Revisiting the photochemical synthesis of [FeFe]-hydrogenase mimics: reaction optimization, mechanistic study and electrochemical behaviour

The photoreaction of [(μ-S)2Fe2(CO)6] and alkenes or alkynes has been optimized to readily obtain functionalized [FeFe]-hydrogenase mimics. Irradiation under low CO pressure in THF produces the corresponding photo-adducts in good/acceptable (alkenes/alkynes) yields, with retention of the starting olefin stereochemistry. DFT-calculations provide plausible reaction pathways in both, singlet and triplet states. The DFT-calculation based in the singlet state is energetically more favorable. The electrochemical behavior of the synthesized compounds is also presented, including studies in acidic media. The electrochemical properties of the products vary in the presence of a double bond (cycloaddition of [(μ-S)2Fe2(CO)6] to alkynes), respect to a single bond (cycloaddition to alkenes).


Introduction
One of the challenges of chemistry in the 21st century is to develop viable methods to produce hydrogen. 1 Hydrogenases are metalloenzymes able to generate molecular hydrogen by reducing protons in anaerobic media. 2 The use of hydrogenase bio-inspired analogues is a promising option for the production of hydrogen. Current research in this eld is mainly focused on two different approaches. The rst one uses whole organisms, inorganic hybrids or supported enzymatic systems. 3 The second approach uses mimics of hydrogenases, namely synthetic small molecules that when coupled to other reagents and catalysts are able to produce hydrogen. 4 Fragment I is the basic motif of the hydrogen production moiety of a [FeFe]-hydrogenase ( Fig. 1). Much synthetic research in this eld targeted the preparation of mimics with a simplied structural motif II similar to I. 4b A second group of [FeFe]-H 2 ase synthetic models has the [(m-SR) 2 Fe 2 (CO) 6 ] motif as the essential core (III in Fig. 1). 4a This second group of mimics has been less studied in the photocatalytic production of hydrogen. 4c The preparation of type III mimics is achieved by thermal reaction of either Fe 2 (CO) 9 or Fe 3 (CO) 12 and sulphides or disulphides to yield compounds having structure 1 (Scheme 1). 4b This approach is versatile and provides access to sophisticated structures. However, again the reaction conditions are not tolerated by several classes of substrates. Additionally, the precursors of the sulphides or disulphides are not always easy to access. 5 An alternative and potentially useful approach to introduce the [(m-S) 2 Fe 2 (CO) 6 ] 2 moiety into substrates not compatible with the conditions used by standard approaches would be the photocycloaddition of [(m-S) 2 Fe 2 (CO) 6 ] and alkenes or alkynes. The photochemical reaction of 2 and simple unfunctionalized substrates has been previously reported. However, yields of photoreactions are usually very poor. Thus, irradiation of [(m-S) 2 Fe 2 (CO) 6 ] 2 and simple olens 6 including 1-and 2-pentene 3a and 3b yielded the corresponding photoadducts 4a and 4b in 6.9% and 8.9% yield, respectively (Scheme 2).
Similar low yields were obtained with both, acyclic 6 and cyclic 7 dienes. The only exceptions are ethylene 8 and p-benzoquinone, 9 that produce the corresponding photoadducts in 65% and 53% yields, respectively. Finally, several C 60 [S 2 Fe 2 (CO) 6 ] n (n ¼ 1-6) and C 70 [S 2 Fe 2 (CO) 6 ] n (n ¼ 1-4) mixtures were obtained from [(m-S) 2 Fe 2 (CO) 6 ] and C60 and C70 fullerenes. The C 60 [S 2 Fe 2 (CO) 6 ] adduct was separated from the mixture with a 52% yield based on recovered C60, while C 70 [S 2 Fe 2 (CO) 6 ] adduct was obtained with a 63% based on recovered C70. 10 The mechanism of these photoreactions remains unexplored. The mechanisms and synthetic applications for organometallic compounds photochemistry are intrinsically different from their all-carbon counterparts, being a subject of general interest. 11 Despite the reported low yields for the photocycloaddition of [(m-S) 2 Fe 2 (CO) 6 ] and alkenes/alkynes, this reaction may be a good alternative to include this [FeFe] moiety into substrates incompatible with the reaction condition used by other synthetic approaches to these classes of compounds. We report herein a useful optimized approach to incorporate the [(m-S) 2 Fe 2 (CO) 6 ] into smooth reaction conditions to different classes of substrates, as well as a proposal for the reaction mechanism using DFT calculations.

