Metal‐only Lewis Pairs of Rhodium with s, p and d‐Block Metals

Abstract Metal‐only Lewis pairs (MOLPs) in which the two metal fragments are solely connected by a dative M→M bond represent privileged architectures to acquire fundamental understanding of bimetallic bonding. This has important implications in many catalytic processes or supramolecular systems that rely on synergistic effects between two metals. However, a systematic experimental/computational approach on a well‐defined class of compounds is lacking. Here we report a family of MOLPs constructed around the RhI precursor [(η 5‐C5Me5)Rh(PMe3)2] (1) with a series of s, p and d‐block metals, mostly from the main group elements, and investigate their bonding by computational means. Among the new MOLPs, we have structurally characterized those formed by dative bonding between 1 and MgMeBr, AlMe3, GeCl2, SnCl2, ZnMe2 and Zn(C6F5)2, as well as spectroscopically identified the ones resulting from coordination to MBArF (M=Na, Li; BArF −=[B(C6H2‐3,5‐(CF3)2)4]−) and CuCl. Some of these compounds represent unique examples of bimetallic structures, such as the first unambiguous cases of Rh→Mg dative bonding or base‐free rhodium bound germylene and stannylene species. Multinuclear NMR spectroscopy, including 103Rh NMR, is used to probe the formation of Rh→M bonds. A comprehensive theoretical analysis of those provides clear trends. As anticipated, greater bond covalency is found for the more electronegative acids, whereas ionic character dominates for the least electronegative nuclei, though some degree of electron sharing is identified in all cases.

Af ascinating class of metal-metal bondedc omplexes that is receiving growing attentiona re those with M!Md ative bonds, also referred as metal-only Lewis pairs (MOLPs). [7] Althoughnoticede arlier, [8] the first authoritative report on such a species dates backt o1 967, when Nowell and Russelle lucidated the solid-states tructure of [(h 5 -C 5 H 5 )(CO) 2 Co!HgCl 2 ]. [9] Numerous studies based on aw ide variety of transition metals were later disclosed, particularly during the last decade. [10] Apart from the fundamentala ppeal of these species, the interest on their study is at the heart of transition metal reactivity. The basicity of at ransition metal site is importantf or small molecule coordination (e.g. borane binding in borylation processes), [11] as well as during oxidativea ddition reactions. In turn, the latter are elementary steps presenti nm ost catalytic cycles, asn oticedf rom early reports. [12] Thus, ab etter understanding of transition metal basicity (i.e. through the examination of metal-only Lewisp airs) [13] may provide important information to be assimilated by bond activation and catalysis research.
In addition, bimetallic dative bonding has implications in many catalytic processes that involvet he participation of two metal fragments of contrasting electronic nature. For instance, as eries of studies on Pd-catalyzed Negishi and Sonogashira cross-coupling reactions revealed the impact on catalytic performance of bimetallic Lewis acid-base interactions between an electron rich Pd II centera nd acidic Zn II or Cu I fragments. [14] Unsupported MOLP compounds have also provedc ompetent in the activation of av ariety of EÀHb onds (E = H, X, N, O) in which their individual monometallic constituents revealed themselves inactive. [15] The incorporation of acidic metalso r metalloids as s-acceptors Z-type ligands in MOLP-type structures permitss tructural and electronic modulation of the basic metal site, [16] whereas the strengtho ft he M!Md ative bonding in thermally induced [17] metal-only frustrated Lewis pairs deeply impactst he reactivity and catalytic performance of the latter systems. [18] In addition, metal-to-metal dative bonding has important implications in supramolecular and molecular engineering, [19] as well as in host-guest chemistry. [20] With all this in mind, it becomes obvious that ad eep understanding of the nature of metal-to-metal bond in these molecular compounds and supramolecular aggregationsw ill have an important impact in ar ange of areas. In fact, this has been a matter of intense debate, which is not surprising considering the set of bonding components that may be involved (i.e. ionic, covalent, dative,dispersion…). As such, unsupported systems in which the bond between the two metals is the sole force holdingt he two fragments together constitute ideal motifs to study,s ince other factors that may obscure bonding analysisa re excluded. In their originalr eport,N owell and Russell postulated that [(h 5 -C 5 H 5 )(CO) 2 Co!HgCl 2 ]c ould be considered am etallicL ewis acid-base adduct, [9] as lately proposed for many other systems, [18a, 21] including those based on d 8 -d 10 interactions (referred to the last filled subshell of the bonding metals). [22] An alternative description proposed by Pyykkç implies dispersion forces as the main component of the bimetallic bonding. [23] However,m ore recentc omputational work speaks in favor of the former assumption,r evealing that dispersion forces contribute to al esser extenti nt hese type of systemsc ompared to the role of electrostatic and orbital interactions. [24] Most studies have either focused on the synthesis and structural characterization of ag roup of severalM OLPso ro nt he computational analysiso fp reviously reported bimetallic architectures of this kind. However,amore comprehensive and combined experimental/computational approach on af amily of unsupported MOLPs is lacking. With this aim, we have selected the electron rich Rh I compound [(h 5 -C 5 Me 5 )Rh(PMe 3 ) 2 ] [25] (1)a saLewis base to investigate av ariety of MOLPs generated by its combinationw ith well-known metallica nd metalloid Lewis acids (Figure 1). We provide not only the spectroscopic (including 103 Rh NMR) and structuralc haracterization of these uncommon compounds, but also ac omputational analysis of their Rh I !Mbonding.

