A new route for the efficient metalation of unfunctionalized aromatics† †Electronic supplementary information (ESI) available. CCDC 1869336–1869340. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c8sc04325f

Efficient metalation of benzene is reported for the first time using reagents based only upon a mixture of a copper amide and a lithium amide.


Introduction
The elaboration of functionalized aromatic compounds underwent little short of a revolution when it was reported, in 1999, that simply prepared lithium zincates could selectively deprotonate these rings whilst exhibiting remarkable levels of ancillary group tolerance. 1 These data led to an explosion of interest in what have become known as 'synergic bases', 2 heterobimetallic reagents of the type R m MðNR 0 2 Þ n M 0 (R, R 0 ¼ organyl; m ¼ 0-3; M ¼ more electronegative metal; NR 0 2 ¼ amide; n ¼ 1-3; M 0 ¼ less electronegative metal). Over the subsequent two decades these bases have afforded levels of reactivity, 3 regioselectivity 4 and functional group tolerance 5 not previously available using traditional main group organometallic bases, and they continue to evolve new applications. 6 Most work in this eld has focused on the directed ortho metalation of functional aromatics, in which M ¼ Zn, 1 Cr, 7 Fe, 8 Mg, 9 Al, 10 and Mn. 11 Oen, in these reactions M 0 ¼ Li and while organyl elimination has been observed and rationalized on thermodynamic grounds, 12 experimental, 13 spectroscopic 14 and computational 15 evidence points to a preference for kinetic amido basicity. Recourse to higher alkali metals (for M 0 ) has allowed other major advances, for example synergic bases capable of offering non-traditional meta 16 and para 16c,17 directed metalation. In particular, whilst synergic bases incorporating Li afford new opportunities in directed deprotonation, much more reactive alkali metals 18 have been required to achieve the highly desirable target of efficiently metalating unfunctionalized aromatic hydrocarbons, such as benzene, 19 toluene 19,20 and naphthalene. 21 Taken together with the inability of traditional main group metal bases to efficiently activate such simple aromatics, 22 it has become clear that opportunities remain for the development of metalating agents capable of activating unfunctionalized feedstock.
One of the most fruitful heterobimetallic combinations employed in the selective activation of functionalized benzenoids 23,24 and heteroaromatics [25][26][27] is that of Cu with Li. This use of so-called lithium cuprates led to the emergence of directed ortho cupration (DoCu) as a synthetic strategy. [28][29][30] These systems have been the subject of recent review, 31 with the importance of employing a sterically demanding amide such as TMP (¼ 2,2,6,6-tetramethylpiperidide) having been noted. 23,32 Though bis(TMP)cuprates themselves have been documented as being capable of DoCu, the inclusion of Li-salts (such as LiCN) to form what have been coined Lipshutz-type cuprates (Fig. 1a) has proven essential for efficient reactivity. 29 This observation, combined with a debate over bonding of the salt anion, 33 has meant that while Li-salt incorporating cuprates have been studied in some detail, 34 Li-salt free cuprates have been subject to less exploration. 35,36 To the best of our knowledge, only two examples of unsolvated bis(amido) Gilman cuprates, (amido) 2 CuLi, have been characterised in the solid state. 30,37 In contrast to its Lipshutz-type analogues (TMP) 2 Cu(X) Li 2 (X ¼ CN, halide, SCN, OCN; Fig. 1a), 23,[25][26][27][28]30,36 the preisolated dimer of this Gilman cuprate is considered not to be reactive in DoCu. 30 We recently reported the isolation of a non-stoichiometric cuprate (TMP) 4 Cu 2. 7 Li 1.3 , 36 which led us to suspect the existence, in solution, of a series of aggregates of TMPLi and TMPCu described by the general formula (TMP) m+n Cu m Li n (m + n ¼ 4). By extension, these data caused us to propose the existence of an isomeric variant on the dimer of a previously reported 30 Gilman cuprate (Fig. 1b) in which metal ions of the same type, instead of being located opposite to one another in the 8-membered metallacyclic core, reside adjacent to one another. 36 Herein we report on the isolation of this species and its characterisation in the solid state. In contrast to the previously reported Gilman cuprate, this new isomer exhibits unexpected reactivity with benzene. The ability to replicate this behaviour using mixtures of TMPLi and TMPCu, but not using either monometallic compound in isolation, points to cooperativity in solution. This process is studied by elucidating a series of Ph(TMP) 3 Cu m Li n (m + n ¼ 4) complexes, with results opening up the possibility of deploying easy-to-handle metal amides 38 to smoothly deprotonate unfunctionalized aromatic feedstock.

