Zintl cluster supported low coordinate Rh(i) centers for catalytic H/D exchange between H2 and D2

Ligand exchange reactions of [Rh(COD){η4-Ge9(Hyp)3}] with L-type nucleophiles such as PMe3, PPh3, IMe4 (IMe4 = 1,3,4,5-tetramethylimidazol-2-ylidene) or [W(Cp)2H2] result in the displacement of the COD ligand to afford clusters with coordinatively unsaturated trigonal pyramidal rhodium(i) centers [Rh(L){η3-Ge9(Hyp)3}]. These species can be readily protonated allowing access to cationic rhodium–hydride complexes, e.g. [RhH(PPh3){η3-Ge9(Hyp)3}]+. These clusters act as catalysts in H/D exchange between H2 and D2 and alkene isomerisation, thereby illustrating that metal-functionalized Zintl clusters are active in both H–H and C–H bond activation processes. The mechanism of H/D exchange was probed using parahydrogen induced polarization experiments.


Introduction
Transition-metal/main-group element alloys (TMMGAs) are an interesting class of materials that combine late transition metals (e.g. nickel, platinum or palladium) with low melting point post-transition metals such as gallium, indium, tin, lead or bismuth. In their molten state, these alloys have been employed as catalysts in a number of challenging processes that involve light alkane valorization, e.g. methane pyrolysis to form graphitic carbon and hydrogen, dry reforming of methane, or propane dehydrogenation. [1][2][3][4][5][6][7] These melts offer access to "solvated" late transition metal atoms accounting for their impressive catalytic performance, however in situ monitoring of reaction proles/mechanisms remains a challenge due to the harsh conditions employed (e.g. operating temperatures of 1040 C) and the lack of suitable spectroscopic probes. This led us to turn our attention to the synthesis of molecular TMMGAs, compounds in which a late transition-metal-ligand fragment is supported on a main-group cluster platform. For this purpose, we have explored the use of Zintl clusters, anionic clusters of the main-group elements, that can be readily functionalized with transition-metals. In principle, these compounds can act as molecular models for TMMGAs. However, poor solubility in non-polar solvents, low yielding syntheses, and a tendency for oxidative coupling have long hindered their application in homogeneous catalysis.
The silylated Zintl ion [Ge 9 (Hyp) 3 ] À (Hyp ¼ Si(SiMe 3 ) 3 ) is an attractive platform for the design of molecular TMMGAs due to its increased solubility in hydrocarbon solvents. 8,9 Recently we demonstrated that the rhodium(I) cluster [Rh(COD){h 4 -Ge 9 (Hyp) 3 }] can be employed as a catalyst for the hydrogenation of cyclic alkenes. 10 This is the rst example of a Zintl cluster being used in catalysis, and a proof-of-concept study illustrating that this class of compound may be used to mimic the impressive reactivity of TMMGAs. Furthermore, the [Ge 9 (-Hyp) 3 ] À cage can readily isomerize to adapt to the steric and electronic requirements of the transition metal. 10 These molecular dynamics have the potential to play a pivotal role in catalysis, where geometric responses to substrate and product binding can provide lower energy pathways in elementary catalytic steps. Herein we demonstrate that, in addition to being active hydrogenation catalysts, metal-functionalized Zintl clusters are also active catalysts for H/D exchange between H 2 and D 2 , and alkene isomerization. While both of these catalytic bond transformations are common in organometallic chemistry they are unknown in Zintl ion chemistry. 11 monodentate L-type ligands to afford a coordinatively unsaturated rhodium(I) metal center that has a formal 16-valence electron count. Consequently, the addition of one equivalent of PMe 3 to [Rh(COD){h 4 -Ge 9 (Hyp) 3 }] led to the reaction shown in Scheme 1. The two hypersilyl environments initially present in the 1 H NMR spectrum of [Rh(COD){h 4 -Ge 9 (Hyp) 3 }] disappeared giving rise to a single resonance (0.50 ppm). In the 31 P{ 1 H} NMR spectrum a doublet resonance was observed at À13.5 ppm, ( 1 J P-Rh ¼ 210. 