Results and Discussion
The precise choice of 1 as the Lewis base to design MOLPs was made on the basis of several features:( i) the basic behavior of 1 has already been well established; [25] (ii)PMe 3 ligands enhance the nucleophilicity [13a] of the Rh I site compared to its more widely exploredc arbonyl analogue [(h 5 -C 5 Me 5 )Rh(CO) 2 ]; [26] (iii)the robustnesso f( h 5 -C 5 Me 5 )l igand prevents undesired reactivity recorded for its unsubstituted (h 5 -C 5 H 5 )a nalogue; [27] (iv) as an eutral Lewis base, itsc ombination with neutrala cids will minimize the ionic and electrostatic components of the Rh I !Mb ond;( v) as ap entacoordinated 18-electron species, insertionr eactions into polar bonds of the Lewis acid, or the formation of intermediate alkyl or hydride bridging species [28] that would cloud analysiso ft he Rh I !M bond, will be less favored;a nd (vi) 103 Rh is NMR active( I = 1/2, 100 %a bundant). With all this in mind, we have combined 1 with av ariety of main group metal precursors as Lewis acids. With the exception of CuCl, we avoided the extensive use of transition metal electrophiles to circumvent more complex bondingp ictures on grounds of their available d orbitals.

Synthesis of Rh I MOLPs with s-Block Acids
The number of compounds exhibitingm etalophilic interactions between transition and alkali metalsi sa bundant. [29] Systems that showi dentical or even reduced MÀMb ond lengths compared to the sum of their corresponding covalent atomic radii [30] presumably present some degree of bond covalency. Althought his is relativelyc ommon in the case of lithium, [31] examplesofi ts heavier congener sodium are less profuse. [32] Considering rhodium, the weaki nteraction of square planar [RhCl 4 ] 3À with an aked Na + cation has been analyzed by computational meansa st he result of orbitalo verlapping. [  as hort RhÀNa bond length of 3.105(2) , [33] only slightly elongated with respect to the sum of their covalent radii (3.08 ). [30] As anticipated, support for covalent bonding was inferred from theoretical studies. It is important to remark that this type of Lewis acid-base interaction with alkali metalsm ay promote interconversion between structural conformations in transition metal complexes, [34] in turn ap owerful tool for designing molecular machines. [35] We decided to explore the possibility of accessing unsupported MOLPs containing lithium and sodium cations. To prevent artificial elongation of the Rh!Mb ond due to steric repulsion, [32b] we focused on lithium and sodium salts of the low-coordinating tetrakis (3,5- , whosee xistence is furthers upported by computational means (vide infra), though weak h 5 -coordination to the empty face of the Cp* ligand cannotb eruled out. At this stage,w ed efer ad efinitive proposal due to the lack of structural data. All our attempts to grow single crystals of these speciesw ere unsuccessful. We recovered in all cases either crystalline M[BAr F ]( M= Li, Na), whichm ay illustrate the weakness of the Rh!Li/Na interaction, or observed the formation of the corresponding Rh III hydride [(h 5 -C 5 Me 5 )Rh(PMe 3 ) 2 H] [BAr F ] [24] (2), the latter formed due to the presence of adventitious water.A nalogous cooperative reactivity has been reported for other MOLPs based on [Pt(PtBu 3 ) 2 ]. [14a,b] For further validation,c ompound 2 could be independently synthesized by addition of equimolar amountso fa mmonium salts to 1 and it has been utilized as ab enchmark species to investigate the bonding.
As noted earlier we aimed to access MOLPs by combining neutralf ragments, aside from the prior Li + and Na + exceptions, to reduce the electrostatic component of the metal-tometal bond. Reaction of 1 with two equivalents of the Grignardr eagent MgMeBrr eadily yielded an ew species 3 (Scheme 1b)c haracterized by as harpd ecrease of the 1 J PRh couplingc onstant to 172 Hz, along with shifts of the 31 P{ 1 H} (d = À10.2 ppm) and pentamethylcyclopentadienyl 1 H( d = 1.87 ppm) NMR signals towards lower frequencies. Despite the high instability of 3,s ingle crystals suitable for X-ray diffraction studies were grown from diluted benzene solutionsa nd revealed the dimerics tructure [(h 5 -C 5 Me 5 )(PMe 3 ) 2 Rh! Mg(Me x Br 1Àx )(m-Br)] 2 ( Figure 2) in which the methyl group boundt om agnesium is mostly exchanged by ab romide nucleus [36] (Me:Br with 15:85 occupancies). Using an equimolar amount of the Grignard reagent did not provide full conversion of 1,w hereast he addition of MgBr 2 or MgMe 2 to access a MOLP without substitutional disorder provedu nsuccessful, partly due to solubility issues.
As expected, MOLP 3 adopts ap iano-stoolc onformation after coordination of the Lewis acid. The RhÀMg bond length accountsf or 2.651(3) ,s hortened by ca. 0.2 with respect to the sum of the covalentr adii (2.83 ), [30] thus indicative of bond covalency (vide infra). Twoo ther parameters, namely d rel [7] (0.94) and fsr (formals hortness ratio) [37] (1.01) ( Table 1), defined as the ratio between the MÀMb ond distance and the sum of either the covalentr adiio rt he metallic radii, respectively,u nderpin this assumption. The most relevant geometric parameters for the X-ray diffraction structures reported in this work are depicted in Table 1. It is worth of note that this exotic structure is the first unambiguous example of an unsupported RhÀMg bond, since the only prior related example contains a metal hydride that exhibitss ome degree of bridging character. [38] Moreover,despite the extensive use of Grignardreagents in organometallicc hemistry,i ti ss urprising that compound 3 seemst ob et he only Mg-based MOLP comprised of neutral fragments. [39] As stated above, the choice of rhodiuma st he Lewis base was in part made attending to its NMR activity (Table 2). To observe chemical shifts associated to 103 Rh centers we employed ac ross polarization approach by means of HMQC experiments  [a] S(r cov ) = sum of the covalent radiio fthe bonded metals. [30] [b] d rel = ratio between d Rh-M and the sum of covalent radii.
