Interplay of hemilability and redox activity in modelsof hydrogenase active sitesShengda Dinga, Pokhraj Ghosha, Marcetta Y. Darensbourga, and Michael B. Halla,1

aDepartment of Chemistry, Texas A&M University, College Station, TX 77843

Edited by Brian M. Hoffman, Northwestern University, Evanston, IL, and approved October 6, 2017 (received for review June 12, 2017)

The hydrogen evolution reaction, as catalyzed by two electro-catalysts [M(N2S2)·Fe(NO)2]

+, [Fe-Fe]+ (M = Fe(NO)) and [Ni-Fe]+

(M = Ni) was investigated by computational chemistry. As nominalmodels of hydrogenase active sites, these bimetallics feature twokinds of actor ligands: Hemilabile, MN2S2 ligands and redox-active,nitrosyl ligands, whose interplay guides the H2 production mech-anism. The requisite base and metal open site are masked in theresting state but revealed within the catalytic cycle by cleavage ofthe MS–Fe(NO)2 bond from the hemilabile metallodithiolate li-gand. Introducing two electrons and two protons to [Ni-Fe]+ pro-duces H2 from coupling a hydride temporarily stored on Fe(NO)2(Lewis acid) and a proton accommodated on the exposed sulfur ofthe MN2S2 thiolate (Lewis base). This Lewis acid–base pair is initi-ated and preserved by disrupting the dative donation throughprotonation on the thiolate or reduction on the thiolate-boundmetal. Either manipulation modulates the electron density of thepair to prevent it from reestablishing the dative bond. Theelectron-buffering nitrosyl’s role is subtler as a bifunctional elec-tron reservoir. With more nitrosyls as in [Fe-Fe]+, accumulatedelectronic space in the nitrosyls’ π*-orbitals makes reductions eas-ier, but redirects the protonation and reduction to sites that post-pone the actuation of the hemilability. Additionally, two electronsdonated from two nitrosyl-buffered irons, along with two exter-nal electrons, reduce two protons into two hydrides, from whichreductive elimination generates H2.

actor ligand | biomimetic | computational mechanism |density functional theory | nitrosyl

Dihydrogen is currently a candidate for energy storage to al-leviate problems from electricity produced intermittently by

photovoltaic cells or wind turbines (1). Hydrogenases (H2ase) (2,3) are Nature’s masterpiece enzymes for H2 production and its useas an energy vector or chemical substrate; they use abundant basemetals in their catalytic active sites. An array of enzymatic andspectroscopic probes, crowned by modern protein X-ray diffractiontechnology (2, 4), provides opportunities for structure-functionanalysis of the intricate H2ase active-site molecular machinery.Strategically placed acid and base functionalities in the active siteguide and store protons and electrons for their efficient processinginto H2, or the reverse, H2 oxidation, reaction. Currently favoredmechanisms are based on earlier proposals from computationalmodeling of [FeFe]- (3, 5–8) and [NiFe]-H2ase (3, 9–12).In the active site of [FeFe]-H2ase (Fig. 1A), a diiron unit takes

up protons via an amine base strategically placed to hold andtransfer that proton to the available open site on a reduced ironto create an iron hydride, whose existence was recently spec-troscopically confirmed (13, 14); the amine then accepts anotherproton (15). Importantly, the H+/H− components of H2 arepositioned within a convenient distance for coupling over a lowbarrier (6) (Fig. 1A). A similar strategy appears to be operativein the [NiFe]-H2ase active site; whether a guanidine base fromR509 (16), which hovers over the NiFe core and is required forfull enzyme activity, is the proton delivery agent itself, or a cys-teine (C546) thiolate sulfur, bound to the Ni (17), facilitates theultimate H+/H− coupling, is not firmly established. Structuralevidence from recent high-resolution X-ray diffraction indicates

the arrangement shown in Fig. 1B, finding a thiol proton nearbya hydride accommodated in a bridge position between Ni and Fe(17), remarkably predicted by density functional theory (DFT)calculations two decades ago (9, 10). Thus, in both hydrogenasesthe hydride-protonation mechanism (HP, also known as het-erolytic coupling) accounts for H2 production (3).Interestingly, while the major function of nitrogenase (N2-ase) is

nitrogen fixation, it is known that a molecule of H2 is an obligatoryside product as one molecule of N2 is fixed into NH3 (18). Fourequivalents of electrons and four protons are required before theH2 is released and the N2 is initially fixed (19, 20); this is Nature’screative mechanism whereby the N2-ase active site can build upsufficient reduction power, stored as hydrides within the expandedFe–S cluster, to reductively activate the strong triple bond of N2.Hoffman and coworkers (20–23) proposed that such H2 releasegoes through a reductive elimination mechanism (RE, also knownas homolytic coupling of two H·) from two hydrides, thus leavingtwo electrons localized within the cluster to fix N2 (Fig. 1C). (Note:The HP mechanism is also applicable to the H2 production on N2-ase in the absence of N2; nevertheless, the capacity for N2 fixationrequires the RE mechanism.)The questions crucial to the development of molecular elec-

trocatalysts are (i) what conditions lead to the preference for REvs. HP mechanisms for H2 production assisted by Fe–S clusters,and (ii) can these conditions be replicated in small biomimeticsof these active sites, using alternate redox-active ligands.

Synthetic AnalogsThe organometallic characteristics of the hydrogenase activesites have led to a rich area of synthetic chemistry aimingto reproduce core features and delineate structure/function

Significance

Segmentation of the bimetallic electrocatalysts under in-vestigation into a metallodithiolate, bidentate S-donor ligand,and a receiver metal is effective for understanding the protonand electron uptake in H2-evolution reactions. Coexisting actor/reaction-involved ligands, i.e., electron-buffering NO and hemi-labile, chelating metallodithiolate, subtly cooperate to controlelectrocatalytic H2-production mechanisms. Two mechanismsemerge in a single catalyst to yield H2: protonation of a hydrideor reductive elimination from a metal dihydride. A Lewis acid–base pair appears by cleaving the hemilabile thiolate from themetal and serves as the reactive centers to process electrons andprotons; a protonation or a reduction on the Lewis pair modu-lates their electron densities and protects these reactive centersfrom converting back to a dative bond.

