Selective C-H activation and functionalization of pyridines and other azines presents special challenges, in part because these heterocycles can function as good ligands towards many transition metals. 1,2 Selectivity for the 3-(or meta-) position is more common with transition metals, 35 but studies of selective 2-position functionalization are also known. 612 In many specific cases, the scope may be limited, and a particular substitution pattern on the azine is often required for selectivity. In 2017, we reported a new approach to the directed activation of C-H bonds in pyridine derivatives using an Ir system supported by a boryl/bis(phosphine) PBP 13 pincer 14,15 ligand. 16 The binding of the pyridine (or quinoline) nitrogen to the Lewis acidic boryl site directs Ir to the 2-position in the heterocycle. This approach is distinct from the more classical directed C-H activation, where the directing group donor binds to the same atom (transition metal) which effects C-H cleavage (Figure 1). [17][18][19][20][21] Pyridine derivatives have played a prominent role in the development of classical directed C-H activation, 19,20 but they typically direct the metal not to the C-H bonds of the pyridine ring itself, but to the more remote C-H bonds in a substituent, such as in the 2-phenyl group. We reasoned that the (PBP)Ir system preferred the C-H activation of the pyridine ring because of the favorability of the Ir/C/N/B trapezoidal four-membered ring formation. 16 Some of the aspects of the mechanism of pyridine activation in our PBP system were recently studied computationally by Ke and coworkers. 22 A similar selectivity was observed by Nakao et al. in the C-H activation of pyridines with a Rh complex 23 supported by a closely related aluminyl/bis(phosphine) PAlP pincer (Figure 1). 24,25 Given Nakao's precedent with Rh, we wished to explore the reactivity with pyridine using the (PBP)Rh system, 26,27 as well as the variations of the Rh and Ir systems with and without the carbonyl ligand. While exploring the analogous reactivity with (PBP)Rh, we came across an unexpected finding. As with Ir, C-H activation of pyridine resulted in the formation of a 2-pyridyl that is bridging the B-Rh bond. However, the connectivity was reversed, with C of the pyridyl attached to B and the N atom of the pyridyl attached to Rh. This prompted us to explore this M-C/M-N isomerism in more systematic detail, as it does not appear to have been considered in the literature. This report describes our analysis of the isomeric preference of the 2pyridyl (or 2-quinolyl) fragment bridging the B-Ir or B-Rh bond in a series of compounds supported by the PBP pincer.
We selected four structural types for analysis (Figure 2). For each type, we considered M-C/M-N isomerism for the Rh and for the Ir version, resulting in sixteen 2-pyridyl compounds that were whose structures were optimized computationally. The compound labels (Figure 2) are derived from the general type (numeral) and the bond present between the metal (Rh or Ir) and C or N. Some of these compounds were isolated or observed experimentally in this (Scheme 1) or the previous report. 16 In addition, we synthesized a few 2-quinolyl analogs of the 2-pyridyl compounds (Scheme 1); they are denoted by adding a "q" to the compound label. The Type 3 and Type 4 compounds are isomeric. We did not attempt the syntheses of the Type 4 compounds because DFT calculations indicated that they are considerably higher in energy than the corresponding Type 3 isomers (vide infra).
