DFT Studies on the Reduction of Dinitrogen to Ammonia by a Thiolate-Bridged Diiron Complex as a Nitrogenase Mimic

We have recently reported that the binuclear iron complexes Cp*Fe(μ-SR)2(μ,η 2-R2NNH)FeCp* (R = Me, Et; R = Me, Ph; Cp* = ηC5Me5) as novel models of nitrogenase could effectively catalyze the N−N bond cleavage of hydrazines, including NH2NH2 (J. Am. Chem. Soc. 2008, 130, 15250−15251). However, the mechanistic aspects involved in the catalytic cycle and the possibility of reducing N2 by these complexes have remained unexplored. In the present study, DFT has been applied for modeling the binding of Cp*Fe(μ-SEt)2FeCp* with N2 and the reduction of the N2 to two NH3 molecules at the diiron centers. The calculations of model system indicate that the hydrogenation (H + e−) energetically prefers to occur at the N atoms rather than the Fe or S atoms. The rate-determining step of the reduction of HNNH could be the isomerization of a μ,η-HN−NH2 moiety to a μ-HN−NH2 form with a NH-bridging feature, which occurred at the diiron centers. Such a transformation of the binding mode of HN−NH2 might be driven by unequal charge populations on the two Fe atoms. The results show an event of net electron transfer from the ancillary ligands to the Fe atoms and the NHNH2 moiety during the rate-determining step. In view of the experimental observations reported previously, the current computations suggest that the diiron complex Cp*Fe(μSEt)2FeCp* is possible to bind N2 and reduce it to NH3 via protonation/reduction. Such a reduction of N2 to NH3 at the diiron centers favorably occurs through the HNNH and HNNH2 forms rather than via the H2NNH2 unit. ■ INTRODUCTION The activation of nitrogenase substrates mediated by transitionmetal complexes bearing sulfide ligands has attracted considerable attention, not only because of the biological interest in nitrogenase but also its industrial potential in ammonia synthesis from N2 and H2. Much effort has been spent on understanding the transformation process from N2 to NH3 in biological systems, 1e,2 and great progress has been made in this framework. It is generally considered that the active site of nitrogenase is most likely the FeMo-cofactor (FeMoco), where N2 binds and is reduced. 3 FeMoco has the stoichiometric formula MoFe7S9X and is bound to the protein through the end Fe atom and the Mo atom. In FeMoco, six of the seven Fe atoms at the core form a prism and are only 3-fold coordinated. The light atom X as a μ6 ligand located at the center of the prismatic cavity is presumed to be N, O, or C but has not yet been identified. The three central sulfur atoms (μ2S) bridge two Fe atoms each, and the remaining six sulfurs (μ3S) bridge three Fe atoms each. It is also considered that diazene (NHNH) and hydrazine (NH2NH2) are nitrogenaserelavent substrates as partially reduced species of N2. Since the biological nitrogen fixation process is rather complicated, it is hard to directly obtain related mechanistic information. This is one reason only very limited knowledge from experiments is available for an understanding of the nitrogen fixation mechanism. An alternative approach is to design and synthesize nitrogenase model compounds and to investigate their reactivity toward N2 or its partially reduced species such as NHNH and NH2NH2. On the basis of the structural characters of FeMoco, much attention has been paid to developing sulfido-ligated transition-metal complexes. For instance, Coucouvanis et al. reported that the cubane-type MFe3S4 (M = Mo, V) clusters showed catalytic activity toward the reduction of hydrazine to ammonia. Hidai et al. reported the catalytic N−N bond cleavage of hydrazine by cubane-type RuMo3S4 and Mo2M2S4 (M = Ir, Rh) complexes. 5 In view of intermetallic cooperation, efforts toward the activation of the N−N bond by binuclear metal complexes were also made. For example, Hidai et al. demonstrated a thiolate-bridged binuclear Ru complex showing catalytic activity toward the disproportionation of hydrazine into NH3 and N2. 