Thermodynamic Modulation of Dihydrogen Activation Through Rational Ligand Design in GeII–Ni0 Complexes

A family of chelating aryl-functionalized germylene ligands has been developed and employed in the synthesis of their corresponding 16-electron Ni0 complexes (PhiPDippGeAr·Ni·IPr; PhiPDipp = {[Ph2PCH2Si(iPr)2](Dipp)N}−; IPr = [{(H)CN(Dipp)}2C:]; Dipp = 2,6-iPr2C6H3). These complexes demonstrate the ability to cooperatively and reversibly activate dihydrogen at the germylene-nickel interface under mild conditions (1.5 atm H2, 298 K). We show that the thermodynamics of the dihydrogen activation process can be modulated by tuning the electronic nature of the germylene ligands, with an increase in the electron-withdrawing character displaying more exergonic ΔG298 values, as ascertained through NMR spectroscopic Van’t Hoff analyses for all systems. This is also shown to correlate with experimental 31P NMR and UV/vis absorption data as well as with computationally derived parameters such as Ge–Ni bond order and Ni/Ge NPA charge, giving a thorough understanding of the modulating effect of ligand design on this reversible, cooperative bond activation reaction. Finally, the utility of this modulation was demonstrated in the catalytic dehydrocoupling of phenylsilane, whereby systems that disfavor dihydrogen activation are more efficient catalysts, aligning with H2-elimination being the rate-limiting step. A density functional theory analysis supports cooperative activation of the Si–H moiety in PhSiH3.


■ INTRODUCTION
The binding, activation, and utilization of dihydrogen has been and remains a central research focus across numerous facets of chemistry, 1 centered perhaps most prominently in hydrogenation and dehydrogenation catalysis. 2Understanding such processes is increasingly more important, in discovering sustainable, efficient catalysts, and more broadly in realizing the effective use of H 2 as an energy carrier. 3Classical heterogeneous systems (e.g., Raney nickel, 4 Pd/C 5 ) remain useful tools for achieving organic reduction chemistry.Still, molecular systems benefit in allowing in situ monitoring and well-defined structure−reactivity relationships, 6 leading to fewer challenges in understanding and tuning, relative to heterogeneous systems.Molecular homogeneous catalysts for hydrogen activation are thus intensely investigated and historically dominated by precious heavier transition metals (TMs). 2 Due to a necessary shift toward utilizing Earthabundant metals for catalysis, 7 homogeneous catalytic systems featuring 3d TMs are becoming increasingly more explored. 7,8n order to tame adverse processes associated with these metals (e.g., single electron mechanisms), and indeed to open new reactive pathways for heavier TMs, noninnocent ligands have played a pivotal role (i.e., metal−ligand cooperativity, MLC). 9or 3d metals, MLC has allowed for the facile activation of a number of catalytically relevant small molecules at the ligand− metal interface, including H 2 , showing particular promise in (de)hydrogenation catalysis employing these Earth-abundant metals.9d,10 Among these, well-defined examples of energetically tunable and reversible dihydrogen activation are rare, which would otherwise allow for a greater depth of understanding of ligand effects on this fundamentally significant process.
