Multi‐Pathway Consequent Chemoselectivities of CpRuCl(PPh3)2/MeI‐Catalysed Norbornadiene Alkyne Cycloadditions

Abstract Chemoselectivities of five experimentally realised CpRuCl(PPh3)2/MeI‐catalysed couplings of 7‐azabenzo‐norbornadienes with selected alkynes were successfully resolved from multiple reaction pathway models. Density functional theory calculations showed the following mechanistic succession to be energetically plausible: (1) CpRuI catalyst activation; (2) formation of crucial metallacyclopentene intermediate; (3) cyclobutene product (P2) elimination (ΔG Rel(RDS)≈11.9–17.6 kcal mol−1). Alternative formation of dihydrobenzoindole products (P1) by isomerisation to azametalla‐cyclohexene followed by subsequent CpRuI release was much less favourable (ΔG Rel(RDS)≈26.5–29.8 kcal mol−1). Emergent stereoselectivities were in close agreement with experimental results for reactions a, b, e. Consequent investigations employing dispersion corrections similarly support the empirical findings of P1 dominating in reactions c and d through P2→P1 product transformations as being probable (ΔG≈25.3–30.1 kcal mol−1).


Mechanisticspecificities:Symmetrical alkynes
Mechanistic structural details and resultant spontaneities for reaction a are summarised in Scheme 2a nd Figure 2, respectively.T herein, two differing pathways (Paths I and II)w ere exploredi nt he competitive generation of dihydrobenzoindoles (P1)a nd cyclobutenes (P2)f rom 7-azabenzonorbornadiene (R1 a)a nd 3-hexyne (R2 a). Paths I and II differ by disparate orientations of Cp and Ig roups with respectt ot he Ru ring plane (Ru-C1-C2-C3-C4, Scheme 2), and competitive sub-pathways to P1 and P2 are denoted by I-1, I-2, II-1, II-2,r espectively.O verall, Path I is more spontaneous, with at endency for the Cp ligand to remain above the Ru ring plane (TS2 a =  18.3 kcal mol À1 ), preferred by 3.6 kcal mol À1 over the sub-plane orientation (TS6 a = 21.9 kcal mol À1 ). This generates Ru-cyclopentene intermediates (INT1 and INT4)t hat subsequently isomerise to INT2 and INT5 by RuÀNc omplexation. This is followed by C5ÀN6 bond breaking via TS4 a and TS8 a,with barriers of 26.5 and 27.5 kcal mol À1 ,r espectively to form rutheniumcyclohexenei ntermediates (INT3 and INT6). This is the rate-determining step (RDS) for pathways I-2 and II-2,a nd overall, it serves to form the cyclohexenem oiety in the benzoindole product (P1).
Alternatively, at INT1 and INT4,t he reaction may pursue cyclobuteneg eneration through reductive eliminationt op roduce dihydrobenzoindole, via TS3 a and TS7 a (affording P2) with barriers of 15.8 and 13.8 kcal mol À1 for paths I-1 and II-1, respectively.H ence, P2 formation is 10.7 and 13.7 kcal mol À1 more spontaneous than P1 formation, on paths I and II respectively.T his is ar easonable explanation for why 63% of P2 a has been isolated experimentally.
Supplementary calculations involving exhaustive attempts to identifyp ossible transition structuresa nd reactionp aths to P1 a and P2 a formation via intermediate III (Scheme 1) were all unsuccessful. Searches dida fford two structures arising from cleavage of as ingle CÀNl inkage in R1 a (TS10 a and TS10 ax), but were prohibitive at 27.2 and 33.4 kcal mol À1 ,r espectively ( Figure S2 in the Supporting Information). Similar product routes arose for reactions b-e,r aising the applicability of potentialr outes for future experimental explorations, and are thus discussed below.