3 Hz). This new species, [Rh(PMe 3 ){h 3 -Ge 9 (Hyp) 3 }] (1a), and free COD are the only two products observable by 1 H NMR spectroscopy. The single hypersilyl resonance observed at room temperature points to a uxional process that makes these groups equivalent on the NMR timescale, as also noted for [Rh(dppe){h 5 -Ge 9 (Hyp) 3 }]. 10 The related clusters [Rh(PPh 3 ){h 3 -Ge 9 (Hyp) 3 6 ]. [16][17][18] Crystallization of 1a, 1c and 1d from saturated n-hexane or npentane solutions allowed for structural characterization of the new clusters by single-crystal X-ray diffraction. All three structures exhibit coordinatively unsaturated rhodium(I) centers, bound to the [Ge 9 (Hyp) 3 ] À cluster in an h 3 mode (Fig. 1), in which the ligand (e.g. PMe 3 ) sits trans to a silylated vertex. This gives the cluster overall, non-crystallographic, C s symmetry. The s 4 values for 1a, 1c and 1d are 0.79, 0.83 and 0.85, respectively, ‡ in line with the value for an ideal C 3v coordination geometry (0.85). 19 Despite there being an apparent vacant coordination site, no structural or spectroscopic evidence was observed for an interaction between the most proximal hypersilyl substituent with the metal center (e.g. the closest Rh/C interatomic distances are greater than 3.65Å). 20 The structures of 1a and 1d each feature a single crystallographically unique cluster in the asymmetric unit which exhibits positional disorder. This disorder is best accounted for by two different orientations of the {RhL} fragment that are related by rotation with respect to the (static) [Ge 9 (Hyp) 3 ] À cluster. By contrast, the structure of 1c reveals two crystallographically unique clusters in the asymmetric unit with near identical bond metrics. The rhodium centers in all the complexes bind to the cluster through three short Rh-Ge bonds (1a: 2.394(1)-2.433(1)Å; 1c: 2.397(1)-2.418(1)Å; 1d: 2.409(2)-2.429(1)Å), and a slightly longer Rh/ Ge contact with the nearest silylated germanium atom (1a: 2.517(7)Å; 1c: 2.667(1)Å; 1d: 2.644(2)Å). The Rh-L distances for 1a and 1c are similar to those reported in related rhodiumÀphosphine, 21 and rhodium-carbene complexes. 22 In the case of 1d, the Rh/W distance of 2.852(1)Å is notably longer than that in [Rh(PPh 3 ) 2 (m-H) 2 W(Cp) 2 ][PF 6 ] (2.721(3)Å), 16 which we hypothesize is a consequence of the steric demands of the hypersilyl substituents of the [Ge 9 (Hyp 3 )] À cage. From an electron-counting perspective, 1a-1d can be viewed as hypercloso-like, and therefore similar to [Rh(CO) 3 (B 9 H 9 )] + . 23 The effect of displacing a four-electron donor ligand (COD) by a two electron donor (e.g. PMe 3 ) is to reduce the overall cluster electron count, and this is consequently accompanied by a structural distortion. The reverse effect has previously been seen when hypercloso-[Ru(PEt 3 ) 2 (C 2 B 7 H 9 )] converts to closo-[Ru(PEt 3 ) 3 (C 2 B 7 H 9 )] on addition of PEt 3 . 24,25 Signicant rhodium d-orbital participation in cluster bonding can be observed for 1b on inspection of frontier orbitals (see ESI †).
Based on these crystallographically-determined structures, one would expect two chemically inequivalent hypersilyl  substituents to be observed in the solution phase 1 H NMR spectra of 1a-1d. The fact that only one 1 H resonance is observed suggests that, in solution, a uxional process is operating that makes the hypersilyl groups equivalent. This process was modelled using density functional theory (DFT) calculations for 1b, and shown to proceed via a C s to C 3v transition ( Fig. 2) whereby the rhodium center adopts a tetrahedral coordination geometry and coordinates to a triangular face of the [Ge 9 (Hyp) 3 ] À cage. This process involves cleavage of the Rh1/Ge7 contact (as per the numbering scheme in Fig. 1) and the contraction of the distance between Ge3 and Ge9 by ca. 0.7 A. Consistent with the observed uxionality, that is not frozen out at À80 C, the optimized geometries for the C s and C 3v symmetric clusters were found to be within 2.4 kcal mol À1 of one another, with a computed transition state barrier of 4.4 kcal mol À1 . The positional disorder observed in the singlecrystal X-ray diffraction structures of 1a and 1d is consistent with the superposition of two of these C s symmetric isomers in the lattice.
