor can feature a Ni–H or Ni–H–Fe moiety, into which CO₂ inserts during the catalytic cycle.^422,423 The similarities between the enzymes discussed here indicate that acid–base and redox reactions are tasks carried out well by FeS-containing enzymes, especially when they are low-spin.

Returning to the H₂ases, each of their unique organoiron sites¹⁵ facilitates rapid H₂ cleavage and/or formation, motivating chemists to unravel the mechanisms by which they operate. Characterizing the many H₂ase intermediates, including key hydride-bearing states, necessitates the use of several complementary experimental and theoretical techniques. These techniques are also applicable to synthetic hydrides inspired by the H₂ase active sites, with small molecules often being more tractable and easily studied. Many catalytically competent examples are known, and the prospect of their use in a future hydrogen economy is attractive.

One reality of H₂ase modeling is that there will always be differences between enzymes and their small molecule analogues. The [FeFe]- and [NiFe]-H₂ases tightly encapsulate their active sites, ruling out bimolecular pathways and permitting only minor solvation changes and ligand reorganization during catalysis.⁴²⁴ This is not the case for model complexes, which are conformationally less restricted, as exemplified by the t-H → μ-H conversion at play in [dppv(OC)Fe(adt)Fe(H)(dppv)-(CO)]⁺ ([17(H)]⁺),²⁵⁸ which does not occur for the [FeFe]-H₂ases, whose cores are held in place by intricate H-bonding networks. Similarly, Ni isomerization in [(dppe)Ni(pdt)Fe-(CO)₃] ([26]) is necessary to accommodate protonation or oxidation,¹⁶⁰ contrasting the [NiFe]-H₂ases, in which Ni is rigidly held in a seesaw coordination. Indeed, preorganized, rigid metal sites and second coordination spheres play key roles in minimizing both overpotential and the possibility of side reactions. On the other hand, some degree of conformational flexibility is required for the systems to find the optimal distances and angles for intramolecular proton or hydride transfer reactions.³⁷⁶

Incorporated into H₂ase protein scaffolds are hydrophobic channels for H₂ propagation to and from their active sites. This is advantageous in selectively introducing substrate to the catalytic metals, but in the case of [NiFe]-H₂ase, its channels may be prone to blockage when H₂ partial pressures are high.³³⁷ The more exposed nature of model catalysts relative to the enzyme active sites ensures that the former will not suffer from this blockage, and access to H₂, protons, and electrons (as well as inhibitors) is enhanced. With compact profiles, model complexes may also bind electrode surfaces closely and densely such that electron transfer might be rapid and catalytic current densities might be high.⁴²⁵ Ideally, synthetic catalysts would operate at high rates and low overpotentials, as well as exhibit a robustness that enables operation over wide temperature and pH ranges that may be necessary for implementation in fuel cells.⁴²⁶

The amino acids surrounding the H₂ase active sites serve to shield mixed-valent or coordinatively unsaturated states from participating in any reactions other than those of the catalytic cycle. Many mimics of coordinatively saturated, inactive enzyme states have been prepared, but certainly the reactive forms have proven more elusive. For example, terminal Fe(II)Fe(II)(t-H) models for [FeFe]-H₂ase have been isolated, yet their 1e⁻ reduced mixed-valent forms have not. Absent from [NiFe]-H₂ase modeling are satisfactory 33e⁻ Ni(I)Fe(II) models for Ni-L, an isolable 32e⁻ Ni(II)Fe(II) model for Ni-SI_a (without its coordinated H₂O), as well as any Ni(III)(μ-H)Fe(II) complex that would resemble the key hydride-bearing Ni-C state. Efforts toward the latter two will involve incorporation of strongly basic ligands to stabilize coordinatively unsaturated and/or high-valent metal sites.

