A matrix of heterobimetallic complexes for interrogation of hydrogen evolution reaction electrocatalysts

Nitrosyls as electron reservoirs guide protons to favorable sites in bimetallic HER catalysts.


Introduction
From protein crystallography the bimetallic active site structures in enzymes such as [NiFe]-, [FeFe]-hydrogenases (H 2 ase), CO-dehydrogenases and acetyl coA synthase (ACS) have been convincingly interpreted in terms of characteristics needed for their organometallic-like functions. 1,2 Whereas most major homogeneous catalytic applications involving redox processes use precious metals that can operate as single site catalysts, the intricate molecular arrangements in nature's biocatalysts harness combinations of at least two rst row transition metals, connected by suldes or thiolates, along with Lewis acid/base sites. [3][4][5] Over the past two decades a rich area in synthetic chemistry inspired by such natural products has developed, yielding biomimetics for insight into enzyme mechanisms. In addition the link between the [NiFe]-and [FeFe]-H 2 ase active sites and base metal, sustainable catalysts for the Hydrogen Evolution Reaction (HER) holds promise for the production of H 2 from "solar" (photovoltaic) electrons via electrocatalysis. 6 Specic efforts have been directed towards the use of metallodithiolates from MN 2 S 2 complexes as bidentate donor ligands (readily deduced from the structure of the ACS enzyme active site), that bind to receiver metal units via bridging dithiolates. [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25] The electronic requirements of the thiolate sulfurs have a steric consequence in the buttery M(m-SR) 2 M 0 cores that are seen in the H 2 ase active sites, placing M and M 0 within close proximity. 26 The advancement of chemistry via structure/function analysis of sets of compounds with well-known differences in composition and structure is a challenge in the complicated area of HER electrocatalysis. Nonetheless the metallodithiolateas-synthon approach, inspired from the ACS active site, permits modular design that includes some features of the bimetallic [NiFe]-and [FeFe]-H 2 ase active sites beyond the obvious dithiolate core structures. An initial foray explored the properties of the diiron, trinitrosyl complex shown in Fig. 1. 8,27 With it we intended to exploit the redox-activity of fFe ðNOÞg 7=8 in the ðNOÞ Fe N 2 S 2 metalloligand bound to a redox-active {Fe(NO) 2 } 9/ 10 , iron dinitrosyl unit. Electrochemical studies of ½ðNOÞ Fe N 2 S 2 $FeðNOÞ 2 þ , ½FeÀFe þ , Fig. 1, found two single-electron, reversible reduction events, À0.78 V and À1.33 V, assigned to {Fe(NO) 2 } 9/10 and fFe ðNOÞg 7=8 couples, respectively. 8 The Fe of the ðNOÞ Fe N 2 S 2 metalloligand is herein distinguished as Fe ; the electron count of the iron nitrosyl units uses the Enemark-Feltham approach. 28 Consistent with the stoichiometric reaction shown in Fig. 1, the {Fe(NO) 2 } 9/10 couple, at À0.78 V, was the catalytically active center for electrochemical proton reduction in the presence of strong acid, HBF 4 $Et 2 O. Although modest in overpotential and TOF, electrocatalysis for H 2 production was observed at this potential; preliminary computational studies indicated that a hydridebound {Fe(NO) 2 } 8 could likely be a transient intermediate, however the overall H 2 releasing mechanism was at that stage incomplete. 8 Interestingly, the second reduction process, related to the more negative fFe ðNOÞg 7=8 couple, showed a current response to added weak acid, however H 2 was not detected. Computational study attributed this to a nonproductive reduction event with protonation on the nitrosyl, which terminates the catalytic cycle. 8 We have made analogues of the diirion trinitrosyl complex making use of NiN 2 S 2 and ðNOÞFe N 2 S 2 metallodithiolates in combination with h 5 -C 5 R 5 derivatives (R ¼ H, CH 3 ), 9,10 of Fe II shown in Fig. 2. The large differences in reduction potential of the MN 2 S 2 ligands, with the d 8 -Ni II being more negative because of a more rigid, less polarizable electronic structure as compared to the delocalized fFe ðNOÞg 7 unit, of greater electronic exibility, inspired the labels of "hard" for the former MN 2 S 2 unit, and "so" S-donor unit for the latter. The Fereceivers also differ in electronic exibility and their ease of electron uptake, the term "so" describing the highly delocalized {Fe(NO) 2 } 9 unit, and the indenite oxidation state of the iron, as compared to the denite Fe II of the h 5 -C 5 R 5 , "hard" receiver derivatives. 8,9 The hard receiver unit, (h 5 -C 5 H 5 )Fe(CO) + , is herein distinguished from the so Fe(NO) 2 unit by Fe 0 and Fe, respectively.