6 The groups of Lee and Holland reported that some diiron complexes bearing thiolate ligands could activate the N−N bond. It was also reported that the electrochemical reduction of a dimolybdenum complex bearing the HNNR (R = Me, Ph) ligand in the presence of an acid led to cleavage of the NN double bond to give aniline and an amido or ammine complex. However, these binuclear complexes could not reduce H2N−NH2 or HNNH to two NH3 molecules in a catalytic manner. The other powerful approach to elucidate the mechanism of N−N bond reductive cleavage related to nitrogen fixation in Received: October 10, 2011 Published: December 27, 2011 Article pubs.acs.org/Organometallics © 2011 American Chemical Society 335 dx.doi.org/10.1021/om200950q | Organometallics 2012, 31, 335−344 nitrogenase is theoretical calculation. A number of computational studies have been reported to discuss the binding of N2 to iron−sulfur clusters modeling FeMoco and their electronic properties. Dance and Rod et al. found that the on-top bonding of N2 to a single Fe atom was the most stable. However, calculations on the activation of the N−N triple bond of N2 or its reductive species, such as NHNH and H2N− NH2, mediated by iron−sulfur model complexes are relatively less. The chemistry associated with N2 reduction is most commonly discussed in the context of Schrock and Chatt mechanisms (Scheme 1a), which were initially proposed for mononuclear metal (Mo, W) complex systems. Despite different theoretical approaches, Stavrev and Nørskov found that the hydrogenation of N2 by FeMoco model compounds occurred through the [M]−NH2NH2 (M = metal) intermediate, as shown in Scheme 1b. However, Kas̈tner and Blöchl recently reported that the reduction of N2 to NH3 at the Fe sites of FeMoco favorably experience the NHNH and NHNH2 forms but not the NH2NH2 fashion. The reduction of HNNH at dimolybdenum centers reported by McGrady was also found to go through the NHNH2 forms (Scheme 1c) rather than the NH2NH2 fashion before NH3 release. Inspired by the results that a series of thiolate-bridged diruthenium complexes have diverse reactivities, two of the authors have recently explored the iron analogues. It was found that, for the first time, a thiolate-bridged diiron complex as a nitrogenase mimic possesses excellent catalytic activity toward N−N bond reductive cleavage of hydrazines. In that work, the catalytic cycle was also proposed, as shown in Scheme 2. In the case of R = Ph, complex A (Scheme 2) was structurally well characterized and could also catalytically reduce NH2NH2 (R = H in Scheme 2) to NH3. 19 As shown in Scheme 2, the HNNH moiety of Cp*Fe(μ-SEt)2(μ,ηHNNH)FeCp* (Cp* = η-C5Me5,) could be reduced to ammonia (R = H in A) and the species Cp*Fe(μ-SEt)2FeCp* (B) might be generated. However, the catalytic mechanism, including the reaction intermediates and whether B could bind N2, remained unknown. This stimulated us to further explore the possibility of B to bind N2 and further reduce it and to investigate the reductive mechanism involved in the reaction observed experimentally. Thiolate-bridged diiron clusters bearing cis-HNNH ligands have recently also isolated and structurally characterized, which show high activity toward the catalytic cleavage of the N−N bond of hydrazines. During our DFT studies on the electronic structure and reactivity of rare earth metal complexes, we also became interested in computing a transition metal complex system. In the present study, we performed DFT calculations to see whether B shown in Scheme 2 can possibly bind N2 and further reduce it to HNNH, a moiety of A (R = H in Scheme 2). The mechanistic details of further reduction of HNNH at the diiron centers was also elucidated. The structures and energetics associated with the reductive process were explored, from which the rate-determining step can be depicted. We also focused on examining the factors affecting