New Scandium Borylimido Chemistry: Synthesis, Bonding, and Reactivity.

We report a combined synthetic, mechanistic, and theoretical study of the first borylimido complex of a rare earth metal, (NacNacNMe2)Sc{NB(NAr'CH)2} (25, Ar' = 2,6-C6H3iPr2, NacNacNMe2 = Ar'NC(Me)CHC(Me)NCH2CH2NMe2). Thermolysis of the methyl-borylamide (NacNacNMe2)Sc(Me){NHB(NAr'CH)2} (18) generated transient imide 25 via rate-determining, first-order methane elimination (KIE ≈ 8.7). In the absence of external substrate, 25 underwent a reversible cyclometalation reaction (sp3 C-H bond addition to Sc═Nimide) with a methyl group of the NacNacNMe2 ligand forming {MeC(NC6H3iPrCH(Me)CH2)CHC(Me)NCH2CH2NMe2}Sc{NHB(NAr'CH)2} (21). In the presence of pyridine or DMAP, reversible sp2 C-H bond activation occurred, forming orthometalated complexes (NacNacNMe2)Sc{NHB(NAr'CH)2}(η2-4-NC5H3R) (R = H or NMe2). In situ reaction of 25 with HCCTol gives irreversible sp C-H bond activation under kinetic control, and with MeCCPh [2+2] cycloaddition to Sc═Nimide takes place. These reactions represent the first substrate activation processes for any metal-bound borylimide. The bonding in 25 and the mechanism and thermodynamics of the reactions have been studied using density functional theory (DFT), supported by quantum theory of atoms in molecules and natural bond orbital analysis. Although the borylimido and arylimido dianions studied here are formally isoelectronic and possess comparable frontier molecular orbitals, the borylimido ligand is both a better π-donor and σ-donor, forming stronger and shorter metal-nitrogen bonds with somewhat reduced ionicity. Despite this, reactions of these types of borylimides with C-H or C≡C bonds are all more exothermic and more strongly activating than for the corresponding arylimides. DFT calculations on model systems of differing steric bulk unpicked the underlying thermodynamic factors controlling the reactions of 25 and its reaction partners, and a detailed comparison was made with the previously described arylimido homologues.


■ INTRODUCTION
Transition metal terminal organoimido complexes of the type (L)MNR (R = alkyl or aryl; L = supporting ligand or ligand set) have given rise to a wealth of new chemistry over the past three decades. 1 Most of this interest has been centered around the activation and/or transformations of unsaturated or saturated substrates at the polar metal−nitrogen multiple bond itself (usually described as a σ 2 π 4 triple bond), mostly for early transition metal complexes of Groups 4 and 5. 2 The most notable exceptions to this general observation are the Schrock olefin metathesis catalyst families in which arylimido groups act as spectator ligands, 3 but also other examples are certain types of Ziegler-type olefin polymerization and ring-opening metathesis polymerization catalysts. 4 Compared with their Group 4 counterparts, well-established terminal organoimido complexes of the rare earth metals (Ln, defined as Group 3 and the lanthanides) are considerably more scarce, 1d,f and their successful synthesis, isolation, and structural characterization was only achieved within the past 7 years, first by Chen et al. (1, Figure 1). 5 Prior to this, the field was limited to dimeric complexes (for example, Hessen's 2, Figure 1 6 ) or higher-nuclearity clusters which contain bridging μ-imido ligands, because of the more ionic nature and larger radii of these metals. 1d Even now, almost all of the isolated rare earth imido complexes are either for scandium (e.g., Piers's 3 ( Figure  1) 7 or Cui's {N(PPh 2 NPh) 2 }Sc(NAr′)(DMAP) 2 (Ar′ = 2,6-C 6 H 3 i Pr 2 )) 8 or other small metals (e.g., in Anwander's 4 (M = Y, Lu)), 9 featuring ligands with pronounced steric profiles. Very recently, Schelter et al. reported a potassium-capped cerium-(4+) imide. 10 Although Chen's (NacNac NMe 2 )Sc(NAr′)(DMAP) (1, Nac-Nac NMe 2 = (Ar′)NC(Me)CHC(Me)NCH 2 CH 2 NMe 2 )) was the first isolated rare earth, terminal imide, evidence of the high propensity to reactivity of the LnNR moiety was provided by Mindiola et al. 's transient imide (PNP)Sc(NAr′) (PNP = N{2-P i Pr 2 -4-C 6 H 3 Me} 2 ) generated in situ from (PNP)Sc(Me)-(NHAr′) by methane elimination. 11 This compound activates the ortho C−H bonds of various pyridines py R via putative adducts of the type (PNP)Sc(NAr′)(py R ), ultimately giving an atom-economical route to substituted pyridines. 12 Nonetheless, with isolated 1 in hand, along with a DMAP-free homologue featuring a tetradentate version of NacNac NMe 2 , 13 Chen et al. have been able to quite comprehensively map out the reactivity of the ScNAr′ functional group with E−H (E = C, B, H) and other saturated bonds, and polar and nonpolar unsaturated substrates. A fairly clear picture of the similarities and differences between ScNAr′ and the more widely studied Group 4 organoimido counterparts is starting to emerge.
Our group has been interested in developing early transition metal-imido and related metal−ligand multiple bond chemistry, 2a,b,d,4b,14 and during the course of these studies we recently reported the serendipitous formation of a very rare 15 titanium borylimido compound 5 ( Figure 2) 16 through the reductive N−N bond cleavage reaction of the hydrazide Cp*Ti{MeC-(N i Pr) 2 }(NNR 2 ) (R = Me or Ph) with 9-BBN. Prior to this, the only other terminal Group 4 borylimide known was Mindiola's 6, formed from the parent imide (i.e., terminal NH) with 2 equiv of NaHBEt 3 . 15a Borylimides of Groups 5 and 6 are also rare ( Figure 2). Wilkinson reported the first borylimido complex of any transition metal (7) 17 by the oxidative addition reaction of WCl 2 (PMe 3 ) 4 with 2 equiv of Mes 2 BN 3 ; comparable routes were used to form 8 and 9 from W(4+) and V(3+) precursors, respectively. 18 Dinitrogen ligand N−N bond cleavage reactions with boranes have also given dinuclear borylimido derivatives, including Fryzuk 's 10. 19 To date there have been no reports of the reactivity of any transition metal borylimido complex with small molecules. Elegant work by Bettinger et al. since 2006, 20 as well as an earlier report from Paetzold, 21 has shown that heteroatomstabilized borylnitrenes such as NB(N i Pr 2 ) 2 , and, in particular, NBCat (Cat = 1,2-O 2 C 6 H 4 ) can be photochemically generated from the corresponding azides X 2 BN 3 , trapped in frozen argon matrices, and shown to undergo coupling reactions with CO and N 2 and insertion reactions into C −H, B−H, H−H, and B− N bonds. 