Results and discussion
Complex [(m-S) 2 Fe 2 (CO) 6 ] 2 and 1-hexene were used to tune up the reaction conditions. Light source and solvent were rst investigated. Thus, irradiation of equimolar amounts of [(m-S) 2 Fe 2 (CO) 6 ] and 1-hexene in anhydrous THF using 4 Â 60 W blue light LEDs did not produce any reaction product. The reagents were recovered unaltered aer 72 hours of irradiation. The use of medium pressure Hg-lamps (Pyrex lter and Pyrex well) produced the desired photoadduct 4c in 31% (400 W) and 47% (125 W) isolated yields. A 6.9% yield for the reaction of 1pentene and [(m-S) 2 Fe 2 (CO) 6 ], using a high-pressure Hg-lamp and quartz glassware, was previously reported. Thus, ltering the UV component of the irradiation source clearly increases the reaction yield. This yield improvement is probably due to a smaller decomposition of the diiron complexes by the COligands photo-removing effect (see below). Other solvents like MeCN (14%), benzene (18%), and Et 2 O (11%) produced lower isolated yields of the adduct 4c (Scheme 3).
Dependence of yields with the choice of solvent pointed to a competitive light-induced CO dissociation leading to, either decomposition or tetrameric species. 12 Thus, THF would ll iron coordination vacants avoiding or retarding competitive undesired reactions. This hypothesis would imply a yield increment under CO-atmosphere. However, it has been reported that complex [(m-S) 2 Fe 2 (CO) 6 ] reacts with CO to form the CO adduct 5 with a 47% yield (Scheme 3). 8 Nevertheless, the reaction of [(m-S) 2 Fe 2 (CO) 6 ] 2 and 1-hexene was repeated under 1 atm (14 psi) of CO and, compound 4c was obtained with a 64% isolated yield. The reaction crude material was cleaner and decomposition of the starting diiron complex 2 was not observed. Therefore, it is clear that CO atmosphere hampers the photo-extrusion of CO and thence the decomposition of the [(m-S) 2 Fe 2 (CO) 6 ], increasing the reaction yields. However, the use of higher pressures of CO (40 psi) resulted in lower yields of the desired product. Competitive CO insertion to produce 5 might be the cause of these lower yields. 8 Fine tuning of the reaction conditions of this photocycloaddition allows a yield increment from the described 6-9% up to 65% in the case of simple aliphatic olens. Functionalized alkenes like N-phenylmaleimide 3d and methyl acrylate 3e were reacted with [(m-S) 2 Fe 2 (CO) 6 ] 2 to form the corresponding photoadducts 4d and 4e with 70% and 86% isolated yields, respectively. These yields were achieved with THF as the choice solvent under 1 atm of CO and a 125 W medium pressure Hglamp (Pyrex lter and Pyrex well) (Scheme 4).