the N−N bond activation. The theoretical results obtained in this study are expected to help experimentalists develop more efficient catalysts for catalytic activation of the N−N bond of hydrazine (H2N−NH2), diazene (HNNH), and perhaps N2 as well. ■ COMPUTATIONAL DETAILS The two-layer ONIOM(TPSSTPSS/6-31G*:UFF) calculations were carried out for modeling the binding of Cp*Fe(μ-SEt)2FeCp* with N2 and the reduction of N2 to the HNNH moiety. In the ONIOM calculation, the methyl groups of η-C5Me5 and μ -SEt ligands are placed in the outside layer treated by the UFF force field (low-level calculation), and the other atoms, including those in HNNH moiety Scheme 1. Mechanisms of N−N Bond Reduction at the Metal Center(s) Scheme 2. Previously Proposed Catalytic Cycle for N−N Bond Reduction at Diiron Centers Organometallics Article dx.doi.org/10.1021/om200950q | Organometallics 2012, 31, 335−344 336 and proton, constitute the inner layer. In these calculations, the steric effect of the methyl groups was considered for the N2 binding. For simplicity, however, the model compound CpFe(μ-SMe)2(μ,η 2-HN NH)FeCp (Cp = η-C5H5) was adopted for pure DFT calculations to model the cleavage of the NN double bond of the HNNH moiety at the diiron centers. The DFT calculation was performed with the TPSS exchange and correlation functional and the all-electron basis set 6-31G*. Normal-coordinate analyses were performed to verify the geometrically optimized stationary points and to obtain the thermodynamic data. To consider solvation effects, single-point selfconsistent reaction field (SCRF) calculations based on the conductorlike polarizable continuum model (CPCM) were performed for the gas-phase optimized geometries from ONIOM or pure DFT calculations. The CPCM model has been widely used for the investigation of solvation effects in various metal complex systems. The tetrahydrofuran was used as solvent, corresponding to the experimental conditions. In the SCRF calculation, the molecular cavity was built up by using UFF radii, and the 6-311+G(d,p) basis set was used for all atoms. To take into account the influence of enthalpy and entropy, the Gibbs free energy contributions from the gas-phase calculations were added to give the final free energy in solution (ΔGsol, 298.15 K, 1 atm). According to the experimental conditions, (LutH)BPh4 was used as the proton source and Cp2Co was used as the reductant. The energies for the reduction and protonation steps are therefore relative to the processes [Cp2Co] → [Cp2Co] + and LutH → Lut, respectively. The energy calculations of the two processes adopted the same strategy described above. That is, the SCRF single-point calculation at the TPSSTPSS/6-311+G** level was performed at the TPSSTPSS/6-31G* geometry. The energy profile was described by the relative free energies in solution. For each stationary point, the most stable spin state was tested. The open-shell species was treated with an unrestricted manner, and the stabilities of the wave functions were tested. The broken symmetry was used during the optimization. All of the calculations were performed with the Gaussian 09 program. ■ RESULTS AND DISCUSSION 1. Effect of Density Functional and Basis Set. Before accessing the mechanistic aspect, we tested the effects of several density functionals and basis sets on the geometry of the model compound CpFe(μ-SMe)2(μ,η 2-HNNH)FeCp and compared with its geometrical parameters with available crystal data of Cp*Fe(μ-SMe)2(μ,η 2-PhNNH)FeCp* observed experimentally. Since the B3LYP, TPSS, and BP86 functionals are often used to compute iron-containing metal complex systems, such functionals were also tested in this study. The results are shown in Table 1, where the atom labeling refers to Chart 1. With the B3LYP functional, the pseudopotential methods (BS1−BS3) produced an average error of more than 0.02 Å in distances and greater than 3° in angles. However, the all-electron basis set (BS4 and BS5) with the same functional gave better results. With the BS5 (6-31G* for all atoms), the TPSS and BP86 gave the best results: viz., an average error of 0.01 in distances and less than 2° in angles (Table 1). Considering that the TPSS functional has been successfully used for calculating an iron−sulfur cluster complex, the TPSS functional was selected in the present study. 