Trigonal pyramidal rhodium(I) compounds are rare, and, in some cases, have been shown to be weakly electrophilic. [26][27][28] For example, Grützmacher has shown that [Rh(trop 2 SiMe)(C 2 H 4 )] (trop 2 SiMe ¼ bis(5H-dibenzo[a,d]cyclohepten-5-yl)methylsilane) will reversibly bind THF to afford a trigonal bipyramidal compound. However, this reactivity is nely balanced, as other nucleophiles, such as PPh 3 , will instead displace the coordinated alkene in the equatorial site rather than bind in the axial position. 26 Reaction of 1a-1d with alkenes indicate that these clusters are not particularly strong electrophiles, as no adduct is observed to be formed to the detection limit of 1 H NMR spectroscopy. Indeed, inspection of the computed frontier orbitals reveals an energetically accessible HOMO (with a signicant rhodium d orbital contribution) suggesting that these complexes are more nucleophilic in character. Thus their reactivity is predicted to mimic that of more traditional organometallic nucleophiles such as Cp*Rh(PMe 3 ) 2 . 29 We probed this chemically by reaction of 1b and 1d with Brookhart's acid [H( Reaction of 1b with 1 equivalent of this acid converts the rhodium(I) center to its conjugate rhodium(III) hydride, [RhH(PPh 3 ) {h 3 -Ge 9 (Hyp) 3 Crystals of 2b[BAr F 4 ]$1.5hex (Fig. 3) and 2d[BAr F 4 ]$Et 2 O suitable for single crystal X-ray diffraction were grown from concentrated n-hexane and diethyl ether solutions, respectively. The cationic cluster 2d exhibits positional disorder (two different cluster orientations related by rotation in a 2 : 8 ratio), consequently only the bond metric data for 2b[BAr F 4 ]$1.5hex will be discussed in detail. The crystal structure reveals a single crystallographically unique cationic cluster in the unit cell, [RhH(PPh 3 ){h 3 -Ge 9 (Hyp) 3 }] + , accompanied by a [BAr F 4 ] À counterion and solvent of crystallization. The cluster core of 2b is comparable to those of 1a and 1c, with the notable exception that the rhodium metal center now adopts a trigonal bipyramidal geometry with a (located) hydride in an axial position, fully consistent with protonation of the HOMO in 1b. On protonation, the formal oxidation state of the rhodium center changes from +1 to +3. This is consistent with the moderate contraction (0.03Å) of the Rh-Ge distances, 2.365(1) to 2.397(1) A that is observed, relative to those observed for 1a and 1c (cf. 2.397(1) to 2.418(1)Å for 1c). As with 1a and 1c, a close Rh/Ge contact to the nearest silylated germanium center is seen (2.677(1)Å).
Having probed the electronic structure of 1b and 1d through protonation studies, we turned our attention to see if they might be viable catalysts. Given the presence of an apparent vacant coordination site, we chose to explore the reactivity of these compounds towards dihydrogen activation, a ubiquitous process in organometallic chemistry that oen occurs at Rh(I) metal centers. 11   16 The reversible nature of H/D exchange at 1d prompted us to re-evaluate the reaction of 1b with H 2 . We reasoned that perhaps H-H bond activation does indeed take place when the cluster is placed under an H 2 atmosphere, but that this process is endergonic, rapid and reversible. In order to probe this, 1b was placed under a mixture of H 2 and D 2 . The immediate formation of HD (dissolved) was observed on time of mixing, indicating fast H 2 and D 2 activation on the NMR timescale, with 1b acting a catalyst for H/D exchange. This reaction was unaffected by addition of mercury to the reaction mixture, suggesting a homogeneous process. 33 To further investigate the H/D exchange process, parahydrogen (p-H 2 ) induced polarization experiments were conducted by reacting 1b and 1d with p-H 2 . Due to the hyperpolarization effects arising from a bias in spin relaxation pathways, such experiments can be used to observe short-lived and low concentration intermediates in reactions involving p-H 2 . [34][35][36][37][38] While these experiments did not allow us to directly observe any rhodiumÀhydride intermediates, two important observations were made allowing us to infer that rapid and reversible H-H bond activation takes place on reaction of 1b or 1d with H 2 . The rst of these is that when solutions of 1b or 1d are exposed to NMR silent p-H 2 (3 bar) in an airtight NMR tube, instant formation of NMR detectable ortho-dihydrogen (o-H 2 ) is observed. The second important observation is that when a dilute solution of 1b (0.0005 mmol ml À1 ) was cooled to 263 K in the NMR spectrometer, removed from the magnet and shaken under p-H 2 (3 bar), on returning to the NMR spectrometer a signal with Partially Negative Lineshape (PNL) was observed at 4.56 ppm. This enhancement has a lifetime of around 5 s at this temperature, aer which, only o-H 2 was observed. The observation of a PNL effect at 263 K, whist no hydride species are seen, suggests a very rapid and reversible transfer of p-H 2 onto the cluster (most likely the rhodium center) in which the two spins become distinct. This would be consistent with the endergonic formation of either a dihydrogen, Rh(H 2 ), or a dihydride, Rh(H) 2 , motif. 39,40 Both are common intermediates in hydrogenation reactions mediated by organometallic complexes. 11 At room temperature p-H 2 destruction to form o-H 2 proceeds so rapidly, by reversible formation of a dihydride complex, that the PNL effect is quenched.