Author contributions: M.Y.D. and M.B.H. designed research; S.D. and P.G. performedresearch; and S.D., M.Y.D., and M.B.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1To whom correspondence should be addressed. Email: hall@science.tamu.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1710475114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1710475114 PNAS | Published online October 30, 2017 | E9775–E9782

CHEM

ISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

0

relationships in bimetallic complexes that are electrocatalysts forH2 production (2, 24, 25). Even without the enzymatic intricatepositioning of proton and electron relay functions, many modelcomplexes show positive responses to appropriate E/C condi-tions (E = electron addition; C = proton addition). Developedfrom systematic alterations in such bimetallics is a series shownin Fig. 2 of minimal synthetic analogs containing dithiolate-bridged Fe–Fe or Ni–Fe cores derived from MN2S2 metal-loligands; all are at least modest electrocatalysts for proton re-duction/hydrogen production. The MN2S2 metalloligand (26, 27)provides a variable platform according to the carbon connectorswithin the N2S2 tetradentate ligand, as well as the M itself.The M, as Ni2+, or particularly in the form of [Fe(NO)]2+ and[Co(NO)]2+, may tune the donor properties of the thiolates. Thenitrosyl, NO, attached to metal, facilitates redox events becauseit features π*-orbitals close in energy to a metal’s d orbitals. Theenergetic proximity enables orbital admixtures so that the M(NO)x moiety shares the electrons during redox activities whilethe delocalization enables the electrons to flow between M andNO easily. Such orbital mixing creates electronic flexibility in theM(NO)x unit during redox processes, but it prohibits the clearassignments of electrons; this ambiguity was defined to be “non-innocence” by Jørgensen (28). The Enemark–Feltham (29) (E-F)electron count (the metal d electrons plus NO π*-electrons) wasintroduced to circumvent the partitioning of electrons betweenthe metal and the nitrosyl(s). The receiver groups, i.e., the sec-ond metal (all are iron in our series) bound to the metalloligandMN2S2, may also be modulated to test and verify the results viastructure-function analysis.Notably, the hemilability, originally defined by Rauchfuss for P–O

bidentate ligands (30), of the metallodithiolate ligands, i.e., theirability to dissociate one arm of the bidentate ligand while main-taining integrity of the bimetallic, was previously determined tocontribute to the catalytic activity of complexes [NiN2S2·Fe(CO)Cp]+

([Ni-FeCO]+, Cp = η5-C5H5) and [Fe(NO)N2S2·Fe(CO)Cp]+

([Fe-FeCO]+) (31) (Fig. 2). The dissociation of one S–Fe(CO)Cp

dative bond cleaves the S-donor (Lewis base) and creates a metalopen site (Lewis acid). If the components coexist within a convenientdistance, the base and acid sites can be used to assist chemicalreactions. In fact, reaction-created Lewis acid–base pairs, suchas this one, handle the hydrides and the protons, respectively,throughout the catalytic cycle, and account for the catalytic activity

of bimetallic models which do not contain obvious built-in Lewisbases as proton shuttles (31). To stabilize the Lewis pair and toavoid the reinstatement of the dative bond, the pair is pro-tected either by reduction of the Lewis acid or protonation on theLewis base.As shown in Fig. 2, multiple electron-buffering NO ligands

have been introduced into the N2S2-based bimetallic models, inattempts to reproduce the electron-buffering function of the[Fe4S4] subcluster of [FeFe]-hydrogenase (32–35). The elec-tronic flexibility introduced by NO raises the optimistic expec-tation that buffer ligands might convert first-row transitionmetals that are 1e− catalysts into 2e− catalysts (36, 37). Anotherprospect is that such electron-buffering ligands prevent dramaticstructural reorganization or a change in coordination number dur-ing redox activities, consistent with the structurally constrained ac-tive sites of enzymes; in this way, they might contribute to catalystlongevity. Precisely how such delocalization might affect themechanistic behavior or the individual steps/events of ourmodels remains a question, and is the topic of this article.In this work, two bimetallic complexes known to be elec-

trocatalysts for proton reduction, [Fe(NO)N2S2·Fe(NO)2]+

([Fe-Fe]+) (32) and [NiN2S2·Fe(NO)2] (Ni-Fe) (38) (Fig. 2) wereinvestigated via computational chemistry as they contain poten-tially hemilabile bridging thiolates as well as multiple electron-buffering NO ligands. For clarity, the two irons in [Fe-Fe]+ aredifferentiated as follows: The former Fe (underlined as shown inFig. 2) refers to the iron in Fe(NO)N2S2, and the latter Fe refersto the receiver unit, Fe(NO)2.) These reaction mechanisms arecompared with a previous theoretical study of [Ni-FeCO]

+ and[Fe-FeCO]

+ (31). The detailed computational mechanistic studydescribed below delineates sequences of protonation and reductionof the bimetallics and the consequent coupling of electrons andprotons to H2. Importantly, the increased electron-buffering ca-pacity conveyed by NO was found to influence the hemilability ofthe bridging thiolates and essentially change the working mecha-nism, especially controlling how H2 is produced: HP, or RE. Theconnections between these two categories of actor ligands, i.e.,hemilabile and redox active ligands, provide insight to the

Fig. 1. Active sites of (A) [FeFe]-hydrogenase (Hhyd state), (B) [NiFe]-hydrogenase (Ni-R state), and (C) the H2-producing intermediate of nitro-genase (E4/H4 state that fixes N2 following RE of H2), displayed with bound Hand indicating coupling routes. Charges are not explicitly assigned. Hydridesand protons are colored red and blue, respectively.