2. N2 Binding and Its Partial Reduction to HNNH. To see whether the diiron complex Cp*Fe(μ-SEt)2FeCp* (B in Scheme 2) binds N2 and reduces the N2 to the NHNH form, a moiety of A (R  H, Scheme 2), we performed the TPSSTPSS/6-311+G**//ONIOM(TPSSTPSS/6-31G*:UFF) calculations. The computed energy profile and some important structures are shown in Figures 1 and 2, respectively. The superscript of the labeling of the stationary point shows the corresponding spin multiplicity and charge. For example, the labeling m0 denotes the neutral species m0 with a spin multiplicity of 3, and the labeling m1b represents structure m1b with a multiplicity of 3 and charge of +1 (protonation product). As shown in Figure 1, among the three N2 complexes of Cp*Fe(μ-SEt)2FeCp*, m0 1 (S = 0 state) and m0a (S = 0 state) are higher in energy then m0 (S = 1) by 8.70 and 7.34 kcal/mol, respectively. The bare complex Cp*Fe(μSEt)2FeCp* has a singlet ground state, and its triplet state is higher in energy by 3.62 kcal/mol. m0 and m0a show sideon/end-on and bridged end-on binding manners of N2, respectively. However, m0 shows an end-on binding fashion of N2. In fact, taking the structure of m0 1 or m0a as an initial structure, geometrical optimization at the multiplicity of 3 (S = 1 state) led to m0. In m0, the binding energy of N2 was computed to be 13.25 kcal/mol (containing a BSSE correction). Such a type of binding energy (BE) was calculated as BE  [E(N2) + E(cat)] − E(cat-N2), where E(N2), E(cat), and E(cat-N2) are the total electronic energies of N2, Cp*Fe(μSEt)2FeCp*, and the coordination complex Cp*Fe(N2)(μSEt)2FeCp*, respectively. These electronic energies were Table 1. Comparison of Computed Geometrical Parameters of Model Compound and Available Experimental Compounds (Interatomic Distances in Å and Angles in deg) functional and basis set Fe···Fe Fe1−S1 Fe1−S2 Fe1−N1 Fe2−N2 N−N Fe−S1−Fe Fe−S2−Fe av error exptl 3.211 2.289 2.300 1.875 1.831 1.337 89.2 88.8 B3LYP/BS1 3.178 2.342 2.341 1.896 1.848 1.287 85.6 85.7 0.04/3.4 B3LYP/BS2 2.992 2.273 2.273 1.834 1.768 1.299 82.5 82.6 0.07/6.5 B3LYP/BS3 3.175 2.335 2.333 1.896 1.844 1.290 85.8 85.9 0.03/3.2 B3LYP/BS4 3.187 2.334 2.332 1.881 1.837 1.301 86.3 86.4 0.02/2.7 B3LYP/BS5 3.194 2.339 2.337 1.879 1.836 1.300 86.3 86.5 0.03/2.6 TPSS/BS5 3.199 2.310 2.309 1.863 1.832 1.328 87.7 87.9 0.01/1.2 BP86/BS5 3.199 2.306 2.305 1.861 1.827 1.330 87.9 88.1 0.01/1.0 Model compound: CpFe(μ-SMe)2(μ,η 2-HNNH)FeCp. BS1: 6-31G* for C, H, N, and S atoms; LanL2DZ and associated poseudopotential for Fe atom. BS2: 6-31G* for C, H, N, and S atoms; LanL2DZ with outer p function and associated poseudopotential for Fe atom. BS3: 6-31G* for C, H, and N atoms; SDD and associated poseudopotential for S and Fe atoms, a single d polarization function (exponent of 0.65) augmented for S atom. BS4: 6-31G* for C, H, N, and S atoms; 6-311G* for Fe atom. BS5: 6-31G* for all atoms. The average error of interatomic distances (Å) and angles (deg, in italics). Taken from the crystal structure of Cp*Fe(μ-SMe)2(μ,η 2-PhNNH)FeCp*.19 Chart 1 Organometallics Article dx.doi.org/10.1021/om200950q | Organometallics 2012, 31, 335−344 337 computed at the level of TPSSTPSS/6-311+G**// TPSSTPSS/6-31G* theory. In a similar way, however, the binding of the solvent molecule THF to the metal centers was computationally found to be energetically unfavorable, possibly due to the steric effect of the THF molecule and the electrostatic interaction between the binding atoms: viz., Fe···O (see Figure S1 in the Supporting Information). The HOMO−LUMO energy gap of m0 is 1.783 eV at the