Series of both, terminal and disubstituted alkynes were next tested as starting substrates. p-Tolyl acetylene 3f formed the corresponding adduct 4f in 45% yield, while methyl propyolate 3g and dimethyl acetylendicarboxylate 3h yielded the corresponding adducts 4g (38%) and 4h (62%). Although functionalized alkynes formed the corresponding cycloadducts in lower yields than those obtained for alkenes, they could be used as substrates for the cycloaddition process. Therefore, the method is general and tolerates a variety of functional groups (vide infra). 13 Photoadducts derived from the reaction of 2 with olens could be formed either as cis-or trans-isomers in the newly formed metallacycle. The symmetry of our molecules avoids the assignation of the cis-trans stereochemistry by conventional NMR techniques. Crystals of compound 4d suitable for X-ray diffraction were grown from a DCM/hexane solution. The Xray structure determination of 4d unambiguously conrms the cis arrangement of the fused bicyclic system (Fig. 2). Molecular structure of 4d shows a [(m-SR) 2 Fe 2 (CO) 6 ] complex with a buttery structure for the [2Fe-2S] cluster. Both iron atoms adopt a distorted square-pyramidal geometry. The Fe-Fe bond length (2.4966(3)Å) lies in the range found for similar ethylenedithiolate-hexacarbonyl-di-iron structures 14 (2.454-2.546Å). Fe-Fe bond length in compound 4d is shorter than in metalloenzymes Hydrogenase DdI (ca. 2.55Å) or CpI (ca. 2.62 A). 15 The dithiolate bridging ligand and both iron atoms form two fused ve-membered metallocycles with the nitrogen substituent N-Ph group bending towards the Fe(1) atom. This conformation implies short intramolecular distances between the nitrogen atom N(1) and the closest carbonyl group C (23) 16 Interactions between N-arene group and the closest carbonyl group has been previously described to produce an enlargement on the C-Fe-Fe angle for the implicated carbonyl group in azadithiolates diirion structures. 5,17 Thus, in compound 4d the C(23)-Fe(1)-Fe(2) angle is 5.25(5) larger than the C(26)-Fe(2)-Fe(1) angle.
Substrates having electroactive moieties were next tested. Methyl trans-2-ferrocenylacrylate 3i reacts with 2 and the Scheme 5 Photochemical reaction between [(m-S) 2 Fe 2 (CO) 6 ] and olefins or alkynes. Compatibility of the process with functionalized substrates. Compound 4i was a racemic material, one single enantiomer is represented for simplicity. corresponding photo-adduct 4i is isolated in 43% yield. The trans stereochemistry of the starting ferrocene derived olen 3i is again maintained in the nal adduct (d ¼ 4.13 and 3.16 ppm, d, J ¼ 6.3 Hz for both CH-S groups). To conrm that the stereochemistry of the starting material is retained in the photocycloaddition, NOE experiments were performed for complex 4i on a 500 MHz NMR spectrometer. Irradiation of the signal at 3.16 ppm, corresponding to the CH-CO proton, showed a main NOE effect with the proton at 3.98 ppm (substituted Cp ring). This observed NOE effect points to a trans relative disposition of the CO 2 Me and the Fc moieties which is in good agreement with the concerted proposed calculated mechanism (see below).
Nucleotide 3j was next tested. In this case the product incorporating the [(m-S) 2 Fe 2 (CO) 6 ] moiety was obtained with a 24% isolated yield. Despite the high functionalization of 3j, no by-products were obtained, and unaltered starting materials could be recovered (Scheme 5).
The possibility of achieving a double photocycloaddition to obtain tetrametallic systems is also addressed. N,N 0 -(1,4-phenylene)dimaleimide 3k reacts with [(m-S) 2 Fe 2 (CO) 6 ], with no further reaction progress observed (tlc) aer 15 hours of irradiation. From the crude reaction mixture, tetrametallic complex 4k was obtained with a 26% isolated yield. An analogous reaction was carried out with ferrocene complex 3l. A mixture of pentametallic complex 4m (single diastereomer, 52%) and trimetallic complex 4l (35%) was obtained. It is worthy to note that the four stereogenic centers of complex 4m are formed in a totally stereoselective way maintaining the conguration of the starting olens (Scheme 6). A high degree of diastereoselectivity has also been achieved in this reaction. An analogous result was obtained from the bis-allyl derivative of hydroquinone 3o which lead to a mixture of dimetallic 4o and tetrametallic complex 4p in 55% and 20% isolated yields, respectively. Although the analysis of the crude mixtures of 4p showed a single product, there are no reasons to believe that a complete stereoselectivity was achieved in this case. Probably, 4p is a mixture of diastereomers but the chiral centres are wellseparated and the differences in their NMR data may be null. Finally, the bis-propargyl derivative of hydroquinone 3n was not able to form the tetrametallic derivative, while bimetallic derivative 4n was obtained with just a 17% yield (Scheme 6). This is in good agreement with the observed lower reactivity of alkynes.