Notable from the computational approach that guided the interpretation of electrochemical events of the [Ni-Fe 0 ] + and ½FeÀFe 0 þ complexes in the presence of acid was the indication of a reductive iron-sulfur bond cleavage (shown in red, Fig. 2 [29][30][31][32][33] adjust their structures to accommodate coupled electron/proton uptake. While the mild potential for the rst EC process for the ½FeÀFe 0 þ complex required both proton/electron uptake for genesis of the pendant base, the more negative potential that reduces the [Ni-Fe 0 ] + labilizes the sulfur and creates an Fe III -H at the rst reduction, Fig. 2. 8,9 In this report we provide a more complete matrix of MN 2 S 2 -Fe complexes of electrocatalytic potential for experimental and computational analysis. Specically a redox innocent ("hard") metalloligand, NiN 2 S 2 , of more negative reduction potential, is incorporated in place of "so" ðNOÞFe N 2 S 2 . The thus generated [Ni II N 2 S 2 $Fe(NO) 2 ] 0/+ , a "hard"/"so" complex may be compared to the other members of the matrix. The solid state structures of [Ni II N 2 S 2 $Fe(NO) 2 ] 0/+ in two redox levels and characteristics as an electrocatalyst (robustness and turnover frequency), for proton reduction are also described. Computational study, addresses the diversity of geometries of di-and poly-metallic compounds containing N 2 S 2 metalloligands by inspecting the versatile bonding orbitals of the metalloligands. The computational mechanisms contrast the working electrocatalysts against a non-working analogue by exploring possible intermediates in the proposed catalytic cycles. Here important roles for hemi-labile and redox active ligands are revealed. . This is consistent with the presence of a single unpaired electron, Fig. S1 and S2. † The EPR spectra for both complexes display the isotropic g ¼ 2.03 signal that is prototypical of the {Fe(NO) 2 } 9 unit, Fig. S21 and S22, † respectively. The 77 K EPR spectrum of the [Ni-Fe] + displayed ne structure requiring two signals for simulation: A major isotropic signal of g ¼ 2.035 showed coupling with nitrogen of A( 14 N) ¼ 32.74 MHz and a minor anisotropic signal had parameters of g xyz ¼ 2.183, 2.012, 1.908 and no observable hyperne coupling, Fig. S21. † X-ray diffraction quality crystals of the oxidized NiFe compound were obtained from the one-pot reaction of equimolar NiN 2 S 2 and (putative) [Fe(CO) 2 (NO) 2 ] + (prepared in situ by reacting [Fe(CO) 3 36 The converging lone pairs (see below) on the cis-dithiolates engage in bidentate binding and impose a hinge angle (the intersection of the best N 2 S 2 plane with the S 2 Fe plane) of ca. 117 , vis-à-vis constricting the :S-Ni-S angle by ca. 4 compared to the free metalloligand. 39 The two nitrosyl units bound to the pseudo tetrahedral iron center are slightly bent towards each other, in an "attracto" orientation; 40 (2)Å, respectively, and are longer than in the [Ni-Fe] 0 reduced complex by ca. 0.5Å. The Ni II maintains a square planar geometry in the reduced and oxidized complexes with a displacement of no more than 0.1Å from the N 2 S 2 best plane. Overall these structures demonstrate the impressive adaptability of the NiN 2 S 2 metalloligands, and their potential to template clusters through S-based aggregation. 7

Computational structural study
This computational section uses density functional theory (DFT) analysis to address the structural question in particular that was raised by the X-ray diffraction report: is there an electronic factor that governs the different m 2 -SR binding modes found in the three forms of NiFe heterometallic aggregates? The functional/basis set combination, TPSS/6-311++G(d,p), and natural bond orbital (NBO) analysis were applied to the computational structural modeling of the free metalloligand NiN 2 S 2 and its derivatives [Ni-Fe] 0 , [Ni 2 -Fe 2 ] 2+ ; more details of the computational methodology is available in ESI. † The divergent or convergent orientation of S lone pairs of NiN 2 S 2 metalloligand and inuences on structures of NiN 2 S 2 $M 0 heterobimetallics. Traditional bidentate ligands such as diphosphines, diamines and bipyridyls have a single lone pair on each donor site. These lone pairs are positioned on orbitals originating from sp x -hybridization and are highly directional. 42 They provide xed binding orientations that correspond one-to-one with the coordination sites of the metal. In contrast, the binding between the sulfurs of the metallothiolate NiN 2 S 2 and an exogeneous metal are more geometrically exible because of the multiple S lone pairs. From NBO bonding analysis, sulfur in the NiN 2 S 2 metalloligand is found to use mainly p orbitals for bonding to Ni and C. 43,44 For example, in a free NiN 2 S 2 , p character makes up 83% and 86% of the S contributions in the S-Ni bonds and S-C a bonds (C a and C b refer to the C 2 H 4 linker connecting S and N where C a is directly bound to S, Fig. 4A), which leaves one lone pair in a p orbital and another in an s-dominated orbital on each S. Because a receiver group, a Fe(NO) 2 unit in our case, may bind to either lobe of the p lone pair(s), whose orientation is determined by the Ni-S-C a torsion angle, a diversity of structures results. 