TPSSTPSS/6-311+G**//ONIOM(TPSSTPSS/6-31G*:UFF) level, which suggests the stability of m0. The N−N bond length (1.141 Å) in m0 is longer than that (1.114 Å) in the free N2 molecule by 0.027 Å. This indicates that the N2 was activated by the diiron complex. The activation of the N2 moiety in m0 3 is also suggested by the NAO bond order (1.70) of the N−N bond. The bond order of 1.70 of the N−N bond indicates that the triple bond of dinitrogen could be reduced at least to a double bond once the dinitrogen is bound to the Fe center. Such a reduction was also previously reported for a hydride-bridged diniobium complex system. The bond order of Fe−Fe in m0 was computed to be 0.02, suggesting almost no Fe−Fe bond interaction in this complex. However, the Fe−Fe bond order of 0.22 was found for the bare complex Cp*Fe(μ-SEt)2FeCp*, suggesting a weak Fe−Fe interaction. An addition of one proton to a nitrogen atom of m0 led to m1b, which has an S = 1 state; the S = 0 state is 3.75 kcal/mol higher in energy. The subsequent reduction of m1b led to m1b with an S = /2 state; the S = /2 state is 5.30 kcal/mol higher in energy. There are two competing pathways for the isomerization of m1b to m1 with the μ-η2-NNH moiety. One is directly through TS, with an energy barrier of 11.81 kcal/mol. The other one is a stepwise process, which occurred via TS1, the intermediate m1a, and the transition state TS. The conversion of m1b to m1a is feasible, as suggested by the energy barrier of 4.02 kcal/mol. The TS1 has an S = /2 state, and attempts to locate its higher spin state (S = /2) were fruitless. Although the stepwise process needs to overcome a slightly larger energy barrier of 15.19 kcal/mol compared to the former process (barrier of 11.81 kcal/mol), the relative energies of TS (16.92 kcal/mol) and TS (15.86 kcal/mol) are similar (Figure 1). In fact, geometrical optimization of the lower spin state structure (S = /2) of TS 4 led to TS. A following intrinsic reaction coordinate, however, confirmed that the TS connects m1a and m1. The m1 has an S = /2 ground state; the S = /2 state is 12.36 kcal/mol higher in energy. In m1b , a spin density analysis shows that the Fe atom connecting to the NNH motif carries a spin population of 0.54, and another Fe atom has a spin density of 1.80. The NNH motif in m1b has a spin population of 0.44. In m1a (S = /2), however, the spin population mainly locates on the Fe centers, as suggested by the spin densities of 0.50 and 0.46 at the two Fe atoms, respectively. The structures and spin states of m1b, m1a, and m1 suggest that the bridging character of the NNH moiety could decrease the spin state to achieve a more stable structure: viz., m1. m1 could undergo protonation to give m2 and subsequent reduction to give m2. The m2 has an S = 0 ground state; the S = 1 state is 28.31 kcal/mol higher in energy. The whole process for the reduction of dinitrogen to the NHNH moiety assisted by the diiron complex is exergonic by 36.05 kcal/mol. Two possible structures, viz., m2a and m2b, which could be obtained by protonation and subsequent reduction of m1 and m1a, respectively, were also located (Figure 2). However, both of them with a NNH2 moiety are higher in energy than m2 by 27.9 and 13.99 kcal/mol, respectively. This suggests that an intermediate with a NNH2 moiety (cf. Scheme 1a) was unlikely to be involved in the current reaction system. m2 has the μ,η2-HNNH moiety and is actually the species A (R = H in Scheme 2). As mentioned above, m2 was experimentally found to experience NN double bond reductive cleavage and finally to give two NH3 molecules. 19 In this sense, our computational results suggest that the diiron complex Cp*Fe(μ-SEt)2FeCp*, which is Figure 1. Computed energy profile for the protonation/reduction of dinitrogen to HNNH. The schematic representation of protonated species is similar to that of the corresponding neutral species and therefore is not included in this figure. The same is true for Figures 2−4. Organometallics Article dx.doi.org/10.1021/om200950q | Organometallics 2012, 31, 335−344 338 computationally found to be a minimum, can possibly reduce N2 to ammonia via protonation and reduction processes. 