Carbenes, which have gained broad attention as both key reactive intermediates and spectator ligands, 11 have also been explored as noninnocent ligands. 12Typically incorporated into a chelating ligand scaffold, noninnocent carbene ligands have allowed for the reversible activation of ammonia, 13 as well as the stoichiometric activation of H 2 , 13,14 in a small number of 3d TM complexes (Figure 1a−c).We note that H 2 activation via this mechanism is known for a considerably greater number of heavier, precious TM systems.12b,15 The utilization of heavier p-block element centers in assisting H 2 activation has gained significant attention in recent years, particularly in Lewis acidic group 13 moieties.10c, 16 The heavier tetrylenes have also garnered some degree of attention here: a handful of silylene complexes of 3d TMs have been shown to activate H 2 across the Si-TM bond, 17 while one example of a Ge II −Pt 0 system even demonstrates the reversible scission of H 2 (Figure 1d), 18 albeit with a precious TM.Such systems remain relatively unexplored.Still, tetrylenes are promising candidates as noninnocent ligands for several reasons: they are readily modifiable, perhaps even more so than classic carbenes, for example, through simple salt metathesis reactions. 19Through this, their electronic nature is readily tunable, as is their steric profile. 20The lessened propensity for sp-mixing on descending group 14 further allows tuning of their ambiphilic character, i.e., the HOMO−LUMO gap for a given system increases down the group. 20Our own work has demonstrated this, whereby tetrylenes can behave as Lewis acidic binding sites while coordinating Ni 0 through their lone electron pair. 21In those systems, no MLC is observed due to the saturated 18electron Ni center.We thus hypothesized that cooperative chemistry may be accessible on moving to a 16-electron Ni 0 system and further that the energetics of this cooperativity may be modulated through ready modification of the germylene ligand.Herein, we describe efforts in this regard toward a deep understanding of factors affecting the energetics of dihydrogen activation in 16-electron Ni 0 complexes bearing noninnocent germylene ligands.We find that electronic modulation of these ligands allows for the fine-tuning of the thermodynamics of H 2 activation, while also allowing for complete inhibition of the H 2 activation process.Experimental kinetic studies show a strong correlation with key NMR and UV/vis spectral data as well as with DFT-derived charge and bonding parameters, giving a clear understanding of ligand effects on H 2 activation.Further, the described 16-electron nickel(0) complexes are effective in catalytic phenylsilane dehydrocoupling, with the catalytic rate also correlating with the catalyst's ability to eliminate H 2 .
Journal of the American Chemical Society 2a−f), featuring strongly electron-withdrawing substituents (e.g., p-NC-Ph, m-(CF 2 ) 2 -Ph) and electron-donating substituents (e.g., p-Me-, p-MeO-, and p-Me 2 N-Ph), as well as the "standard" Ph system.In addition, the [Me 2 N] ligand 2g stands as a highly electron-rich germylene, expected to have a considerably stabilized LUMO through N → Ge π-donation.All systems are formed essentially quantitatively as ascertained by 31 P NMR spectroscopy but are highly soluble in aliphatic solvents, including pentane.This made the isolation of all systems somewhat challenging, although large-scale synthesis of 2b and 2c allowed for their isolation as analytically pure solids in 54 and 78% yields, respectively, and for the growth of X-ray quality crystals (Figure 2b,c). 22A high-field shift for the signal relating to the ligand's [Ph 2 P] moiety is observed in the 31 P{ 1 H} NMR spectra on moving from 2a to 2g, in line with increased electron density on moving from the p-NC-Ph moiety (δ = 5.9 ppm) to the strongly π-donating [Me 2 N] moiety (δ = −24.4ppm).This already indicates that the electronic nature at Ge II can be easily tuned through such ligand modifications.Reminiscent of our earlier reports, 21,23 addition of ligands 2a−2g to 1:1 mixtures of Ni(cod) 2 and IPr (IPr = [(H)CN(Dipp)] 2 C:) in toluene led to immediate formation of deep purple (2a), red (2b−f), or green (2g) solutions, indicative of complex formation (3a−3g, Figure 2a).
In all cases, 31 P NMR spectroscopic analysis of crude reaction mixtures suggested the formation of single reaction products, with resonances low-field shifted relative to the free ligands (Figure 3b).All complexes could be isolated as intensely colored crystalline solids in moderate yields of between 46 and 59%.Single-crystal X-ray diffraction analysis revealed all seven complexes to be the target 16-electron Ni 0 systems, featuring acyclic two-coordinate (aryl)(amido)-or bis(amido)-germylene ligands through the insertion of the nickel center into the Ge−P bond of the "free" ligands (Figures 4, and S129−S133 in SI). 24The Ni centers in these complexes bear a trigonal planar geometry, in all cases supported by an agostic C−H•••Ni interaction with one ligand Si-i Pr group, leading to a boat-conformation of the 6-membered core of these complexes (Figure S145 in Supporting Information).Indeed, such an interaction is maintained in solution, with one C(H)−CH 3 doublet shifted upfield, 25 with an increased shift aligning with increased Ge-Ar electron-withdrawing nature (Figures S58−S59 in SI), indicative of increased Ni → Ge back-donation.Through this, the Ge centers in all complexes deviate slightly from trigonal planarity, i.e. through slight pyramidalization.Again, this pyramidalization becomes more prominent with the electron-withdrawing nature of the Ge-Ar substituent (Table 1), and it is most likely caused by an increased Ni → Ge back-donation.This is further borne out by the 31 P NMR spectra of 3a−3g, in which a deshielding of the ligand's P-center is observed along the same series (Figures 3b  and S60 in SI), indicative of increased P → Ni donation.These points seem to have little effect on the experimental Ge−Ni bond lengths in these systems, which show no correlation with their aryl substituents.As one might expect, however, these are shorter than those in related N-heterocyclic germylene-Ni 0  complexes, 26 but longer than observed for the strongly electron-withdrawing [(F 5 C 6 ) 2 Ge] ligand. 27The modulation of the electronic nature of this bond across this series is nevertheless clearly observed upon comparison of their UV/vis spectra (Figure 3a).Here, a gradual but significant red shift is observed on moving from the most electron-rich ligand system (3g: λ = 623 nm) to the most electron-deficient system (2a: λ = 799 nm).This absorption relates to the HOMO−LUMO gap of these complexes, 28 late the frontier orbitals in 3a−3g, having a distinct effect on reactivity (vide infra).