7,26 The orientation of this remaining p lone pair in the NiN 2 S 2 metalloligand is determined by the NiN 2 S 2 metalloligand's Ni-S-C a -C b -N ve-membered rings that adopt a non-planar envelope conformation like cyclopentane. The C a carbon, the "ap" of the envelope conformation, puckers towards one side or the other of the N 2 S 2 plane, Fig. 3. Fig. 4 shows how this puckering tilts the remaining 3p-lone pair on each sulfur from the normal to the N 2 S 2 plane. This tilt causes two p-orbital lobes (green lobes) to converge on the side to which the ap puckers, while the red lobes diverge on the opposite side. The orientation of the added Fe(NO) 2 receiver unit(s), will be thus determined by such directional property of the donor p lone pairs. The structure of the reduced monomer [Ni-Fe] 0 shows the Fe(NO) 2 fragment is on the same side as the ap; while in the oxidized dimer [Ni 2 -Fe 2 ] 2+ the ap and the Fe(NO) 2 fragment(s) appear on different sides of each N 2 S 2 plane, thus, binding to the other end of the p lone pair. Based on the analysis above, the converging lobes of the p donor lone pairs maximize contact to the Fe(NO) 2 unit in the monomer [Ni-Fe] 0 , while the diverging lobes are preferred by two bridging Fe(NO) 2 units between two metalloligands in the dimer [Ni 2 -Fe 2 ] 2+ . The utilization of the divergent lobes apparently lessens the steric repulsion between Fe(NO) 2 units. In summary, the binding position of the Fe(NO) 2 unit with respect to the ap in the Ni-S-C a -C b -N vemembered rings are correlated by the competition between chemical bond directionality of the binding sulfurs and steric repulsion of the receiver units.

Electrochemistry
The cyclic voltammograms of  Fig. 5 (inset). A second rise in cathodic current at À1.10 V, commences upon addition of >12 equivalents of the acid, which continues to rise as the catalytic current response, Fig. 5. The rst response is attributed to the reduction of [Ni-Fe] + followed by a protonation. The second response is assigned to the up-take of another electron by the reduced and protonated counterpart of [Ni-Fe] + . The mechanism below connects the successive protonation to the production of H 2 , thus closing the catalytic cycle. Overlays of this response of the NiFe complex in presence of 50 equivalents of HBF 4 $Et 2 O (0.1 M), as well as the CV of the free acid, are shown in Fig. 5. The catalytic H 2 produced was conrmed by applying a constant potential at À1.12 V for 60 min (black bold line in Fig. 5), and analysis of the headspace by gas chromatography. The H 2 was quantied by an average of two consistent constant potential coulometry experiments with subtraction of the H 2 produced from the free acid. 9,10 The nitrosylated compounds [Ni-Fe] + and ½FeÀFe þ were found to have low turnover numbers and faradaic efficiencies, 68 AE 2% and 58 AE 1%, respectively, for H 2 production. In contrast the [Ni-Fe 0 ] + gave a faradeic efficiency of ca. 96%. We assume that the former involves alternate protonation pathways, particularly at NO, that lead to degradation and hence low F.E. In addition the TON for the robust Ni-Fe 0 complex in 50 equiv. of TFA, measured at À1.73 V and over a period of 8 h was found to be 6.7, assuring catalytic proton reduction. The electrocatalytic response of the reduced complex, [Ni-Fe] 0 in the presence of HBF 4 $Et 2 O, is, as expected, the same as [Ni-Fe] + and is shown in Fig. S20-B. † Following the approach of Helm and Appel, 45 and Wiese, 46 the turnover frequency (TOF) as calculated from the CV experiment for [Ni-Fe] + , was 39.7 s À1 , which is slightly better than the ½FeÀFe þ complex, 26.7 s À1 , calculated under similar experimental conditions. The [Ni-Fe] + shows a saturation of the more negative catalytic current upon addition of 80 equivalents of HBF 4 $Et 2 O, i.e., $0.16 M CH 2 Cl 2 solution. Notably, the behavior of the ½FeÀFe þ complex is similar, and further addition of acid leads to decomposition of both catalysts. The precise calculation of overpotential is indeterminable as the thermodynamic potential (E HBF 4 /H 2 ,BF 4 À ) of 0.1 M HBF 4 $Et 2 O in CH 2 Cl 2 is unavailable. 47 Using the thermodynamic potential of HBF 4 $Et 2 O in acetonitrile (À0.26 V), 48,49 an estimate of the overpotential of [Ni-Fe] + and ½FeÀFe þ were 711 mV and 660 mV, respectively, which are lower than those of the [Ni-Fe 0 ] + and ½FeÀFe 0 þ electrocatalysts by over 220 mV.