Fig. 2. Structural representations of electrocatalysts for proton reduction:[Ni-FeCO]

+, [Fe-FeCO]+ (31), Ni-Fe (38), [Fe-Fe]+ (32). The background of each

species shows the cyclic voltammograms (CVs) before (blue) and after (red)the addition of acid. The current enhancement in the red scan is determinedto relate to H2 production. For Ni-Fe, the CVs were obtained from dimeric[Ni-Fe]2

2+ which was calculated to dissociate into [Ni-Fe]+ in solution.

E9776 | www.pnas.org/cgi/doi/10.1073/pnas.1710475114 Ding et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

0

question as to why Nature settled on the unique configurationsthat are found in redox-active, metalloenzyme active sites.

General Computational MethodologyAll structures were fully optimized with the crystal structures ofNi-Fe, [Fe-Fe]+ and the reduced form Fe-Fe, imported as geo-metric starting points in the computational investigations byDFT. Natural bond orbital (NBO) method was applied to certainspecies for bonding analysis. Further details of the methodologyand optimized coordinates are provided in SI Appendix. Thecomparison of experimental and computed metric data forNi-Fe, [Fe-Fe]+, and Fe-Fe in SI Appendix, Table S1 and selectedexperimental and calculated IR frequencies given in SI Appendix,Table S2 validate the calculations. The bond-distance error isgenerally less than 1%, with the maximum less than 2%. Oneexception is the metal–metal distance as here there is no co-valent bond between them. It is noteworthy that the oxidizedstate of Ni-Fe crystallizes in the dimeric form [Ni-Fe]2

2+, but thedimer was calculated to dissociate in solution (SI Appendix, Ta-ble S4); hence the oxidized monomer [Ni-Fe]+ is considered aspart of the catalytic cycle and its structure is calculated to beanalogous to the reduced monomer Ni-Fe.To stoichiometrically produce a molecule of H2 on an elec-

trocatalyst, two protons (i.e., chemical, C, steps) and two elec-trons (i.e., electrochemical, E, steps) must be introduced to thecatalytic site. After each reduction, any immediate geometricreorganization is treated as part of the corresponding E step, i.e.,a so-called “concerted” (39) E step. Generally, the E and C stepsalternate to avoid the accumulation of like charges (31, 40). Thelikelihood of each C step is evaluated in our computations bycomparing the acidities of the protonated species vs. the protonprovider HOEt2

+ (as the acid is HBF4·OEt2 or HOEt2·BF4)(32, 38). A positive ΔpKa (ΔpKa = pKa(CatH) − pKa([HOEt2]

+)indicates a thermodynamically favorable C step. Each E step hasa calculated redox potential E1/2 (vs. Fc

+/Fc), which is comparedwith the experimentally applied electrode potential derived fromcyclic voltammetry. As a result, the mechanisms will be pre-sented as a set of equilibrium values (G/ΔG, ΔpKa, and E1/2) forevaluating the thermodynamic preference of each step. In ad-dition, transition-state barriers (GTS) for steps other than protonand electron transfers are calculated to determine whether sucha step is kinetically allowed. The geometric representations ofspecies within these electrochemical cycles were based on theoptimized structures.The mechanisms for H2 production by Ni-Fe and Fe-Fe, as

described below, start in parallel with the previous study of[Ni-FeCO]

+ and [Fe-FeCO]+ (31). But, the mechanisms soon di-

verge as the effects of multiple redox-active NO ligands on Ni-Feand Fe-Fe appear to redirect the protonations and the reductions todifferent recipients.

Mechanism of H2 Production on the Ni–Fe Model ComplexThe First Reduction and the First Protonation. Fig. 3, I shows the firstreduction at −0.77 V (exp. −0.72 V) on the oxidized monomer[Ni-Fe]+, dissociated from the dimer [Ni-Fe]2

2+ in solution (seeSI Appendix, Table S4 for the ΔG of dissociation reactions) andthe successive protonation. The Fe(NO)2 moiety of [Ni-Fe]+

accepts the electron and increases the E-F electron count from{Fe(NO)2}

9 to {Fe(NO)2}10. Although [Ni-Fe]+ itself cannot be

protonated by HBF4·OEt2 (38), the reduced species [Ni-Fe]accepts a proton (ΔpKa = 12.9 with respect to HBF4·OEt2).Other possible protonation sites are listed in SI Appendix, TableS3. Protonation reduces the E-F electron count of the Fe(NO)2moiety to eight as two electrons are consumed by the Fe–Hbond, but the overall electron count of the iron in [Ni-FeH]+

remains at 18. The [Ni-FeH]+, with reduced basicity after thefirst protonation, cannot accept a second proton.

The Second Reduction and Associated Geometric and ElectronicReorganization. The reduction of [Ni-FeH]+ at −1.28 V and itsfurther geometric changes are reported in Fig. 3, II. The incomingelectron is initially shared by both metals of Ni-FeH, as Ni’s onlyvacant orbital dx2−y2 is heavily destabilized and Fe already has18 electrons. Nevertheless, the electron-buffering effect of NO li-gands in the Fe(NO)2 moiety, along with the electron depletion bythe hydride, facilitates the acceptance of the second electron at amoderate potential. However, to lower the energy of this electron-rich species, the hemilabile bridging thiolate easily dissociates theS–Fe bond with a concomitant shift of the added electron to Fe(NO)2, now a 17-electron species, Ni-Fe#H (S# denotes the S ofthe S–Fe bond that will be broken and, in the text, Fe# denotes theFe with the broken S–Fe bond). The Ni-Fe#H then rotates thehydride beneath the Ni and Fe, creating a semibridging hydride, d(Ni-H) = 1.691 Å and d(Fe-H) = 1.587 Å, and inverting the C2linker, to produce Ni-H-Fe#′ (′ indicates the inverted C2 linker).The nickel in Ni-H-Fe#′ has a distorted trigonal bipyramidal ge-ometry (τ = 0.48) (41), which stabilizes a high-spin state andresults in the antiferromagnetic coupling between the high-spinNiII(d8) and the {Fe(NO)2}