In order to access a Rh species capable of C-H activation, the previously reported 5 was treated with NaBEt 3 H followed by the removal of volatiles. Although the stoichiometry suggests the formation of "(PBP)RhH 2 ", we have not established the nature of the resultant species; from the in-situ NMR observations, it appears that a mixture of a few complexes forms (Figure S1). Nonetheless, thermolysis of this mixture in the presence of cyclohexene and either pyridine or quinoline led to the formation of complexes 1RhN and 1RhNq, with an isolated yield of 70% and 59% respectively. The corresponding Type 1 Ir compound 1IrCq was prepared by the treatment of (PBP)IrHCl with NaN(SiMe 3 ) 2 in the presence of quinoline. Compound 1IrCq exists in equilibrium with the minor isomer 1IrNq (1.00:0.055 ratio at 25 °C and 1.00:0.095 at 65 °C). Attempts to prepare 1IrC in a pure form were not successful. Unlike 1RhN, 1IrC appears to bind an extra equivalent of pyridine, which resulted in a mixture of products when one equiv. of pyridine was used. Utilization of 3 equiv. of pyridine permitted observation of the pyridine adduct of 1IrC as the dominant product by NMR spectroscopy, but we did not pursue its isolation in a pure solid form (compound 7, Figure S4). The conversion of the hydride complexes 1RhN and 1IrCq to the bromide derivatives 2RhC and 2IrCq was effected by Please do not adjust margins Please do not adjust margins thermolysis with NBS. Good isolated yields (70% and 69% respectively) were obtained after workup. No evidence of the presence of 2RhN or 2IrNq was noted. The carbonyl adduct 3RhN was prepared by exposure of 1RhN to carbon monoxide and characterized in situ in solution after 10 min. After removing carbon monoxide under vacuum, thermolysis of the solution of 3RhN in C 6 D 6 for 1 h at 65 °C resulted in the formation of a mixture of 3RhN and 3RhC in a 1.0:0.08 ratio. Extended thermolysis for 24 h at 65 °C led to the formation of multiple complexes along with 3RhN and 3RhC, but in that mixture 3RhN was still present in a much higher concentration than 3RhC. The synthesis of the analogous Ir complex 3IrC was previously reported. The synthesis involved extended thermolysis at 100 °C and no evidence of the presence of 3IrN was noted.
The compounds explored in this study are rich in NMR active nuclei ( 1 H, 13 C, 31 P, 11 B, and 103 Rh). All of the compounds possess C s -symmetry on the NMR time scale. The M-C/M-N isomers can be distinguished based on the relative 1 H NMR chemical shift of the Rh/Ir-H signal. Since N of pyridyl is less trans-influencing than C of 2-pyridyl, a hydride trans to N appears at a more upfield frequency vs a hydride trans to C. For the Rh compounds 1RhN, 1RhNq, and 3RhN with a hydride trans to N, its 1 H NMR chemical shift falls into a narrow range of -15.7 to -17.3 ppm, but for 3RhC, the hydride resonates considerably upfield at δ -11.04 ppm. The contrast is even greater for the Ir pair 1IrCq (δ -0.20 ppm) and 1IrNq (δ -17.10 ppm). The shape of the 13 C{ 1 H} NMR resonance corresponding to the boron-or metal-bound carbon of the 2-pyridyl or 2-quinolyl unit is also telling. In compounds 1RhN, 1RhNq, and 3RhN, this carbon is bound to boron and the corresponding 13 C NMR resonances in these compounds possess some broadness. In compounds RhBr-C and IrH-Cq, this carbon is bound do the metal and displays coupling to the two equivalent 31 P nuclei, as well as to 103 Rh in RhBr-C.
The structures of the 16 molecules shown in Figure 2 were optimized using the B97D3/LANL2DZ/6-31G(d) method (see details in the SI). Figure 4 summarizes the results of the calculations, showing the Wiberg bond indices (WBI) within the four-membered rings, as well as the calculated free energies of the isomerization from the M-C to the M-N isomer. The metric details of the DFT-optimized geometries matched those from the XRD structures reasonably well. Calculations indicate that the isomers with the carbon bound to the transition metal are more favorable for all Ir complexes and for the Rh complexes of Types 2 and 4. For the other Rh complexes, the isomer with the nitrogen bound to Rh is preferred. Across all four types, the relative free energy preference of Ir for the metal-carbon bonded isomer is very consistently 5-7 kcal/mol higher than that of Rh. Overall, the calculated thermodynamic parameters are consistent with the experimental observations we have for the Rh and Ir compounds of Types 1-3. Moreover, the calculated free energy preferences for 1IrC (over 1IrN) and for 3RhN (over 3RhC) are <2 kcal/mol, suggesting that both isomers in these two pairs should be present at observable concentrations. This is precisely what we observed for 1IrCq/1IrNq and for 3RhN/3RhC (vide supra), with the isomer predicted to be more fav orable by DFT present in a higher proportion. Type 3 (CO trans to B) compounds are isomeric to Type 4 (H trans to B), and DFT calculations predict that any of the four Type 3 compounds (3IrC, 3IrN, 3RhC, 3RhN) is lower in free energy than their corresponding Type 4 analog (4IrC, 4IrN, 4RhC, 4RhN, respectively) by 13-19 kcal/mol. This is consistent with the lack of observation of 4RhC or 4RhN in the thermolysis of the 3RhC/3RhN mixture.