In contrast to the NiFe complexes, addition of HBF 4 $Et 2 O to a 2.0 mM CH 3 CN solution of [Ni 2 -Fe] + , (the N 2 S 2 ligand used in this electrochemical study is bme-dach) did not show an increase in the cathodic current at À0.75 V, the reversible {Fe(NO) 2 } 9/10 redox event. Instead, a new reversible redox event at E 1/2 ¼ À0.52 V, appeared upon addition of two equivalents of HBF 4 $Et 2 O with a concomitant disappearance of the original redox process, Fig. 6. Further addition of acid resulted in electrode fouling, Fig. S20-A. † A possible explanation, from computational chemistry, vide infra, for the positive 230 mV shi is that [Ni 2 -Fe] + can be protonated on its exposed thiolate sulfur by HBF 4 $Et 2 O, vide infra. Such would account for the greater ease of reduction for the {Fe(NO) 2 } 9/10 couple, compared to the [Ni 2 -Fe] + complex. Supporting this conclusion is that addition of 1 equivalent of HBF 4 $Et 2 O to a CH 3 CN solution of [Ni 2 -Fe] + produced a small but denite shi of the n(NO) in the IR spectrum from 1787 and 1734 cm À1 to 1793 and 1737 cm À1 , Fig. S27. † In addition, the irreversible oxidation event at 0.07 V, which is assumed to be sulfur-based oxidation, shows a decrease in the anodic current upon addition of acid, indicating disulde formation is inhibited under acidic conditions.

Computational mechanistic study
The electrochemical study points to additional questions for computational study: (A) how do the calculated electrocatalytic mechanisms compare for the hard-so vs. so-so donor/ receiver adducts? (B) Can computational analysis clarify those cases of non-catalytic electrochemical responses to added protons? Modeling is extended to [Ni 2 -Fe] + , along with [Ni-Fe] 0 , [Ni-Fe] + , in various oxidation states and with multiple added protons to answer these questions. Note that no hemi-lability of the metallodithiolate ligand 9 is necessary here as the mechanism does not entail hydride/proton coupling to H 2 , but rather reductive elimination from two hydrides. 7 The [Ni-Fe] + and its reduced counterpart [Ni-Fe] 0 are determined to be electrocatalysts at À0.73 V for H 2 production with HBF 4 $Et 2 O, Fig. 5. [Note: The computational study nds that the [Ni 2 -Fe 2 ] 2+ , whose dimeric structure was established in the solid state by crystallography, nds greater stability in solution as the monomeric form, [Ni-Fe] + . Experimental evidences including ESI-MS and determination of m eff support this thesis, vide supra.] The catalytic cycle is thus initiated with the monomer [Ni-Fe] + (Fig. 7B). As indicated in panel B of Fig. 7, the {Fe(NO) 2 } 9 in the [Ni-Fe] + unit accepts the rst incoming electron, followed by the rst proton, to create a hydride on the now {Fe(NO) 2 } 8 unit. Addition of a second electron activates the hemi-lability of the bridging thiolate to break one S-Fe bond, while the terminal hydride becomes bridging between Fe and Ni. The cleavage of the S-Fe dative bond essentially releases one p lone pair of the thiolate so that S can act as a pendant base to accept the second proton and guide it to a coupling position with the hydride and produce H 2 . Details of the full catalytic cycle with energetics and analysis of electronic structure evolution for both ½FeÀFe þ and [Ni-Fe] + are presented in a separate report. 50 Explanation for the absence of catalytic activity of [Ni 2 -Fe] + . While one might have expected the dangling thiolates in the 2 : 1 complex [Ni 2 -Fe] + to act as a pendant base, in fact this complex does not show any catalytic activity in the presence of strong acid, HBF 4 $Et 2 O, within the solvent potential window. A computational study, summarized in Fig. 8, reveals that while reduction still occurs on the Fe(NO) 2 unit, the protonation process is diverted from the Fe(NO) 2 unit. In this 2 : 1 complex, the computations show that only one thiolate from each NiN 2 S 2 binds to Fe(NO) 2 , while the other thiolate, is "free" to interact with other electron acceptors; thus it may also be protonated, even before the reduction of the {Fe(NO) 2 } 9 unit occurs, see   7 Comparative catalytic cycles for H 2 production catalyzed by ½FeÀFe þ and [Ni-Fe] + . All pK a , thermodynamic, and metric data for the two mechanisms are available in a separate report. 50 Table S9, † and shis the its potential, which is supported by experiment, vide supra.