9 (SI Appendix, Fig. S3 shows spin-density changes during the geometric reorganizations). The bridg-ing hydride on the mimics of [NiFe]-H2ase has long been featured inthe literature (3, 42–45) with a recent interpretation of a model withthe hydride closer to Ni, as it appears to be in the enzyme (46). Afterthese geometric changes, the second reduction is fully assigned toFe(NO)2 as the concomitant actuation of hemilability facilitatesthe accommodation of the incoming electron, as in the case of[Fe-FeCO]

+ (31).

The Second Protonation and the Production of H2. The secondprotonation on Ni-H-Fe#′ and successive H2 production arepresented in Fig. 3, III. The S# of Ni-H-Fe#′ is an ideal target forprotonation (ΔpKa of 14.4) producing [Ni-H-Fe#′-S#H]+. Thethiol–hydride pair in [Ni-H-Fe#′-S#H]+ is already in spatialproximity (2.773 Å) and they exothermally couple to H2 over abarrier of 7.4 kcal/mol without formation of a σ-complex (η2-H2)intermediate. The H2 release restores the Fe–S# bond in [Ni-Fe′]+and the inverted C2 linker reverses to regenerate the catalyst[Ni-Fe]+. Thus, the [ECEC] catalytic cycle in Fig. 3, III closes withan HP step.The calculations predict that second reduction event at −1.28 V

(calculated) should produce the catalytic wave. However, this cat-alytic wave appears experimentally at approximately −0.70 V andthe current increases with additional equivalents of HBF4•OEt2(38). The early appearance of the catalytic wave indicates that thesecond reduction in the mechanism is a proton-coupled electrontransfer (PCET) as in an [ECEC] cycle. The calculated standardpotential (see SI Appendix, Fig. S1 for more information) of thisproton-coupled reduction from [Ni-FeH]+ (Fig. 3, I) to the restingstate [Ni-H-Fe#′-S#H]+ (Fig. 3, III) is −0.32 V, less negative than thecalculated potentials for both standalone reduction events, −0.77and −1.28 V. Of course, the resting state [Ni-H-Fe#′-S#H]+ differsfrom [Ni-FeH]+ significantly; thus, the actual PCET to [Ni-FeH]+

cannot generate [Ni-H-Fe#′-S#H]+ without geometric reorga-nizations over barriers (Fig. 3). The calculations essentially seta range, from −0.32 to −1.28 V, for the reduction potential fora PCET process to [Ni-FeH]+; thus, the existence of an un-resolved PCET process may explain the appearance of the cat-alytic wave at a less negative potential than that calculated.The intermolecular protonation by the acid on the hydride of

Ni-H-Fe#′ of Fig. 3, II to produce H2 directly from the addedacid is ruled out by a barrier ∼10 kcal/mol higher than that of theprotonation on S# (SI Appendix, Fig. S4). An alternate hydride-bearing species is presented in SI Appendix, Fig. S5. As in theprevious work with [Ni-FeCO]

+ (31), a third electron could beadded to [Ni-H-Fe#′-S#H]+ before it releases H2, which wouldrender an E[CECE] catalytic cycle (SI Appendix, Fig. S6). However,

Ding et al. PNAS | Published online October 30, 2017 | E9777

CHEM

ISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

0

for this particular catalyst [Ni-Fe]+, the third electron would notappear to accelerate the catalytic cycle; see SI Appendix for moreinformation.

Mechanism of H2 Production on the Fe-Fe Model ComplexThe First and Second Reduction and the Roaming of the First Proton.The [Fe-Fe]+ complex, described here with NO ligands on bothirons, is also an effective electrocatalyst for H2 production withHBF4·OEt2 (32). Fig. 4, I depicts the first two steps of themechanism, the reduction and protonation of the {Fe(NO)2}

9

moiety in the [Fe-Fe]+, consistent with our earlier experimentaland theoretical study (32). Fig. 4, II shows the second reductionevent and associated geometric changes. The second electronreduces the iron-mononitrosyl in [Fe-FeH]+ to {Fe(NO)}8, bestdescribed as antiferromagnetically coupled high-spin FeII (S = 2)and high-spin NO− (S = 1) which results in a linear NO (173.1°)(32). Next, hydrogen in Fe-FeH migrates from the iron-dinitrosylto the iron-mononitrosyl while an electron migrates in the op-posite direction, which results in the intermediate Fe-H-Fe with asemibridging hydride. The six-coordinate iron in the{Fe(NO)}7

has a vacant dz2 orbital due to the strong axial hydride ligand,which forces the unpaired electron into dx2−y2; thus, the nitrosylhas no tendency to bend (175.7°) to mix its π*-orbital with Fe dz2.An alternate, but less favorable, roaming route and product issummarized in SI Appendix, Fig. S7.