The calculated Wiberg bond indices (WBI) allow a way to analyze the changes in the nature of the bonds in the fourmembered cycle for the pairs of isomers. The WBI for the M-B bond in any Ir compound is 0.08-0.13 higher than for the exact Rh analog. Higher WBI values in Ir (vs Rh) compounds are also notable for the M-C and M-N bonds (by 0.04-0.08). This is in general expected for a 5d metal (Ir) compared to its 4d congener (Rh).
Within each M-C/M-N isomeric pair with the same metal, the M-B bond WBI values differ only by 0.04 or less, except for the 4RhC/4RhN pair (0.08 difference). The WBI vary even less for the C-B bonds (0.79-0.82 range) and for the N-B bonds (0.57-0.59) throughout the whole array of compounds. It can be concluded that the changes in the M-B, C-B, and N-B bonding contribute little to the thermodynamic preferences for the M-C vs M-N isomers.
The WBI values for the CN bond vary within a range of 1.22-1.31 for all 16 compounds. Within every M-C/M-N isomeric pair, this value is higher for the N-M bound isomer, by 0.02-0.08, suggesting that coordination to Ir or Rh strengthens the C-N bond slightly, but to a similar degree across all four types of compounds.
Considering the M-C bonds, there appears to be a surprisingly linear correlation (Figure 5) between the WBI values and the thermodynamic isomeric preference, that covers both the Rh and the Ir examples. Higher M-C WBI corresponds to higher preference for the M-C isomer, with ergoneutrality of the isomerization predicted at ca. 0.65 M-C WBI. The WBI values of the M-N bonds trend in the same direction. However, the correlation is more diffuse and not as steep, likely reflecting the intrinsically weaker nature of the M-N bond and its lesser dependence on the environment about the metal center (see Figure S5).
Please do not adjust margins Next, we examined the mechanism 29 of the interconversion between 3IrC and 3IrN as a representative example (Figure 6). From 3IrC, the reaction proceeds via dissociation of the pyridine N from B with concomitant ca. 90° rotation about the Ir-C bond, resulting in 3IrX. The structure of the intermediate 3IrX evinces no bonding interactions between the pyridyl fragment and B, but a full-fledged Ir-C bond. The transition state connecting it with 3IrC (TSCX) possesses both a similar energy and geometry, with an incomplete rotation. The migration of the pyridyl from Ir in 3IrX to B in intermediate 3IrY proceeds via TSXY. In 3IrY, the pyridyl C is connected to the B by means of well-developed C-B bond, which is even 0.023 Å shorter than the calculated C-B distance in 3IrN. The pyridyl C in 3IrY can also be viewed as weakly interacting with Ir. We did not locate a transition state for the conversion of 3IrY into 3IrN; this process is also simply a rotation of the pyridyl with coordination to Ir. It is clear that most of the barrier for the interconversion between 3IrC and 3IrN is owing to the dissociation of N from B/Ir, corresponding to the rotation of N away from B/Ir. Once the N is free, the barrier for the migration of the C-pyridyl between B and Ir is only a few kcal/mol. This likely also applies to the other Types presented in this paper.