20,22 Computational studies of the hypothetical 20valence-electron (CO) 4 FeNBCat and (CO) 4 FeNBH 2 were also reported, with NBCat being found to be a better 4-electron donor to the iron center than BNH 2 because 2pπ(O)→2pπ(B) donation reduces the boron atom's π-acidity with respect to the nitrogen atom 's lone pair. 22a Given the current interest and novelty of rare earth organoimido compounds, and the absence of any borylimido complexes of these metals, or reactivity studies of any metalbound NBX 2 moiety, we were interested to develop new scandium borylimido chemistry. None of the previously reported routes to borylimides (i.e., hydrazido N−N bond cleavage, oxidative addition to lower oxidation state precursors, transformation of a parent imide) are suitable for these metals because either the necessary starting material does not yet exist (Group 3 terminal hydrazide or parent imide), or are rather rare (1+ oxidation state), and so we report new, alternative routes based on hydrocarbon elimination protocols. In addition, building on Bettinger's initial bonding analysis of the hypothetical 20-valence-electron (CO) 4 FeNBCat and (CO) 4 FeNBH 2 compounds, we describe a molecular orbital and quantum theory of atoms in molecules (QTAIM) study of scandium aryl-and boryl-imido complexes as well as comparative combined experimental and computational studies of the C−H bond activation and alkyne coupling reactions of these types of imido complex. Figure 1. Selected bridging and terminal rare earth imido complexes (Ar′ = 2,6-C 6 H 3 i Pr 2 ). [5][6][7]9 Figure 2. Some previously reported terminal borylimido complexes. 15a, 16−19 Journal of the American Chemical Society Article ■ RESULTS AND DISCUSSION Entry Points to Scandium Borylimido Chemistry. Given that Chen et al. have already demonstrated the effectiveness of donor-functionalized "NacNac"-type ligands, namely NacNac NMe 2 (cf. 1 in Figure 1) and a tetradentate homologue MeC(NAr′)CHC(Me)NCH 2 CH 2 N(Me)-CH 2 CH 2 NMe 2 , in accessing new scandium arylimido chemistry we chose to use these same supporting platforms. Likewise, drawing upon Bettinger's suggestion 22a that heteroatomfunctionalized NBX 2 (X = heteroatom group) moieties may be better overall donors than their NBR 2 (R = σ-only group) analogues we targeted the two borylamines H 2 NB(N i Pr) 2 C 6 H 4 (11) and H 2 NB(NAr′CH) 2 (12, Ar′ = 2,6-C 6 H 3 i Pr 2 ) 23 as entry points to scandium borylimido chemistry. We reasoned that a borylimido ligand of the type NB(N i Pr) 2 C 6 H 4 formally derived from 11 would be sterically and electronically similar to the NAr′ moieties used in the work of Chen, Mindiola, Piers, and others, and also related to Bettinger's reactive NBCat (we did not target H 2 NBCat itself because of the poor steric protection offered by the 1,2-O 2 C 6 H 4 moiety). Borylamine 12, with its large flanking N−Ar′ groups provides even greater steric protection than the N− i Pr substituents of 11. DFT studies of model compounds found that there was no significant electronic effect on ScNBX 2 bonding upon replacing the 1,2-C 6 H 4 moiety in 11 (chosen for synthetic expedience) with the simple 1,2-C 2 H 2 linkage in 12.
Our initial attempts to access a scandium borylimide focused on H 2 NB(N i Pr) 2 C 6 H 4 (11) as described fully in the Supporting Information (SI). Reaction of (NacNac NMe 2 )ScCl 2 (13) with 2 equiv of MeLi in THF, based on previous protocols, 5 formed (NacNac NMe 2 )ScMe 2 (14). In situ reaction of 14 with 1 and 2 equiv of 11 produced the methyl-borylamide (NacNac NMe 2 )Sc-(Me){NHB(N i Pr) 2 C 6 H 4 } (15) and the bis(borylamide) (NacNac NMe 2 )Sc{NHB(N i Pr) 2 C 6 H 4 } 2 (16), respectively. Thermolytic alkane elimination from primary amido-alkyl compounds is a well-established protocol for the synthesis of imido compounds for both the rare earths (vide supra) and the early transition metals. 1a,b,2a,c Disappointingly, thermolysis reactions of 15 and 16 on the NMR tube scale formed complex mixtures of unidentified products. In a reaction on the preparative scale, although no single product could be properly isolated, some diffraction-quality crystals of [{MeC(NAr′)CHC(CH 2 )-NCH 2 CH 2 NMe 2 }Sc{NHB(N i Pr) 2 C 6 H 4 }] 2 (17) were obtained. This is a dimeric complex formed by intermolecular C−H bond activation of the methyl group at the β-position of the NacNac ligand by a second scandium center (see Figure S2 of the SI).
In an attempt to direct the reactivity of these borylamido systems we turned to the more sterically encumbering borylamine H 2 NB(NAr′CH) 2 (12). 2 3 Reaction of (NacNac NMe 2 )ScMe 2 (14) with 1 equiv of H 2 NB(NAr′CH) 2 (12) for 60 h at room temperature gave the methyl-borylamide (NacNac NMe 2 )Sc(Me){NHB(NAr′CH) 2 } (18) which was isolated as an off-white powder in 53% yield (Scheme 1). This reaction time is significantly longer than that required for quantitative formation of 15 and Chen's (NacNac NMe 2 )Sc-(Me)(NHAr′) (19) (16 and 24 h at room temperature, respectively), 5 which may reflect the likely steric hindrance in the approach of the bulkier borylamine to the scandium center. When followed by 1 H NMR spectroscopy on the NMR tube scale, the formation of 18 was quantitative, and CH 4 evolution was observed (signal at 0.16 ppm). The 1 H NMR spectrum of 18 displays an Sc−Me resonance at −0.90 ppm and a broad NH singlet at 3.90 ppm. The 11 B resonance of the borylamido ligand is 23.3 ppm. Scheme 1. Synthesis of (NacNac NMe 2 )Sc(X){NHB(NAr′CH) 2 } (X = Me (18) or Cl (20))

Journal of the American Chemical Society
Article Diffraction-quality crystals of 18 were grown from a benzene solution at room temperature. The solid-state structure is shown in Figure 3, along with key bond distances and angles.
Additionally, while we prepared a borylamide chloride analogue of 18, (NacNac NMe 2 )Sc(Cl){NHB(NAr′CH) 2 } (20), through reaction of (NacNac NMe 2 )ScCl 2 (13) with 1 equiv of the lithium borylamide LiNHB(NAr′CH) 2 on the NMR tube scale (Scheme 1), we were unable to convert it into a borylimido complex in situ. Full details are provided in the SI. Other attempts, all ultimately unproductive, to enter borylimido compounds through the use of CH 2 SiMe 3 or CH 2 CMe 3 groups are described in the SI, along with additional experiments using a tetradentate version of NacNac NMe 2 .
Thermolysis Reactions of (NacNac NMe 2 )Sc(Me){NHB-(NAr′CH) 2 } (18). With and without Lewis Bases. Thermolysis screening reactions of 18 were performed on the NMR tube scale in C 6 D 6 , both in the presence of 1 equiv of pyridine or DMAP, as well as without a Lewis base. After ∼16 h at 60°C, each reaction gave quantitative conversion to a different new species, together with the evolution of methane. The three products were each isolated on the preparative scale (Scheme 2).

Journal of the American Chemical Society
Article as a broad singlet at 2.83 ppm in the 1 H NMR spectrum, and the diastereotopic hydrogens of the metalated methylene group (ScCH 2 ) appear as an apparent triplet (intensity 1 H) at 0.22 ppm and a doublet of doublets (intensity 1 H) at −0.01 ppm due to vicinal coupling to the adjacent methine (CHMe) in addition to geminal coupling. The Sc-bound carbon atom gives a 13 C signal at 32.0 ppm, and the 11 B NMR spectrum shows a signal at 23.3 ppm.
Diffraction-quality crystals of 21 were grown from a benzene solution at room temperature. The solid-state structure is shown in Figure 4, along with key bond distances and angles.