Results and discussion
Synthesis and characterization of (TMPCu) 2 (TMPLi) 2 1 In previous work, Lipshutz-type cuprates have typically been the focus of study due to their superior reactivity in DoCu when compared to pre-isolated Gilman cuprate. 29 Accordingly, reaction conditions have been sought which favour their formation. These have been found to vary substantially with the choice of Cu(I) precursor. In the case of Lipshutz-type cuprate (TMP) 2 -Cu(Cl)Li 2 (Et 2 O), appropriate conditions have been reported to be 2 : 1 LiTMP : CuCl in toluene followed by the introduction of Et 2 O prior to recrystallisation. 23 We have now found that when the same reaction is performed in toluene containing limited Et 2 O (1 eq. wrt Li), a different crystalline material can be obtained. X-ray diffraction reveals a cyclic mixed-metal aggregate 1 which is isomorphous with previously reported mixed-metal aggregate (TMP) 4 Cu 2.7 Li 1.3 , 36 but with an overall composition of (TMP) 4 Cu 2 Li 2 (i.e. that of a Gilman cuprate) (Fig. 2). Elemental analysis indicated that the bulk composition of the product was consistent with the empirical formula (TMP) 2 CuLi.
Crystallographic renement of 1 revealed that Cu2 and Li4 in Fig. 2 were almost perfectly disordered according to the pseudo two-fold symmetry of the molecule. This suggests an aggregate best formulated as (TMPCu) 2 (TMPLi) 2 (Fig. 1c), which can be reasonably interpreted as an isomer of the previously characterized dimer of Gilman cuprate TMPCu(m-TMP)Li 2 (Fig. 1b). 30 In recent work, we have proposed the existence of this isomer from solution-state data. 36 Evidence that this species corresponds to 1 as isolated in this work was gathered from 1 H NMR spectroscopy. When crystalline 1 was dissolved in C 6 D 6 , the key spectroscopic features observed previously 36namely the presence of three TMP-Me resonances, at d 1.76, 1.57 and 1.39 ppm, in a 1 : 2 : 1 integral ratiowere reproduced (in addition to minor reformed TMPH, identied by a TMP-Me resonance at d 1.06 ppm; see ESI, Fig. S1 †). 1 H, 1 H-NOESY reinforces the view that the solid-state structure of 1 is robust in solution by demonstrating a lack of exchange peaks between Me-groups at room temperature. 7 Li NMR spectroscopy revealed a dominant singlet, at d 1.64 ppm, alongside a very minor resonance, at d 0.95 ppm. The former of these is consistent with the single Li-environment in 1 whereas the minor resonance has previously been interpreted as belonging to a Cu-rich species such as (TMP) 4 Cu 3 Li. 36 13 C NMR spectroscopy revealed a spectrum consistent with the three different TMP environments in 1, most clearly evidenced by the TMP-2,6 carbon resonances, at d 56.9, 54.2 and 52.0 ppm. These correspond to TMP ligands bonded to two Cu centres, one Cu and one Li centre and two Li centres, respectively.

Reactivity of 1 towards unfunctionalized aromatics
In previous work, it was hypothesised that 1 might be the kinetic form of (TMP) 4 Cu 2 Li 2 and that it should convert to the thermodynamic isomer (Gilman dimer 2 2 ) under suitable conditions. 36 To test this, crystalline 1 was dissolved in C 6 D 6 and heated to 50 C. The composition of the mixture was monitored at regular intervals by in situ NMR spectroscopy. This established the expected progressive loss of 1. However, instead of revealing the anticipated proportional growth of Gilman cuprate, a number of additional Li-containing species were observed, characterised by unusual high-eld 7 Li resonances (Fig. 3). Furthermore, the quantity of free amine present in the reaction mixture (observed by 1 H NMR spectroscopy) was far in excess of what could reasonably be attributed to hydrolysis by adventitious moisture. In any case, the absence of an N-H signal (typically observed by 1 H NMR at d 0.3 ppm) suggested that this may be deuterated free amine TMPD rather than TMPH. These observations suggested the intriguing possibility of a metal-deuterium exchange reaction occurring between 1 and the NMR solvent. Logically, this would result in the generation of aryl anions (C 6 D 5 À ), whose p-type interaction with Li + could then account for the observed high-eld 7 Li NMR resonances (as previously reported in aryl(amido)cuprates). [39][40][41] The 7 Li NMR spectrum of the nal reaction mixture (Fig. 3, top) reveals a subsidiary peak at d 0.90 ppm, identied as 2, and two dominant resonances, at d 1.41 and À2.93 ppm (in a 1 : 1 integral ratio). The former of these two signals falls within the chemical shi range expected for a Li + center bridging two amido ligands (as observed previously in amidocuprates). 36 On the other hand, the resonance d À2.93 ppm indicates some degree of Li/p interaction. [39][40][41] These data point towards the presence of a well-dened organo(amido)cuprate, though  symmetry indicates this not to be a simple homodimer of the type [ArCu(m-TMP)Li] 2 (Ar ¼ C 6 D 5 ). This view is substantiated by data obtained from the reaction of 1 with C 6 H 6 solvent under similar conditions, for which 1 H NMR spectroscopy on the reaction mixture shows both TMP ligands and a number of aromatic hydrogens (d 7.9-6.7 ppm), the splitting pattern of the latter suggesting a Ph group. Integration suggests an amido :phenyl ratio of 3 : 1. Furthermore, removal of the volatiles, followed by recrystallisation from toluene provided material free from TMPH, establishing that this potentially Lewis basic by-product of aromatic deprotonation does not coordinate the organometallic product in this system. 7 Li NMR spectroscopy on the same material reveals singlets, at d 1.41 and À2.86 ppm, in a 1 : 1 integral rationearly identical to those observed in the deuterated system (Fig. 3) overall suggesting the formula of the major product to be Ph(TMP) 3 Cu 2 Li 2 3.