The Second Protonation and the Production of H2. The secondprotonation step is summarized in Fig. 4, III. Protonation on oneof the two bridging thiolates (ΔpKa = 7.9) of Fe-H-Fe initiallybreaks the S#–Fe(NO) bond but this species rearranges bybreaking the S#–Fe(NO)2 bond, restoring the S#–Fe(NO) bond,and inverting the C2 linker. The product, [Fe-H-Fe#′-S#H]+,

does not couple thiol-hydride to generate H2 but transfers theproton on S to the Fe(NO)2 to create the intermediate [Fe-H-FeH]+, featuring one terminal and one (semi)bridging hydride.This dihydride may also be created by the direct protonation ofFe-H-Fe on Fe(NO)2, ΔpKa = 13.9. Either protonation has anegligible barrier (SI Appendix, Fig. S8) and leads to the sameproductive process.Fig. 4, IV shows the final step, the reductive elimination of H2

from two hydrides on [Fe-H-FeH]+ over a low barrier. The [Fe-Fe]+ is regenerated after the exothermic release of H2 and this[ECEC] catalytic cycle closes with an RE step. The reduction of[Fe-FeH]+ (at −1.29 V, calculated) is expected to produce acatalytic wave. With only a few equivalents of added acid, theexperimental catalytic wave appears as early as −0.8 V, whichoverwhelms the shoulder peak at ∼−0.7 V representing the firstelectron reduction (32). Again, the discrepancy may be attributed to aPCET process. This standard reduction potential (SI Appendix, Fig.S1) is −0.28 V from [Fe-FeH]+ (Fig. 4, I) to the resting state [Fe-H-Fe#′-S#H]+ (Fig. 4, III), if the necessary geometric reorganizations inbetween are ignored. This value, −0.28 V, is the high limit of thethermodynamic potential of the PCET process with −1.29 V asthe low limit. The addition of a third electron to [H-Fe-Fe-H]+

before H2 release is also possible, yielding an E[CECE] cat-alytic cycle, as in the previous report of [Fe-FeCO]

+ (31). Moreinformation is provided in SI Appendix, Fig. S9.

DiscussionThe Factors Controlling the Actuation of Hemilability. The conditionswherein the hemilability can be triggered in these bimetallicswith bridging thiolates as core structures provide interestingcomparisons. As summarized in Fig. 5, as few as one and as manyas four steps may be required. Here, we note that the number of

Fig. 3. Computational mechanism of electrocatalytic H2 production on [Ni-Fe]+ in the presence of HBF4•OEt2: The Gibbs free energies in kcal/mol are scaledto the reference point (G = 0), which resets after every reduction or protonation. The reduction potentials (E1/2) are reported in volts with reference to thestandard redox couple Fc+/0 = 0.0 and the relative acidities (ΔpKa) are reported versus [HOEt2]

+.

E9778 | www.pnas.org/cgi/doi/10.1073/pnas.1710475114 Ding et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

0

steps correlates with the number of NO ligands installed on thecatalyst. In the [Ni-FeCO]

+ (Fig. 5), the NiII (d8) within themetalloligand is unable to hold onto the incoming electron,resulting in an internal electron transfer to the 6-coordinate, 18-electron FeII (d6) in Fe(CO)Cp+, which concomitantly breaks one of

the two S–FeI(CO)Cp bonds to reduce the 19-electron count for ironto 17, as indicated by an irreversible reduction event in the absenceof the acid (31). See SI Appendix, Fig. S10 for the spin-density plotafter the bond dissociation. This type of reduction-actuated hemil-ability of a multidentate ligand has precedent in the tridentate

Fig. 4. Computational mechanism of electrocatalytic H2 production on [Fe-Fe]+ in the presence of HBF4•OEt2. Note that one transition state was marked witha star as its Gibbs free energy is lower than its immediate precursor, which is caused by the error of solvation and thermal corrections. See the legend of Fig. 3for more explanation.

Fig. 5. Conditions required to realize the hemilability of the bridging thiolates on different models. E indicates an electrochemical step (reduction) and C, achemical step (protonation).

Ding et al. PNAS | Published online October 30, 2017 | E9779

CHEM

ISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

0

trispyrazolborate ligand of [Rh(CO)(PPh3)TpMe2]+/0 and derivatives

by Geiger et al. (39) and Connelly et al. (47, 48).The complex [Fe-FeCO]

+ (Fig. 5) places the incoming electronon {Fe(NO)}7 in the N2S2 site, which has electron-bufferingcapacity and does not initiate bond dissociation. Here a secondstep, the protonation of S, which is quite basic after the re-duction, removes electron density from S, reduces the donationto Fe(CO)Cp+, and leads to the S–Fe bond rupture (31). NBOanalysis in SI Appendix, Fig. S11 shows that both the S–Fe bond(before the protonation) and the S–H bond (after the pro-tonation) primarily use the 3p orbital of the three-coordinateS, while the remaining lone pair is dominated by the less ac-cessible 3s orbital.In contrast to [Ni-FeCO]

+ and [Fe-FeCO]+, the [Ni-Fe]+ and

[Fe-Fe]+ complexes, whose H2 production mechanisms are pre-sented in this work, require additional electrons and protons todissociate the S–Fe(NO)2 bond. The reason lies in the nitrosylligand’s electron-buffering capacity. In both [Ni-Fe]+ and [Fe-Fe]+,the {Fe(NO)2}

9 moiety accepts the first electron to produce{Fe(NO)2}

10; this reduction does not require the dissociation of aS–Fe(NO)2 bond to accommodate the incoming electron. Sincethe NO’s electron-buffering capacity supports the electron-rich Fe,the first proton goes to the {Fe(NO)2}

10 moiety to create an ironhydride, rather than to the sulfur. Still these two steps do not elicitthe S–Fe(NO)2 bond dissociation. However, the second reduction to[Ni-Fe]+ triggers the S–Fe(NO)2 bond dissociation as neither thenow-saturated, five-coordinate {Fe(NO)2}

8-hydride nor the square-planar NiII can accept the electron easily. Thus, the S–Fe(NO)2bond now cleaves after three steps, ECE, so that the FeH(NO)2moiety becomes four-coordinate and has the vacancy for the second

incoming electron. Since for [Fe-Fe]+ the second electron is againbuffered by the Fe(NO), as one saw in the first reduction of[Fe-FeCO]

+, a fourth step, the second protonation, which occurson thiolate sulfur this time, is needed for the S–Fe(NO)2 cleavage,as in the first protonation for [Fe-FeCO]

+.

