Download PDF
Synlett 2023; 34(10): 1169-1173
DOI: 10.1055/a-1937-9296
cluster
Dispersion Effects

Activation Barriers for Cobalt(IV)-Centered Reductive Elimination Correlate with Quantified Interatomic Noncovalent Interactions

Lucas Loir-Mongazon
,
Carmen Antuña-Hörlein
,
Christophe Deraedt
,
Yann Cornaton∗
,
Jean-Pierre Djukic  ∗

Centre national de la Recherche Scientifique, University of Strasbourg, GENCI-IDRIS grant 2021-A0100812469, the HPC Center of the University of Strasbourg grant g2021a248c.
› Further Information
 
 PDF Download Permissions and Reprints All articles of this category


Abstract

In this joint theoretical and experimental study, an analysis of weak interligand noncovalent interactions within Co(IV) [Cp*Co(phpy)X]+ cobaltacycles (phpy = 2-phenylenepyridine, κ C,N ) was carried out by using the independent gradient model/intrinsic bond strength index (IGM/IBSI) method to evaluate the dependency of the catalytically desired reductive elimination pathway (RE) on the nature of the X ligand. It is shown that the barrier for activation of the RE pathway correlates directly with the IBSI of the X-to-carbanionic chelate’s carbon. This correlation suggests that in silico prediction of which X ligand is more prone to operate an efficient Cp*Co-catalyzed directed X-functionalization of an aromatic C–H bond is attainable. A set of experiments involving various sources of X ligands supported the theoretical conclusions.


#

Key words

metallacycles - cobalt catalysis - density functional theory - noncovalent interactions - reductive elimination - C–H bond activation

Analyzing the role of the dispersion force and noncovalent interactions (NCI) in a broader sense is crucial to understanding the roots of molecular cohesion, but also in comprehending the chemical reactivity of organometallic molecules.[1] Recently, a few reports[2] have examined this issue in depth by establishing the existence of a favorable NCI coding in agostic reactant complexes as a prerequisite for effective C–H bond activation by a metal center in the concerted metalation/deprotonation (CMD) mechanism,[3] also formulated as ambiphilic metal-ligand activation (AMLA)[4], i.e. one variant of the existing C–H bond-activation mechanisms,[5] such as base-assisted intramolecular electrophilic substitution (BIES).[6] The upsurge of interest in 3d metal complexes,[7] particular those containing Co, as catalysts for C–H bond functionalization[8] poses a wider mechanistic challenge: the engineering of new catalysts requires control of both the chemoselectivity and the durability of the catalytically active species. Indeed, two recent studies[9] have outlined the central role of highly reactive transient Co(IV) metallacycles in C–H bond functionalization (Scheme [1]), revealing that the nature of the X ligand, often introduced to functionalize the aromatic C–H bond,[10] is crucial.[9b] Such Co(IV)–reactant complexes might indeed evolve by at least two main pathways: (1) the reductive elimination[9a] (RE) desired in catalysis[10] or (2) the undesired collapse[9b] of the key Cp*Co motif by a cyclocondensation (CC) of the carbanionic chelating ligand with the Cp* ligand (Scheme [1]).

Zoom Image
Scheme 1 Proposed C–H bond-functionalization catalytic cycle involving Cp*Co complexes with a cyclocondensation (CC) quenching step, the extent of which depends on the nature of X. Note that for X = Cl in 1i, cation 3 is also the dominant product, similarly to 1a.[9b] Cp* = η5-pentamethylcyclopentadienyl.

We have shown that quantitatively evaluating the strength of noncovalent interactions between reactive centers in those pseudotetrahedral Co(IV) intermediates with the help of the independent gradient model (IGM)[11]/intrinsic bond strength index (IBSI)[12] analytical methods revealed propensities[1a] [9b] toward either RE or CC. Here, we report on a theoretical and experimental joint study covering the choice of X ligands in the RE pathway by using the IGM/IBSI analysis, correlating its output to computed activation barriers, and evaluating those trends by experimental assessment. Note that computed structures are denoted by roman numerals and experimentally probed compounds by Arabic ones throughout this report.