Finally, assessing the electronic structure of the studied systems using Density Functional Theory (DFT) further bolsters the experimentally observed trends.First, the HOMO of all complexes is a distorted π-bond, showing significant Ge− Ni bonding character (e.g., Figure 3c).For all aryl systems, the LUMO represents the antibonding part of this orbital, that is, a distorted π*-orbital, with delocalization across the Ge-Ar substituent becoming more prominent with increasing electron-withdrawing character.This is not the case for bis(amido)germylene complex 3g, where the LUMO is largely ligand-centered, and the LUMO + 1 represents the abovedescribed Ge−Ni antibonding orbital.The calculated HOMO−LUMO gaps correlate well with those found experimentally, i.e., using UV/vis spectroscopy (Table 1), although generally underestimating this value.An NBO analysis of the Ge−Ni bonding interactions in 3a−3g (Tables S22−S28) shows only one σ-type bond between the Ge and Ni atoms, in addition to empty p-orbitals on the Ge centers with considerable partial occupation (∼0.5e).This NBO-derived bonding picture is consistent with calculated Mayer Bond Orders (MBOs) which also indicate a single bond with additional secondary interactions (MBO ∼ 1.1).Natural population analysis (NPA) shows a partial negative charge on the Ni center (−0.512 to −0.546) and a positive charge on the Ge center (+1.116 to +1.160).Throughout the series of ligands, we see a clear trend in the MBO of the Ni−Ge bond with the partial charge of the Ge and Ni atoms, which also correlates with measured UV/vis and H 2 activation parameters (vide infra).Among the series of 3a−3g, 3g provides a considerable difference in the Mayer bond order for the Ni− Ge bond (1.23) and partial charges (Ni: −0.59; Ge: + 1.31), indicating significant tunability of the Ge−Ni interaction through the described ligand design.
Reversible Dihydrogen Activation.With complexes 3a− g in hand, we aimed to investigate the effect of the described electronic modulation on the reactivity of these systems.The (amido)(aryl)germylene complexes 3a−3f readily activate dihydrogen under just 1.5 atm of pressure at ambient temperature, each leading to a single activation product as ascertained by 1 H and 31 P{ 1 H} NMR spectroscopy (4a−4f, Scheme 1). 29Evidence for cooperativity in these reactions came from variable-temperature (VT) NMR spectroscopic studies.Between −40 and −60 °C, a clear resonance for the Ge-H can be seen for all complexes between δ = 5.22 (3f) and 6.03 (3a) ppm.The corresponding Ni-H is observed as a broad doublet between δ = −9.76(3f) and −10.83 (3a) ppm.Both resonances integrate to 1H, and couple with each other as shown via 2D-COSY NMR experiments for 3b (Figure S72 in SI).Though these peaks are too broad to ascertain 3 J HH coupling constants, the signals observed for the Ni-H at low temperatures give 2 J PH coupling constants between 105 and 109 Hz, in keeping with known trans-phosphine nickel hydride complexes. 30As a whole, this demonstrates the activation of H 2 across the Ge−Ni bond, forming novel germyl-nickel species with a [HGeNiH] core.Confirming this, these signals are not observed on the addition of D 2 to 3c, while the remainder of the spectrum is unchanged.At ambient temperature no clear signals are observed for the described hydride ligands in the H 2 activation products, while at 60 °C all species exhibit a resonance at approximately δ = −2 ppm, integrating to 2H.These observations would suggest that a dynamic hydride exchange is in play in these complexes, as has been observed previously in the activation of silanes by Ni 0 complexes. 