The Lewis Acid–Base Pair Generated from the Cleavage of the S–FeBond. Fig. 6 summarizes the key mechanistic aspects of the ac-tuation of the hemilability of the bridging thiolate in the reactivecenters to assist the electrocatalytic process of H2 production.The first reaction (Fig. 6, Left) shows a simple dative-bond dis-ruption, which would be expected to be short-lived as it is ther-modynamically advantageous to reestablish the dative bond. Inour mechanistic study, it was discovered that this bond dissoci-ation is synergic with either reduction or protonation. In otherwords, either of these manipulation modulates the electrondensity on the Lewis acid or the Lewis base to quench theiracidity or basicity and prevent the reformation of the dative bond(Fig. 6, Right).Inspection of the correlation between the number of steps to

trigger the hemilability and the number of NO ligands (Fig. 5) leadsto the conclusion that the electron-buffering capacity demonstratedby NO ligand prevents an earlier S–Fe bond dissociation as it in-terferes with either endeavor to make the bond dissociation irre-versible, by redirecting the reduction or protonation. The morenitrosyls the complex has, the more electron-buffering capacity itreceives such that more reduction/protonation steps are neededbefore the dative-bond dissociation occurs.

H2 Production Step: HP vs. RE. The increasing number of nitrosylson the model also makes formation of the hydride(s) easier. For[Fe-Fe]+, two external electrons are enough to reduce two protonsand create two hydrides, one on each iron of [Fe-H-Fe-H]+, withthe two additional internal electrons coming from the Fe(NO)xfragments buffered by NO π*-orbitals. In comparison, other[FeFe]-H2ase models, such as (dppv)(CO)Fe(μ-pdt)Fe(CO)(dppv)(49, 50) and (dppv)(CO)Fe(μ-edt)Fe(CO)3 (edt = ethane-1,2-dithiolate) (51) that lack an electron reservoir such as NO, ex-perimentally achieve a dihydrido derivative by the addition of anexogenous hydride.The coupling to produce H2 varies mechanistically. The HP

mechanism only needs one hydride, which means it can occur onless electron-rich metal complexes, while the RE needs two hy-drides to produce H2. Thus, theHP process operates in the [ECEC]and E[CECE] cycles of both [Ni-FeCO]

+ and [Fe-FeCO]+ (31). For

Fig. 6. Actuation of the hemilability of the thiolate. Either the reduction onthe dissociated metal, or the protonation on the dissociated sulfur, finalizesthe dative-bond dissociation and preserves the reactive sites.

Fig. 7. H2 production by either HP or RE coupling and their immediate precursors.

E9780 | www.pnas.org/cgi/doi/10.1073/pnas.1710475114 Ding et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

0

[Ni-Fe]+, the [ECEC] cycle involves the HP coupling, while withone more electron in the E[CECE] cycle [Ni-Fe]+switches to theRE coupling process (SI Appendix, Fig. S5). In other words, theRE coupling is made possible by the electronic flexibility of NOligands attached to metals. Finally, the RE mechanism takes overcompletely in either the [ECEC] cycle or the E[CECE] cycle (SIAppendix, Fig. S7) of the trinitrosyl species [Fe-Fe]+. Fig. 7 sum-marizes the H2 production steps and corresponding immediateprecursors of all four electrocatalysts. From the aspect ofstructure-function analysis, the RE coupling requires one moreLewis acid to park the additional hydride while eliminating thenecessity of a Lewis base to store the proton, which is howeverrequired by HP coupling. Although the hemilability of the thi-olate may still be important in molecular isomerization and theresting states, the Lewis base required to hold a proton is nolonger mandatory for [Fe-Fe]+, as the incoming proton can al-ways be stored as a hydride, whose production uses the electronsheld by the irons and their nitrosyl(s).The relevance between electron availability and H2 coupling

mechanism is also present in the enzymes: [NiFe]-H2ase (with noimmediate buffering) and [FeFe]-H2ase (with the [Fe4S4] sub-cluster to buffer one electron) can provide two and three elec-trons on the most reduced Ni-L and Hsred states, respectively. Inother words, they are unable to generate two hydrides, even ifthere are enough vacant sites for two hydrides; therefore, theycan only proceed through the HP mechanism (2). Intriguingly,nitrogenase, whose electron buffering can be attributed to theextensive delocalization of the FeS cluster (Fig. 1), takes in fourelectrons to create two hydrides on the E4 state and it can exe-cute an RE step to produce H2 concomitantly with N2 fixation.This RE step strategically deposits two reduction equivalencesneeded for the initial uptake and activation of N2 in the nitro-genase FeMo cofactor (19, 21–23). The E4 state of nitrogenasecan actually be discharged with an HP step to produce H2, but inthe absence of N2 (20). In conclusion, such electron delocalization is

a strategy used by Nature to enable nonnoble metals to domultielectron chemistry.

SummaryOur theoretical investigation highlights the role of the hemil-ability of the bridging dithiolate and the electron-buffering ca-pacity of ligands such as NO in bimetallic electrocatalysts. Uponthe actuation of the hemilability by dative-bond dissociation,Lewis acid and base sites are created and serve as reactive cen-ters. To maintain the availability of the Lewis acid–base pair, itmust be protected from reformation of the dative bond. Theprotection for our systems is achieved by the modification ofthe electron densities either on the acid (by reduction) or on thebase (by protonation) to prohibit a stable donation from the baseto the acid.The electron buffering by the NO provides metal sites with

flexibility such that the protonation and reduction are directed tosites other than the potential Lewis base and acid sites. Therealization of the hemilability of the bridging thiolate is modu-lated by its interplay with the NO ligand. Thus, NO can interferewith the early (upon reduction) hemilability of the MN2S2 ligandand postpones the creation of the Lewis acid–base pair. Theelectron-buffering NO ligands also provide an electron-rich metalsite that facilitates conversion of protons to hydrides, pivoting themechanism from HP to RE as multiple hydrides become available.The ultimate role of NO in the models is recognized to be a bi-functional electron reservoir and it (at least partially) reproducesthe function of the [Fe4S4] subcluster of [FeFe]-H2ase active site.