A series of [Co(III)Cp*(phpy)X] (I) complexes, as well as their [Co(IV)Cp*(phpy)X]+ [I]+ counterparts, have been optimized at the DFT ZORA-PBE-D4(EEQ)/all electron TZP level of theory by using COSMO(CH2Cl2) as a standard for implicit solvation [see the Supplementary Information (SI) for computational details and references]. The transition states corresponding to the first and limiting step of the RE (i.e., the formation of the C–X bond) and to the suggested first and limiting step of one of the most favorable mechanism of CC[9b] (i.e., the one involving the prior formation of the C–C bond between phpy and Cp*) have also been searched for and optimized at the same level of theory (see SI, Table S1). Previous studies suggested that the oxidation of the Co(III) complex I to a Co(IV) [I]+ is required to trigger the RE.[9a] To clarify whether or not this oxidation is mandatory, the RE activation energies at the Co(III) and Co(IV) states were compared. Note that, whereas the Co(IV) transition states were readily located by using a simple linear transit procedure, the Co(III) ones required more-careful exploration of the potential-energy surface. Even though, in some cases, the theoretical RE energy barrier in the Co(III) state suggests that the reaction is feasible at room temperature (e.g., ΔG ‡ = 14 kcal/mol for X = CH=CH2 in Ih), the activation energy barrier drastically drops in the Co(IV) state (e.g., ΔG ‡ < 2 kcal/mol for X = CH=CH2 in [Ih]+ ). The Co(IV) complex, e.g. [Ih]+, can then be considered as a reactant complex for the RE because the configuration of the system allows a lower energy barrier. The CC activation energy in the Co(IV) state is obviously weakly sensitive to the nature of the X ligand. Indeed, all activation barriers lie around +20 kcal/mol. In contrast, the RE activation energy in the Co(IV) state varies from ΔG ‡ < +2 kcal/mol (X = CH=CH2; [Ih]+) to ΔG‡ = +28 kcal/mol [X= OC(O)CF3; [Ie]+.

Investigating the geometries of the different complexes in both the Co(III) I and Co(IV) [I]+ states reveals slight variations in the distances between the three ligands (SI; Table S2). On one hand, for all studied systems, oxidation brings the phpy and Cp* ligands closer to one another, as evidenced by the decrease in either or both distances between the carbanionic C2′ or/and the N of phpy and the respective closest C of the Cp* ligand (C4 and C3, respectively; Figure [1]), suggesting a favorable rearrangement toward the CC reaction. On the other hand, the distance between C2′ of phpy and the atom of the X ligand bound to Co decreases or increases, depending on the nature of the X ligand. This variation in the C2′–X distance correlates with the ΔG ‡(RE) in the Co(IV) state: the closer the atoms move toward each other, the lower the energy barrier (SI; Figure S1).

Zoom Image
Figure 1 (a) and (b) are IGM-Δg inter isosurface plots (cutoff: Δg inter = 0.01 a0 –4) of the interaction between the phpy and X ligands: (a) X = I, (b) X = CH=CH2. (color code: blue = attractive interaction, green = nonbonding interactions, red = repulsive interactions). (c) Correlation plot between the RE Gibbs free energy of activation (ΔG ‡) and the C2′–X intrinsic bond strength index (IBSI) for a series of different X ligands in [I]+ species. The blue dashed line represents a trend curve of the equation ΔG ‡ = a.exp(–b.IBSI) (a = 69 kcal/mol; b =42; r² = 0.94). The red and green dots outline the cases for which experimental results are reported in this work (X = NHAc, OAc, OC(O)Ph, OC(O)CF3, Ph, CH=CH2) or in the literature (X = I, Cl, SCF3), respectively.[9]

The IGM-Δg inter descriptor,[11] quantifying the drop in electron-density gradient between two user-defined fragments in the real system compared with a corresponding noninteracting reference, gives a measure of the electron sharing between fragments. In this study, this was used to investigate the interactions between different ligands (SI; Table S3). The integrated value of this descriptor, the Δg inter score, gives a quantification of the interaction. The derived IBSI descriptor measures any interatomic pairwise interaction against the covalent bond in H2 that is used as a reference.[12] Even though multicenter interactions cannot be properly analyzed by using IBSI,[12b] a comparison of pairwise interactions between adjacent atoms bound to a common metal center + ligand retinue in a series of related species differing only by one variation in the structure (X in this case) had not previously been attempted. Assuming the electron-density variations at the metal would not be significant, we speculated that IBSI could be used to quantify interactions in atom pairs involved in processes occurring at the Co-centered reactive site (SI; Table S4), and might help in predicting reactivity. Here, from the variation in the Δg inter values between phpy and X ligands upon Co(III)-to-Co(IV) oxidation no clear feature could be inferred; an identical conclusion was drawn for the variation of the Δg inter values between phpy and Cp* (SI; Figure S2).