31This dynamic exchange is further demonstrated by the addition of D 2 to in situ generated hydride complex 4c, leading to the formation of H−D (Figure S83 in SI).Notably, degassing reaction mixtures of in situ generated H 2 activation complexes 4a−4f leads to quantitative regeneration of 3a−3f (e.g., Figure 5), indicative of complete reversibility in this H 2 activation process.Importantly, bis(amido)germylene complex 3g does not activate dihydrogen under the given conditions, or when heated under an atmosphere of H 2 , already indicating that electronic modulation of these complexes allows for tailored bond-activation capacities. 32he structure of the described complexes was confirmed through single-crystal X-ray diffraction analysis of 4a and 4c (Figure 6), which confirms the proposed 1,2-dihydride structure observed in low-temperature 1 H NMR spectra. 33iven the aforementioned reversibility of this H 2 activation reaction, crystals were grown under an atmosphere of H 2 .Attaining X-ray quality crystals of all species was thus not possible, but microcrystalline solids could be isolated for all systems except 3b, through precipitation of these complexes from their concentrated solutions at ambient temperature.This allowed the collection of ATR-IR spectra for these H 2 activation products, with Ge−H and Ni−H stretching frequencies observed between 1934 and 1949, and 1878 Thermodynamic Analysis of H 2 Activation.To gain further insights into the H 2 activation reaction, we used DFT to calculate the thermodynamics and reaction profile for complexes 3a−3g.The calculated reaction profile for 3c can be found in Figure 7, whereas the energetics of the same reaction steps for all synthesized complexes can be found in Table S29.We find that the initial coordination of an H 2 molecule at Ni is a barrierless process, forming an η 2 -complex (INT1, 16.1 kcal• mol −1 ).The activation of the H 2 bond occurs with the concurrent coordination of one of the H atoms to the Ge center (TS1, 24.7 kcal•mol −1 ) forming a square-pyramidal Ni-dihydride intermediate, the germylene ligand at the apex (INT3, 17.5 kcal•mol −1 ).The reaction concludes with the insertion of the germylene into one Ni−H bond (TS3, 23.8 kcal•mol −1 ) forming near thermoneutral product 4c (−0.3 kcal•mol −1 ).The activation barrier for a number of the reaction steps in this mechanism is rather similar for all ligands, i.e., within a few kcal•mol −1 .In addition to these differences being within the expected accuracy of DFT methods (∼5 kcal/mol), the formal rate-determining step may well change from ligand to ligand.Nevertheless, that calculated ΔG values are essentially thermoneutral for all arylgermylene systems is in line with the experimentally observed reversibility for this reaction.
A closer inspection of 31 P{ 1 H} NMR spectra for reactions of complexes 3a−3f with H 2 indicates that hydride complexes 4a−4f are not quantitatively generated under the given conditions (i.e., 1.5 atm of H 2 , 298 K).Further, the ratio of complexes 3 to complexes 4 varies on the basis of the Ge−Ar substituent.Generating a Hammett plot for these equilibria clearly demonstrates that the reaction becomes more favorable for increasingly electron-withdrawing substituents (e.g., K eq for 4b: 10,307 L•mol −1 ; for 4f: 449 L•mol −1 ; Figure 8a).A K eq value of 2129 L•mol −1 was found for the activation of D 2 by 4c (for H 2 : 1761 L•mol −1 ), giving an inverse kinetic isotope effect (KIE) of 0.83, as is known for the activation of H 2 /D 2 in TM systems. 34The thermodynamics of H 2 activation by 3a−3f could be further probed using VT 31 P{ 1 H} NMR spectroscopy (Figure 8a,b).Integration of relative concentrations of complexes 3 and 4 between 293 and 305 K in 2K intervals allowed for the extrapolation of thermodynamic data for all complexes through Van't Hoff analyses (Tables S2−S9 in SI).Reaction enthalpies become more favorable with a decreasing HOMO−LUMO gap, i.e., correlating with a red shift in experimentally observed UV/vis absorptions.That is, 4a shows the most exothermic reaction (λ = 799 nm; ΔH = −44.39± 1.14 kJ•mol −1 ), and 4e shows the least exothermic reaction (λ = 741 nm; ΔH = −33.09± 1.20 kJ•mol −1 ).We note that, although 4a has the largest ΔH value, this does not result in the largest K eq at 298 K, due to the much higher ΔS value of −77.6 J•mol −1 •K −1 for this system, leading to a greater influence of the temperature on K eq .We hypothesize that this entropic value may be due to the coordinative nature of the p-NC-Ph group at Ge in 3a/4a, which could allow for the formation of higher-order complexes for product 4a in   solution, further lowering its entropy.Nevertheless, ΔG 298 values also follow the trend described for enthalpic values. 