ACKNOWLEDGMENTS.We are grateful for computing resources provided bythe Laboratory for Molecular Simulation and the High-Performance Re-search Computing Facility at Texas A&M University. This work was fundedby the National Science Foundation (CHE-1300787, CHE-1664866 to M.B.H.and CHE-1266097, CHE-1665258 to M.Y.D.) and the Robert A. Welch Foun-dation (A-0648 to M.B.H. and A-0924 to M.Y.D.).

1. Cook TR, et al. (2010) Solar energy supply and storage for the legacy and nonlegacyworlds. Chem Rev 110:6474–6502.

2. Lubitz W, Ogata H, Rüdiger O, Reijerse E (2014) Hydrogenases. Chem Rev 114:4081–4148.

3. Siegbahn PEM, Tye JW, Hall MB (2007) Computational studies of [NiFe] and [FeFe]hydrogenases. Chem Rev 107:4414–4435.

4. Mulder DW, et al. (2011) Insights into [FeFe]-hydrogenase structure, mechanism, andmaturation. Structure 19:1038–1052.

5. Cao Z, Hall MB (2001) Modeling the active sites in metalloenzymes. 3. Density func-tional calculations on models for [Fe]-hydrogenase: Structures and vibrational fre-quencies of the observed redox forms and the reaction mechanism at the diiron activecenter. J Am Chem Soc 123:3734–3742.

6. Fan H-J, Hall MB (2001) A capable bridging ligand for Fe-only hydrogenase: Densityfunctional calculations of a low-energy route for heterolytic cleavage and formationof dihydrogen. J Am Chem Soc 123:3828–3829.

7. Liu Z-P, Hu P (2002) Mechanism of H2 metabolism on Fe-only hydrogenases. J ChemPhys 117:8177–8180.

8. Liu Z-P, Hu P (2002) A density functional theory study on the active center of Fe-onlyhydrogenase: Characterization and electronic structure of the redox states. J AmChem Soc 124:5175–5182.

9. Niu S, Hall MB (2001) Modeling the active sites in metalloenzymes 5. The heterolyticbond cleavage of H(2) in the [NiFe] hydrogenase of Desulfovibrio gigas by a nucleo-philic addition mechanism. Inorg Chem 40:6201–6203.

10. Niu S, Thomson LM, Hall MB (1999) Theoretical characterization of the reaction in-termediates in a model of the nickel−iron hydrogenase of Desulfovibrio gigas. J AmChem Soc 121:4000–4007.

11. Siegbahn PEM (2004) Proton and electron transfers in [NiFe] hhydrogenase. Adv InorgChem 56:101–125.

12. Pavlov M, Siegbahn PEM, Blomberg MRA, Crabtree RH (1998) Mechanism of H−Hactivation by nickel−iron hydrogenase. J Am Chem Soc 120:548–555.

13. Mulder DW, Guo Y, Ratzloff MW, King PW (2017) Identification of a catalytic iron-hydride at the H-cluster of [FeFe]-hydrogenase. J Am Chem Soc 139:83–86.

14. Reijerse EJ, et al. (2017) Direct observation of an iron-bound terminal hydride in[FeFe]-hydrogenase by nuclear resonance vibrational spectroscopy. J Am Chem Soc139:4306–4309.

15. Adamska A, et al. (2012) Identification and characterization of the “super-reduced”state of the H-cluster in [FeFe] hydrogenase: A new building block for the catalyticcycle? Angew Chem Int Ed Engl 51:11458–11462.

16. Evans RM, et al. (2016) Mechanism of hydrogen activation by [NiFe] hydrogenases.Nat Chem Biol 12:46–50.

17. Ogata H, Nishikawa K, Lubitz W (2015) Hydrogens detected by subatomic resolutionprotein crystallography in a [NiFe] hydrogenase. Nature 520:571–574.

18. Simpson FB, Burris RH (1984) A nitrogen pressure of 50 atmospheres does not preventevolution of hydrogen by nitrogenase. Science 224:1095–1097.

19. Hoffman BM, Lukoyanov D, Yang Z-Y, Dean DR, Seefeldt LC (2014) Mechanism ofnitrogen fixation by nitrogenase: The next stage. Chem Rev 114:4041–4062.

20. Hoffman BM, Lukoyanov D, Dean DR, Seefeldt LC (2013) Nitrogenase: A draftmechanism. Acc Chem Res 46:587–595.

21. Lukoyanov D, et al. (2016) Reductive elimination of H2 activates nitrogenase to re-duce the N≡N triple bond: Characterization of the E4(4H) Janus intermediate in wild-type enzyme. J Am Chem Soc 138:10674–10683.

22. Lukoyanov D, et al. (2016) Reversible photoinduced reductive elimination of H2 fromthe nitrogenase dihydride state, the E(4)(4H) Janus intermediate. J Am Chem Soc 138:1320–1327.

23. Lukoyanov D, et al. (2015) Identification of a key catalytic intermediate demonstratesthat nitrogenase is activated by the reversible exchange of N2 for H2. J Am Chem Soc137:3610–3615.

24. Tard C, Pickett CJ (2009) Structural and functional analogues of the active sites of the[Fe]-, [NiFe]-, and [FeFe]-hydrogenases. Chem Rev 109:2245–2274.

25. Schilter D, Camara JM, Huynh MT, Hammes-Schiffer S, Rauchfuss TB (2016) Hydrogenaseenzymes and their synthetic models: The role of metal hydrides. Chem Rev 116:8693–8749.