[ c] fsr = formal shortness ratio = ratio between d Rh-M and the sumo fm etallic radii. [37] [d] d Rh-P = averageR h ÀPb ond length. [e] d Rh-Cp* = distance between Rh and the centroid of C 5 Me 5 .
through its coupling to 31 Pn uclei (see Experimental Section for details). Considering its low sensitivity andr ather wide chemical shift range (ca. 12 000 ppm), [40] this strategy enormously facilitates the acquisition of 103 Rh NMR data. The new MOLPsa re characterized by 103 Rh{ 1 H} NMRr esonances shiftedt ol ower frequencies comparedt op recursor 1 (À9165 ppm), with 3 exhibiting as ignal at À9404 ppm and the products derived from the addition of alkali metals resonating at around À9262 ppm ( Figure 3).

Synthesis of Rh I MOLPs with p-Block Acids
Movingt ot he p-block we examined the reactivity of 1 with widely used metalloid precursors of the group 13 and 14, more precisely GeCl 2 ·dioxane, SnCl 2 ,G aCl 3 ,A lCl 3 and AlMe 3 .
Whereas tricoordinated group 14 species has been widelye xploiteda sL ewis acids, heaviert etrylenes (i.e. :GeCl 2 ,: SnCl 2 )e xhibit ambiphilic behavior due to the joint presence of al one electron pair and an empty p orbital. We thought of interest to access both types of MOLPst ol ater provide ac omparison of the bonding scheme between each other.R eaction of 1 with either GaCl 3 or AlCl 3 resulted in the precipitation of ah ighly insoluble materialo rt he formation of intractable mixtures, respectively.T he latter is not surprising considering previously reported difficulties to accessR h-alane MOLPs by direct combination of the two metal fragments. [41] However,a ddition of one equivalent of AlMe 3 (toluene solution,1m)t oab enzene Figure 2. ORTEP diagram of compounds 1·SnCl 2 , 1·AlMe 3 , 1·Zn(C 6 F 5 ) 2 , 1·GeCl 2 ,1·ZnMe 2 and 3;for the sake of clarity hydrogen atoms and solventm olecules are excluded, while thermal ellipsoidsare set at 50 %p robability.  solution of 1 resulted in clean formation of the corresponding 1·AlMe 3 MOLP.T he same occurs by adding GeCl 2 ·dioxane or SnCl 2 to bromobenzene solutions of the rhodium precursor to yield 1·GeCl 2 and 1·SnCl 2 ,r espectively,t hought he former required three hours forc ompletion while the tin MOLP formed immediately.I nt he case of germanium, two equivalents of GeCl 2 ·dioxane were required to achieve full consumption of 1, presumably because the second germanium may facilitated ioxane withdrawal from the coordinatingG eCl 2 terminus (Scheme 2). Multinuclear NMR spectroscopic analysis illustrates the formation of the new MOLPse xhibiting the same distinctive features commented above (Table2), that is, am arked decrease of the 1 J PRh coupling constanto fc a. 40 Hz and ad isplacement to lower frequencies of the 1 HNMR signal due to the pentamethylcyclopentadienyl ring. For the tin analogue we could also detect ab road 119 Sn{ 1 H} NMR signal at 810.7 ppm, whereas 1·AlMe 3 provides ad istinctive 1 HNMR singletatÀ0.1 ppm due to the Al-bound methyl termini, with ac orresponding 13 C{ 1 H} NMR signal at 1.0 ppm. Interestingly, 103 Rh{ 1 H} NMR resonances due to the tetrylene MOLPsa ppear upshifted by ca. 400 ppm (d = À8756, 1·GeCl 2 ; À8836 ppm, 1·SnCl 2 )c ompared to 1 (d = À9165 ppm), contrasting with all otherm ain-group based MOLPs reported herein ( Table 2).
Single-crystals of compounds 1·GeCl 2 , 1·SnCl 2 and 1·AlMe 3 amenable to X-ray diffraction studies where grown by slow diffusion of pentanei nto their benzene or bromobenzene solutions, once more revealing the piano stool configuration aroundt he rhodium centera fter coordination to the Lewis acids ( Figure 2, Ta ble 1). The unsupported MÀMb ond lengths for 1·GeCl 2 (2.501(1) )a nd 1·SnCl 2 (2.687(3) )a re slightly shorter than the sum of covalentr adii (r Rh + Ge = 2.62; r Rh + Sn = 2.81 ), [30] whereas that of 1·AlMe 3 (2.635(4) )i si dentical to the expected theoretical value for ac ovalenti nteraction (2.63 ). [30] The asymmetric unit of structure 1·GeCl 2 contains four independentm olecules of the MOLP,b eing the aforementioned Rh-Ge bond length the average for all of them. The solid-state structures of 1·GeCl 2 and 1·SnCl 2 unveil as trong pyramidalization of the tetrel moiety,a ss een in other related systemsb ased on platinum. [42] However,t his is not the case in other metallic complexes with bound tetrels and ap lanar dispositiona round the group 14 element. [43] It has been noticed that pyramidalization requires both coordinationt os trongly Lewis basic metals and an on-directional lone pair, [42d] features fulfilled for 1·ECl 2 (E = Ge, Sn). Sincet he lone pair on stannylene dichloride has more pronounced s-character than that in its germylenea nalogue,t he directionality of the former is decreaseda nd as such ah igher pyramidalizationi sa nticipated for 1·SnCl 2 .I nf act, the pyramidalization angle estimated by the POAVm ethod of Haddon [44] for 1·SnCl 2 (26.2) surpass that of 1·GeCl 2 (24.4).