Complex 21 possesses a distorted square-based pyramidal geometry (τ = 0.35 and 0.37 for the two crystallographically independent molecules) and is similar to that of 18 in which one vertex of the basal plane is now a cyclometalated methylene carbon (C(40)). This forms a new six-membered metallacycle with a distorted boat conformation. The borylamide group is again positioned at the apical site of the square pyramid. The average Sc(1)−N(1) bond length of 2.063(3) Å is the same as that in 18 (2.0629(8) Å). The average Sc(1)−C(40) distance of 2.248(2) Å is slightly shorter than for the Sc−Me bond in 18 (2.2730(10) Å) which may be due to the chelating nature of this moiety. 21 is the analogue of Chen's anilide {MeC-(NC 6 H 3 i PrCH(Me)CH 2 )CHC(Me)NCH 2 CH 2 NMe 2 }Sc-(NHAr′) (22), obtained by abstraction of DMAP from 1. 25 Mechanistic and computational (DFT) aspects of the formation of 22 and 21 are discussed later on.
The scaled-up thermolysis reactions of 18 in the presence of pyridine or DMAP at 60°C gave the pale yellow, orthometalated products (NacNac NMe 2 )Sc{NHB(NAr′CH) 2 }(η 2 -4-NC 5 H 3 R) (R = H (23), NMe 2 (24)) in 56 and 85% isolated yields, respectively (Scheme 2). Although diffraction-quality crystals could not be obtained, their 1 H and 13 C NMR spectra confirmed that they each have a borylamido ligand and an η 2bound pyridyl ligand or its DMAP counterpart, apparently formed from C−H bond activation at the 2-position of the heterocycle. For example, the 1 H NMR spectrum of 23 has four mutually coupled multiplets for the η 2 -pyridyl ligand at 6.89, 6.80 (overlapping), and 6.48 ppm, and a broad NH singlet at 3.12 ppm for the borylamide ligand. The 13 C NMR resonance for the Sc-bound pyridyl carbon atom appears at 217.2 ppm. These spectral features are comparable to those of Mindiola's crystallographically characterized pyridyl-anilido complex (PNP)Sc(NHAr′)(η 2 -NC 5 H 4 ) formed by reaction of the methyl-anilide (PNP)Sc(Me)(NHAr′) with pyridine at room temperature. 11 DFT calculations on 23 (vide inf ra) support the η 2 -coordination of the pyridyl ligand in this complex (discussed later). The 1 H and 13 C NMR spectra of the DMAP counterpart, 24, are analogous to those of 23.
In a further attempt to trap the putative transient imide 25 (see inset to Scheme 2) formed (vide inf ra) en route to the C− H activation products 21−24, thermolysis of 18 was carried out in the presence of THF and 2,6-dimethylpyridine at 60°C in C 6 D 6 . Unfortunately, in each case this resulted only in quantitative formation of 21. Therefore, under the conditions necessary to effect methane elimination from 18, these weaker Lewis bases offer insufficient kinetic and/or thermodynamic stabilization against formation of 21. Furthermore, it was also found that further heating of 23 for 16 h at the higher temperature of 90°C in C 6 D 6 resulted in conversion to the cyclometalated complex 21 and free pyridine. Labeling and DFT studies below support the mechanism shown in Scheme 3. This conversion from the kinetic product 23 (formed at 60°C) to the thermodynamic product 21 and pyridine (at 90°C) is favored by the entropic gain arising from the elimination of pyridine at higher temperatures. The equivalent process does not take place for the DMAP analogue, 24, for which heating at temperatures above 60°C on the NMR tube scale results in Scheme 3. Conversion of (NacNac NMe 2 )Sc{NHB(NAr′CH) 2 }(η 2 -NC 5 H 4 ) (23) to {MeC (NC 6 H 3 i PrCH(Me)CH 2 )CHC(Me)NCH 2 CH 2 NMe 2 }Sc{NHB(NAr′CH) 2 } (21)

Journal of the American Chemical Society
Article conversion to a complex mixture of products which cannot be identified.
As mentioned, Chen recently reported the anilido analogue of 21 (i.e., 22) by reaction of 1 with t BuCH 2 CH 2 BC 8 H 14 , at room temperature or 50°C. This reaction is proposed to proceed via transient (NacNac NMe 2 )Sc(NAr′). 25 It was also reported that heating (NacNac NMe 2 )Sc(Me)(NHAr′) (19) itself (the analogue of 18) gave no reaction after 48 h at 70°C, which is in contrast to our observation of the ready conversion of 18 to 21 and methane at 60°C within several hours. In addition, it was not clear from the published work whether 1 could undergo further reaction on heating, converting for example to the anilido analogue of 24 or to 22. To avoid mechanistic uncertainties in connection with the DFT studies reported later on in this paper, we therefore independently prepared 19 and 1 and studied their thermolysis chemistry. Despite extended heating at 90°C in C 6 D 6 , 1 was found to be thermally stable. In contrast, however, heating 19 at 70°C for 8 h gave complete consumption of the anilide and formation of 22 as the major product. The half-life for the decay of 19 was found to be ∼1.1 h, which is comparable to the half-life of 18 in conversion to 21 under the same conditions (∼1.2 h).
Mechanism. On their own, the thermolysis reactions of 18 (Scheme 2) do not establish that the mechanism of formation of 21, 23, and 24 proceeds via an imido intermediate such as 25 (inset to Scheme 2) at all. Mechanisms proceeding via direct σbond metathesis at the Sc−Me bond in systems of this type are precedented. 26 Furthermore, for the reactions with py or DMAP, the data do not establish whether or not Lewis base assistance occurs, again as found in previous systems. 26b These three thermolysis reactions of 18 were monitored by 1 H NMR spectroscopy at 343 K in duplicate for at least three half-lives (>85% consumption of 18). All followed first-order kinetics as judged by linear semilogarithmic plots of −ln([18]/ [18] 0 ) vs time (see Figure S3 of the SI). The overall first-order rate constants (k) for all three reactions are listed in Table 1. All three reactions have comparable k values, implying that they share a common rate-determining step. The very similar rate constants for pyridine and DMAP (DMAP being a much stronger nucleophile than pyridine 27 ) show that base-assisted methane elimination is not involved in a rate-determining step, in contrast to the corresponding reaction of related systems and specifically the anilido analogue of 18. 5 Deuterium labeling was used to gain additional information on the mechanism. Selectively ND-labeled (NacNac NMe 2 )Sc-(Me){NDB(NAr′CH) 2 } (18-d 1 ) was prepared using D 2 NB-(NAr′CH) 2 (12-d 2 ). The 1 H NMR spectrum of the labeled borylamide is identical to that of 18 except for the absence of an NH singlet. Thermolysis of 18-d 1 gave conversion exclusively to 21 containing a hydrogen (relative intensity 1 H) at the borylamide (NH) site, with the N-bound deuterium of 18-d 1 being eliminated as MeD to apparently form the imide 25 prior to the intramolecular C−H bond activation. An alternative mechanism 26b in which the metalation occurs by σbond metathesis between Sc−Me and the isopropyl C−H bond would necessarily release CH 4 and form a species (21-d 1 ) containing a ScNDB(NAr′CH) 2 moiety.
The corresponding thermolysis reaction of 18 in the presence of pyridine-d 5 was performed under the same conditions. The product was found to be exclusively (NacNac NMe 2 )Sc{NDB(NAr′CH) 2 }(η 2 -NC 5 D 4 ) (23-d 5 ), and a singlet assigned to CH 4 (0.16 ppm) was observed in the 1 H NMR spectrum of the reaction mixture. The 1 H NMR spectrum of 23-d 5 is identical to that of 23, other than the absence of the borylamide NH singlet and all of the pyridyl multiplets. Consistent with Scheme 3, thermolysis of 23-d 5 at 90°C in C 6 D 6 or C 6 H 6 formed exclusively 21 and pyridine-d 5 , with no incorporation of hydrogen into the eliminated Lewis base.