26. Denny JA, Darensbourg MY (2015) Metallodithiolates as ligands in coordination, bi-oinorganic, and organometallic chemistry. Chem Rev 115:5248–5273.

27. Denny JA, Darensbourg MY (2016) Approaches to quantifying the electronic andsteric properties of metallodithiolates as ligands in coordination chemistry. CoordChem Rev 324:82–89.

28. Jørgensen CK (1966) Differences between the four halide ligands, and discussion re-marks on trigonal-bipyramidal complexes, on oxidation states, and on diagonal ele-ments of one-electron energy. Coord Chem Rev 1:164–178.

29. Enemark JH, Feltham RD (1974) Principles of structure, bonding, and reactivity formetal nitrosyl complexes. Coord Chem Rev 13:339–406.

30. Jeffrey JC, Rauchfuss TB (1979) Metal complexes of hemilabile ligands. Reactivity andstructure of dichlorobis(o-(diphenylphosphino)anisole)ruthenium(II). Inorg Chem 18:2658–2666.

31. Ding S, et al. (2016) Hemilabile bridging thiolates as proton shuttles in bioinspired H2

production electrocatalysts. J Am Chem Soc 138:12920–12927.

Ding et al. PNAS | Published online October 30, 2017 | E9781

CHEM

ISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

0

32. Hsieh C-H, et al. (2014) Redox active iron nitrosyl units in proton reduction electro-catalysis. Nat Commun 5:3684.

33. Hsieh C-H, Darensbourg MY (2010) A Fe(NO)310 trinitrosyliron complex stabilized byan n-heterocyclic carbene and the cationic and neutral Fe(NO)2(9/10) products of itsNO release. J Am Chem Soc 132:14118–14125.

34. Hsieh C-H, Chupik RB, Pinder TA, Darensbourg MY (2013) Dinitrosyl iron adducts of(N2S2)M(NO) complexes (M = Fe, Co) as metallodithiolate ligands. Polyhedron 58:151–155.

35. Pulukkody R, Darensbourg MY (2015) Synthetic advances inspired by the bioactivedinitrosyl iron unit. Acc Chem Res 48:2049–2058.

36. Lyaskovskyy V, de Bruin B (2012) Redox non-innocent ligands: Versatile new tools tocontrol catalytic reactions. ACS Catal 2:270–279.

37. Luca OR, Crabtree RH (2013) Redox-active ligands in catalysis. Chem Soc Rev 42:1440–1459.

38. Ghosh P, et al. (2017) A matrix of heterobimetallic complexes for interrogation ofhydrogen evolution reaction electrocatalyst. Chem Sci, 10.1039/C7SC03378H.

39. Geiger WE, Ohrenberg NC, Yeomans B, Connelly NG, Emslie DJH (2003) Reversiblesequence of intramolecular associative and dissociative electron-transfer reactions inhydrotris(pyrazolylborate) complexes of rhodium. J Am Chem Soc 125:8680–8688.

40. Surawatanawong P, Tye JW, Darensbourg MY, Hall MB (2010) Mechanism of elec-trocatalytic hydrogen production by a di-iron model of iron-iron hydrogenase: Adensity functional theory study of proton dissociation constants and electrode re-duction potentials. Dalton Trans 39:3093–3104.

41. Addison AW, Rao TN, Reedijk J, van Rijn J, Verschoor GC (1984) Synthesis, structure,and spectroscopic properties of copper(II) compounds containing nitrogen–sulphurdonor ligands; the crystal and molecular structure of aqua[1,7-bis(N-methyl-

benzimidazol-2’-yl)-2,6-dithiaheptane]copper(II) perchlorate. J Chem Soc DaltonTrans, 1349–1356.

42. Ogo S, et al. (2007) A dinuclear Ni(mu-H)Ru complex derived from H2. Science 316:585–587.

43. Ogo S, et al. (2013) A functional [NiFe]hydrogenase mimic that catalyzes electron andhydride transfer from H2. Science 339:682–684.

44. Barton BE, Whaley CM, Rauchfuss TB, Gray DL (2009) Nickel-iron dithiolato hydridesrelevant to the [NiFe]-hydrogenase active site. J Am Chem Soc 131:6942–6943.

45. Barton BE, Rauchfuss TB (2010) Hydride-containing models for the active site of thenickel-iron hydrogenases. J Am Chem Soc 132:14877–14885.

46. Brazzolotto D, et al. (2016) Nickel-centred proton reduction catalysis in a model of[NiFe] hydrogenase. Nat Chem 8:1054–1060.

47. Connelly NG, Emslie DJH, Metz B, Orpen AG, Quayle MJ (1996) Redox-induced κ2-κ3

isomerisation in rhodium hydrotris(3,5-dimethylpyrazolyl)borate chemistry: The sta-bilisation of square-pyramidal rhodium(II). Chem Commun, 2289–2290.

48. Connelly NG, et al. (2001) Redox-induced κ2-κ3 isomerisation in hydrotris(pyrazolyl)borator-hodium complexes: Synthesis, structure and ESR spectroscopy of stabilised rhodium(II) species.J Chem Soc Dalton Trans, 670–683.

49. Barton BE, Rauchfuss TB (2008) Terminal hydride in [FeFe]-hydrogenase model haslower potential for H2 production than the isomeric bridging hydride. Inorg Chem 47:2261–2263.

50. Wang W, Rauchfuss TB, Zhu L, Zampella G (2014) New reactions of terminal hydrideson a diiron dithiolate. J Am Chem Soc 136:5773–5782.

51. Heiden ZM, Zampella G, De Gioia L, Rauchfuss TB (2008) [FeFe]-hydrogenase modelsand hydrogen: Oxidative addition of dihydrogen and silanes. Angew Chem IntEd Engl 47:9756–9759.

E9782 | www.pnas.org/cgi/doi/10.1073/pnas.1710475114 Ding et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

0