To the best of our knowledge, compounds 1·GeCl 2 and 1·SnCl 2 represent the first examples of rhodium-bound germylene and stannylene non-stabilized by the coordination of a base. All prior structures containing RhÀE(II) (E = Ge, Sn) bonds involvet etrel centersb earing an additional intra-or intermolecular Lewis donor. [45] As such, those escape the definition of MOLP investigated in this work, since base-stabilized tetrylenes do not behave as acidic fragment any more, but as s-donating ligands. For its part, earlier reports describe base-free rhodium adducts of SnCl 2 ,b ut their dimeric nature preclude ac lear understanding of the bondings ituation. [46] As introduced earlier, the preparation of aR h-alane adduct by direct combination of the two metal fragments, as reported herein, had so far been unsuccessful. The first crystallographycally characterizedR halane adduct was reported by Braunschweig relyingo nt he transmetalation of the alane from [(PCy 3 ) 2 Pt!AlCl 3 ]t o[ ( h 5 -C 5 H 5 )Rh(PMe 3 ) 2 ]. [47,48] The RhÀAl bond length in 1·AlMe 3 is considerably elongated by around0 .2 relative to the two previously reported Rh-alane adducts based on AlCl 3 , [41,47] as expected for the lessa cidic AlMe 3 .T his diminished acidity may explain the absence of previous unsupported transition metal MOLPs containing trimethylaluminum,b eing 1·AlMe 3 the first of its kind. [49] Once more, this is an unexpected finding considering the extensive use of AlMe 3 as am ethylating agent or in transition metal catalyzed polymerization.

Synthesis of Rh I MOLPs with d-Block Acids
Turning into the d-block and keeping our aim to prepare Rh I MOLPS with neutral main group metal Lewis acids we decided to check the reactivity of 1 with two common zinc precursors, more precisely ZnMe 2 and Zn(C 6 F 5 ) 2 .F or the sake of completeness, we also examined the formation of metal adducts with simple forms of copper and silver.C omplexes 1·ZnMe 2 and 1·Zn(C 6 F 5 ) 2 were immediately formed after addition of one equivalent of the organometallic zinc substrate over ab enzene solution of 1 (Scheme 3). These complexesexhibit sharp 31 P{ 1 H} NMR signals at d = À6.9 ( 1 J PRh = 192 Hz) and À7.2 ppm ( 1 J PRh = 167 Hz), respectively.T he noticeable decrease of the 1 J PRh coupling constantsr elative to 1 evidencesf ormation of Rh!Zn MOLPs. Their corresponding 103 Rh{ 1 H} NMR resonances appear downshifted to À9212 (1·ZnMe 2 )a nd À9355 (1·Zn(C 6 F 5 ) 2 )ppm. Other relevant NMR spectroscopic parameters are collected in Table 2and in the Experimental Section.
Reaction with group 11 precursors, whose acidity is also well-recognized, proved more problematic.R eaction with CuOTf( OTf À = CF 3 SO 3 À )o rA gNTf 2 (NTf 2 À = (CF 3 SO 2 ) 2 N À )r esulted in complexm ixtures that involveanumber of rhodium compounds as inferredf rom the presence of several doublets in the corresponding 31 P{ 1 H} NMR spectra. In contrast, addition of one equivalent of CuCl over ab romobenzene solution of 1 cleanly provided an ew species (1·CuCl)c haracterized by a 31 P{ 1 H} NMR doublet at À3.0 ppm ( 1 J PRh = 144 Hz), once again suggesting the formation of ad ative bond between the two metals (Scheme 3). The corresponding 103 Rh{ 1 H} signal resonates at À8540 ppm, shiftedt oh igherf requencies compared to 1.T his contrasts with all other MOLPs described herein except those containing ambiphilic tetrylenes, which speaks in favor of some differences in the bondings ituation between the MOLPsi nvolving purely acidic fragments and those where some degree of back-donation may be anticipated (i.e. those based on Ge, Sn and Cu).
Crystals of 1·ZnMe 2 and 1·Zn(C 6 F 5 ) 2 where grown by slow diffusiono fp entane into their benzene solutions. The larger acidity of the fluorinated zinc moiety is reflected in as horter Rh-Zn bond length of 2.484(1) in 1·Zn(C 6 F 5 ) 2 comparedt o that in 1·ZnMe 2 (d RhZn = 2.618(1) ), attesting as well that steric effects may be less relevant( Figure 2). Nonetheless, both RhÀ Zn distances account for less than the sum of the corresponding covalent radii (2.64 ), [30] suggesting as trong metal-metal interaction. These two complexes constitute the first unsupported MOLPse xhibiting ad ative Rh!Zn bond and constructed around neutral fragments. [38,50] Structures alike these are presumably relevant intermediates during Rh I -catalyzed Negishi coupling reactions. [50e, 51] Mechanistic studies have permitted to isolate aR h/Zn complex derived from insertion of the rhodiumc enter into one of the ZnÀCb onds in diphenylzinc, [51b] whose likely precursor consist in aL ewis adduct akin to 1·ZnMe 2 or 1·Zn(C 6 F 5 ) 2 .R elated to this, formation of a [Rh I ]!ZnCl 2 MOLP was postulated as ad eactivation product during catalysis, although their molecular formulation could not be elucidated.