Probing Ph(TMP) 3 Cu 2 Li 2 3: synthesis and characterization of PhCu(m-TMP)Li 4 Crystals obtained directly from the reaction mixture described above proved to be of poor quality. An alternative method of preparing 3 for structural characterisation was therefore sought. The realisation that 3 is a heterodimer of TMPCu(m-TMP)Li and PhCu(m-TMP)Li suggested that the combination of these reagents might result in equilibration to the desired product. Testing this hypothesis necessitated the synthesis of solvent-free PhCu(m-TMP)Li. This compound has been isolated previously as a monomeric tris-(THF) solvate, synthesised through the combination of PhCu in THF with TMPLi. 29 Given the known difficulty of removing THF from such preformed complexes, an alternative method was devised. Solvent-and salt-free PhCu is known to be very difficult to isolate, so we opted to synthesise PhLi and then combine it with in situgenerated TMPCu in a hydrocarbon solvent. Remarkably, the insoluble PhLi dissolved very rapidly to give a yellow solution, which, aer concentration and chilling, yielded prismatic orthorhombic crystals. X-ray crystallography revealed a homodimer of the expected heterocuprate, [PhCu(m-TMP)Li] 2 4 2 (Fig. 4). The coordination of Li differs signicantly from that seen in the small number of previously reported cuprates of this type. While homodimer cuprates [MesCu(NR 2 )Li] 2 (R 2 N ¼ TMP, N(CH 2 Ph) 2 ) exhibit exclusively h 6 -hapticity towards Li, 39-41 this is not observed in 4 2 . In this complex, Li1A is canted away from the Ph-group (Cu1-C10-Li1A 81.33 (13) ) and appears to interact primarily with ipso carbon C10 (Li1A-C10 2.153(4)Å). Meanwhile, interaction with the ortho carbons C11 and C15 is weaker and distinctly asymmetrical (Li1A-C11 2.594(4), Li1A-C15 2.826(4)Å), suggesting that the Ph-group in 4 2 is best regarded as h 1 . Though unusual, DFT calculations (see later) nonetheless conrm that h 1 -arene coordination by lithium is energetically preferred in this case. 1 H NMR spectroscopy on 4 in C 6 D 6 solution revealed a single set of resonances attributable to Ph and TMP moieties, though the observation of two TMP-Me resonances (1 : 1 integral ratio) suggested retention of a static conformation for the amido ligand. Meanwhile, a dominant resonance was observed at d À2.43 ppm by 7 Li NMR spectroscopy, consistent with retention of the low-hapticity Li/p interactions found in the solidstate dimer. The combination of crystalline 4 with TMPCu(m-TMP)Li 2 (1 : 1 molar ratio) in C 6 D 6 was followed by screening of the reaction mixture for the formation of Ph(TMP) 3 Cu 2 Li 2 3 using in situ NMR spectroscopy. This failed to reveal any changes upon mixing at room temperature, but indicated substantial levels of  conversion aer 24 h at 50 C. As a result, the same reaction was then attempted on a larger scale, using toluene as the solvent. Replacement of this solvent with hexane, and chilling of the resulting solution to 5 C gave a crop of block-like crystals. X-ray diffraction revealed a metallacycle, featuring Ph and TMP ligands in the expected ratio, conrming the successful fabrication of 3 (Fig. 5).