In general, it is debatable whether characterizing catalysts in their many states is necessary or even possible if the species in question is highly active. But it is undeniable that the native H₂ases are significantly faster and more energy-efficient than model complexes. While many models show reactivity toward protons, the discrepancy is most apparent when considering H₂ activation.⁴²⁷ Despite considerable progress over the past decade, the few H₂ase models that heterolyze H₂ do so very slowly. Thus, bidirectional synthetic catalysts are scarce, and those that exist exhibit slow rates for electrocatalytic H₂ oxidation and/or evolution arising from significant overpotentials associated with their mechanisms. The bidirectionality of many [FeFe]- and [NiFe]-H₂ases has evolved not only from a biological need to both consume and dispense reducing equivalents, but also to operate in very narrow potential ranges and at low concentrations of H₂ and H^+.6 Furthermore, for a H₂-processing catalyst to be bidirectional, it follows that interconversion of H₂ and hydrides must be rapid and close to isoergic.⁴²⁸ Such a transformation has been identified as a major challenge in organometallic chemistry,⁴²⁹ and the viability of H₂ as an energy vector is reliant on both its efficient production and consumption. Electrocatalysis is certainly not the only means of H₂ processing, but it is notable that the reaction of H₂ and O₂⁴³⁰ in fuel cells is potentially more efficient than traditional combustion, in which energy losses from the Carnot cycle are unavoidable.

Regardless of the motivation for modeling the H₂ases—be it fundamentally learning more about enzymes or perhaps developing base-metal hydrogen-processing catalysts—this work is invariably instructive, and intricacies associated with apparently simple reactions are continually uncovered. At the heart of H₂ases and their functional models are metal hydrides,⁴³¹ which are more pervasive and of broader relevance than many would have guessed. A deeper understanding of these hydrides, obtained through a confluence of biological and organometallic chemistry research, will place us closer to a sustainable future.

ABBREVIATIONS

Arg

arginine residue

adt²⁻

azadithiolate (⁻SCH₂N(H)CH₂S⁻)

(adt^R)²⁻

substituted azadithiolate (⁻SCH₂N(R)-CH₂S⁻)

BAr^F₄⁻

(3,5-(CF₃)₂C₆H₃)₄B⁻

bdt²⁻

1,2-benzenedithiolate

Bn

benzyl

Bu

butyl

CODHase

carbon monoxide dehydrogenase

Cp⁻

cyclopentadienide

Cp*⁻

pentamethylcyclopentadienide

CV

cyclic voltammetry

Cy

cyclohexyl

Cys

cysteine residue

dcpe

1,2-bis(dicylcohexylphosphino)ethane

DBU

1,8-diazabicycloundec-7-ene

DFT

density functional theory

dmpe

1,2-bis(dimethylphosphino)ethane

dppbz

1,2-bis(diphenylphosphino)benzene

dppe

1,2-bis(diphenylphosphino)ethane

dppn

1,8-bis(diphenylphosphino)naphthalene

dppv

1,2-bis(diphenylphosphino)ethane

edt²⁻

1,2-ethanedithiolate (⁻S(CH₂)₂S⁻)

ENDOR

electron nuclear double resonance

EPR

electron paramagnetic resonance

Et

ethyl

Fc

ferrocene

Fc*

decamethylferrocene

Fe-GP

Fe-guanylylpyridinol cofactor

FLP

frustrated Lewis pair

IR

infrared

H₂ase

hydrogenase

HER

hydrogen evolution reaction

Hmd

H₂-forming methylenetetrahydromethanopterin dehydrogenase

HYSCORE

hyperfine sublevel correlation

Lys

lysine residue

Me

methyl

methenyl-H₄MPT⁺

methenyltetrahydromethanopterin

methylene-H₄MPT

methylenetetrahydromethanopterin

N₂ase

nitrogenase

NHE

normal hydrogen electrode

NMR

nuclear magnetic resonance

ⁿPr

n-propyl

NRVS

nuclear resonance vibrational spectroscopy

OTf⁻

triflate (CF₃SO₃⁻)

PCET

proton-coupled electron transfer

pdpp²⁻

1,3-propanedi(phenylphosphide) (⁻PhP-(CH₂)₃PPh⁻)

pdt²⁻

1,3-propanedithiolate (⁻S(CH₂)₃S⁻)

Ph

phenyl

ppm

parts per million

Ser

serine residue

Tf

trifluoromethanesulfonyl

THF

tetrahydrofuran

TOF

turnover frequency

Ts

4-toluenesulfonyl

Biographies

David Schilter was educated at The University of Sydney, from which he received B.Sc. (Hons.) and Ph.D. degrees in 2005 and 2009, respectively. These studies were followed by postdoctoral research at the University of Illinois at Urbana–Champaign and at the IBS Center for Multidimensional Carbon Materials at UNIST, where he has served as a research scientist since 2014. His interests include coordination, organometallic, and materials chemistry.