What remains undetermined at this stage is the mechanism of H/D exchange by 1b. Oxidative addition of H 2 /D 2 at the rhodium(I) center is supported by the studies using p-H 2 . As this would result in an 18-valence electron rhodium complex, association of a second molecule of H 2 /D 2 would necessitate a change in the hapticity of the [Ge 9 (Hyp) 3 ] À cage (for example from h 3 to h 1 ), 41 or phosphine dissociation. The resulting dihydride/s-dihydrogen compound could undergo H/D exchange via a s-CAM type mechanism. 42,43 DFT calculations, however, show that cluster isomerization from h 3 to h 1 is unfavorable (42.4 kcal mol À1 ), as is phosphine dissociation (32.8 kcal mol À1 ). An alternative is a proton-catalyzed mechanism, as previously invoked by Brookhart and co-workers for the oxidative-addition of H 2 by the d 8 iridium(I) complex [Ir(PONOP)(CH 3 )] (PONOP ¼ 2,6-bis(di-tert-butylphosphinito) pyridine). 44 Here H/D exchange would occur by initial reversible protonation of 1b by adventitious water to form trace amounts of formally 16-electron 2b [OH], which would then undergo H/D exchange. To explore such reactivity, 2b[BAr F 4 ] was exposed to a H 2 /D 2 mixture. While these studies were hampered by the low solubility of 2b[BAr F 4 ] and its propensity to precipitate out of solution, they showed that C 6 D 6 solutions of 2b[BAr F 4 ] generate HD when exposed to a mixture of H 2 and D 2 . Moreover, such H/ D scrambling occurs on a similar timescale to that observed when using 1b (i.e. approx. 15 minutes). Thus, it is possible that adventitious moisture in the solvent or gas mixture could give rise to trace amounts of 2b[BAr F 4 ] on dissolving 1b and that this species is, in fact, the active catalyst. Arguing against this hypothesis is that addition of proton sponge to 1b H 2 /D 2 mixture did not suppresses H/D exchange, as might be anticipated for a proton-transfer mechanism. However, we cannot discount the formation of trace [OH] À under these conditions that may act to deprotonate intermediate dihydrogen or hydride complex to form 2b. 45 While the precise details of the mechanism remain to be resolved, clear is that rapid H/D exchange does occur, a rst for a Zintl cluster.
The oxidative addition of H 2 to 1b, and the ability to subsequently bind an additional ligand, is further demonstrated by isotope scrambling experiments when 1b is used to catalyze the deuteration of 1-hexene. Addition of 1-hexene to 1b (2 mol%) under a D 2 atmosphere (1 bar, 16 hours) resulted in the formation of a mixture of the deuterated alkenes d n -1-hexene (major) and d n -2-hexene (minor), as well as d n -hexane, as measured by 1 H and 2 H NMR spectroscopy. The d n -1-hexene has deuterium incorporated into both geminal positions ($75% D total) as well as the vicinal position ($90% D) of the alkene. HD (dissolved) is also observed as pictured in Scheme 4A. Recharging with D 2 results in only d n -hexane being observed aer a further 16 hours, in which deuterium has been incorporated into the 1-, 2-and 3-positions. These observations suggest the formation of di-deuteride intermediate (consistent with both H 2 /D 2 exchange and p-H 2 experiments), followed by reversible coordination of 1-hexene, and reversible insertion into either of the alkene positions of 1-hexene, followed by a rate-determining reductive elimination of hexane, Scheme 4B. The observation of 2-hexene and d-incorporation into positions 1, 2 and 3 of the nal product, hexane, indicates a slower isomerization process also occurs, likely via non-degenerative belimination from a 2 alkyl-hydride intermediate. As for H 2 /D 2 exchange we cannot discount that catalysis occurs by a protoncatalyzed mechanism, via an (undetected) analogue of monohydride 2b[BAr F 4 ].