The crystal structure of 3 supports its interpretation as a heterodimer and establishes a higher arene coordination mode than observed in 4 2 . Li4 interacts in a p-fashion with Phcarbons in an asymmetrical manner best described as h 3 . Retention of this feature in solution would explain the spectroscopic observation of reduced shielding of the arene-bound Li in 3 ( 7 Li NMR: d À2.86 ppm) relative to the h 6 -coordination seen previously in lithium aryl(amido)cuprates. [39][40][41] The Li-p interactions in 3 have a wide-reaching inuence upon the structure. In contrast to 4 2 , the C-Cu-N unit deviates signicantly from linearity, allowing the Ph to incline towards Li4. The framework twists to accommodate this geometry, with a mean deviation from the plane of the metallacyclic ring (which is taken to incorporate C28) of 0.3046Å (ESI, Fig. S3a †). This contrasts signicantly with 4 2 , which was much more planar (mean deviation from the plane of the metallacycle of 0.0565Å). Lastly, it was noted previously that there is a preference for the so-called endo orientation of TMP in cuprates incorporating this ligand 37 and it is interesting to note that this feature is also preserved in the dimer of 4 and heterodimer 3.
Beyond Ph(TMP) 3 Cu 2 Li 2 3: synthesis and characterization of We next considered whether other species with the same general structure as 3, but with different metal content, might exist and whether these might potentially account for some of the minor products present in the 1-C 6 H 6 reaction mixture. Logically, this could be achieved by combining TMPCu or TMPLi with PhCu(m-TMP)Li 4 in a 2 : 1 ratio. Preliminary in situ NMR experiments conrmed that re-aggregation took place. When TMPCu was used, heating to 50 C for 24 h was necessary to achieve this. However, heating was not necessary for the analogous reaction using TMPLi.
In the same manner as described above for the synthesis of 3, a 2 : 1 combination of TMPCu and 4 in toluene provided a crystalline material aer concentration of the solution. X-ray diffraction revealed a Cu-rich cuprate with composition Ph(TMP) 3 Cu 3. 12 Li 0.88 , which can be regarded as Ph(TMP) 3 Cu 3 Li  5, subject to a small amount of Cu-Li substitution 42 at the sole Li-containing site (Fig. 6).
The metallacycle in 5 is similar to that in 3, whilst the Ph coordination mode is clearly more akin to that of 4 2 . In 5, C-Li bond distances (Li4-C28 2.204(5) and Li4-C33 2.656(6)Å) suggest h 1 -coordination to the ipso carbon (cf. 4 2 ). Regarding substitutional Cu-Li disorder, even though site Li4 in the present structure contains only a small amount of Cu (12% by crystallography with NMR spectroscopy suggesting similar levels in bulk samples), its presence implies the existence of a purely Cu-based homologue: Ph(TMP) 3 Cu 4 6. It was speculated that trace amounts of 6 could arise from Cu-Li exchange between 5 and TMPCu, expelling TMPLi (Scheme 1, eqn (1)). To attempt to prepare 6 in isolable quantities, we opted to conduct this exchange using the previously reported strategy of Cu I -O/ Li-O bond metathesis. 43 In situ NMR spectroscopy on a 1 : 1 mixture of t BuOCu with 5 in C 6 D 6 (which developed a green colouration upon heating to 50 C) revealed formation of 6 and   from which colourless crystals of 6 could be isolated in low yield. X-ray crystallography on 6 ( Fig. 7) reveals a structure that is nearly superimposable with that of 5. The major difference between the two lies in the position of Ph, which adopts a more symmetrical bridging mode in 6 than in 5. The M1-C28-M4 (M1 ¼ Cu; M4 ¼ Li 5, Cu 6) angle is correspondingly reduced (Cu1-C28-Li4 81.28 (11) in 5 and Cu4-C28-Cu1 75.45 (15) in 6), bringing it well within the range known for tetrameric arylcopper species (70.1-77.5 ). 44 This presumably reects a greater level of covalency in the C-Cu bond compared to that of C-Li, a suggestion which is supported by the signicant shortening of the Cu1-C28 interaction when Ph bridges Cu and Li as opposed to Cu only (Cu1-C28 1.895(2)Å in 3, 1.915(4)Å in 5 and 1.971(4)Å in 6).
In contrast to the previous reactions, when TMPLi was combined with PhCu(m-TMP)Li 4 at room temperature in hexane a viscous oil was obtained which repeatedly failed to deposit crystals; multiple attempted crystallisations from pentane also failed. However, removal of all volatiles gave a white solid which was conrmed to be Ph(TMP) 3 CuLi 3 7 by 1 H NMR spectroscopy through the observation of Ph and TMP resonances in a 1 : 3 ratio, in addition to 7 Li NMR signals at d 2.09 and À4.26 ppm in a 2 : 1 ratio.