James M. Camara earned his Ph.D. in chemistry from Columbia University in 2009, where he studied the polymerization and transmetalation chemistry of zirconium and aluminum complexes. Following the completion of his Ph.D., he was a postdoctoral fellow at the University of Illinois at Urbana–Champaign where he conducted synthetic and mechanistic studies on models of the [FeFe]-H₂ase active site. He is currently an assistant professor of chemistry at Yeshiva University in New York City. His research interests are centered around various aspects of mechanistic organometallic chemistry including the use of redox-active ligands in catalysis, development of new transmetalation reactions, and small molecule activation by metal centers.

Mioy T. Huynh received his B.S. in chemistry from the University of California, Los Angeles, in 2012. He is currently a graduate student at the University of Illinois at Urbana–Champaign in Sharon Hammes-Schiffer’s research group. His research interests include the investigation of proton-coupled electron transfer reactions in catalytic systems.

Sharon Hammes-Schiffer received her B.A. in chemistry from Princeton University in 1988 and her Ph.D. in chemistry from Stanford University in 1993, followed by two years at AT&T Bell Laboratories as a postdoctoral research scientist. She was the Clare Boothe Luce Assistant Professor of Chemistry and Biochemistry at the University of Notre Dame from 1995 to 2000 and spent the next 12 years at The Pennsylvania State University, initially as the Shaffer Associate Professor of Chemistry and later as the Eberly Professor of Biotechnology. In 2012 she became the Swanlund Chair and professor of chemistry at the University of Illinois Urbana–Champaign. Dr. Hammes-Schiffer’s research centers on the investigation of electron, proton, and proton-coupled electron transfer reactions in chemical, biological, and interfacial processes. Her work encompasses the development of analytical theories and computational methods, as well as applications to a wide range of experimentally relevant systems.

Thomas B. Rauchfuss was born in 1949 in Baltimore, MD. He received his undergraduate degree from the University of Puget Sound (1971) and his Ph.D. from Washington State University (1976). After a postdoc with David Buckingham at the Australian National University, he started his independent career at the University of Illinois at Urbana–Champaign in 1978, where he has remained. He has also studied at the following institutions: the University of Auckland, University of Louis Pasteur, and the Technical University of Karlsruhe. His research focuses on synthetic inorganic and organometallic chemistry, with an emphasis on environmentally motivated themes. For the past decade, his group has conducted research on synthetic modeling of the active sites of the hydrogenase enzymes.

APPENDIX

A.1. Proposed Pyramidal Inversion of Azadithiolate During Catalysis

Figure 58 shows proton transport to and from the [2Fe] site in [FeFe]-H₂ase.

A.2. Determining the pK_a of a Metal Hydride

If titrating a metal hydride MH⁺ with a base B^50,51

$$\begin{array}{l}{\text{MH}}^{+}+\mathrm{B}\rightleftharpoons {\text{BH}}^{+}+\mathrm{M}\hfill \\ K_{\mathrm{eq}}=\text{unknown equilibrium constant}\hfill \\ {\text{MH}}^{+}\rightleftharpoons \mathrm{M}+{\mathrm{H}}^{\text{+}}\hfill \\ K_{\mathrm{a}}\left(\mathrm{M}{\mathrm{H}}^{+}\right)=\text{unknown acid dissociation constant}\hfill \\ {\text{BH}}^{+}\rightleftharpoons \mathrm{B}+{\mathrm{H}}^{\text{+}}\hfill \\ K_{\mathrm{a}}\left(\mathrm{B}{\mathrm{H}}^{+}\right)=\text{known acid dissociation constant }\left(\text{ref}51\right)\hfill \\ {\text{BH}}^{+}\text{+B}\rightleftharpoons {\text{BHB}}^{\text{+}}\hfill \\ K_{\mathrm{f}}\left(\mathrm{B}{\mathrm{H}}^{+}\right)=\text{known acid homoconjugation}\hfill \\ \text{constant }\left(\text{ref}50\right)\hfill \end{array}$$

1. A known amount of base B (=[B]_total) is added to a known amount of MH⁺. The values [MH⁺] and [M] in solution are determined, typically using IR or NMR spectroscopy with an internal standard.