In addition, the rate of formation of 21 by thermolysis of 18d 1 was found to be significantly slower than that of 18, with a kinetic isotope effect (KIE) at 343 K of 8.7 ± 0.5 (Table 1, average of two independent values). The KIE value obtained here, albeit large compared to classical organic systems, is very comparable to those reported by Bergman, Wolczanski, and others for α-abstraction reactions in Group 4 and 5 systems which release methane or other alkanes to yield organoimido functional groups, 28 as well as a number of other examples. 28h,i The magnitudes of these isotope effects, their origins, and any contributions (or not) from nonclassical mechanisms have been discussed in these previous papers and reviews, 28h,i and the reader is referred to them for further discussion. The observed KIE value reported here is indicative of a process having a transition state (TS) in which the transfer of H is relatively linear. 29 We discuss the geometry of the DFT computed TS later on in this paper. In contrast, thermolysis of 18 in the presence of pyridine-d 5 gave no significant rate difference to the reaction with protio-pyridine.

Journal of the American Chemical Society
Article the ortho-metalated isomers, unlike in the case of Chen's arylimido analogue of 24_int, (NacNac NMe2 )Sc(NAr′)(DMAP) (1). These results contrast with the base-assisted thermolysis of (NacNac NMe2 )Sc(Me)(NHAr′) in the presence of DMAP to form the stable 1. 5 In that case an adduct (NacNac NMe2 )Sc-(Me)(NHAr′)(DMAP) is observed prior to formation of 1, whereas in the case of 18 none is seen. Related C−H bond activation reactions of coordinated pyridine by a scandium terminal imide have been reported by Mindiola and co-workers, 11,12 as have a number of similar reactions by other early transition or rare earth metal systems. 30 Heating (PNP)Sc(Me)(NHAr′) in the presence of pyridine generates a transient scandium imide (PNP)Sc(NAr′)(py) which then (rapidly) activates a C−H bond of pyridine to form the pyridyl complex (PNP)Sc(NHAr′)(η 2 -NC 5 H 4 ). Mechanistic studies showed that pyridine coordinates to (PNP)Sc(Me)-(NHAr′) before α-abstraction. 11b Significantly, the reaction with DMAP was almost 2 orders of magnitude faster than that with pyridine. This contrasts with the formation of 23 and 24, which have comparable rates because pyridine and DMAP coordinate only after the rate-determining, α-abstraction step. The mechanistic difference with regard to base assistance between the systems of Chen and Mindiola and the one reported here is proposed to have a steric basis, and DFT studies (vide inf ra) support this conclusion.
As mentioned, 23 and 23-d 5 can be converted to 21 and py or py-d 5 , suggesting a number of accessible equilibria, as indicated in Scheme 4. Consistent with this it was found that addition of 1 equiv of pyridine-d 5 to 23 on the NMR tube scale in C 6 D 6 resulted in an equilibrium mixture of 56% 23 and 44% 23-d 5 after heating at 60°C for 2 h, indicative of a modest equilibrium isotope effect and showing that interconversion between 23_int and 23 is facile under the experimental conditions required to form the latter from 18 (10 h at 60°C under the same reagent concentrations).
The overall equilibrium isotope effect observed in the intermolecular exchange reaction of pyridine-d 5 with 23 involves a series of interconverting, individual equilibria between 23 and 23_int (and their isotopomers) as well as 25 (assuming that exchange between 23_int and free pyridine (or their isotopomers) is a dissociative process). In order to isolate the specific C−H(D) bond activation step between 23_int and 23, an intramolecular competition reaction was performed through thermolysis of 18 in the presence of 2-pyridine-d 1 (Scheme 5). After 8 h at 60°C in C 6 D 6 this resulted in an isotopomer mixture of 60% (NacNac NMe 2 )Sc{NHB-(NAr′CH) 2 }(η 2 -6-NC 5 H 3 D) (23-d 1 _C) and 40% (NacNac NMe 2 )Sc{NDB(NAr′CH) 2 }(η 2 -NC 5 H 4 ) (23-d1_N). The equilibrium isotope effect of 1.5 is comparable with the value found in the intermolecular competition reaction. The Trapping of a Transient Scandium Borylimide with Alkynes. To further probe the transient borylimide 25, we performed trapping reactions starting from the methylborylamide 18 in the presence of various terminal and internal alkynes. On the NMR tube scale, the reaction of 1 equiv of either HCCTol or MeCCPh with 18 at 60°C gave quantitative conversion to a single new product in each case after 8 h. On the preparative scale, the reaction with HCCTol in toluene gave the terminal acetylide (NacNac NMe 2 )Sc{NHB(NAr′CH) 2 }-(CCTol) (26) as an off-white powder in 50% yield (Scheme 6). The corresponding reaction with MeCCPh gave the azascandacyclobutene (NacNac NMe 2 )Sc{N{B(NAr′CH) 2 }C-(Me)C(Ph)} (27) as a yellow powder in 48% yield. Compounds 26 and 27 represent the products of alkyne C− H bond activation (1,2-addition) or [2+2] cycloaddition across the Sc−N multiple bond of transient 25. Attempted reactions with two further internal alkynes were unsuccessful. NMR tube scale thermolysis in the presence of MeCCMe resulted in a mixture of unknown products, whereas thermolysis in the presence of PhCCPh gave only the cyclometalated complex 21, the bulkier internal alkyne being too sterically demanding to compete with the intramolecular bond-activation process.
The 1 H and 13 C NMR spectra of 26 are consistent with the presence of an acetylide group in addition to a borylamido ligand. The NH singlet is found at 3.58 ppm in the 1 H NMR spectrum, and the two acetylide carbons give 13 C resonances at 155.4 and 124.6 ppm for the carbons bonded to the metal and to the p-tolyl group, respectively. The 13 C NMR spectrum of 27 is consistent with the azascandacyclobutene unit having the regiochemistry illustrated in Scheme 6, with the metal-bound ScC(Ph) carbon appearing at 172.9 ppm, and the CCMe counterpart at 154.5 ppm. Diffraction-quality crystals of complex 27 were grown from a benzene solution. The solid-state structure is shown in Figure 5, along with key bond distances and angles.

Journal of the American Chemical Society
Article and co-workers by abstraction of the DMAP ligand from 1 using 9-BBN in the presence of MeCCPh. 25 The metallacycle N(1)−Sc(1)−C(48) angle of 69.92 (13)°in 28 is comparable to that in 27. The N(1)−C(49) and metallacycle C(49)− C(50) bond lengths are also comparable and consistent with N−C single and C−C double bonds. A related scandium anilide-acetylide, ( tBu NacNac)Sc(NHAr′)(CCPh), was recently prepared by Piers and co-workers via a C−H bond activation reaction of HCCPh with 3 ( Figure 1) at 70°C and was crystallographically characterized. 7 Mechanistic studies showed that, like Chen's 28, this also formed after initial dissociation of the DMAP ligand.
The reactions of 18 with HCCTol or MeCCPh were monitored by 1 H NMR spectroscopy at 343 K, as for the previous thermolysis experiments. Both followed first-order kinetics (see Figure S4 of the SI for semilogarithmic plots of −ln([18]/[18] 0 ) vs time). The observed first-order rate constants are listed in Table 2. These values are similar to each other and to those in Table 1, consistent with the bondforming events for 26 and 27 occurring after initial α-hydrogen abstraction to release methane as the rate-determining step. As expected, reaction of 18 with DCCTol gave elimination of CH 4 and exclusive incorporation of deuterium at the borylamide site in (NacNac NMe 2 )Sc{NDB(NAr′CH) 2 }(CCTol) (26-d 1 ), and no KIE was observed between the reactions of HCCTol and DCCTol. A competition reaction between 18 and a mixture of HCCTol and DCCTol (1:5:5 ratio, so as to determine any isotope effect in the alkyne C−H bond activation reaction of transient 25) was not clean, probably due to competing side reactions in the presence of an excess of terminal alkyne.