Mechanisticspecificities:Unsymmetrical alkynes
Asymmetric substitution of the alkyne results in an additional splitting of the four paths described in reaction a for asymmetric alkyne. This forms an octet of paths to be investigated in the competitive generation of P1 b, P2 b and P3 b (I-1, I-2, II-1, II-2, III-1, III-2, IV-1, IV-2). From the outset,P ath I-1 dominates firstly in the generation of ruthenium-cyclopentene intermediate (TS2 b = 14.6 kcal mol À1 )a nd subsequently in reductive elimination ( Figure 3). Mechanistic dimensionality is immediately reduced throughp reclusion of the latter two paths (IV-1 and IV-2)i nvolving C4ÀC2 binding (TS6 b-n,F igure 3a nd Scheme S1 in the Supporting Information), due to their being 2.2 kcal mol À1 less spontaneous than C3ÀC1 binding (TS2 b, Figure3and Scheme 2).
Although formation of INT1 b-n and INT1 b are competitive due to near-identical barriers of TS2 b-n (14.5 kcal mol À1 )a nd TS2 b (14.6 kcal mol À1 ), path I-1 is more spontaneouso verall, with am aximal barriero f1 6.9 kcal mol À1 at reductivee limination (TS3 b). This TS3b is 1.1 kcal mol À1 thermodynamically Scheme2.Putative formation mechanismsfor dihydrobenzoindoles (P1) and cyclobutenes (P2)f rom 7-azabenzonorbornadiene (R1)a nd symmetric alkynes (R2 a, c, e;S cheme 1), investigated at the IDSCRF-B3LYP/BS1 level in dioxanesolvent.  Thus, path I-1 dominates in the generation of P2 b and is in good thermodynamic agreement with experimental yields of 98 %. This differs from Ta m's conclusions for dominance of ap athway similar to path IV-1,b ased on gas-phase computations using the inferior LAN L2DZ basis set, involving as tatic general potential to describe all core electrons. [14] 3.2 Chemoselectivities

Chemoselectivities of reaction a
To resolve the structuralb ases for the observed chemoselectivities and corresponding energetics, key structures and their corresponding Wiberg bond indices (WBI) along P2 a and P2 b formation pathways forr eactions a and b are presented in Figures 4a nd 5, respectively.T he reduced spontaneity of surmounting the TS6 a relative to TS2 a barriers, detailed in section 3.1.2, renderst he contributiono fp ath II-1 to P2 a production negligible and is attributed to al ater transition state. This is evidencedb yb oth the shorter (cat-complex) C1ÀC3 (product) bond length (2.04 vs. 2.14 )a nd bigger Wiberg bond index (WBI) (i.e.,s tronger bonding, as per WBI % 0.454 vs. 0.406) of TS6 a, relative to those of TS2 a.T he inverse is observed in the subsequent step, in which the earliert ransition structure of TS7 a raises its free energy by 1.3 kcal mol À1 relative to that of TS3 a, furtherh ampering P2 a formation alongp ath II-1.T his is evidenced by the longer (2.28 vs.2.03 )a nd correspondingly weaker (cat-complex) C2ÀC4 (product) bond (i.e.,s maller WBI % 0.318 vs. 0.491), relative to that in TS3 a.
Crowding around Ru by the Cp, Et and CO 2 Me groups is alleviated by transfer of electronic density to the Ia tom, facilitating its departure. This is evidenced by an aturalb ond order (NBO) chargeo fÀ0.911ē and the decreasing WBI of the RuÀI linkage at the INT1 a!TS4 a step (0.780!0.018) (see Figure S3 in the Supporting Information). This step is 10.7 kcal mol À1 less spontaneous than the path via TS3 a,e ffectively making path I-2 (thus P1 a production) improbablei nt his manner.S imilarly, P1 a productionb yp ath II-2 is less probable than its corresponding path II-1,w ith the INT5 a!TS8a RDS step for the former being 13.7 kcal mol À1 less spontaneoust han the RDS of the latter with INT4 a!TS7 a.T hese trends support our proposal for path I-1 dominancei nt he observed 63 %y ield of P2 a at 363 K( Scheme 1).