Conclusions
To conclude, we have shown that the cluster [Rh(COD) {Ge 9 (Hyp) 3 }] can be modied resulting in coordinatively unsaturated rhodium(I) containing Zintl clusters. Protonation results in a Rh-hydride, while rapid, catalytic H/D exchange occurs between H 2 and D 2 in the presence of such clusters. The organometallic chemistry of rhodium-phosphine-hydrides is well establishedparticularly with supporting cyclopentadienyl ligandsand has been critical in the development of our collective understanding of important concepts in the eld, such as structure/property relationships, reactivity patterns and catalysis. 46 However, the "inorganometallic" chemistry of analogous species with Zintl ions as supporting ligands is essentially unexplored. 47,48 This contribution shows that molecular TMMGAs offer similar rich structural and bond activation chemistry that suggests further study into their uses in catalysis is warranted.

Synthetic procedures
Synthesis of [Rh(PMe 3 ){h 3 -Ge 9 (Hyp) 3 }] (1a). [Rh(COD) {Ge 9 (Hyp) 3 }] (100 mg, 0.062 mmol) was dissolved in toluene (20 ml) and cooled to À77 C. PMe 3 in toluene (0.031 M, 2 ml, 0.062 mmol) was added dropwise over the course of 10 minutes. The brown solution was stirred for 90 minutes before warming to room temperature and stirring for 4 hours. The toluene was removed in vacuo and the product was extracted in n-hexane (10 ml) as a brown solution. The product was dried under a dynamic vacuum and lyophilized from benzene (4 ml). Yield: 91 mg, 93%. CAUTION: ne dry powders of 1a ignite spontaneously in air. Black-brown crystals suitable for X-ray crystallography were obtained by dissolving 50 mg of 1a in 0.5 ml of nhexane and cooling to À40 C. Elemental analysis calcd for C 30  Synthesis of [Rh(PPh 3 ){h 3 -Ge 9 (Hyp) 3 }] (1b). Toluene (20 ml) was added to a mixture of [Rh(COD){Ge 9 (Hyp) 3 }] (250 mg, 0.16 mmol) and PPh 3 (41 mg, 0.16 mmol) at room temperature. The brown solution was stirred for 3 hours before heating to 70 C for 4 days. The toluene was removed in vacuo and the product was extracted in n-pentane (10 ml) as a brown solution. The product was dried under a dynamic vacuum and lyophilized from benzene (3 ml). The resulting ne brown powder was puried by sublimation of any excess (typically $5%) PPh 3 at 85 C over 12 hours. This was then again extracted with npentane (5 ml) and lyophilized from benzene (3 ml) to yield 1b as a brown powder (240 mg, 84%). CAUTION: ne dry powders of 1b ignite spontaneously in air. Black-brown crystals suitable for X-ray crystallography were obtained by dissolving 200 mg of the powder in 0.5 ml of n-hexane and cooling to À40 C. Elemental analysis calcd for C 45 3 }] (100 mg, 0.062 mmol) was dissolved in n-pentane (5 ml) and added to a suspension of IMe 4 (8 mg, 0.062 mmol) in n-pentane (5 ml) at room temperature. The brown suspension was stirred for 16 hours before ltering. The solution was concentrated to $1 ml and cooled to À40 C producing black crystals of compound 1c (60 mg, 58%) which were suitable for single crystal X-ray diffraction. Elemental analysis calcd for C 34 29 Si NMR (C 6 D 6 , 298 K, 99.32 MHz): d (ppm) À97.7, À9.7.
Synthesis of [Rh{h 3 -Ge 9 (Hyp) 3 }(m-H) 2 W(Cp) 2 ] (1d). [Rh(COD) {Ge 9 (Hyp) 3 }] (300 mg, 0.19 mmol) was dissolved in n-hexane (20 ml) and cooled to À78 C. (Cp) 2 WH 2 (60 mg, 0.19 mmol) was dispersed in n-hexane (10 ml). The stirring (Cp) 2 WH 2 mixture was then added dropwise to the solution of [Ge 9 (Hyp) 3 ]Rh(COD) over the course of 20 minutes. This was allowed to stir and warm to room temperature for 18 hours. The solution was then reduced in volume to $10 ml in vacuo and stirred at room temperature under a static vacuum for 4 days, producing a brown suspension. Upon standing for 3 hours, the mixture was ltered, affording 1d as a brown powder. The product was dried in vacuo and lyophilized from benzene (5 ml) to yield a ne brown powder of 1d (240 mg, 71%). CAUTION: powders of 1d are highly pyrophoric. Black crystals suitable of X-ray crystallography were obtained by dissolving 30 mg in a 1 : 10 mixture of toluene and n-pentane ($0.5 ml total volume) and cooling to À40 C. Attempts to heat the reaction to above 45 C