Mechanistic studies
According to a previous theoretical study, 18 irradiation of the starting diiron complex 2 with UV-light would generate two buttery isomers or a rhombus isomer by breaking one or both of the Fe-Fe and S-S bonds (see Fig. 4 in ref. 18). This study concludes that photochemical reactions of complex 2 should proceed through Fe-Fe buttery biradical Fe 2 (CO) 6 S 2 intermediates (Fig. 3).
However, optimization of the Fe-Fe buttery using unrestricted uBP86 functional together with the command guess(mix, always) or restricted BP86 yielded the same energy minimum. Careful examination of the spin densities and bond distances in the output les did not match a biradical species in any case. Optimization of the reaction pathway was calculated for the cycloaddition between starting complex 2 and both, Scheme 6 Photochemical reaction between [(m-S) 2 Fe 2 (CO) 6 ] and olefins or alkynes. Synthesis of polymetallic systems. Compounds 4l, 4m were racemic mixtures. Only one enantiomer is depicted for clarity. methyl propiolate 3g (Scheme 7) and methyl acrylate 3e (Scheme 8).
We rst tested the possibility of reacting methyl propiolate with the FeFe-buttery intermediate in the singlet state. In order to contemplate the possibility of biradicals implied in the process, broken spin symmetry [UBP86 + guess(mix, always)] was compared to restricted RBP86. Both calculations for the rst TS of the pathway converged to the same minimum which should, in principle, discard a triplet diradical reaction pathway. Restricted singlet-state calculations for a concerted cycloaddition reaction are shown in Scheme 7. An alternative pathway involving triplet excited states was also contemplated. This process should involve a stepwise cycloaddition having a DDG ‡ ¼ 30.1 kcal mol À1 in the rate-determining step, which makes the process less probable than the concerted pathway (DDG ‡ ¼ 9.9 kcal mol À1 ). Moreover, the calculated nal product 6 for this alternative mechanism has a structure different to the experimentally isolated complex 4g. These species, lacking one Fe-S bond were not observed in any of the experiments carried out in this work. 19 The reaction of complex 2 and methyl acrylate 3e was also calculated in the singlet and triplet spin states. Results for the singlet state are similar to those obtained for the methyl propiolate. A concerted reaction pathway with a low activation barrier (DDG ‡ ¼ 8.7 kcal mol À1 ) drives the reaction to the formation of the experimentally isolated product 4e. Unrestricted UBP86 singlet state was also tested and again it converged to the same energy minimum obtained with restricted BP86 one. While the triplet initial state of the reagents was found to be only 5.7 kcal mol À1 over the singlet, the two steps process was found to have an overall DDG ‡ of 35.2 kcal mol À1 which makes this process unfavorable when compared to the singlet concerted cycloaddition mechanism (Scheme 8). Therefore, we can safely conclude that the photoreaction of [(m-S) 2 Fe 2 (CO) 6 ] with alkenes and alkynes is a concerted process, which additionally accounts for the observed retention of the stereochemistry of the starting olens into the obtained nal products. 20  Table 1). Alkyne-derived compounds present a double bond within the metallacycle that is not present for alkene-derived compounds.