In a competition reaction of 18 with HCCTol and MeCCPh (1:5:5 molar ratio) at 60°C in C 6 D 6 , quantitative conversion to the acetylide complex 26 was observed, indicating that the C− H bond activation of HCCTol by 25 is at least 2−3 orders of magnitude faster than [2+2] cycloaddition with MeCCPh. In a separate experiment, addition of HCCTol to 27 gave no significant reaction even at 90°C. Addition of 1 equiv of MeCCPh to 26 and heating up to 90°C in C 6 D 6 likewise gave no exchange with the added alkyne, and 26 appeared to be thermally stable at this temperature. Therefore, although HCCTol reacts faster than MeCCPh with the proposed transient imide 25, we have no information on their relative thermodynamic stabilities, or on whether a hypothetical [2+2] cycloaddition product (not seen with HCCTol) is thermodynamically competitive with 26.
Trapping of the base-free borylimide 25 by MeCCPh was also achieved from the cyclometalated 21. When heated at 100°C in the presence of 1 equiv of MeCCPh, conversion to 27 took place over 16 h to reach completion. At this temperature 21 is therefore able to re-form the borylimide 25 before being trapped. In contrast, heating 23 in the presence of 1 equiv of MeCCPh at 60°C did not result in conversion to 27, despite earlier exchange reactions showing that the reaction of pyridine with 25 is reversible at this temperature. In an analogous competition reaction of (NacNac NMe 2 )Sc(Me){NHB-(NAr′CH) 2 } (18) between pyridine and MeCCPh (1:5:5) at 60°C, exclusive formation of 23 was also observed. It appears

Journal of the American Chemical Society
Article that at 60°C, the reactions of 25 with pyridine and MeCCPh are under kinetic control. At 90°C, the system comes under thermodynamic control, giving quantitative conversion of 23 to 27 over 7.5 h.
Reaction of cyclometalated 21 with HCCTol to form 26 also proceeded on the NMR tube scale in C 6 D 6 but at the lower temperature of 80°C and within 5 h, suggestive of a different mechanism. Repeating this reaction with DCCTol indicated incorporation of deuterium into the isopropyl methine group of 26, and the NMR spectra showed an NH (as opposed to the expected ND) moiety in the product. These observations are consistent with a σ-bond metathesis reaction as opposed to reforming the transient borylimide, 25.
DFT Studies. Electronic and Molecular Structure. To gain a full understanding of the similarities and differences between the bonding and reactivity characteristics of borylimido compounds and their isoelectronic arylimido analogues, we carried out a detailed DFT study involving all of the experimental systems discussed above. Three closely related series of quantum chemical models were required to interrogate properly the various aspects. Table 3 summarizes how the imido/amido and NacNac NMe 2 ligand substituents were handled in each case, together with the labeling scheme adopted to help the reader follow the discussion.
In order to extract underlying electronic trends in the absence of significant steric effects and dispersive interactions, the sterically minimized series prefixed "1q" (borylimido systems) or "2q" (arylimido systems) were used in which the imido/amido groups were treated as NB(NHCH) 2 and NPh.
The supporting ligand was approximated as Ph NacNac NMe 2 in which the real 2,6-C 6 H 3 i Pr 2 group has been replaced with Ph. For mechanistic and thermodynamic studies, two DFT series with labels prefixed as "Q" and "Q*" were used. The "Q" series are identical to the experimental systems, except that the 2,6-C 6 H 3 i Pr 2 substituents of the imido/amido groups are represented by 2,6-C 6 H 3 Me 2 (Xyl): this gives systems which are sterically quite close to the real ones but yet small enough to evaluate for TSs and mechanistic pathways at an appropriate level of DFT. The "Q*" series have the full experimental substituents and were used to test the full thermodynamic effects of the steric bulk in these very large systems. Details of the calculations are given in the SI. By way of example, the sterically minimized borylimide pyridine adduct ( Ph NacNac NMe 2 )Sc{NB(NHCH) 2 }(py) (1q_E_py) and the full model of the experimentally inferred, base-free borylimide (NacNac NMe 2 )Sc{NB(NAr′CH) 2 } (1Q*_C) are illustrated in Figure 6.
Turning first to the electronic structure of the new scandiumborylimido compounds we compared the Sc−N imide bonding in the model compounds ( Ph NacNac NMe 2 )Sc{NB(NHCH) 2 }(py) (1q_E_py) and ( Ph NacNac NMe 2 )Sc(NPh)(py) (2q_E_py). Compound 2q_E_py is a model of Chen's (NacNac NMe 2 )Sc-(NAr′)(DMAP) (1), and 1q_E_py is a model of the presumed intermediates formed en route (cf. Scheme 4) to the metalated pyridine complexes 23 and 24. The geometries of 1q_E_py ( Figure 6) and 2q_E_py are very similar, as expected. Table 4 summarizes key electronic parameters regarding the electronic structures of these systems which we discuss later on. Figure 7   (

Journal of the American Chemical Society
Article shows the πand σ-type three highest energy frontier orbitals of the [NB(NHCH) 2 ] 2− dianion, and Figure 8 shows the Sc−N πand σ-bonding molecular orbitals of 1q_E_py. The σand πtype MOs of [NPh] 2− and the frontier MOs of 2q_E_py are given in Figures S6 and S8 of the SI, and their energies are summarized for convenience in Figures 7 and 8, respectively. The HOMO of [NB(NHCH) 2 ] 2− is a nitrogen 2p AO (denoted π h ) lying in the horizontal plane of the B(NHCH) 2 ring. The HOMO−1 (denoted π v ) lies in a vertical orientation and is a nitrogen 2p AO stabilized relative to π h by delocalization to the 2p AO of boron. However, this latter orbital is also involved in π-bonding with the two adjacent N(H) atoms and in π-bonding within the aromatic B(NHCH) 2 ring (cf. Figure S5 of the SI) in general. The electropositive nature of B makes MOs containing contributions from this element less stable. Therefore, although [NB(NHCH) 2 ] 2− is isoelectronic with [NPh] 2− , the Ph ring in the latter better stabilizes π v as judged by the larger HOMO-(HOMO−1) separation in [NPh] 2− (0.68 eV) compared to 0.32 eV in [NB(NHCH) 2 ] 2− . Because of the electropositive nature of boron compared to carbon, the average energy of the HOMO and HOMO−1 of the borylimide dianion is 0.65 eV higher (less stable), suggesting that this borylimide is a better π-donor than phenylimide. This effect is amplified in the HOMO−2 of the anions which are the σ-donor orbitals. Here the borylimide orbital is ca. 2 eV less stable, suggesting an even greater improvement in σ-donor ability. This molecular orbital treatment is supported by a Natural Bond Orbital (NBO) 31 analysis of the anions ( Figure S7 of the Supporting Information) which find the same trends in orbital energies, including a greater degree of π v delocalization in [NPh] 2− and more 2p atomic orbital character in the σ-donor natural localized molecular orbital (NLMO) of [NB(NHCH) 2 ] 2− .
The Sc−N imide π-MOs of 1q_E_py shown in Figure 8 (the σbonding counterpart (not shown) is mainly in HOMO−6) are consistent with a ScN triple bond (σ 2 π 4 configuration) as expected. The frontier MOs of 2q_E_py are similar in general terms ( Figure S8 of the SI) and resemble those reported by Chen. 5 The frontier MOs for the borylimide are destabilized relative to these by ca. 0.6 eV on average because of the different properties of the B(NHCH) 2 ring. The Sc−N π v bonding is distributed between the HOMO and HOMO−3 of 1q_E_py and 2q_E_py because of additional π-interactions (antibonding and bonding, respectively) with the B(NHCH) 2 or Ph rings. The bonding in 1q_E_py and 2q_E_py is quite ionic with only ca. 6−12% Sc contribution to HOMO, HOMO−1, and HOMO−3 in the Mulliken population analysis.