In situ synthesis of 3, 5 and 7 Comparison of the NMR spectra of 3, 5, and 7 with the in situ data for the nal reaction mixture of 1 in C 6 D 6 revealed that d 5 -5 and d 5 -7 constitute minor products in this system. In addition, there were two other Ph-containing species, though the proportion of these species is too small to permit condent assignment from NMR data (Fig. 8).
Our attention next turned to eliciting whether 3, 5 and/or 7 could be generated selectively via deprotonative means. It was clear from in situ work that the Cu : Li ratio (of 1 : 1) in 3 re-ected that of the starting material 1. Moreover, our interpretation of 1 as a molecular mixed aggregate of TMPLi and TMPCu raised the possibility that reactivity akin to that shown by 1 might be exhibited by a physical mixture of these two amides. Such an approach would remove the restrictions in the Cu : Li stoichiometry imposed by 1. To test this hypothesis, preisolated, crystalline TMPCu and TMPLi 45 were combined in C 6 D 6 (in a sealed NMR tube) and heated to 50 C for ca. 24 h. The TMPCu : TMPLi molar ratio in the starting mixture was varied in each of the ve experiments, 3 : 1, 2 : 1, 1 : 1, 1 : 2, and 1 : 3. Analysis of the resulting reaction mixtures by 7 Li NMR spectroscopy revealed multiple productsincluding d 5 -3, d 5 -5, and d 5 -7and a gradual transition of the dominant organo(amido)cuprate from d 5 -5, through d 5 -3, to d 5 -7 as Cu was Scheme 2 Reactions explaining the production of 3, 5 and 7.  replaced by Li in the reaction mixture ( Fig. 9; see also ESI, Fig. S17a and b †). While 1 in C 6 D 6 showed low levels of d 5 -3 aer 24 h at 50 C, all physical mixtures of TMPCu and TMPLi generated much higher levels of organometallic products d 5 -3, d 5 -5 and/or d 5 -7 aer the same period of time at 50 C, implying a faster rate of reaction. Nonetheless, the presence of residual TMPCu-TMPLi aggregates suggested that longer reaction times might be benecial to achieve optimal conversion (especially in deuterated solvents, which might be expected to react more slowly than their protic counterparts).
Further experiments on a larger scale reinforced our hypothesis: equimolar TMPCu and TMPLi were combined in C 6 H 6 and heated to 50 C for ca. 5 days, aer which time the volatiles were removed. 7 Li NMR spectroscopy on the residue (in C 6 D 6 ) indicated a composition of mainly 3 (85%), the remainder being 2 (4%) and other Ph-containing species (11%), including 4 and 5 (see ESI, Fig. S18c †). This contrasts with the 1 : 1 reaction in Fig. 9, where the levels of d 5 -3, d 5 -5 and d 5 -7 were much more evenly distributed.
Considering the observations outlined above and in the preceding sections, it seems likely that the reaction pathways leading to the formation of 3, 5 and 7 involve at least two steps. This is suggested by the fact that 4 is explicitly demonstrated to react with TMPLi-TMPCu aggregates but is not itself observed in signicant quantities in the in situ reaction mixtures examined. A plausible sequence of reactions would therefore involve (1) reaction of 1 or components thereof with benzene and (2) subsequent re-equilibration with the remaining 2, TMPCu or TMPLi to generate 3, 5, or 7, respectively (Scheme 2). Though the nature of the active base in these reactions is not certain, control experiments in which either TMPCu or TMPLi was heated in C 6 D 6 to 50 C revealed no substantial changes by NMR spectroscopy aer 24 h (see ESI, Fig. S14a-c and S15a-b †), suggesting synergistic activity instead. Furthermore, preisolated 2 (known to be a dimer in the solid state) was also inactive under these conditions (see ESI,  5, and expected Ph(TMP) 3 CuLi 3 7. This was done with a view to predicting the structural features of 7, for which no crystals could be obtained. Efforts focussed rst on the two structural isomers of the formulation (TMP) 4 Cu 2 Li 2 , amide aggregate 1 and Gilman cuprate dimer 2 2 . Analysis conrmed the Gilman cuprate dimer to be the thermodynamically preferred isomer, but by only 2.5 kcal mol À1 (see ESI, Fig. S8 and S9 †).