Chemoselectivities of reaction b
For reaction b,s imilar structure-spontaneity trends to those in reaction a support the observed predominance for P2 b formation, with predominance of path I-1 and additional contributions from path III-1,s ince the free energy barrier of TS3 b-n (16.4 kcal mol À1 )i sc ompetitive with that of TS3 b.P aths II-1 and IV-1 are avoided due to their higher barriers at TS6 b and TS6 b-n,e ffectively slowing down reactions along these paths by 1.5 10 5 and 28 times, respectively,w ith respect to path I-1 (kinetic calculations detailed in Figures 3a nd S4 in the Supporting Information). The 7.9 kcal mol À1 reduction in spontaneity of TS6 b,relative to TS2 b,a rises from its later transition structure;t his is evidenced by its contracted (cat-complex) C1À C3 (product) linkage( 2.07 vs. 2.13 )a nd correspondingly larger WBI (0.452 vs. 0.416). The situation is similarf or TS6 b-n and TS2 b-n,w herein the latter has an extended (catÀcomplex) C2À C4 (product) bond length (2.00 vs. 2.21 )a nd correspondingly larger WBI (0.350 vs. 0.416) ( Figure 5).
Confidence in the spontaneity-bonding-Wiberg trends and the crucial product stoichiometry role of the atomic cohesion of C1ÀC3 and C2ÀC4 linkages is further procured by an identi-  cal WBI value (0.416) for the C1ÀC3 and C2ÀC4 linkages in TS2 b and TS2 b-n.T his translates to an egligible free energy differenceo f0 .1 kcal mol À1 (Figures 3a nd 5). Similarly for TS3 b and TS3 b-n,f or whicht he C1ÀC3 and C2ÀC4 WBIv alues are 0.509 and 0.520, respectively,c orrespondingly bear near-identical free energy barriers of 12.6 and 13.7 kcal mol À1 ,r espectively (Figures 3and 5).

Chemoselectivities of reaction c
Resultsf or reaction c are presented in Figure S5 (Supporting Information). Path II is avoided due the reduced spontaneity of TS6 c relative to that of TS2 c along path I (9.6v s. 2.8 kcal mol À1 ). An 88.0 %y ield of P1 c was observed experimentally (Scheme 1), without any P2 c formed. However, computations predict P2 c dominance if only paths of types I and II are considered, due toa voidance of TS4 c (path I-2,2 7.3 kcal mol À1 ) and a9 .7 kcal mol À1 free energy preference for TS3 c (path I-1, 17.6 kcal mol À1 ). This highlights the importance of alternative path calculations.
The overwhelming thermodynamic stabilityo fP1 c (À60.2 kcal mol À1 ), relative to that of P2 c (À35.1 kcal mol À1 ), hinted at the possibility of ak inetic!thermodynamic product transformationp athway.An ovel type of path (V)i nvolving the CpRuI catalysto xidativelyi nserting into aC ÀNb ond on the P2 c product, was thus tested to rectify this experimental-computational discord (Scheme 3a nd Figure S6 in the Supporting Information). Differing constitutional orientations of Cp and I ligandsd ivides the path into two differing channels (V-1 and V-2). The initial oxidative insertion step is locally demanding at 41.0 and 40.8 kcal mol À1 ,f or TS11c and TS13 c,r espectively. However,t hese barriers are globally surmountable due to sufficient energyr emaining in the reactione nsembler elative to the original starting materials;t he transformationsa re + 5.9 and + 5.7kcal mol À1 ,r elative to the starting materials, respectively.S ubsequently,t he (cat-complex) C2ÀC4 (product) bond is cleaved at TS12 c and TS14 c,( 31.5 and1 0.9 kcal mol À1 ,r espectively), the latter providing ap ossible route to P1 c dominance, althoughthe oxidative insertion remains prohibitive.
Furthere xamination of the structures along these pathways revealed weak OÀH···O and OÀH···Ii nteractions, which are specific to reaction c ( Figure S7 in the SupportingI nformation) and are poorly described by the B3LYP method. [15] Subsequent calculations employing the dispersion-corrective B3LYP + D3 methodr endered the oxidative insertion barriers to be manageablev alues of 25.7 (TS11c)a nd 25.3 kcal mol À1 (TS13 c). More encouraging was the reduction of the TS14 c barriert o 15.1 kcal mol À1 ,a llowing the C2ÀC4 linkaget ob eeasily cleaved at 323 Ka nd support P1 c dominance ( Figure 6). Once again as pontaneity-Wiberg correlation is apparent, with the O8ÀH···O9 and O9ÀH···O8 connections in TS11c and TS13c having near-identical WBI values of 0.053 versus 0.057.T his indicatest hat the barrier lowering in the latter TS arises from elsewhere in the transition structures. Indeed, the WBI of the C5ÀH···I interaction in TS11c (0.033), is half that of the O8ÀH···I link in TS13 c (0.071);t his relativelys trong hydrogen-halide interaction is responsible for the barrier elevation. Similarly, TS14 c has an advantage with the O8ÀH···O9 and O9ÀH···I intramolecular interactions, helping effect its observed spontaneity and manageable barriera tt he B3LYP + D3 level.
Overall, the B3LYP + D3 results recover the experimentalcomputational agreement, through its resolution of the product stereoselectivity (two bridged cis hydrogens) observed in most experiments. [1,10,[12][13] Kineticc alculations from the RDS barriers in reaction c,f urther strengthens the agreement with the time for the entiret ransformation anticipated at~48 h (contrasted to the 8a nd 20 hf or reactions a and b, respectively), again in agreement with experimental trends.