Further insight into the relative bonding capabilities of the borylimide and arylimide systems was gained from the data summarized in Table 4 which include the results of a quantum theory of atoms in molecules (QTAIM) 32 analysis of 1q_E_py and 2q_E_py. These data are again indicative of quite ionic systems, with large QTAIM atomic charges (Q) for Sc and sums of charges (∑Q) for the imido and Ph NacNac NMe 2 ligands. The relative ∑Q values reflect the formally di-and monoanionic nature of the respective groups. The comparatively small magnitudes of the electron density (ρ) and total energy density (H) at the bond critical points (BCPs), and the positive value of the electron density Laplacian (∇ 2 ρ), all imply covalent, very polar bonds. The Mayer bond indices of 1q_E_py and 2q_E_py are comparable to each other and to the value of 1.56 reported by Mindiola et al. for Sc(PNP)-(NAr′)(py). 11b The QTAIM delocalization indices (δ, which relates to the relative bond order for similar atom-pairs) are likewise comparable. The data also point to the NB(NHCH) 2 ligand being a stronger donor than NPh. The larger magnitudes of ρ and H, and less positive ∇ 2 ρ, demonstrate more covalency in the Sc−N imide bond of 1q_E_py, and the less negative ∑Q(NR) value for NB(NHCH) 2 is consistent with this. Note that the reduced ∑Q(NR) value is accompanied by a more negative ∑Q( Ph NacNac NMe 2 ) due to the imido group's  . DFT computed Sc−N imide π-type molecular orbitals of ( Ph NacNac NMe 2 )Sc{NB(NHCH) 2 }(py) (1q_E_py) with the values for ( Ph NacNac NMe 2 )Sc(NPh)(py) (2q_E_py) in parentheses. Isosurfaces are drawn at the 0.0360 au contour level.

Journal of the American Chemical Society
Article labilizing effect on the other ligands present (the ∑Q value for pyridine in 1q_E_py is also slightly more negative than in 2q_E_py). The larger Mayer bond order and δ value for 1q_E_py follow this trend in donating ability. The value of ε (ellipticity at the BCP) for 1q_E_py is lower than in 2q_E_py because the two π-components of the Sc−N imide bond are more similar (cf. the difference in energy of the HOMO and HOMO−1 of the imide dianions in Figure 7, which follows from the Ph ring being a better π-acceptor than B(NHCH) 2 ).
The shorter Sc−N imide bond length for 1q_E_py compared with 2q_E_py is not unexpected given the electronic structure analysis. The value of 1.835 Å in the latter is less than in the real compound 1 (1.881(5) Å) due to the different steric bulk in the two systems. However, it is also important to comment on the very "soft" nature of the Sc−N imide bonds (shallow potential energy wells) in 1q_E_py and 2q_E_py. We found that adjusting these bonds by ±0.04 Å gave changes in SCF energy of less than 1 kcal mol −1 . Despite this, there is a significant difference in the overall Sc−N imide bond dissociation energy (ΔE BD ) to geometry-relaxed fragments for the two systems. As shown in Table 4, the difference (ΔΔE BD ) in these values for the two compounds is 10.7 kcal mol −1 in favor of 2q_E_py (i.e., the Sc−N imide bond is weaker in the arylimido system). As a test of the implied ability of the NB(NHCH) 2 fragment to form stronger bonds than its NPh analogue we computed the bond dissociation energy (ΔE BD ) of the simple parent amines H 2 NB(NHCH) 2 and H 2 NPh to the respective NR fragments (S = 1) and H atoms. For the borylamine ΔE BD = 204.9 kcal mol −1 , whereas ΔE BD = 188.2 kcal mol −1 for H 2 NPh (ΔΔE BD = 16.7 kcal mol −1 ). This variation in general donor ability as judged by bond strength is a very important, key difference between the two NR moieties which permeates throughout the reaction chemistry described above and is analyzed further below.
Thermodynamic Trends and Mechanism in the Reaction Chemistry. We turn now to the reaction chemistry of the borylimido and arylimido systems. The computed ground-state and TS results (using small range of substrates) for the sterically minimized "q" series are given in Scheme 7, results for the intermediate steric bulk series "Q" (cf. Table 3) are summarized in Scheme S3 of the SI, and those for the full systems "Q*" are given in Scheme 8. It is necessary to draw on Scheme 7. DFT Computed Reactions of ( Ph NacNac NMe 2 )Sc(Me){NHB(NHCH) 2 } (1q_A) and ( Ph NacNac NMe 2 )Sc(Me)(NHPh) (2q_A) with Relative Gibbs Free Energies (G rel , kcal mol −1 ) at 298.15 K and H rel Values in Parentheses a a Black text is used for 1q-based systems and red for 2q.

Journal of the American Chemical Society
Article all three sets to gain a complete picture of the reaction chemistry. Scheme 7 gives good initial insight into the underlying electronic trends of the two types of imido compound in the absence of significant steric or dispersion energy effects. Gibbs free energies are given (with the corresponding enthalpies) relative to the imido compounds 1q_C and 2q_C. Note that because the "q" series necessarily uses the sterically minimized Ph NacNac NMe 2 , Scheme 7 is not able to give direct insight into the cyclometalation reactions of the real systems that involve activation of the Ar′ group of the experimental NacNac NMe 2 supporting ligand.
For both the borylamide 1q_A and arylamide 2q_A, methane elimination to form the imido intermediates nq_C (n = 1 or 2) is both endergonic and endothermic in these model systems. From the imides nq_C, C−H bond activation follows the expected thermodynamic trend: sp (most exothermic) > sp 2 > sp 3 . 28i For reactions with the model terminal alkyne, HCCPh, formation of nq_H (C−H activation) is slightly less exothermic than nq_I ([2+2] cycloaddition), in contrast to experimental outcome, although in terms of Gibbs free energy there is little difference. This is in keeping with the experimental and computational trends for reactions of early transition metal imides with terminal alkynes. 33 As noted for related transition metal systems, 34 the [2+2] cycloaddition reaction with MeCCPh to form nq_J is less exothermic than with HCCPh due to the electron-releasing methyl group which destabilizes the electron-rich azametallacycle. An important factor that emerges from all of the C−H activation or cycloaddition reactions is that those of the borylimide 1q_C are all more exothermic by between −4.2 and −5.3 kcal mol −1 compared to the phenylimide 2q_C, consistent with the ΔE BD calculations on nq_E_py and H 2 NR (R = B(NHCH) 2 or Ph).
Pyridine and DMAP involvement and/or activation are important factors in the real experimental systems, and it was important to pay attention to this aspect. There is negligible thermodynamic difference between pyridine coordination to either methyl-amide starting compound nq_A (ΔΔG = 0.2 kcal mol −1 for nq_A + py → nq_B), and both reactions are endergonic. In contrast, pyridine addition to the fourcoordinate imides nq_C to give nq_E_py is moderately exergonic, as expected from the lower coordination number. DMAP addition, forming nq_E_DMAP, is more favorable than for pyridine as expected. For both nq_E_py and nq_E_DMAP, Lewis base addition is less exothermic for the borylimido homologue due to the better donor ability of NB(CHCH) 2 compared to NPh. From all four adducts nq_E, pyridine or Scheme 8. DFT Computed Reactions of (NacNac NMe 2 )Sc(Me){NHB(NAr′CH) 2 } (1Q*_A) and (NacNac NMe 2 )Sc(Me)(NHAr′) 2Q*_A, with G rel Values (kcal mol −1 ) at 343.0 K a a The numerical values in bold type are for experimentally-observed species. The dotted arrows are for reactions known from experiment to be reversible. The transition state (TS) G rel values (in italic) are based on the ΔG ⧧ values for the "Q" series (cf. Scheme S3 in the SI) expressed relative to the energy of 1Q*_A and 2Q*_A. Black text is used for 1Q* based systems and red for 2Q*.