35ncorporating DFT calculations, plots of ΔG 298 for H 2 activation by 3a−3f against related Ge and Ni NPA charges (Figure 9a), as well as against Mayer bond orders for Ge−Ni interactions (Figure 9b) all show a strong correlation.Here, a decrease in negativity at Ni, an increase in negativity at Ge, and a lower MBO all favor H 2 activation.These data can be used to estimate the ΔG 298 value for bis(amido)germylene complex 3g, which does not activate H 2 under the given conditions and would therefore be expected to have an endergonic value for this process.Using the linear regression equation in combination with the calculated NBO for 3g, a ΔG 298 value of 9.0 kJ•mol −1 is found, in line with experimental observations.Although this is lower than the corresponding calculated value for this system (i.e., 21.6 kJ•mol −1 ), the difference between the former ΔG 298 value (i.e., 9.0 kJ•mol −1 ) and that found experimentally for 3c (−18.7 kJ•mol −1 ; d ΔGexp = 27.7 kJ• mol −1 ) is similar to the difference in the related calculated values for the same systems (d ΔGcalc = 22.8 kJ•mol −1 ).This observation further reinforces the utility of DFT calculations in predictive ligand design.As a whole, these results demonstrate that electronic modulation of easily accessible germylene ligands can be utilized in tuning the thermodynamics of the reversible, cooperative H 2 activation reaction, employing an abundant 3d transition metal.Given the correlation between both experimental and calculated descriptors with thermodynamic values, this opens the door for utilizing high-throughput modeling methodologies (e.g., machine learning) to further define and discover such cooperative bond activation processes, rooted in computational ligand screening.
Silane Dehydrocoupling.Having a handle on tuning the thermodynamics of a fundamentally important process such as H 2 activation may have profound implications in catalysis, with ligand design being a cornerstone of catalyst development. 36s such, we sought to couple the observed reversible H 2 activation systems to a dehydrocoupling process.For this, we focused our efforts on the dehydrocoupling of phenyl silane, given that this relies on H 2 elimination, and is known for lowvalent nickel systems.
In an initial reaction, a slight excess of PhSiH 3 was added to 3c, which was seen to generate the dihydride complex 4c, [Ph(H) 2 Si] 2 , and H 2 (Scheme 2).From these reaction mixtures, analytically pure 4c precipitates from the solution and is readily isolated in up to 77% yield.We then sought to ascertain whether the initial activation of phenyl silane proceeds via a similar mechanism to dihydrogen activation, that is, involving both the Ni and Ge centers.The most feasible calculated mechanism for this process is given in Figure 10.We find that the initial coordination of the Si−H bond in PhSiH 3 occurs at the Ni center in a barrierless step (INT1, 13.9 kcal•mol −1 ).The scission of the Si−H linkage occurs with the help of the Ge center, through a Ge•••Si bridge (TS1, 30.5 kcal•mol −1 ), resulting in 5 with a [Ge(Si)-Ni(H)] core (14.1 kcal•mol −1 ), which is favored over the corresponding Ge-H intermediated (17.5 kcal•mol −1 ; Figure S153 in SI).This latter point is interesting, given that the Ge center is more electropositive than Ni and thus may be expected to be a stronger electrophile in this case.