In summary, we have examined the unusual isomerization of a bridging 2-pyridyl unit in an array of Rh and Ir complexes supported by a PBP pincer ligand. The main factor governing the thermodynamic preference appears to be the strength of the M-C bond in the M-C bonded isomer. It was observed that the thermodynamic preference for the M-N vs M-C bond depends both on the nature of the metal center and on the nature of the ligand trans to the M-N/M-C bond. The M-C isomer is favored for the 5d metal Ir vs Rh and by the presence of a more weakly trans-influencing ligand trans to the M-N/M-C bond. For some of the complexes, both isomers were observed experimentally, in close agreement with theoretical analysis. The interconversion between isomers of similar (PB Qu P)Rh(H) (1RhNq). In a 50 mL PTFE screw-capped reaction tube, NaEt3BH (0.52 mL, 0.52 mmol, 1.0 M in toluene) was added to a toluene solution (5 mL) of 5 (0.27 g, 0.50 mmol). The solution changed color from orange to dark-red immediately. After stirring at room temperature for 5 min, volatiles were removed under vacuum to remove BEt3 that was S6 generated. The resulting red solid was dissolved in toluene (10 mL), to which were added quinoline (62 μL, 0.52 mmol) and cyclohexene (53 μL, 0.52 mmol). The reaction mixture was stirred at 80 °C for 12 h, and gradually changed from dark-red to brown color. The solution was filtered through Celite, and volatiles were removed under vacuum. The resulting brown solid was recrystallized in toluene/pentane 1:3, yielding a pale-yellow solid (0.19 g, 59%). 1 (PBP)RhPyBr (2RhC). In a J. Young tube, N-bromosuccinimide (0.17 mmol, 30 mg) was added to a solution of 1RhN (100 mg, 0.17 mmol) in 0.40 mL C6D6. This reaction mixture was stirred at room temperature for 24 h and heated at 100 °C for 12 h until all 1RhN were consumed. (monitored by 31 P{ 1 H} NMR). The resulting mixture was filtered over a pad of silica gel and washed through with dichloromethane. All volatiles were removed under vacuum and the resulting solid was recrystallized from toluene/pentane (1:3), yield an orange solid (80 mg, 70%) S10 1RhN. A light yellow, multi-faceted block of suitable size (0.46*0.22*0.16 mm 3 ) was selected from a representative sample of crystals of the same habit using an optical microscope and mounted onto a nylon loop. Low temperature (110 K) X-ray data were obtained on a Bruker APEXII CCD based diffractometer (Mo sealed X-ray tube, Kα = 0.71073 Å). All diffractometer manipulations, including data collection, integration and scaling were carried out using the Bruker APEXII software. 2 An absorption correction was applied using SADABS. 3 The space group was determined on the basis of systematic absences and intensity statistics and the structure was solved by direct methods and refined by full-matrix least squares on F 2 . The structure was solved in the monoclinic P 21/c space group using XS 4 (incorporated in Olex). All non-hydrogen atoms were refined with anisotropic thermal parameters. All hydrogen atoms were placed in idealized positions and refined using riding model with the exception of the hydrogen bound to rhodium which was located from the difference map. The structure was refined (weighted least squares refinement on F 2 ) and the final least-squares refinement converged. No additional symmetry was found using ADDSYM incorporated in PLATON program. 5 CCDC 2014200 contain the supplementary crystallographic data.
1RhNq. A Leica MZ 75 microscope was used to identify a suitable yellow block with very welldefined faces with dimensions (max, intermediate, and min) 0.368 x 0.36 x 0.164 mm 3 from a representative sample of crystals of the same habit. The crystal mounted on a nylon loop was then placed in a cold nitrogen stream (Oxford) maintained at 110 K. A BRUKER APEX 2 X-ray (threecircle) diffractometer was employed for crystal screening, unit cell determination, and data collection. The goniometer was controlled using the APEX2 software suite, v2008-6.0. 6 The sample was optically centered with the aid of a video camera such that no translations were S17 observed as the crystal was rotated through all positions. The detector was set at 6.0 cm from the crystal sample (APEX2, 512x512 pixel). The X-ray radiation employed was generated from a Mo sealed X-ray tube (K = 0.70173Å with a potential of 40 kV and a current of 40 mA). Sixty data frames were taken at widths of 1.0. These reflections were used in the auto-indexing procedure to determine the unit cell. A suitable cell was found and refined by nonlinear least squares and Bravais lattice procedures. The unit cell was verified by examination of the h k l overlays on several frames of data. No super-cell or erroneous reflections were observed. After careful examination of the unit cell, an extended data collection procedure (4 sets) was initiated using omega scans.