Chemoselectivities of reactions da nd e
Reactionm echanismsa nd profiles for reactions d and e are similar to those presented for reaction c,w ith P1 d and P2 e predicted as the corresponding major products in each case, in  agreement with experimental observations of 42 %a nd 97 %, respectively (Scheme 1). P1 d emerges from path I-1 generation of P2 d and eventual P2 d!P1 d conversion along at ype V pathway,w hereas P1 e forms along a I-1 path only.C omplete free energy profiles for reactions d and e are presented in Figures S8, S9, S10 and S11( Supporting Information).K ey steps of these transformationsa re compared in Figure 7, right side, against their matching transitions in reactions a-c.T he RDSs on path type V forr eactions d and e emergea sTS13 d and TS13 e,m ediated by barriers of + 30.1 and + 38.1 kcal mol À1 , respectively.T he former supporting the 42 % P1 d product yield experimentally observed at 333 Ko ver 19 h, the latter effectively blocking P1 e generation.

Influence of computational methods
In determining the probabilityo fP2 a(b, e)!P1 a(b, e) transformation,b yp aths V-1 or V-2,a nd to more fully explore the influence of the methodo nr elative free energies, calculations employing both B3LYP and B3LYP + D3 functionals were carried out for reactions a and b.T he resultsi nd ioxanes olvent are summarised in Tables 1, 2a nd S1. Corresponding results for reactions c-e are also listed for sake of comparison. It is clear that path V-1 is more difficult to overcome than V-2 in all five reactions, featured by higher free energy barriers ranging from 37.6 to 43.0 kcal mol À1 at the IDSCRF-B3LYP + D3/BS1 computational level and even higher free energy barriers at the IDSCRF-B3LYP/BS1 computational level (ranging from 55.0 to 63.8 kcal mol À1 ). When the B3LYP method is employed,t he activation free energy barriers corresponding to RDS in path V-2 (TS13)i sp redicted to be 48.7, 53.4, 40.8, 45.2, and 56.1 kcal mol À1 respectively for reactions a-e, suggesting no detectable transformation from P2 to P1 and this does not support the predominant formation of P1 c and P1 d.
The B3LYP + D3 methodd oes generate reduced RDS (TS13) free energy barriers of 34.5, 36.7, 25.3, 30.1 and 38.1 kcal mol À1 for reactions a-e,r espectively,a ligningw ell with the experimentally observed yields of P1.For reactions b and e,their corresponding half-lives of 4.880 10 7 h( ca. 5571 years) and 4.048 10 8 h( ca. 46210 years) nullify any corresponding P2! P1 conversion via TS13;i na greementw ith experimental data. For reaction a,i ts lower TS13 barrier is perhaps surmountable under the reaction conditions, providing ar oute to the~5% isolated yield of P1 a.F or H-bound systems( c and d), these RDS free energy reductions correspond to representative halflives of 5.42 and 2.275 10 3 h( ca. 3months) and their experimentally observed chemoselectivities of 88 %a nd 42 %, respectively.T he latter time of 3months is within one order of magnitude( ca. 1.50 kcal mol À1 )o ft he experimental reactiont imes of 6d ays. Thus, P2!P1 through path V-2 seems possible for reaction a,p robable for reaction c and, at least to some extent, also so for reaction d.I ns ummary,i ti sd eemed necessary to include dispersion correction on the B3LYP method for systems demonstrating weak interactions. [15][16]