Journal of the American Chemical Society
Article DMAP orthometalation (C−H addition) across the Sc−N imide bond is electronically favorable, but less so (i) for DMAP (at least in part because of the stronger bonding in nq_E_DMAP) and (ii) for the phenylimides 2q_E_py and 2q_E_DMAP for the reasons already highlighted above. In terms of electronic preferences, the model systems suggest that formation of 2q_E_DMAP is the most favorable product starting from the methyl-amides nq_A, and least likely to form an orthometalated pyridine derivative nq_G, consistent with experiment.
Scheme 8 shows the computed reactions for the full experimental models (NacNac NMe 2 )Sc(Me){NHB(NAr′CH) 2 } (1Q*_A) and (NacNac NMe 2 )Sc(Me)(NHAr′) (2Q*_A). Gibbs free energies relative to nQ*_A (G rel ) are given at 343 K for comparison with experiment. The G rel values in bold type are for experimentally observed species; the dotted arrows are for reactions known to be reversible. Scheme S3 of the SI shows the corresponding data for the intermediate series "Q" for which TS energies could be evaluated for the key reaction steps in most cases. Both schemes generally express the same trends and outcomes, but more accurate ground-state energies (as judged by correlation with experimental outcomes) are found with the full model systems.
The underlying thermodynamic trends in the larger models follow those determined on electronic grounds for the "q" series with respect to activation reactions of the Sc−N imide bonds of the intermediate imides: for example, reactions are consistently more exergonic for the borylimide, and C−H activation reactions follow the expected sequence. In the following discussion we focus mainly on the borylimide/ borylamide reactions, drawing attention to the key differences between these and the arylimide where required.
The calculated ΔG for forming 1Q*_C is slightly endergonic despite the positive ΔS for the elimination reaction (38.8 cal mol −1 K −1 ). The activation energy (ΔG ⧧ = 27.4 kcal mol −1 ) is in good agreement with that measured experimentally (26.0 ± 0.1 kcal mol −1 derived from the rate constants in Tables 1 and  2) at the same temperature. The slightly negative entropy of activation (ΔS ⧧ = −8.8 cal mol −1 K −1 ) is due to the rather early (with respect to overall methane elimination) nature of this TS (Sc−N amide and Sc−C = 1.874 and 2.400 Å compared to 2.027 and 2.237 Å in 1Q_A). Comparable negative experimental ΔS ⧧ values have been reported by Wolczanski for loss of alkane from titanium amide−alkyl compounds forming a transient imide. 28a In addition, the quite linear N−H···C Me interaction (158.5°), the planar four-center C Me −Ti−N−H arrangement, and the similar amounts of N−H bond breaking (N···H = 1.357 Å) and C−H bond making (C···H = 1.467 Å), as judged by the bond distances, are all consistent, 28a,h,i,k with the large primary KIE = 8.7 ± 0.5 observed for the elimination reactions of 18 and 18-d 1 . From 1Q*_C pyridine or DMAP addition is exergonic, ΔG for the latter being more negative. The activation ΔG ⧧ for adduct formation could not be located (the reaction surface being characteristically rather flat) but is expected 16,34,35 to be low, and certainly in comparison to that which forms the cyclometalated 1Q*_D. This, and the rather low ΔG ⧧ barrier for pyridine orthometalation (15.9 kcal mol −1 ), is consistent with 23 and 24 being the experimentally observed products (modeled by 1Q*_G_py and 1Q*_G_DMAP). Experimentally, thermolysis of 18 in the presence of THF formed only 21. We found that the THF analogue of 1Q*_E_py (namely, 1Q*_E_THF, not shown in Scheme 8) has ΔG = −2.7 kcal mol −1 relative to 1Q*_A at 343 K (a similar value was found for the "Q" series counterpart 1Q_E_THF) and is therefore thermodynamically unstable with respect to forming 1Q*_D and THF. Although a THF adduct of this type could be kinetically favored, under the reaction conditions required to form the imido intermediate 1Q*_C (ΔG ⧧ = 27.4 kcal mol −1 from 1Q_A), the barrier to forming the cyclometalated product from the THF adduct is only 17.0 kcal mol −1 , explaining why such an adduct is not experimentally observed. An alternative, base-assisted pathway to form 1Q*_E_DMAP from 1Q*_A via the adduct 1Q*_B was found to have a significantly higher barrier (ΔG ⧧ = 34.6 kcal mol −1 ). This is consistent with experiment that found the rates of methane elimination from 18 were unaffected by a Lewis base. In contrast, the barriers either with or without DMAP precoordination for methane elimination from 2Q_A* to form 2Q*_E_DMAP are more similar, consistent with Chen's observation of DMAP assisted methane elimination. The differences between the reactions of the boryl-and arylamido systems in this case are attributed to steric effects, the required intermediate nQ*_B being 5.7 kcal mol −1 less stable for the borylamide system (in the sterically pruned systems there was no substantial difference in energy for the two methyl-amides nq_B). In the absence of added DMAP, thermolysis of Chen's 19 experimentally formed the cyclometalated product 22, consistent with the calculated formation of 2Q*_D from 2Q*_A.
The pyridine metalation processes modeled by nQ*_E → nQ*_F → nQ*_G are in good agreement with the experimental observations and the underlying trends already discussed for the small model ("q") systems. The arylimide 2Q*_E is stable with respect to 2Q*_G, as found experimentally. 2Q*_E is also computed to be stable to DMAP elimination to form 2Q*_D (even at 90°C the process is endergonic (ΔG = 1.6 kcal mol −1 ) despite the strong entropic driving force), again in agreement with experiment. Pyridine or DMAP C−H activation (orthometalation) from the borylimide adducts 1Q*_E easily form kinetic products (1Q*_F), and subsequent in-place rotation of the metalated ligand (TS too shallow to be located) forms the isomers 1Q*_G. The DFT computed structures of (NacNac NMe 2 )Sc-{η 2 (C,N)-NC 5 H 4 } (1Q*_F_py) and (NacNac NMe 2 )Sc-{η 2 (N,C)-NC 5 H 4 } (1Q*_G_py) are shown in Figure S9 of the SI, together with selected bond distances. In the former, the anionic carbon of the metalated pyridine is placed trans to one of the anionic donors of the NacNac NMe 2 ligand. Rearrangement allows shortening of two of the Sc−N bonds of this supporting ligand and of Sc−N py . The very similar energies of 1Q*_E, 1Q*_F, and 1Q*_G, and the relatively low barriers between them and the base-free imide 1Q*_C (for example, G rel = 11.0 kcal mol −1 cf. 1Q*_G_py at 343 K), account for the facile pyridine exchange reactions found experimentally (Scheme 4).
Experimentally, formation of 23 and 24 is kinetically favored compared to cyclometalation to give 21 in the thermolysis reactions of 18 at 60°C. However, at 90°C, conversion to 21 and pyridine is observed for 23 but not for 24. The calculations are consistent with this being a process under thermodynamic control at this temperature (ΔG 363 K = −1.6 and +1.5 kcal mol −1 respectively for 1Q*_G_py or 1Q*_G_DMAP → 1Q*_D). This reaction is strongly endothermic (ΔH = 12.7 and 13.3 kcal mol −1 ) as expected, but entropically very favorable. The overall activation ΔG ⧧ for formation of 1Q_D and pyridine (relative to 1Q_G_py) is 24.9 kcal mol −1 .