Regarding the coppera dduct,a ttempts to grow single crystals of 1·CuCl were unfruitful, partly because of the low solubility of the adduct which caused rapid precipitation in most cases. This fact, along with non-definitive diffusion spectroscopic studies, prevented us to obtain ac lear picture of its molecular structure. In principle, both am onomeric or dimeric nature could be proposed. To discern between theset wo possibilities, we made use of DFT calculations. However,a ttempts to optimize ad imeric specieso ft ype [(h 5 -C 5 Me 5 )Rh(PMe 3 ) 2 Cu(m-Cl)] 2 resulted in cleavage of the chloride bridges,s upporting an unbridged formulation for 1·CuCl.I ti s interesting to note that this species represents ar are case of Rh!Cu MOLP,w ith prior complexes bearing aR h ÀCu bond typicallyr elyingo nt he stability conferred by bridgingl igands, [52] the use of cationic copperf ragments [53] or the coordination of the neutral copperh alide as ab ridging motif. [54] Computational analysis of Rh!Mb onding in Rh I MOLPs Insight into the nature of the Rh!Mi nteractions in the Rh I ÀM adducts has been obtained from DFT calculations, analysiso f the calculated electron densitieso ft he adducts within the Atoms In Molecules theory (AIM) [55] and Natural Bonding Orbitals (NBO) analysis. [56,57,58] Optimized geometries of the adducts in bulk solvent were obtained by DFT methods (SMD-wB97XD/ 6-31 g(d,p)/SDD level) [59,60,61,62,63,64] with the Gaussian09 software. [65] Although it can be argued that DFT-optimized geometries with as olvent model may not represent appropriately the solid state structures,i tm ust be highlighted that our model is in good agreement with the X-ray diffraction geometries available (RMSD for all geometries is 0.58 )a nd particularly that the calculated RhÀMd istances remaine qual or below the sum of the covalentr adii of the two atoms. [7,29] Optimized geometries for the Na, Li and Cu adducts were also calculated in halogenatedb enzene. In the case of the Li and Na species, the BAr F À anion was excluded from the calculations to yield RhÀM distances of 2.46 and 2.76 respectively.W hen the BAr F À was introduced in the Na system, the RhÀNa distance increased only slightly to 2.77 ,s till shorter than the sum of the covalent radii of Rh andN a. The CuCl adduct was considered as a monomeric species and the calculations afforded aR h ÀMd istance of 2.37 (AE cov radii = 2.74 ).
To pological analysis of the electron density was carried out with the AIM methods and the Multiwfn software [66,67] from wavefunctions calculated at the SMD-wB97XD/6-311 ++ g(2d,p)/Sapporo-TZPl evel [68,69,70,71,72] with the previously optimized geometries.T his study located bond critical points (BCPs)i nt he electron density and unique bond pathsc onnecting the Rh and Matoms for all adducts (Figure 4and SC1).
The existence of BCP and bond paths between two atoms has been interpreted as the necessary condition for them to form ac hemical bond and several indicators based on the electron density have been used in the literature to characterize interatomici nteractions. [55,73] Namely,t he Laplacian of the electron density at the BCP, r 2 1 b ,a nd the total energyd ensity, H b ,a st he sum of the electronic potential and kinetic energy densities, G b and V b . Thus, for open-shell interactions (pure covalent bonds) r 2 1 b < 0( the electron density is locally concentrated) and for closed-shell interactions r 2 1 b > 0( the electron density is locally depleted). Closed-shell interactions are also characterized by electron densities at the BCPs, 1 b ,o ft he order of 0.01 a.u.,a tl east one order of magnitude smaller than in open-shell interactions. Moreover,i th as been argued that the sufficient condition for ab ond to be considered covalenti s H b < 0, independently of the sign of the Laplacian. [74,75] Ac lass of intermediate or partially covalent bonds [76] have thus been characterized as having 2 > j V b j /G b > 1. Shared (metal-metal) and donor-acceptor (metal-ligand)i nteractions fall within this class. [77] As shown in Table 3( and Table S3 in the Supporting Information),t he valueso f1 b foro ur RhÀMi nteractions are small, ranging from 0.020 a.u. for 1·Na to 0.072 a.u. for 1·GeCl 2 .T his, in addition to positive values for r 2 1 b ,i si na greement with closed-shell interactions between the Rh and Ma toms. [55] For the sake of comparison, RhÀPB CPs' have 1 b values close to 0.1 a.u and r 2 1 b > 0. Also, 1 b at the RhÀHbond of the Rh III hydride 2 has av alue of 0.150 a.u. and r 2 1 b > 0. Arguably, [24a] the magnitude of 1 b and H b can be used to assess the strength of an interaction. [79] In this case, 1 b follows the order Na + < Li + < MgBr 2 < ZnMe 2 % AlMe 3 < Zn(C 6 F 5 ) 2 % SnCl 2 < CuCl % GeCl 2 ! H, and it correlates with H b , [80] which interestingly, is negative for all speciese xcept for that with the smallest 1 b , 1·Na ( Figure 5).