3. Reductive Cleavage of the NN Bond of the HN NH Moiety. The reductive cleavage of the NN bond of the HNNH moiety at the diiron centers has been experimentally observed. The computed energy profile of such a process is shown in Figure 3. The free energies in solution are relative to the complex 1 and the protonation (or reduction) processes illustrated in Computational Details (vide ante). The superscript of the labeling of the stationary point (Figure 3) shows the corresponding spin multiplicity and charge, which is similar to that for stationary points in Figure 1. As shown in Figure 3, the starting complex has two almost isoenergetic spin states, S = 0 (1) and S = 1 (1), with an energy difference of 0.15 kcal/ mol. The protonation of 1 is endergonic by 13.48 kcal/mol, leading to 2 with an S = 1 state; the S = 0 state is 4.34 kcal/ mol higher in energy. The subsequent reduction of 2 gives 2 with an S = /2 state; the S = /2 state is 13.97 kcal/mol higher in energy. The first protonation/reduction process is slightly endergonic by 1.80 kcal/mol. We also explored the possibilities of protonation/reduction occurring at a Fe or S site in 1. It was computationally found that the resulting species CpFe(μSMe)2(μ,η 2-HNNH)Fe(H)Cp (S = /2 ground state) and CpFe(μ-SMe)(μ-S(H)Me)(μ,η2-HNNH)FeCp (S = /2 ground state) are higher in free energy by 11.97 and 17.95 kcal/mol, respectively, in comparison with 2. This suggests that the first protonation/reduction process preferably occurred on a nitrogen atom, such as N1 (see Chart 1 for the atom labeling), and led to the complex 2. The protonation/ reduction occurring at the N1 atom slightly changed the core structure of 2. For example, the N−N bond length of 1.457 Å (Figure 4) and the Fe···Fe distance of 3.245 Å (Table 2) are longer than that in 1 (1.328 and 3.236 Å, respectively). In 2, the Fe−NH2 contact (2.012 Å) is also therefore longer than the Fe−NH bond length (1.833 Å). Unlike the case for 1, the two Fe−S bonds in the Fe−S−Fe connection of 2 are no longer equal. The populations of Mulliken charge and spin density are also changed during the protonation/reduction process (Table 2). The 2 has unequal charge population on the two Fe atoms (0.45 and 0.58). The same is true for the two N atoms. It is found that the unpaired electron of 2 (S = /2) is mainly distributed on the Fe2 (0.79) and N2 (0.22) atoms (Table 2). 2 may coexist with the NH-bridged structure 3, and their interconversion occurs via the transition state TS[2-3] (Figure 3). The conversion of 2 to 3 needs to overcome a free-energy barrier of 22.07 kcal/mol, and 3 is slightly higher in energy than 2 by 1.83 kcal/mol. However, the energetically favorable protonation and subsequent reduction of 3, which led to 4 Figure 2. Optimized stationary points for N2 binding and its partial reduction. The relative free energies in solution are given in kcal/mol. The distances are in Å and angles are in deg. The methyl groups of the Cp* ligand are omitted for clarity. Organometallics Article dx.doi.org/10.1021/om200950q | Organometallics 2012, 31, 335−344 339 and then to 4, could facilitate the conversion of 2 to 3 via TS[2-3]. The higher spin states (S = /2) of the three stationary points involved in this isomerization are higher in energy by 3−18 kcal/mol (Figure 3). The 4 (S = /2) and 4 (S = 1) are about 2.5 kcal/mol higher in energy compared with their corresponding lower spin states. An isomeric structure of 3, viz., CpFe(μ-SMe)2(μ-NNH3)FeCp, which was obtained via hydrogen transfer in 3, was computationally found to be 23.70 kcal/mol higher than 3. This suggests that the N−NH3 form (cf. the Chatt pathway in Scheme 1a) was unlikely to be involved in the reduction process of HNNH. Several isomeric structures of 4 were also computationally located, viz. 4a−d, as shown in Chart 2. 