Compounds derived from alkenes behave like the analogous derivatives having [(m-SR) 2 Fe 2 (CO) 6 ] structures. 21 However, compounds derived from alkynes show a strongly displaced anodic wave (even lower than those derivatives of type II in Fig. 1), together with the new reversible wave (see Fig. 5 for comparison). A similar behaviour has been reported for complex 4h. 22 DFT calculations (BP86/Def2tzvpp/SCRF, CPCM-MeCN) were performed for further understanding this anodic displacement and the electrochemistry of these complexes. The LUMO in complex 4e is clearly centered in the [FeFe] moiety, while the LUMO of complex 4g having a double bond has a strong component in the organic moiety of the metallacycle (Fig. 6). Therefore, the strong anodic displacement caused by the presence of one double bond in the metallacycle may be explained by the reception of the electron by the organic moiety. Contrary to the complexes having one double bond those complexes having a saturated moiety receipt the electron into the metallic moiety, which accounts for a reduction potential in the À1.44 to À1.61 range. This situation is maintained in the radical-anions 4ec À , and 4gc À . The LUMO orbital of 4ec À is still localized across the [FeFe] fragment while radical anion 4gc À has the LUMO located in the organic moiety, which is easily reducible (still with strong anodic displacements, giving lower reduction potentials than their saturated congeners, due to the presence of the metals). 23 This proposal nicely explain the "strong effect of the dithiolene and (in their case) tetrachloro-biphenyl dithiolate groups on the level of the LUMO" reported by Gloaguen and Schollhammer. 22 The electrochemical behavior in acidic media of complexes 4e and 4g (as representative examples of complexes having either a saturated or double bond in the bridge joining the sulfur atoms) was next studied. None of these complexes showed electrocatalytic behavior in their rst reduction wave in the presence of increasing amounts of acetic acid (pK a $ 22.3 in MeCN) 24 (up to 20 eq., see Fig. 7), while a peak appears around À1.80 V which increases its intensity with the concentration of acid. These results are fully consistent with those reported in the literature for related compounds. 25 It should be noted that   the rst reduction wave at À0.42 V for compound 4g (the one attributed to the reduction of the double bond) remains quasireversible, while the second reduction wave at À0.60 V losses its quasi-reversibility in the presence of AcOH as previously reported. 22 The behavior of complexes 4e and 4g towards a stronger acid (CF 3 COOH, pK a $ 12.6 in MeCN) 24 was next studied. Fig. 8 shows the behavior of these complexes upon increasing additions of CF 3 COOH. The reduction of 4e (Fig. 8, top) becomes irreversible upon addition of less than 1 eq. of CF 3 COOH, as reported in the literature for related compounds. 26 However, the intensity of the reduction wave at À0.97 V steadily increases with the concentration of acid and slightly shis towards more negative values (up to 20 eq. of added CF 3 COOH). Therefore, the species generated in the electrochemical reduction are able to catalyze the proton reduction. 26a In addition, reduction of protons is also observed around À1.60 V, even at low acid concentrations (<0.5 eq.).
Complex 4g (Fig. 8, bottom) behaves differently. In this case, both waves (À0.42 V and À0.60 V) increase their intensity with the acid concentration until the ratio 4g/acid exceeds 4 equivalents. This supports the participation of the double bond in the rst reduction event, generating species that are able to catalyze the reduction of protons.    6 ] and alkene/alkynes under medium-low CO pressures produce the corresponding photoadducts in good (alkenes) or acceptable yields (alkynes). The formation of photoadducts derived from alkenes occurs with retention of the stereochemistry of the starting olen, as demonstrated by NOE measurements and X-ray diffraction. The process is compatible with substrates having ferrocene moieties, as well as functional groups like imides and esters. The photocycloaddition occurs through a concerted reaction pathway as demonstrated by extensive DFT-calculations. The stereochemistry of these reactions is compatible with the computed pathway. Alternative reaction pathways involving triplet states are considerable higher in energy, and, for alkynes, predict the formation of products that have not been detected experimentally. Photoadducts formed from alkynes present a double bond within the metallacycle, that strongly affect the electrochemistry of these compounds. Thus, in the presence of this double bond two strongly anodic displaced quasi-reversible reduction waves appear. These reduction events are compatible with the one electron 17 reduction of the double bond "conjugated" with the [FeFe]-moiety. This one electron reduction, forms a radicalanion with a formal [Fe 0 Fe I ] state, which facilitate the second reduction to form the [Fe 0 Fe 0 ] state.