Extending the described stoichiometric activation of phenyl silane, we found that (aryl)germylene-Ni 0 complexes 3a−3f are capable of catalyzing the dehydrocoupling of PhSiH 3 , akin to earlier reported Ni complexes (Table 2). 37Using a 2.5 mol % catalytic loading of 3c, full consumption of PhSiH 3 is observed after 48 h, with the visible generation of H 2 in reaction mixtures.Here, oligo-silanes can be observed in 1 H NMR spectra of reaction mixtures. 37GPC analysis of polymers isolated from these reactions yielded Mw values of between 795 and 1052 g•mol −1 , with average PDIs of 1.67 (Figures  S129−S131 in SI).A key point here is the observation of 4c in stoichiometric reaction mixtures, which would suggest that the loss of H 2 to regenerate 3c is rate-determining.It follows that more endergonic H 2 activation values should favor the dehydrocoupling reaction.That is, 3f should be the most active catalyst, and 3a the least.Comparing these systems demonstrated that this is indeed the case: dehydrocoupling of PhSiH 3 with 1 mol % of complexes 3a−f at ambient temperature for 3 h led to 48% conversion for 3b (ΔG 298 = −23.00kJ•mol −1 ), in contrast to 64% for 3f (ΔG 298 = −15.27kJ•mol −1 ).Although being a small data set, this gives early evidence that the electronic modulation in these complexes may allow for catalyst fine-tuning in further key synthetic processes.In this regard, we continue to investigate the stoichiometric silane activation using these and related complexes, to further elucidate key mechanistic aspects and allow for further catalyst improvements.

■ CONCLUSIONS
In summary, we present a novel family of (amido)(aryl)germylene ligands 2, which are employed in the synthesis of 16-electron Ni 0 complexes 3.All systems are capable of reversible activation of dihydrogen, leading to 1,2-dihydride complexes (4) featuring HGe-NiH cores.Van't Hoff analyses of equilibria formed upon addition of H 2 to 3 allow for the accurate extrapolation of thermodynamic data for this activation process, showing a strong correlation with the electronic nature of the germylene ligands: generally, more electron-withdrawing systems favor the H 2 activation reaction.This is further supported by computational investigations, allowing for the identification of calculable descriptors that correlate with experimental data in linear regression analyses.These observations could then be applied in a catalytic context: all Ge II −Ni 0 complexes demonstrate the ability to catalyze the dehydrocoupling of phenyl silane.As the ratelimiting step in this process is the loss of H 2 , we have shown that catalysts that disfavor H 2 activation lead to a more rapid dehydrocoupling reaction.Taken as a whole, this study further establishes tetrylenes as readily tunable noninnocent ligands, giving key insights into the effects of ligand modifications on modulating both stoichiometric and catalytic processes.Ligand modification has a predicable effect on catalyst performance, having exciting implications for the broader use of low-valent p-block ligands in catalytic systems.Taking this forward, we now aim to develop models for computationally exploring effective tetrylene-TM complexes as catalysts for hydrogencentric processes.

* sı Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.4c08297.Synthetic and characterization data for all new compounds, including spectra; details of kinetic and thermodynamic studies; and full details and references for the crystallographic and computational data (PDF)

■ AUTHOR INFORMATION
Corresponding Authors

Figure 1 .
Figure 1.(a−c) Known examples of cooperative H 2 activation with 3d metal carbene complexes; (d) reversible H 2 activation in a Ge−Pt complex; (e) this work.

Figure 7 .
Figure 7. DFT-derived mechanism for the cooperative activation of H 2 by 3c.

Figure 8 .
Figure 8. Analysis of the equilibrium for the H 2 activation in 3a−f; (a) Hammet plot of Log(K eq(Ar) /K eq(Ph) ) vs. Hammet parameters (σ); (b) Plot of experimental ΔG 298 values vs. 31 P NMR shift for the "free" germylene ligands 2a−2f; (c) A summary of experimental thermodynamic parameters for the H 2 activation reaction in 3a−3f; a determined by NMR spectroscopy, see Supporting Information for details; b in kJ•mol −1 ; c in J•K −1 •mol −1 .

Table 1 .
− Selected Experimental and Calculated Metrical Data for Compounds 3a−3g

Table 2 .
Catalytic Phenylsilane Dehydrocoupling Catalyzed by Complexes 3a−3g Determined by relative integration of Si-H peaks vs an internal standard (see the SI for details).b Values given are averaged from three runs.c This experiment was run for 48 h to ensure full consumption of PhSiH 3 . a CCDC 2312482−2312494 and 2363825 contain the supplementary crystallographic data for this paper.These data can be obtained free of charge via www.ccdc.cam.ac.uk/data_request/ cif, or by emailing data_request@ccdc.cam.ac.uk, or by contacting The Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, U.K.; fax: +44 1223 336033.