These results suggest that the least electronegative atoms (Li, Na, andM g), with the smallest 1 b and H b close to zero, form predominantly ionic interactions with Rh (although with some degree of electron sharing as it shall be discussed below), whereas the covalentc haracter becomes more prominent as the electronegativity of the element bound to Rh increasesa nd their electronegativity differenced ecreases (Dc p = c P (M or H)-c P (Rh)). Indeed, reasonable correlations have been found between Dc p and 1 b or H b as shown in Figure 5f or 1 b  (and FigureS4f or H b ). These correlations highlight ag eneral trend,b ut they obviously fail to account for the complexity of the interactions. For example, they do not reflect the different acidity of the two Zn fragments and do not include the RhÀH bond of 2,s ince its associated H b relative to those of the RhÀM bondsi sh igher than the correspondinge lectronegativity difference.
Another parameter that has been considered in this study is the delocalization index between the Rh and M, or Ha toms, d(Rh,M), which accountsf or the extent of electron sharing between the atomicb asins [81,82] and can be considered an AIM equivalent to orbital-based bond orders. For ac ovalent bond, shown. [78] Distances are in . Table 3. QTAIM indicators at Rh I ÀMB CPs. All data are in atomic units. electron density, 1 b (e·bohr À3 ); total energyd ensity H b (hartree·bohr À3 ); Laplacian of the electrond ensity r 2 1 b (e·bohr À5 ); ratio betweent he absolute electronic potential energy and kinetice nergyd ensities j V b j /G b ;d elocalizationi ndex between Rh and Ma toms, d(Rh,M) [e].  such as the HÀHo rC ÀHb onds, d(C,H) is close to 1, whereas purely ionici nteractions have delocalization indicesc lose to zero. Ta ble 3s hows d(Rh,M) values stretching from less than 0.022 electrons for the adducts with s-block metals to closet o 0.8 electrons for the adducts with the two tetrylenes, attesting the higher covalent character of the latter interactions. The value calculated for the RhÀHb ond of 2,acovalentb ond, is 0.91 electrons. Thus, the same trendsa st hose emerging from 1 b and H b are observedf or d(Rh,M) including al inear dependence with Dc p (Figure S4). When the Laplacian of the electron density is considered, all adducts yield positive values at the RhÀM( and RhÀH) BCPs, which is indicative of close-shell interactions. In this case, no correlationsa rose between the Laplacian and other magnitudes derived from the electron density.S ome correlations between the Laplacian and the electron density or the electronegativityd ifferenceh ave been found in coordinationc ompounds [79] and their absence in this case may reflect the different nature of the various RhÀMi nteractions of this work, as shall be discussed below from an orbital perspective.N evertheless, we can classify these interactions in at least two groups according to j V b j /G b values (vide supra). One includes the adducts with s-block metals,w hichh ave j V b j /G b values close to 1( H b % 0), characteristic of interactions with very low covalentc haracter (for 1·Na j V b j /G b = 0.98 a.u.), and as econd group contains the remaining adductsw ith p-a nd d-block metal, with j V b j /G b values that range from 1.30 a.u. for 1·ZnMe 2 to 1.83 a.u. for 1·GeCl 2 ,t ypical of more covalent, intermediate interactions.F or the sake of comparison Rh ! P bonds in theses ystems, classicald onor-acceptor interactions, have associated j V b j /G b values of about 1.5-1.6 a.u. The higher values for j V b j /G b have been found for the three p-block metals,w ith the value for the RhÀGe interaction approaching the j V b j /G b ! 2( r 2 1 b 0) limit for shared-shell( pure covalent) interactions. The j V b j /G b and r 2 1 b values for the RhÀHb ond of 2 are 1.975 and 0.010 a.u. respectively. Natural bonding orbital (NBO) analysis was performed at the DFT SMD-wB97XD/6-311g(2d,p)/def2-TZVP(ECP)l evel. [83,84] The NBO methodc reates ap attern of localized bonds and lone pairs that is aL ewis-type description for the molecule. These natural bonding orbitals mayn ot achieved oubleo ccupancy, and the departures from the "electron pair" can be rationalized in terms of partial occupation of "non-Lewis" orbitals and donor-acceptor interactions between molecular fragments. Each NBO can be associated with an atural localized molecular orbital( NLMO), which is exactly doubly occupieda nd results from incorporation of mixingsw ith non-Lewis orbital. [85] Thus, when an NBO is identified as the donor orbitali na ni nteraction, the correspondingN LMO informs aboutt he degree of mixing with the acceptor orbital. In addition, donor-acceptor stabilization energies( DE ij )c an be calculated that are related to the strengthoft he interaction.
Ta ble 4s ummarizes relevant donor-acceptor interactions and Wiberg bond orders (WBO) from the NBO analysis of the Rh I !Mb onds. Typical NBO terminologyh as been used to name the different types of NBOs, such as LP for lone pair,a nd LV for lone vacancy,w hich refers to an empty valence orbital localized on one atom. Also, the main atomic orbitalc ontribution to the LVsh as been included in parenthesis. TheN LMO columni ndicates the percentage of non-Lewis orbitalsf rom the acceptora tom that are mixed with the parentd onor NBO. This section does not aim at being comprehensive, buttoi llustrate representative interactions and to offer aq ualitative picture of the RhÀMb onding. For instance, more than one LP Rh ! LV interaction has been locatedf or most systemsw hereas only the most important is shown. Data for s Rh-P !LV interactions correspond to the average values of the interaction with the two RhÀPb onds in each adduct. Finally,b ack donation is by far one minor contribution to the Rh-tetrylene interactions, but it has been highlighted to illustrate the ambiphilic behavior of GeCl 2 andS nCl 2 in these adducts. Figure 6a nd Figure 7 show examples of relevant NBOs andN LMOs fort he above interactions.