4a,b, having newly formed Fe− H bonds, are higher in energy than 4c,d with newly formed HN−H and H2N−H bonds (Chart 2). However, all of them are significantly less stable by more than 24 kcal/mol in comparison with 4 with (μ2-NH)···NH3 moiety (Chart 2 and Figure 3). Therefore, the formation of 4c with (μ,η-H2N− NH2) moiety (cf. the FeMoco pathway in Scheme 1b) also is unlikely to occur during the current reduction process. The Fe···Fe distance in 3 is shortened by 0.631 Å compared with 2 (Table 2). The structure of TS[2-3] shows a Fe···S contact of 2.548 Å, which is significantly longer than that (∼2.30 Å) in 2 and 3. This indicates that the Fe−S bond may serve as a “switch” to assist the interconversion between 2 and 3. It is noteworthy that 2 has an unequal charge population on the two Fe atoms, and the same is true for the spin density on the Fe atoms of 2 (Table 2). Such unequal populations may be a driven force for the conversion of 2 to 3 having equal populations of both charge and spin density on the two Fe atoms. In 4, as the product of second protonation and subsequent reduction processes, the N−N bond has completely cleaved, and the resulting NH3 molecule to be released interacts with the bridging μ2-NH moiety via a hydrogen bond (see structure 4 in Figure 4). Attempts to locate the N−N bond cleavage transition state during the second protonation/ reduction processes were fruitless. Actually, relaxed scans of the N−N contact against the energy for the second protonation and reduction steps show no significant transition state region on the potential energy surface considered (Figure 5). This result suggests that the N−N bond cleavage has no energy barrier with respect to the current calculation. The release of NH3 from 4 1 led to the intermediate 5 with an NH-bridging feature. The S = 1 state of this structure is slightly higher in energy by 2.16 kcal/mol. The 5 underwent the third and further the fourth protonation/reduction processes to give 7 with the newly formed NH3 (Figure 3). The formation of 7 3 is exergonic by 80.65 kcal/mol. The binding energy of NH3 in 7 3 was computed to be −23.63 kcal/mol (containing BSSE correction), while the binding energy of HNNH in 1 was computed to be −103.58 kcal/mol. This suggests that the release of NH3 from 7 3 via an access of HNNH is energetically favorable, and the catalytic cycle could be achieved. As shown in Figure 3, the isomerization of 2 to 3 via TS[23] needs to overcome an energy barrier of 22.07 kcal/mol, which is greater than the energy required for the first protonation step (13.48 kcal/mol). The former could be therefore the rate-determining step for the reduction of the HNNH moiety to two NH3 molecules. For this reason, we further analyzed the stationary points involved in the ratedetermining step. To get more accurate results, such an analysis is based on single-point calculations (tpsstpss/6-311+G**) on the optimized geometries. The results are shown in Tables 3 Figure 3. Computed energy profile for the reductive cleavage of the NN double bond. Organometallics Article dx.doi.org/10.1021/om200950q | Organometallics 2012, 31, 335−344 340 and 4. As shown in Table 3, the Wiberg bond indexes (WBI) for Fe···Fe contacts significantly increased from 0.06 in 2 to 0.23 in 3. This suggests that the very weak interaction between the two Fe atoms was strengthened by the bridging of the μNH−NH2 moiety. However, the strength of the N−N bond did not significantly change during the isomerization of 2 to 3, as suggested by the similar bond indexes (∼1.0) in 2, TS[23], and 3. The WBI of Fe1···S1 in TS[2-3] (0.57) is smaller Figure 4. Optimized stationary points involved in the HNNH reduction. The distances are in Å, and angles are in deg. Table 2. Mulliken Charge, Spin Density, and Fe···Fe Distances (Å) in the Computed Stationary Points Mulliken charge spin density stationary point Fe1 Fe2 N1 N2 