2. [BH⁺] is determined according to

   $$\begin{array}{l}{\left[{\text{BH}}^{+}\right]}^2+\{{\left[\mathrm{B}\right]}_{\text{total}}-2\left[\mathrm{M}\right]+1/K_{\mathrm{f}}({\text{BH}}^{+})\}\left[{\text{BH}}^{+}\right]\hfill \\ -\left[\mathrm{M}\right]/K_{\mathrm{f}}({\text{BH}}^{+})=0\hfill \\ \Rightarrow \left[{\text{BH}}^{+}\right]=\{\{{\{{\left[\mathrm{B}\right]}_{\text{total}}-2\left[\mathrm{M}\right]+1/K_{\mathrm{f}}\left({\text{BH}}^{+}\right)\}}^2\hfill \\ {+4\left[\mathrm{M}\right]/K_{\mathrm{f}}\left({\text{BH}}^{+}\right)\}}^{1/2}-{\{\left[\mathrm{B}\right]}_{\text{total}}\hfill \\ -2\left[\mathrm{M}\right]+1/K_{\mathrm{f}}({\text{BH}}^{+})\}\}/2\hfill \end{array}$$
3. [BHB⁺] is determined using mass balance on M:

   $$\begin{array}{l}{\left[{\text{MH}}^{+}\right]}_{\text{initial}}=\left[\mathrm{M}\right]+\left[{\text{MH}}^{+}\right]\hfill \\ {\left[{\text{MH}}^{+}\right]}_{\text{initial}}=\left[{\text{BH}}^{+}\right]+\left[{\text{MH}}^{+}\right]+\left[{\text{BHB}}^{+}\right]\hfill \\ \left[\mathrm{M}\right]=\left[{\text{BH}}^{+}\right]+\left[{\text{BHB}}^{+}\right]\hfill \\ \Rightarrow \left[{\text{BHB}}^{+}\right]=\left[\mathrm{M}\right]-\left[{\text{BH}}^{+}\right]\hfill \end{array}$$
4. [B] is determined using mass balance on B:

   $$\left[\mathrm{B}\right]={\left[\mathrm{B}\right]}_{\text{total}}-\left[{\text{BH}}^{+}\right]-2\left[{\text{BHB}}^{+}\right]$$

5. All the values [B], [BH⁺], [M], and [MH⁺] are now known such that K_eq and thus pK_a(MH⁺) = pK_eq + pK_a(BH⁺) can be computed.

   If titrating a metal complex M with a weak acid HA:

   $$\begin{array}{l}\mathrm{M}+\text{HA}\rightleftharpoons {\text{MH}}^{+}+{\mathrm{A}}^{-}\\ K_{\text{eq}}=\text{unknown equilibrium constant}\\ {\text{MH}}^{+}\rightleftharpoons \mathrm{M}+{\mathrm{H}}^{+}\\ K_{\mathrm{a}}\left(\mathrm{M}{\mathrm{H}}^{+}\right)=\text{unknown acid dissociation constant}\\ \text{HA}\rightleftharpoons {\mathrm{H}}^{+}+{\mathrm{A}}^{-}\\ K_{\mathrm{a}}\left(\mathrm{H}\mathrm{A}\right)=\text{known acid dissociation constant}\\ \text{HA}+{\mathrm{A}}^{-}\rightleftharpoons {\text{AHA}}^{-}\\ K_{\mathrm{f}}\left(\mathrm{H}\mathrm{A}\right)=\text{known acid homoconjugation constant}\end{array}$$

1. A known amount of acid HA (=[A⁻]_total) is added to a known amount of M. The values [MH⁺] and [M] in solution are determined, typically using IR or NMR spectroscopy with an internal standard present.

2. [A⁻] is determined according to

   $$\begin{array}{l}{\left[{\mathrm{A}}^{-}\right]}^2+\{{\left[{\mathrm{A}}^{-}\right]}_{\text{total}}-2\left[{\text{MH}}^{+}\right]+1/K_{\mathrm{f}}(\text{HA})\}\left[{\mathrm{A}}^{-}\right]\\ -\left[{\text{MH}}^{+}\right]/K_{\mathrm{f}}\left(\text{HA}\right)=0\\ \Rightarrow \left[{\mathrm{A}}^{-}\right]=\{{\{\{{\left[{\mathrm{A}}^{-}\right]}_{\text{total}}-2\left[{\text{MH}}^{+}\right]+1/K_{\mathrm{f}}(\text{HA})\}}^2\\ {+4\left[{\text{MH}}^{+}\right]/K_{\mathrm{f}}(\text{HA})\}}^{1/2}-\{{\left[{\mathrm{A}}^{-}\right]}_{\text{total}}-2\left[{\text{MH}}^{+}\right]\\ +1/K_{\mathrm{f}}(\text{HA})\}\}/2\end{array}$$
3. [AHA⁻] is determined using mass balance on A⁻:

   $${\left[{\mathrm{A}}^{-}\right]}_{\text{total}}=\left[{\mathrm{A}}^{-}\right]+\left[\text{HA}\right]+2\left[{\text{AHA}}^{-}\right]$$
   One does not directly know [HA], but the H⁺ from reacted HA formed either [MH⁺] or [AHA⁻] (mass balance on H⁺):

   $$\left[{\text{MH}}^{+}\right]+\left[{\text{AHA}}^{-}\right]={\left[{\mathrm{A}}^{-}\right]}_{\text{total}}-\left[\text{HA}\right]$$

   and

   $$\left[\text{HA}\right]={\left[{\mathrm{A}}^{-}\right]}_{\text{total}}-\left[{\text{MH}}^{+}\right]-\left[{\text{AHA}}^{-}\right]$$
   Substituting this expression for [HA] into the mass balance above:

   $$\left[{\text{AHA}}^{-}\right]=\left[{\text{MH}}^{+}\right]-\left[{\mathrm{A}}^{-}\right]$$
4. [HA] is then computed from the mass balance:

   $$\left[\text{HA}\right]={\left[{\mathrm{A}}^{-}\right]}_{\text{total}}-\left[{\mathrm{A}}^{-}\right]-2\left[{\text{AHA}}^{-}\right]$$

5. All the values [A⁻], [HA], [M], and [MH⁺] are now known such that K_eq and thus pK_a(MH⁺) = pK_a(HA) – pK_eq can be computed.

A.3. Expressions for Hydricity and H Atom Donation

The hydricity of a metal hydride MH⁺ can be determined by thermodynamic cycles.^53,58,59 If the potentials E_1/2[M]^+/0 and E_1/2[M]^2+/+ are known, one can sum the four reactions

$$\begin{array}{l}{\text{MH}}^{+}\rightleftharpoons \mathrm{M}+{\mathrm{H}}^{+}\\ \mathrm{M}\rightleftharpoons {\mathrm{M}}^{+}+{\mathrm{e}}^{-}\\ {\mathrm{M}}^{+}\rightleftharpoons {\mathrm{M}}^{2+}+{\mathrm{e}}^{-}\\ {\mathrm{H}}^{+}+2{\mathrm{e}}^{-}\rightleftharpoons {\mathrm{H}}^{-}\end{array}$$

This affords

$${\text{MH}}^{+}\rightleftharpoons {\mathrm{M}}^{2+}+{\mathrm{H}}^{-}$$ (A9)

Summing the free energy changes for the four reactions gives the change for eq A9:

$$\begin{array}{l}\mathrm{\Delta }G°({\mathrm{H}}^{-})=(RT/{\text{log}}_{10}\mathrm{e})\mathrm{p}{K}_{\mathrm{a}}+F{E}_{1/2}({\left[\mathrm{M}\right]}^{+/0})\\ +F{E}_{1/2}({\left[\mathrm{M}\right]}^{2+/+})-2F{E}_{1/2}({\left[\mathrm{H}\right]}^{+/-})\end{array}$$ (A10)

where F = 96 485 C mol⁻¹, R = 8.314 J K⁻¹ mol⁻¹, e = 2.718… and T is reported in kelvin.