The NBO analysis locates 4LPs, almostp ure d orbitals, on the Rh atoms of all adducts,e xcept for 1·GeCl 2 and the hydride 2,f or which only 3 d LPs where found. This agreeswith a Rh I formulation and d 8 electron count for most adducts and the expected Rh III , d 6 ,f ormulation for the hydride. In the case of 1·GeCl 2 ,aR h III /Ge 0 formulation cannot be assumed.I nstead, we propose that the [(h 5 -C 5 Me 5 )Rh(PMe 3 ) 2 ]m oiety forms one dative covalent bond with GeCl 2 ,t he lattera cting effectively as an Z-type ligand, [22a] as will be discussed in more detail later.
Inspection of the Rh d LPs of the formally Rh I adducts shows that at least one of them is populated by 1.82 electrons or less, except for 1·Li and 1·Na,f or which the relevant lowesto c-cupiedR hL Ps have 1.95 electrons each. The occupancy is higherf or the adducts of the more electropositive elements and lower for the adductso ft he more electronegative ones, with the lowest occupancy found for 1·SnCl 2 at 1.69 electrons. This reflects, once more, ah igherd egree of electron sharing in the adducts with the electronegative atoms. These Rh LPsa re delocalized onto LV NBOs of the acceptorm etal atoms. Thus, for s-block atoms, the acceptorL Vi sm ostly av alence so rbital, and the corresponding interaction can be described as d (Rh) ! s (M) .T he occupancy of the acceptoro rbital and the major donor-acceptor stabilization (or delocalization) energies (DE ij ) for these interactions are:0 .05 electrons and 6.00 kcal mol À1 for 1·Li;0 .05 electrons and 6.71 kcal mol À1 for 1·Na;a nd 0.40 electrons and 28.7 kcal mol À1 for the MgBr 2 adduct, 3.
However,i nt he above species, as well as in the remaining adducts considered, the Rh!Mi nteractioni sd ominated, at least in terms of delocalization energies, not by Rh-localized d orbitals, but by electron donation from the s (Rh-P) bonds, [31b] which have about 72 %P( sp)a nd 28 %R h( sd)c haracter.F or the Li, Na and MgBr 2 adducts the s (Rh-P) !s (M) interaction have DE ij of 22.5, 19.3, 41.3 kcal mol À1 respectively.
The NBO description of the RhÀMb onding in the adducts with p-block acceptora toms is more varied than above.T hus, d (Rh) !sp 3 (Al) and s (Rh-P) !sp 3 (Al) donor-acceptor interactions were located for 1·AlMe 3 ,w ith the latter being the major contribution in terms of delocalization energy (DE ij are 33.7 and 80.6 kcal mol À1 respectively). The occupancies of the donor NBOs are 1.75 and 1.84 electrons for the Rh LP and the Al LV,a valence sp 3 hybrid, respectively.T he RhÀMi nteractions in the adducts with the two tetrylenes, 1·SnCl 2 and 1·GeCl 2 ,w hich were assigned the highest covalent character according to the AIM analysis, are described very differently by the NBO analysis:w hereas the donor-acceptor description is used for the former,o ne bonding NBO was localized between Rh and Ge in the latter (Figure 7). Close inspectiono ft he NLMO associated with the donor NBO of the d (Rh) !p (Sn) interaction in 1·SnCl 2 (DE ij = 55.2 kcal mol À1 )r eveals that it has the highestm ixingo f acceptorm etal orbitals of all analogous NLMOs in this study, with 81.9 %R ha nd 15 %S nc omposition, [86] whereas the NLMO associated with the s (Rh-Ge) NBO of 1·GeCl 2 has an even higher mixing of Ge orbitals, although it is heavily weighted towards the Rh atom:7 1% Rh (sd 2 )a nd 23 %G e( p), with about 2% mixingf rom each Pa tom. This can be compared with the s (Rh-H) NBO of 2,w hich has about 55 %R hc haracter and 45 %H character.T he bonding in this case is pure covalent from the localized orbital perspective.N evertheless, the s (Rh-P) !p (Sn) interaction is also dominant in 1·SnCl 2 , [31b] with DE ij = 62.8 kcal mol À1 .T he involvement of the RhÀPb onds in the RhÀGe interaction of 1·GeCl 2 is described in terms of donor-acceptor interactions: s (Rh-Ge) !s* (Rh-P) and s* (Rh-Ge) ! s (Rh-P), and in the mixing of Po rbitals in the NLMO associated to the s (Rh-Ge) NBO. According to these results, the interaction in 1·GeCl 2 is best de-  scribed as ad ative covalent bond with the Rh fragment acting as an Ll igand, and as imilar description, with al ower degree of electrond onation/sharing, could be used for 1·SnCl 2 ,t hat is, both can be equallyd escribed as MOLPs.
In addition, it is interesting to note that both tetrylenes have LPs which are mostly filled valence s orbitals, which back donate electron density onto antibonding s* (Rh-P) NBOs. [31b] Back-donation to the [(h 5 -C 5 Me 5 )Rh(PMe 3 ) 2 ]i sam inor contribution to the RhÀGe and it es negligible, when detected (DE ij 1kcal mol À1 ), in the remaining cases.