Integrated intensity information for each reflection was obtained by reduction of the data frames with the program APEX2. 6 The integration method employed a three-dimensional profiling algorithm and all data were corrected for Lorentz and polarization factors, as well as for crystal decay effects. Finally, the data was merged and scaled to produce a suitable data set. The absorption correction program SADABS 3 was employed to correct the data for absorption effects.
Systematic reflection conditions and statistical tests of the data suggested the space group P-1. A solution was obtained readily using XT/XS in APEX2. 4,6 Hydrogen atoms were placed in idealized positions and were set riding on the respective parent atoms. All non-hydrogen atoms were refined with anisotropic thermal parameters. Absence of additional symmetry and voids were confirmed using PLATON (ADDSYM). 5 The structure was refined (weighted least squares refinement on F 2 ) to convergence. 4,7 CCDC 2014201 contain the supplementary crystallographic data.
1IrCq. A Leica MZ 75 microscope was used to identify an orange plate with very well-defined faces with dimensions (max, intermediate, and min) 0.528 x 0.213 x 0.046 mm 3 from a representative sample of crystals of the same habit. The crystal mounted on a nylon loop was then placed in a cold nitrogen stream (Oxford) maintained at 110 K. A BRUKER APEX 2 Duo X-ray S18 (three-circle) diffractometer was employed for crystal screening, unit cell determination, and data collection. The goniometer was controlled using the APEX3 software suite, v2017.3-0. 8 The sample was optically centered with the aid of a video camera such that no translations were observed as the crystal was rotated through all positions. The detector (Bruker -PHOTON) was set at 6.0 cm from the crystal sample. The X-ray radiation employed was generated from a Mo sealed X-ray tube (K = 0.71073Å with a potential of 40 kV and a current of 40 mA). 45 data frames were taken at widths of 1.0. These reflections were used in the auto-indexing procedure to determine the unit cell. A suitable cell was found and refined by nonlinear least squares and Bravais lattice procedures. The unit cell was verified by examination of the h k l overlays on several frames of data. No super-cell or erroneous reflections were observed. After careful examination of the unit cell, an extended data collection procedure (16 sets) was initiated using omega scans.
Integrated intensity information for each reflection was obtained by reduction of the data frames with the program APEX3. 8 The integration method employed a three-dimensional profiling algorithm and all data were corrected for Lorentz and polarization factors, as well as for crystal decay effects. Finally, the data was merged and scaled to produce a suitable data set. The absorption correction program SADABS 3 was employed to correct the data for absorption effects.
Systematic reflection conditions and statistical tests of the data suggested the space group P-1. A solution was obtained readily (Z=4; Z'=2) using XT/XS in APEX2. 4,8 A molecule of toluene was found solvated (1/2 a molecule per Ir-complex). Hydrogen atoms were placed in idealized positions and were set riding on the respective parent atoms. All non-hydrogen atoms were refined with anisotropic thermal parameters. Elongated ellipsoids and nearby residual electron density peaks for C64 -C66 indicated disorder which was modeled between two positions with an occupancy ratio of 0.77:0.23. Appropriate restraints and constraints were added to keep the bond S19 distances, angles, and thermal ellipsoids meaningful. Absence of additional symmetry and voids were confirmed using PLATON (ADDSYM). The structure was refined (weighted least squares refinement on F 2 ) to convergence. 4,7 CCDC 2014205 contain the supplementary crystallographic data.