According to the computations, in the reduced [Ni 2 -Fe] 0 the "free" thiolate competes with the reduced {Fe(NO) 2 } 10 unit for the incoming proton (Fig. 8A); in addition, by rotation around an Fe-S bond, the two NiN 2 S 2 ligands may orient their "free" thiolate sulfurs to pinch the proton, i.e., consequently forming a strong hydrogen bond ( Fig. 8A and B). Spectroscopic evidence supports protonation on S even before reduction, i.e., in [Ni 2 -Fe] + , Fig. S27. † Two geometries of the pinched proton by two "free" thiolates, [Ni 2 -Fe-SHS-1] + and [Ni 2 -Fe-SHS-2] + can be achieved by either translating or rotating one NiN 2 S 2 unit of [Ni 2 -Fe], respectively. Precedent in Dubois' Ni(P 2 N 2 ) 2 catalysts, 51 a proton pinched between two amine N bases is relatively stable; in our case, the pinched proton is even more stable than a hydride on Fe(NO) 2 (Fig. 8A). However, the mechanistic clue from the [Ni-Fe] complex 50 indicates the requirement for a proton to be reduced into a hydride, by {Fe(NO) 2 } 10 , before the H 2 can be produced by the proton-hydride coupling mechanism. Therefore, the formation of a stable pinched proton likely prevents the generation of the hydride and cuts off the catalytic cycle. The thiolate already bound to Fe(NO) 2 also helps stabilize the proton on a "free" thiolate, to a smaller extent, with the example of [Ni 2 -Fe-SHS-3] + (Fig. 8A).

Conclusions
Our collection of hydrogen evolution reaction catalysts is summarized in Fig. 9. While the small differences in donor units and acceptor units do not inuence the overall structures of the S-bridged bimetallics; all have buttery-like [M(m-SR) 2 Fe] core and the potential for opening up sites for proton addition via the hemi-lability of the metallothiolate donors. Nevertheless, demonstrable and explicable differences are seen in their catalytic performances as indicated by catalytic potential, required acid strength, and TOF.
Analogous to the HSAB (Hard and So (Lewis) Acids and Bases) concept, we offer an electronic parallel, "so vs. hard donor/receiver units", in this case directed towards the number of NO ligands in the bimetallics ranging from 0 to 3, with increasing exibility (i.e., so) of electronic structure within each unit. The non-innocence of the NO ligand confers electron uptake at milder potentials, which we have seen in both the donor units and acceptor units. Thus the incorporation of NO ligands on the acceptor units, the 'hard-so' and 'so-so' electrocatalysts lead to energetically more accessible catalytic current, however, at the cost of a stronger acid and a lower TOF in comparison to the bimetallics with hard acceptor units.
While these electrocatalysts are only moderately efficient for H 2 production, they are well-behaved and have demonstrated reproducibility. Two of the catalysts, c and d, with so receivers, are isolated and crystallized in both oxidized and reduced forms at ambient conditions lending condence to the presumed catalytic cycle.
Features in the electrochemical scans may be reasonably ascribed to protonation products whose identities are further described by computational chemistry. The resulting computational mechanisms identify key features that may guide future synthetic targets. For example, the hemi-lability of the S-donors may be optimized by steric constraints; the usefulness of the Fe(NO) 2 unit as electron depot and protonation site with low redox potential, should encourage explorations with other redox-active, so acceptors. The computations also suggest a mechanistic paradigm of heterolytic H À /H + , hydride-proton, coupling for bimetallics a, b and c from the chart, and reductive elimination from d arising in the so-so construct. Such a supposition derives from extreme electron delocalization in the trinitrosylated ½Fe ÀFe þ complex and argues that suitably constructed rst row, bimetallic complexes may take on twoelectron processes that emulate noble metals.

Conflicts of interest
There are no conicts of interest to declare.