In the adducts with d-block acceptora toms, the donor-acceptor interaction description has also been chosen. The Zn and Cu atoms of 1·ZnMe 3 , 1·Zn(C 6 F 5 ) 2 and 1·CuCl accept electron density onto their 4s valence orbitals from Rh LP (d)a nd s (Rh-P) NBOs. The degree of interaction, based on bond order, occupancy of the donor and acceptor orbitals and donor-acceptor stabilization energiesi si ntermediate between those of s-block metals containing adductsa nd those of p-block metal containing adducts, for which is greatest.
The magnitude of the RhÀMo rbitali nteractions is reflected in the WBOs, which roughlyf ollow the same trends as 1 b and d(Rh,M), described above. Interestingly,s omeo ft hese trends can be used to explain, at least qualitatively the variation of the RhÀPd istances, which are shorter for adducts with smaller WBOs for their RhÀMb onds (or 1 b and d(Rh,M)) andl ongerf or adducts with larger WBOs (Figure 8). As the Rh I !Mi nteractions become more important, there is ag reater involvement of s (Rh-P) orbitals (and in some cases weak back donation onto s* (Rh-P) ), therefore weakening the RhÀPb onds.

Conclusions
The choice of [(h 5 -C 5 Me 5 )Rh(PMe 3 )a saLewis base for the synthesis of unsupportedM OLPs has proved highly successful. We have prepared up to nine Rh-based bimetallic compounds of this kind, providing X-ray diffractions tructures fort hose containing fragments MgMeBr,Z n(C 6 F 5 ) 2 ,Z nMe 2 ,G eCl 2 ,S nCl 2 and AlMe 3 .I ti ss urprising that despite the wide use of some of these Lewis acidic fragments, their corresponding MOLPs represent highly unusual examples of unsupported MÀMbonding, particularly in cases like those with aR h!Mg (3)o raRh!Al (1·AlMe 3 )d ative bonds. The growing interesto nM OLPs is reflectedb yi ncreased number of studies focusing either on accessingn ew structures or computationally investigating families of compounds already prepared, whereas ac ombined effort on as eries of MOLPs is still lacking. Wep rovide here a comprehensive computational investigation on the RhÀM bondingo ft he preparedR hM OLPs, with severals ound correlationsf ound for relevant parameters associated to the metalto-metal bond. For instance, the more electronegative atoms (Ge, Sn, Al) tend to form more covalentb onds with rhodium, whereas the ionic character becomes more prominent in the least electronegative (Li, Na, Mg). Nevertheless, we have quantified some degree of electron sharing for all investigated MOLPs. Curiously,t he Rh!Mb ond is dominated by electron donation from the RhÀP s-bonds rathert han from af illed Rh d-orbital to the acidic site, which resultsi nother relevant correlationb etween Wiberg Bond Orders and RhÀPb ond lengths. Overall,w ebelieve that this combined experimental/computational approacht oR h-based MOLPs will aid in the development of other related systems for the advancemento ft his growing field.

Experimental Section
General considerations:A ll preparations and manipulations were carried out using standard Schlenk and glove-box techniques, under an atmosphere of argon and of high purity nitrogen, respectively.A ll solvents were dried, stored over 4 molecular sieves, and degassed prior to use. To luene (C 7 H 8 )a nd n-pentane (C 5 H 12 ) were distilled under nitrogen over sodium. [D 6 ]Benzene were dried over molecular sieves (4 ). Tind ichloride was dried by vigorous stirring with acetic anhydride, while copper(I) chloride by co-evaporation with toluene and drying under vacuum. Other chemicals were commercially available and used as received. For elemental analyses aL ECO TruSpec CHN elementary analyzer,w as utilized.
NMR Spectroscopy:S olution NMR spectra were recorded on Bruker AMX-300, DRX-400 and DRX-500 spectrometers. Spectra were referenced to external SiMe 4 (d:0ppm) using the residual proton solvent peaks as internal standards ( 1 HNMR experiments), or the characteristic resonances of the solvent nuclei ( 13 CNMR experiments), whereas 31 Pw as referenced to H 3 PO 4 .S pectral assignments were made by routine one-and two-dimensional NMR experiments where appropriate. 103 Rh NMR was acquired at 15.9 MHz using an observe 5mmt riple resonance broadband probe (broadband inner coil and doubly tuned 1 H/ 31 Po uter coil) with 908 pulses of 37.5 msa nd 30.0 msf or 103 Rh and 31 P, respectively. 103 Rh chemical shifts, d,a re given in ppm relative to X = 3.186447 [87] (reference compound Rh(acac) 3 ,w here acac stands for [CH 3 COCHCOCH 3 ] À ) and derived indirectly from the 31 P- 103 Rh HMQC experiments by four pulse 31 PÀ 103 Rh HMQC experiments with 1 Hd ecoupling during acquisition. Note that despite the fact that IUPAC recommends the use of Rh(acac) 3 as the reference, the alternative Xi value X = 3.160000 for Rh metal has been commonly employed in the literature. The experiments were optimized using the 1 J RhP values obtained from the corresponding 31 P{ 1 H} spectra. The transmitter frequency offset and the spectral width were varied to ensure that no signals were folded. 2D data were zero filled and processed with exponential line broadening of 10 Hz in the direct F2 dimension, and unshifted sine-bell window function in the indirect F1 dimension.