Subsequently, the remainder of the structures were assessed, whereby the energy of all possible conformers (generated by individually ipping each amido ligand from exo to endo or vice versa) 23 was evaluated (see ESI, Fig. S10 to S13 †). For models based on the formulation Ph(TMP) 3 Cu 3 Li (5), the preferred conformer was found to be 6.3 kcal mol À1 lower in energy than the most unstable one, and corresponded exactly to that observed experimentally by X-ray diffraction. The experimentally seen low-hapticity coordination mode of the arene is reproduced satisfactorily and differs little between Ph(TMP) 3 -Cu 3 Li conformers. In a similar vein, for models based on the formulation Ph(TMP) 3 Cu 2 Li 2 (3), the energy difference between the conformers was relatively small (5.5 kcal mol À1 ). Again, the conformer observed in the solid-state proved to be the most stable calculated one. As for attempts to model 5, modelling of 3 led to the coordination mode of the arene (approximately h 3 ) again being quite accurately reproduced by theory. In spite of the demonstrable ability of DFT calculations to replicate the experimentally observed structures of 1, 3 and 5, attempts to predict the conformation of Ph(TMP) 3 CuLi 3 (7) proved difficult. In this case, DFT work revealed a much smaller range of energies associated with the different conformers (only a 2.0 kcal mol À1 difference between the highest and lowest energy forms). However, in all modelled conformers, a high hapticity arene coordination mode was predicted (approximately h 6 ) and this is consistent with the signicant shielding of the Li-centre seen by NMR spectroscopy.

NMR spectroscopy
Further insights into the structures of 3, 5 and 7 in hydrocarbon solution have been gathered using 2D NMR spectroscopy. Analysis of 1 H, 1 H-NOESY and 1 H, 7 Li-HOESY data corroborates the maintenance, in solution, of structure-types exemplied by 3 and 5 in the solid-state. However, extension of this analysis to 7 (for which solid-state data could not be obtained) suggests a degree of uxionality not observed in 3 or 5.
In contrast to the low-temperature data, 1 H, 1 H-NOESY results on 3 at ambient temperature (25 C) revealed exchange correlations between the axial and equatorial hydrogens of TMP 1 (in both its Me groups and its ring; Fig. 12 and S3h †). This required (at least in part) ring inversion of TMP 1 , which is notable since DFT calculations predicted the resulting conformation of 3 to be the most energetically accessible conformer above the ground state (see ESI, Fig. S10a †).
Turning to Cu-rich 5 (Fig. 13 and S5d †), 1 H, 7 Li-HOESY data obtained in C 6 D 6 at room temperature revealed one correlation at d( 1 H, 7 Li) ¼ (1.21, À1.46) ppm (see ESI, Fig. S5g †), allowing assignment of TMP 3 -Me eq (viz. C26/Li4 2.801(6)Å in Fig. 6). The remaining Me-groups could be assigned by their proximity to the o-Ph hydrogens and to each other, using 1 H, 1 H-NOESY data ( Fig. 14a and ESI Fig. S5e †). Aside from TMP 3 -Me eq (C26/C33 4.097(6)Å), one other TMP-Me group showed a strong correlation to o-Ph, at d( 1 H, 1 H) ¼ (7.88, 1.99) ppm, which arose from TMP 1 -Me eq (C8/C29 4.357(6)Å). Unlike in 3, this ligand adopted an exo orientation with respect to the aryl-bonded Cu center (in the solid state), leaving TMP 1 -Me ax unsuitably orientated to generate a nOe with the aromatic ring. This is consistent with the failure to observe exchange correlations for TMP 1  suggesting thermally induced ring inversion (cf. 1 H, 1 H-NOESY on 3 at room temperature). Interestingly, DFT calculations suggest that the molecular conformer in which TMP 2 is ring-ipped with respect to the ground-state structure shown in Fig. 13 is the second lowest energy structure, whilst the third lowest in energy is generated by inverting TMP 1 instead (see ESI, Fig. S12a and S12g †). In other words, spectroscopic data suggest that increasing the temperature experimentally accesses higher energy conformers, and these appear in the order predicted by DFT calculations. Unlike for 3 and 5, solid-state structural data on 7 could not be obtained. At ambient temperature, 1 H NMR spectroscopy on 7 revealed broadly similar features to those of 3 and 5, namely, six TMP-Me resonances. These could therefore reasonably be grouped by comparison with spectra of 3 and 5 (Fig. 15, ESI  Fig. S7d †). Additionally, 1 H, 7 Li-HOESY of 7 revealed a d( 1 H, 7 Li) ¼ (0.93, À4.25) ppm correlation presumably originating from TMP 3 -Me. However, rapid exchange of Li 1 and Li 2 was implied by the observation of not three but of only two 7 Li resonances (d 2.09 (br), À4.26 ppm) in a 2 : 1 integral ratio. More surprisingly, 7 Li, 7 Li-NOESY revealed exchange of Li 1 and Li 2 centers with phenyl-bound Li 3 (Fig. 16), implying that, overall, all Li sites are exchangeable in solution. 