The electrochemistry of complexes 4 in the presence of acids reveals a different behaviour between complexes having a double bond in the dithiametallacycle and those lacking this insaturation. Thus, complexes lacking the insaturation in the metallacycle behave like the analogous products reported in the literature. For this compounds, in the presence of so acids (AcOH) the species derived from the quasi-reversible reduction wave around À0.97 V are electrocatalytically inactive, and a new electrocatalytically active band appears at À1.80 V. Complexes 4 having a double bond in the metallacycle behave similarly towards so acids. However, in the presence of strong acids (CF 3 COOH) the species formed upon reduction in the wave around À0.97 V are able to reduce protons. For these unsaturated complexes, a new reduction wave appears around À1.60 V that is also catalytically active. Therefore, for complexes having a double bond, both waves (À0.42 V and À0.60 V) become catalytically active, showing the participation of dithiolene ligand in its structure.
Further work to apply these smooth methodologies to prepare more sophisticated [FeFe]-mimics, together with post functionalization of the photoadducts, is now underway in our laboratories.

Experimental section
General Flame-dried glassware was used for moisture-sensitive reactions, and anhydrous solvents were taken from a Pure Solvent PS-MD-5 apparatus. Silica gel (Merck: 230-400 mesh) was used as stationary phase for purication of crude reaction mixtures by ash column chromatography. NMR spectra were recorded at 25 C in DMSO-d 6 or CDCl 3 on a 300 and 500 MHz spectrometers. IR spectra were taken on a MIR (8000-400 cm À1 ) spectrometer using the attenuated total reectance (ATR) technique. HRMS experiments were recorded on an Agilent 6500 accurate mass apparatus with a Q-TOF analyzer. Cyclic voltammograms were recorded using a Metrohm Autolab Potentiostat model PGSTAT302N with a glassy carbon working electrode, Ag/AgCl 3 M as reference and a Pt wire counter electrode. All the measurements were performed under Ar, at room temperature from CH 3 CN solutions containing 0.1 M [N n Bu 4 ]PF 6 as supporting electrolyte, with analyte concentrations of 1 mM (scan rate 0.1 V s À1 ). When needed, an ultrasound bath was used to promote solubilization in those samples were a suspension was initially obtained.

Computational details
Theoretical calculations have been performed using the Gaussian 09-D.01 soware package 27 at the BP86/Def2tzvpp 28  level for all atoms. A SCRF, CPCM 29 solvent model for THF was also used. Compounds 4e, 4g and their corresponding radical anions were also calculated using MeCN as solvent in order to match the conditions used in the electrochemical experiments An ultrane-grid was used as integration grid for all the calculations as implemented in the G09 soware suite.
General procedure for the synthesis of [FeFe]-hydrogenase mimics Photochemical reactions. Photochemical reactions were conducted by using a 125 W or 400 W-medium pressure mercury lamp through a pyrex lter/pyrex well. Starting materials were dissolved in dry and degassed (vacuum-Ar, four cycles) THF in a rubber septum-sealed Pyrex tube purged with argon. In a typical experiment, an equimolecular solution of [(m-S) 2 Fe 2 (CO) 6 ] 2 and the corresponding alkene or alkyne in dry THF (200 mL mmol À1 ) was bubbled with CO for 5 minutes and was irradiated overnight under CO pressure (1 atm, balloon). The solvent was then removed under reduced pressure, and the product was puried by SiO 2 column chromatography.