Fe1 Fe2 N1 N2 Fe···Fe dist 1 (S = 1) 0.56 0.56 −0.54 −0.54 0.73 0.72 0.29 0.29 3.236 2 (S = 1) 0.53 0.60 −0.67 −0.60 1.10 0.86 −0.03 0.21 3.209 2 (S = /2) 0.45 0.58 −0.64 −0.60 0.02 0.79 −0.01 0.22 3.245 TS[2-3] (S = /2) 0.56 0.50 −0.61 −0.61 0.59 0.13 0.02 0.20 3.102 3 (S = /2) 0.47 0.47 −0.54 −0.67 0.48 0.48 0.00 0.07 2.614 4 (S = /2) 0.51 0.51 −0.93 −0.74 0.45 0.44 0.00 0.02 2.488 4 (S = 0) 0.51 0.51 −0.90 −0.79 2.747 5 (S = 0) 0.53 0.53 −0.76 2.762 6 (S = 1) 0.48 0.48 −0.93 0.99 0.99 0.04 2.526 6 (S = /2) 0.45 0.45 −0.92 0.48 0.48 0.07 2.621 7 S = /2) 0.34 0.47 −0.95 1.11 −0.06 0.00 2.548 7 S = 1) 0.42 0.53 −0.93 2.01 0.00 0.00 3.395 4a (S = 1) 0.49 0.50 −0.63 0.57 0.94 0.89 −0.02 0.10 3.695 4b (S = 1) 0.47 0.57 −0.67 −0.56 1.06 0.88 −0.02 0.11 3.631 4c (S = 0) 0.47 0.47 −0.66 −0.66 3.303 4d (S = 1) 0.49 0.54 −0.96 −0.75 0.03 1.30 0.02 0.68 3.438 Atom labeling referring to Chart 1. The data were computed at the level of TPSSTPSS/6-31G*. Organometallics Article dx.doi.org/10.1021/om200950q | Organometallics 2012, 31, 335−344 341 than that in 2 and 3 (0.76), suggesting that the Fe1−S1 bond serves as a “switch” during the isomerization of 2 to 3. The bonding interaction between the Fe1 and N2 atoms is almost absent in 2 but is observed in 3 (bond index of 0.66). Also, the Fe1···N1 bonding in 2 (bond index of 0.57) almost disappeared in 3 (bond index of 0.03). On the other hand, Fe2···N2 bonding exists in both 2 and 3. These results support the μ-NH−NH2 bridging character of 3. An analysis of Mulliken charge population suggests an event of net electron transfer from the Cp and SMe2 ligands to the Fe atoms and the NHNH2 moiety during the conversion of 2 2 to 3 (Table 4). Also, the two SMe2 ligands served as major electron donors. This suggests that increasing electron density on sulfide and ancillary Cp ligands may facilitate such an electron transfer event and therefore be beneficial to the subsequent hydrogenation of the NHNH2 moiety in 3 . In light of the process discussed above, the reductive cleavage of the HNNH double bond at the diiron centers could occur along with the sequences [M]NHNH → [M]NHNH2 → ([M]NH + NH3) → ([M]NH2 + NH3) → ([M] + 2NH3) (M = metal centers). The feature of this cleavage mechanism is that the release of the first NH3 molecule is via the protonation/ reduction of the [M]NHNH2 intermediate. Such a process is in contrast to either the Chatt mechanism (Scheme 1a), where the release of the first NH3 molecule is directly from the [M]NNH2 form, or the mechanism proposed for N2 reduction on the FeMoco model compounds (Scheme 1b), where the release of the first NH3 molecule occurred through the hydrogenation of [M]NH2NH2 species. This is because the [M]NNH3 (cf. Chatt pathway in Scheme 1a) and [M]NH2NH2 (cf. FeMoco pathway in Scheme 1b) species are higher in energy than 3 and 4, respectively, and therefore was unlikely to be involved in the reduction process (vide ante). Although the reductive mechanism obtained in this study is similar to the Schrock route (Scheme 1a), where the release of the first NH3 is via the hydrogenation of [M]NHNH2, our computation suggests that the N2 reduction at the diiron centers occurs through the [M]NHNH intermediate rather than the [M]NNH2 species (Figures 1 and 2) proposed in the Schrock mechanism and therefore follows the mechanism recently proposed for N2 reduction at the Fe sites of FeMoco. 15 In this sense, it is of importance to take the thiolate-bridged diiron complex investigated here as a nitrogenase model compound to study the mechanism of N−N bond reduction. According to the computational results obtained in this study, one feature of the reduction of N2 to NH3 assisted by the diiron complex is to occur via HNNH and HNNH2 forms rather than the H2NNH2 unit. The other feature of such a process is that the ratedetermining step could be the isomerization of (μ,η-HN− NH2) moiety to the bridging form (μ-HN−NH2), and such an isomerization is reversible.