2RhC. A Leica MZ 75 microscope was used to identify a suitable yellow block with very welldefined faces with dimensions (max, intermediate, and min) 0.222 x 0.19 x 0.182 mm 3 from a representative sample of crystals of the same habit. The crystal mounted on a nylon loop was then placed in a cold nitrogen stream (Oxford) maintained at 110 K. A BRUKER APEX 2 X-ray (threecircle) diffractometer was employed for crystal screening, unit cell determination, and data collection. The goniometer was controlled using the APEX2 software suite, v2008-6.0. 6 The sample was optically centered with the aid of a video camera such that no translations were observed as the crystal was rotated through all positions. The detector was set at 6.0 cm from the crystal sample (APEX2, 512x512 pixel). The X-ray radiation employed was generated from a Mo sealed X-ray tube (K = 0.70173Å with a potential of 40 kV and a current of 40 mA). Sixty data frames were taken at widths of 1.0. These reflections were used in the auto-indexing procedure to determine the unit cell. A suitable cell was found and refined by nonlinear least squares and Bravais lattice procedures. The unit cell was verified by examination of the h k l overlays on several frames of data. No super-cell or erroneous reflections were observed. After careful examination of the unit cell, an extended data collection procedure (4 sets) was initiated using omega scans.
Integrated intensity information for each reflection was obtained by reduction of the data frames with the program APEX2. 6 The integration method employed a three-dimensional profiling algorithm and all data were corrected for Lorentz and polarization factors, as well as for crystal decay effects. Finally, the data was merged and scaled to produce a suitable data set. The S20 absorption correction program SADABS 3 was employed to correct the data for absorption effects.
Systematic reflection conditions and statistical tests of the data suggested the space group P43. A solution was obtained readily using XT/XS in APEX2. 4,6 Hydrogen atoms were placed in idealized positions and were set riding on the respective parent atoms. All non-hydrogen atoms were refined with anisotropic thermal parameters. Elongated thermal ellipsoids on C27 and C28 suggested disorder, which was modeled successfully between two positions with an occupancy ratio of 0.42 to 0.58. Appropriate restraints were added to keep the bond distances, angles and thermal ellipsoids meaningful. Absence of additional symmetry and voids were confirmed using PLATON (ADDSYM). 5 The structure was refined (weighted least squares refinement on F 2 ) to convergence. 4,7 CCDC 2014203 contain the supplementary crystallographic data.
3RhN. A Leica MZ 75 microscope was used to identify a suitable colorless block with very welldefined faces with dimensions (max, intermediate, and min) 0.774 x 0.706 x 0.429 mm 3 from a representative sample of crystals of the same habit. The crystal mounted on a nylon loop was then placed in a cold nitrogen stream (Oxford) maintained at 110 K. A BRUKER APEX 2 Duo X-ray (three-circle) diffractometer was employed for crystal screening, unit cell determination, and data collection. The goniometer was controlled using the APEX3 software suite, v2017.3-0. 8 The sample was optically centered with the aid of a video camera such that no translations were observed as the crystal was rotated through all positions. The detector was set at 6.0 cm from the crystal sample (APEX2, 512x512 pixel). The X-ray radiation employed was generated from a Mo sealed X-ray tube (K = 0.70173Å with a potential of 40 kV and a current of 40 mA). 45 data frames were taken at widths of 1.0. These reflections were used in the auto-indexing procedure to determine the unit cell. A suitable cell was found and refined by nonlinear least squares and Bravais lattice procedures. The unit cell was verified by examination of the h k l overlays on several S21 frames of data. No super-cell or erroneous reflections were observed. After careful examination of the unit cell, an extended data collection procedure (24 sets) was initiated using omega and phi scans. Integrated intensity information for each reflection was obtained by reduction of the data frames with the program APEX3. 8 The integration method employed a three-dimensional profiling algorithm and all data were corrected for Lorentz and polarization factors, as well as for crystal decay effects. Finally, the data was merged and scaled to produce a suitable data set. The absorption correction program SADABS 3 was employed to correct the data for absorption effects.
Systematic reflection conditions and statistical tests of the data suggested the space group P21/n.
A solution was obtained readily using XT/XS in APEX2. 4,8 Hydrogen atoms were placed in idealized positions (the one connected to Rh was refined independently) and were set riding on the respective parent atoms. All non-hydrogen atoms were refined with anisotropic thermal parameters.
Absence of additional symmetry and voids were confirmed using PLATON (ADDSYM). 5 The structure was refined (weighted least squares refinement on F 2 ) to convergence.
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 5