1 (Fig. 17). The rst and last pairs of correlations indicate conformational uxionality associated with each of TMP 1 and TMP 3 (i.e. ring inversion-type behaviour, as seen in 3 at 25 C). Whilst it is likely that TMP 2 also behaves in this way,  any potential correlations are obscured by the diagonal. The remaining pairs of correlations indicate that TMP 2 and TMP 3 undergo chemical exchange; this necessitates the involvement of a dissociative pathway. Most probably, exchange proceeds via a metallacyclic (TMPLi) 2 intermediate of a type well established in amidolithium chemistry. 46 Its symmetry would permit exchange of all Li centers and of TMP 2 and TMP 3 (but not TMP 1 , see Scheme 3). This thesis also explains the appearance of the 7 Li NMR spectrum of 7 if it is suggested that the two exchange paths illustrated proceed at different rates, with path B occurring faster and causing the broadness of the singlet at d 2.09 ppm. Upon cooling to À20 C, 1 H, 1 H-NOESY fails to show exchange correlations, whilst at À10 C only those associated with TMP 3 -Me eq /TMP 3 -Me ax interconversion are seen (see ESI, Fig. S7g and f †). Clearly therefore, TMP 2 /TMP 3 exchange is not established at either temperature. Similarly, 7 Li, 7 Li-NOESY at these temperatures fails to reveal exchange correlations (ESI, Fig. S7j and k †). Taken together, these data suggest that the exchange of inequivalent amido ligands and Li sites in 7 is mediated by a common intermediate, and this lends support to the mechanism proposed in Scheme 3.

Conclusions
Recently, the importance of Li-Cu interchange has been established in lithium amidocuprates. The relevance of this to their reactivity has now been substantiated through the observation of the dramatically different action of two metalinterchange isomers, metal amide aggregate (TMPCu) 2 -(TMPLi) 2 1 and Gilman lithium cuprate [TMPCu(m-TMP)Li] 2 2 2 . While isolated 2 2 proves incapable of metalating benzene, 1 is able to do so smoothly under mild conditions, yielding predominately Ph(TMP) 3 Cu 2 Li 2 3, a novel mixed organo(amido) cuprate aggregate. This reactivity is not limited to preformed 1: simple mixtures of TMPLi and TMPCu also exhibit this behaviour in benzene solution, and in so doing offer both ease of handling and greater possibilities in terms of the products accessible. Critically, the failure of either monometallic amide to efficiently metalate benzene conrms that ring metalation has a synergistic origin. Manipulating the reactant Cu : Li stoichiometry when combining monometallic amides leads to the additional formation of either Ph(TMP) 3 Cu 3 Li 5 or Ph(TMP) 3 CuLi 3 7, whose presence dominates in Cu-or Li-rich reaction mixtures, respectively. Additionally, the formation of 5 is accompanied by that of minor amounts of Ph(TMP) 3 Cu 4 6, which is believed to result from Cu-Li exchange between 5 and TMPCu. In the reaction pathway that leads to 3, 5 and 7, the generation of putative PhCu(TMP)Li 4 from reaction of a base of formula (TMP) 2 CuLi with benzene is suggested. The absence of 4 from the nal reaction mixtures is explained by its rapid in situ combination with metal amides. This notion is borne out by the observation that pre-isolated 4 equilibrates with 2, TMPCu or TMPLi to give 3, 5 and 7, and that these can subsequently be isolated independently of one another. Both solid-(3-5) and solution-state (3-5, 7) measurements expose Li/p interactions as a prominent structural feature and these interactions are reproduced satisfactorily by DFT calculations. Additionally, DFT analysis predicts the ground-state molecular conformations of 3-5, each of which was also exposed by crystallography. The relatively small differences in energy associated with the molecular conformers of each lead ostensibly to the emergence of exchange correlations in 1 H, 1 H-NOESY experiments, while Lirich 7 exhibits additional exchange that is suggested by 7 Li, 7 Li-NOESY and requires the dissociation of a lithium amide moiety. These results expose signicant new possibilities for functionalising weakly acidic aromatic hydrocarbons using easy-tohandle and readily accessible lithium and copper amide reagents. Investigations are currently underway to assess the suitability of different solvents for handling and applying 1 and the substrate scope of the deprotonation reaction, particularly as applied to hydrocarbons such as toluene and naphthalene, which could offer more complex regioselectivity. Meanwhile, further DFT studies have been initiated to probe the reaction pathway with a view to explaining the substantially different behaviour of 1 and 2 2 and to shed light on reaction intermediates accessible from both 1 and mixtures of TMPLi and TMPCu.