Journal of the American Chemical Society
Article Experimentally, with respect to the reactions with alkynes (Scheme 6), thermolysis of 18 forms only the C−H activation product 26 with HCCTol, and the [2+2] cycloaddition product 27 (one isomer only) with MeCCPh. No rate differences were found with DCCTol. These products are modeled by 1Q*_H and 1Q*J for the full DFT systems and 1Q_H and 1Q_J for the intermediate steric bulk models. An alternative [2+2] cycloaddition product (modeled by 1Q*_I in Scheme 8) was not observed in the reaction with HCCTol. Thermolysis of 21 at 100°C in the presence of MeCCPh formed 27. Therefore, although 21 is the final thermodynamic product (at least at elevated temperatures) in the reactions with pyridine, with alkynes the C−H activation or [2+2] cycloaddition products appear to be the thermodynamic products.
The DFT results, in full agreement with experiment, offer insights into these observations. Activation of the sp hybridized C−H bond of HCCTol (ΔG = −31.4 kcal mol −1 at 343 K for forming 1Q*_H from 1Q*_C) is strongly thermodynamically preferred to sp 3 C−H bond activation on electronic grounds (as discussed for the model systems "q") despite being entropically less favorable. The TS for this process is also lower for the alkyne reaction than for cyclometalation (ΔΔG ⧧ ca. 12 kcal mol −1 ), consistent with methane-elimination being the rate-determining step in all cases. ΔG ⧧ for re-forming the imido intermediate 1Q*_C from 1Q*_D is 26.2 kcal mol −1 at 343 K consistent with the reversibility observed experimentally. Interestingly, the observed C−H bond activation product 26 (as modeled by 1Q*_H) is not the thermodynamic product. Instead the [2+2] cycloaddition species 1Q*_I is the most stable isomer by ca. 4 kcal mol −1 (as found for the small model systems 1q_H and 1q_I, in terms of enthalpic (electronic) preferences). The formation of 26 is therefore under kinetic control as confirmed by the higher TS for formation of 1Q_I compared to 1Q_H (ΔΔG ⧧ = 9.6 kcal mol −1 at 343 K). The high stability of 1Q*_H compared to the base-free imide 1Q*_C and HCCTol (ΔG = 31.4 kcal mol −1 at 343 K for the alkyne elimination process) ensures that, even at elevated temperatures, 26 remains the only product formed.
The DFT compound 1Q*_J (model for 27) is ca. 11 kcal mol −1 less stable than 1Q*_I, consistent with factors already discussed for 1q_J and 1q_I. An alternative isomer (1Q*_K) with the cycloaddition occurring in the opposite sense is less stable than 1Q*_J. In imidotitanium chemistry, for example, this same preference is attributed to the ability of the aryl group to better stabilize the negative charge accumulated on the metal-bound metallacycle carbon. 34 We were not able to locate computationally the TS leading to 1Q*_J but its energy can be estimated as being higher by at least 2−3 kcal mol −1 than that for 1Q*_H because, experimentally, a 1:5:5 competition reaction between 18 and HCCTol and MeCCPh gave exclusively 26. On the other hand, the cycloaddition TS is probably lower in energy than that leading from the base-free imide to 21 since none of the cyclometalated compound is formed on thermolysis of 18 with MeCCPh and 21 only reacts back with MeCCPh to form 27 at 90°C. These experimentally guided estimates are broadly consistent with ΔG ⧧ calculated for the [2+2] cycloaddition reaction of HCCTol (11.3 kcal mol −1 at 343 K). A barrier of this magnitude would be also consistent with the observation that 27 does not react even under thermally forcing conditions with HCCTol to give the thermodynamically preferred 26.
Compound 25 undergoes trapping reactions with alkynes. Thus, thermolysis of 18 in the presence of HCCTol produced the acetylido complex (NacNac NMe 2 )Sc{NHB(NAr′CH) 2 }-(CCTol) (26) via C−H bond activation at the sp center, and in the presence of the internal alkyne MeCCPh formed the azascandacyclobutene complex (NacNac NMe 2 )Sc{N{B-(NAr′CH) 2 }C(Me)C(Ph)} (27) by [2+2] cycloaddition across the Sc−N imide multiple bond of 25, thus giving excellent evidence for its existence. The first-order rate constants of these two reactions were found to be comparable to those described previously, so confirming the common rate-determining step.
A combined DFT and QTAIM study on model scandium alkylimides and borylimides confirmed that borylimides are expected to be comparable to arylimides structurally, due to the similar π effects arising from the aryl ring, and aromatic boryl ring. However, scandium borylimides are calculated to have stronger scandium−nitrogen multiple bonds than their arylimido counterparts due to the strong σand π-donating nature of the borylimido ligand. DFT calculations performed on the reaction products of a model base-free borylimide, and its arylimido counterpart, with a number of substrates resulted in an analysis of the relative energetics of these processes. It was shown that C−H bond activations are more favorable in the borylimido case, due to the higher strength of the N−H bond formed from a borylimido ligand compared with an arylimido ligand. Other bonds formed to a borylimido fragment, such as the C−N bond in the metallacyclic product of a [2+2] cycloaddition, are also found to be stronger than those to arylimido fragments, as shown by the more favorable enthalpic contribution to free energy of the process in which it is formed.
Overall, these first detailed computational and experimental studies of borylimido systems indicate the potential for enhanced reactivity compared to their arylimido analogues, and the synthetic metholodogy developed here should be useful for developing borylimido chemistry of early transition metals, as well as other rare earth metal complexes. Further studies are currently underway in our laboratory.

■ GENERAL METHODS AND INSTRUMENTATION
All manipulations were carried out using standard Schlenk line or drybox techniques under an atmosphere of argon or dinitrogen. Solvents were either degassed by sparging with dinitrogen and dried by passing through a column of the appropriate drying agent 36 or refluxed over sodium (toluene), potassium (THF), Na/K alloy (Et 2 O), or CaH 2 (pyridine) and distilled. Deuterated solvents were dried over potassium (C 6 D 6 ) or CaH 2 (pyridine-d 5 ), distilled under reduced pressure, and stored under argon in Teflon valve ampules. Unless otherwise stated, NMR samples were prepared under dinitrogen in 5 mm Wilmad 507-PP tubes fitted with J. Young Teflon valves. 1 H, 13 C{ 1 H}, 11 B{ 1 H}, and 2 H spectra were recorded on a Bruker Ascend 400 NMR spectrometer, a Bruker Avance III 500 NMR spectrometer or on a Bruker AVC 500 spectrometer fitted with a 13 C cryoprobe. Unless otherwise stated, all NMR spectra were recorded at 298 K. 1 H and 13 C{ 1 H} spectra were referenced internally to residual protiosolvent ( 1 H) or solvent ( 13 C) resonances and are reported relative to tetramethylsilane (δ = 0 ppm). 2 H NMR spectra were referenced to the natural abundance deuterium resonance of the protio solvent. 11 B NMR spectra were referenced externally to Et 2 O·BF 3 respectively. Assignments were confirmed as necessary with the use of twodimensional 1 H− 1 H and 13 C− 1 H correlation experiments. Chemical shifts are quoted in δ (ppm) and coupling constants in Hz. Unless otherwise stated, IR spectra were recorded on a Thermo Scientific Nicolet iS5 FTIR spectrometer and samples prepared in a drybox using NaCl plates as a Nujol mull. The data are quoted in wavenumbers (cm −1 ). Mass spectra were recorded by the mass spectrometry service of Oxford University's Department of Chemistry. Elemental analyses were carried out by the Elemental Analysis Service at the London Metropolitan University.