Note that the last term accounts for the hydricity of H₂ itself. When data are collected in MeCN at T = 298 K, and E_1/2 values are each relative to Fc^+/0, eq A10 becomes eq A10′, in which the free energy change, in kcal mol⁻¹, is

$$\begin{array}{l}\mathrm{\Delta }G°({\mathrm{H}}^{-})=1.37\mathrm{p}{K}_{\mathrm{a}}+23.06E_{1/2}({\left[\mathrm{M}\right]}^{+/0})\\ +23.06E_{1/2}({\left[\mathrm{M}\right]}^{2+/+})+76.6\end{array}$$ (A10′)

When data are instead collected in H₂O vs NHE, eq A10″ is appropriate:

$$\begin{array}{l}\mathrm{\Delta }G°({\mathrm{H}}^{-})=1.37\mathrm{p}{K}_{\mathrm{a}}+23.06E_{1/2}({\left[\mathrm{M}\right]}^{+/0})\\ +23.06E_{1/2}({\left[\mathrm{M}\right]}^{2+/+})+34.2\end{array}$$ (A10″)

Alternatively, if the potential for the 2e⁻ couple E_1/2[M]^2+/0 is known, then the following equations are summed:

$$\begin{array}{l}{\text{MH}}^{+}\rightleftharpoons \mathrm{M}+{\mathrm{H}}^{+}\\ \mathrm{M}\rightleftharpoons {\mathrm{M}}^{2+}+2{\mathrm{e}}^{-}\\ {\mathrm{H}}^{+}+2{\mathrm{e}}^{-}\rightleftharpoons {\mathrm{H}}^{-}\end{array}$$

This once more gives eq A9, the free energy change of which is now equated in terms of

$$\begin{array}{c}\mathrm{\Delta }G°({\mathrm{H}}^{-})=(RT/{\text{log}}_{10}\mathrm{e})\mathrm{p}{K}_{\mathrm{a}}+2F{E}_{1/2}({\left[\mathrm{M}\right]}^{2+/0})\\ -2F{E}_{1/2}({\left[\mathrm{H}\right]}^{+/-})\end{array}$$ (A11)

Again, when data are collected in MeCN at T = 298 K, and E_1/2 values are each relative to Fc^+/0, eq A11 becomes eq A11′, in which the free energy change, in kcal mol⁻¹, is

$$\mathrm{\Delta }G°({\mathrm{H}}^{-})=1.37\mathrm{p}{K}_{\mathrm{a}}+46.1E_{1/2}({\left[\mathrm{M}\right]}^{2+/0})+76.6$$ (A11′)

When data are instead collected in H₂O vs NHE, eq A11″ is appropriate:

$$\mathrm{\Delta }G°({\mathrm{H}}^{-})=1.37\mathrm{p}{K}_{\mathrm{a}}+46.1E_{1/2}({\left[\mathrm{M}\right]}^{2+/0})+34.2$$ (A11″)

One can also consider MH⁺ as a source of H^• instead of H⁺. In this case, summing the reactions

$$\begin{array}{l}{\text{MH}}^{+}\rightleftharpoons \mathrm{M}+{\mathrm{H}}^{+}\\ \mathrm{M}\rightleftharpoons {\mathrm{M}}^{+}+{\mathrm{e}}^{-}\\ {\mathrm{H}}^{+}+{\mathrm{e}}^{-}\rightleftharpoons {\mathrm{H}}^{•}\end{array}$$

affords the net conversion:

$${\text{MH}}^{+}\rightleftharpoons {\mathrm{M}}^{+}+{\mathrm{H}}^{•}$$ (A12)

The free energy change of this reaction is

$$\begin{array}{c}\mathrm{\Delta }G°({\mathrm{H}}^{•})=(RT/{\text{log}}_{10}\mathrm{e})\mathrm{p}{K}_{\mathrm{a}}+F{E}_{1/2}({\left[\mathrm{M}\right]}^{+/0})\\ -F{E}_{1/2}({\left[\mathrm{H}\right]}^{+/0})\end{array}$$ (A13)

When data are collected in PhCN/MeCN at T = 298 K, and E_1/2 values are each relative to Fc^+/0, eq A13 becomes eq A13′, in which the free energy change, in kcal mol⁻¹, is

$$\mathrm{\Delta }G°({\mathrm{H}}^{•})=1.37\mathrm{p}{K}_{\mathrm{a}}+23.06E_{1/2}({\left[\mathrm{M}\right]}^{+/0})+54.9$$ (A13′)

When data are instead collected in H₂O vs NHE, eq A13″ is appropriate:

$$\mathrm{\Delta }G°({\mathrm{H}}^{-})=1.37\mathrm{p}{K}_{\mathrm{a}}+23.06E_{1/2}({\left[\mathrm{M}\right]}^{+/0})+52.8$$ (A13″)