General synthetic and analytical details
Reactions were carried out under dry nitrogen, using double manifold and glove-box methods. Solvents were distilled off sodium (toluene) or a sodium-potassium amalgam (THF, Et 2 O, and hexane) immediately before use. 2,2,6,6-tetramethylpiperidine (TMPH) was purchased from Alfa Aesar and stored over molecular sieves (4Å). Other chemicals were used as received. obtained at 25 C (unless otherwise stated) using deuterated solvent stored over molecular sieves (3Å). For 1 H and 13 C, chemical shis were internally referenced to deuterated solvent and calculated relative to TMS. For 7 Li, an external reference was used (1 M LiCl in D 2 O). Chemical shis are expressed in d ppm. The following abbreviations are used: br ¼ broad, m ¼ multiplet, s ¼ singlet, sh ¼ shoulder.

Crystallographic details
For details of data collections see Table 1. Crystals were transferred from the mother liquor to a drop of peruoropolyether oil mounted upon a microscope slide under cold nitrogen gas. 48 Suitable crystals were attached to the goniometer head via a MicroLoop™, which was then centred on the diffractometer. Data were collected on a Bruker D8 Quest (Cu-Ka, l ¼ 1.54184Å) or a Nonius Kappa CCD diffractometer (Mo-Ka, l ¼ 0.71073Å), each equipped with an Oxford Cryosystems low-temperature device (T ¼ 180(2) K). Structures were solved using SHELXT, 49 and renement (based on F 2 , by the full-matrix least squares method) was performed using SHELXL. 50 Non-hydrogen atoms were rened anisotropically (for disorder, standard restraints and constraints were employed as appropriate) and a riding model with idealized geometry was employed for the renement of H-atoms. Data have been deposited with the Cambridge Crystallographic Data Centre as supplementary publications CCDC 1869336-1869340. †

Synthesis and characterisation of 1
To a stirred solution of TMPH (0.34 mL, 2 mmol) and Et 2 O (0.21 mL, 2 mmol) in toluene (2 mL) at À78 C was added n BuLi (1.6 M in hexanes, 1.25 mL, 2 mmol). The solution was warmed to room temperature and then added dropwise to a suspension of CuCl (0.1 g, 1 mmol) in toluene (1 mL) at À78 C. The mixture was warmed to room temperature whereupon a grey discolouration occurred. LiCl was removed by ltration to give a yellow solution, the storage of which at 5 C for 48 h yielded a small crop of colourless block-like crystals. Yield

Synthesis and characterisation of 3
PhCu(TMP)Li 4 (290 mg, 1 mmol) and (TMP) 2 CuLi 2 (350 mg, 1 mmol) were combined in toluene (6 mL) and heated to 50 C for ca. 24 h. The toluene was removed in vacuo, and the residue was dissolved in hexane (ca. 5 mL). This was ltered and the resulting colourless solution was stored at 5 C for 24

Synthesis and characterisation of 4
To a stirred solution of TMPH (0.68 mL, 4 mmol) in hexane/THF (8 mL/8 mL) was added n BuLi (1.6 M in hexanes, 2.5 mL, 4 mmol) at À20 C. The pale-yellow solution was warmed to room temperature and transferred to a stirred suspension of CuCl (0.4 g, 4 mmol) in hexane/THF (4 mL/4 mL) at À20 C. The mixture was warmed to room temperature and stirred for 15 min, whereupon a thick, cream-coloured suspension of TMPCu was formed. The solvents were removed in vacuo and the residue was treated with toluene (24 mL). Hot ltration of the resulting suspension to remove LiCl gave a tan coloured solution, which precipitated upon standing. PhLi was prepared by dropwise treatment of a solution of PhI (0.44 mL, 4 mmol) in hexane (6 mL) with n BuLi (1.6 M in hexanes, 2.5 mL, 4 mmol) at room temperature. The solvent and BuI were removed in vacuo to leave a white powder. TMPCu was re-dissolved by gentle heating and transferred to the residue of PhLi at room temperature, whereupon the PhLi dissolved to give a bright yellow solution. This was concentrated (ca.  19.5 (TMP-4). 7 Li NMR spectroscopy (194 MHz, 298 K, C 6 D 6 ): d 2.09 (s, 2Li), À4.26 (s, 1Li).

Conflicts of interest
There are no conicts to declare.