Conclusion
Density functional theory (DFT) calculations employing the IDSCRF-B3LYPa nd IDSCRF-B3LYP + D3 methods, with 2d iffering basis sets, have been performed to probe the experimentally observed chemoselectivities of five CpRuCl(PPh 3 ) 2 /MeI cat- Figure 7. Relative free energies (kcal mol À1 )o fkey transitions tructures in reactions a-e,with selected interatomic distances (in ). All hydrogen atoms apart from those of OH groups are omitted for clarity. Our work has depicted the importance of manifoldm echanisms to accurately andreproducibly resolve experimental chemoselectivities of azabenzonorbornadienes and alkynes. This work has led to complementary explorations of reactive regio-selectivites of unsymmetrical alkynes, as wella st he diastereoselective formation of dihydrobenzoindoles in related systems.

Computational Methods
All models involved the full-sized systems (i.e.,n ot runcations) to accurately represent the real chemical transformations under investigation. Stable structures along the mechanistic profiles were initially optimised, their identities verified and relative free energies determined in solvent (see below) using the B3LYP method, as implemented in Gaussian 09 Program Package (G09), [17] employing ab asis set labelled 'BS1' for convenience. BS1 employs the 6-31G(d,p) Pople basis set [18] for C, H, O, Na toms and the standard double-z valence polarized (DZVP) all-electron basis set for the Ru atom. [19] For the Ia tom, diffuse 1s,1pa nd 1d functions, taken from the aug-cc-pVTZ-PP basis set, [20] have been added to the standard 6-311G(d) basis set. [21] Asecond basis set combination, labelled 'BS2', differing from BS1 only in its use of the 6-311 ++G(d,p) Pople basis set for C, H, O, Na toms, was also employed for selected computations. Experimental solvent effects (dioxane, e = 2.21), were addressed using the default self-consistent reaction field (SCRF) polarisable continuum model (PCM), [22] employing IDSCRF atomic radii [23] to define the molecular cavity;d enoted IDSCRF-B3LYP. All free energies reported throughout the work have been corrected to include translational entropy contributions in the condensed phase (S trans(l) )u sing the THERMO method [24] towards avoiding the pitfalls associated with default gas-phase calculations of S trans originating from S trans(g) .I ntrinsic reaction coordinate (IRC) [25] calculations were carried out on selected reaction pathways to confirm key transition states (TSs) and connect two corresponding adjacent minima. Furthermore, the dispersion-corrected DFT-D [26] method (denoted IDSCRF-B3LYP + D3) was chosen to characterise selected stationary points and reaction channels when necessary.N BO analyses, [27] as implemented in G09, was also performed on selected stationary points at IDSCRF-B3LYP/BS2//BS1 or IDSCRF-B3LYP + D3/BS2//BS1 level, to investigate their electronic properties and bonding characteristics. To dispel the spectres of methodological uncertainty and anomaly in the B3LYP results, single-point energies using the more modern M062X, X3 LYPa nd CAM-B3LYP functionals, as well as MP2 and B2PLYP methods were carried out on selected paths in reaction a; these are presented in Ta ble S2 in the Supporting Information. For concision, only B3LYP or B3LYP + D3 results are discussed in the text.