However, for all considered complexes, the Co(III)-to-Co(IV) oxidation led to a stronger C2′–C4 interaction, as evidenced by an increase in the IBSI. This suggests that the C2′–C4 bond formation between phpy and Cp* is easier in the Co(IV) state compared with the Co(III) state. Furthermore, systems for which the RE activation barrier is ΔG ‡ < 10 kcal/mol in the Co(IV) state undergo an increase in the C2′–X interaction strength. Conversely, systems for which the RE activation barrier in the Co(IV) state is ΔG ‡ > 10 kcal/mol undergo a decrease in the same interaction (SI; Figure S3). By focusing exclusively on the Co(IV) reactant complexes [I]+, a first qualitative investigation was carried out, looking at the IGM–Δg inter isosurfaces (Figures [1a] and 1b and SI; Figures S8–S35), which materialize attractive, repulsive, or nonbonding interaction areas between two user-defined fragments. A slightly attractive isosurface can be observed between C2′ and C4 for all systems. The attractive feature between C2′ and X, however, appears only for systems that present a lower activation barrier, the extent of the attractive area seemingly varying with the barrier height. This observation might be regarded as qualitative evidence of the propensity of the Co(IV) reactant complex to undergo an RE: the larger the area, the more likely the reactant complex is to evolve toward RE. In all studied cases, the Δg inter score for the phpy–Cp* interaction exceeds the value for the phpy–X interaction. This might be attributed to the size of the contact area between the ligands; indeed, Cp* could approach phpy to bind to two sites in the CC scenario, whereas X only interacts significantly with one site in the RE case (SI; Table S3).

In an attempt to check the reliability of the IBSI as a reactivity descriptor, we studied the correlation between the activation energies toward RE and CC in the Co(IV) state and the IBSI values for the interactions between atoms involved in the respective reaction site. Even if systems with smaller CC energy barriers tend to present higher C2′–C4 IBSI values, no clear correlation could be drawn between the IBSI and the reactivity toward CC (SI; Figure S6). The only information that could be extracted is that systems for which IBSI(C2′–C4) > 0.05 tend to present CC activation energies ΔG ‡(CC) > 20 kcal/mol. This lack of correlation might be attributed to the potential involvement of the N atom of phpy in the first step of the CC process, as some doubts remain regarding its actual mechanism. IBSI, as a pairwise interaction strength descriptor is irrelevant for multicenter delocalized interactions.[12b] For this reason, its use for the CC process, even in the mechanistic scenario considered here, is subject to caution. In the case of the RE process, IBSI values of C2′–X correlate remarkably well with the RE activation barriers, the RE activation barriers decreasing with increasing C2′–X interaction strength (IBSI), and coming close to zero for the strongest IBSI values (Figure [1c]). From a mathematical point of view, such a behavior is best reproduced by using a decreasing exponential formula. We therefore considered the possibility of correlating ΔG ‡ to the IBSI through a decreasing exponential relation, [i.e., ΔG ‡ = a.exp(–b.IBSI) (r² = 0.94)] (Figure [1c]). The IBSI therefore appears to be a decent predictive descriptor of the reactivity of [I]+ toward RE. From Figure [1c], another trend can be identified: the weaker the interaction [i.e., a low IBSI(C2′–X) value], the likelier it is that any other opportunistic pathway will supplant the RE, in a manner similar to that suggested by our experimental results (see below). A frontier emerges at around IBSI(C2′–X) = 0.05: those Co(IV) [I]+ systems with higher IBSI values tend to have ΔG ‡(RE) < 10 kcal/mol and are, therefore, more likely to undergo RE. On comparing the C2′–X and C2′–C4 IBSI values, the general trend appears to be that the two interactions vary in an opposite manner: [I]+ systems with stronger C2′–X interactions tend to have a weaker C2′–C4 interaction compared with other systems, and vice versa (see the SI). In particular, systems with IBSI(C2′–X) < 0.03, all show IBSI(C2′–C4) > 0.05. These systems possess an RE activation barrier of ΔG ‡(RE) > 20 kcal/mol and a CC activation barrier of ΔG ‡(CC) < 20 kcal/mol. They are, therefore, more likely to undergo CC than RE in the Co(IV) state.

To verify the trends suggested by the IGM/IBSI study, a limited series of Cp*Co(III) metallacycles derived from 2-phenylpyridine and containing various X ligands were synthesized with the aim of identifying which pathway is favored under oxidizing conditions (Scheme [2]). A first set of compounds was synthesized by displacement of the iodo ligand by treatment of complex 1a with either an alkali salt or a silver salt. Co(III) compounds 1c–f were all isolated in a pure form and submitted to oxidative conditions by using Ag(I) salts to initiate the formation of the key Co(IV). In all cases, the potential products of RE were absent. The dominating product was that of the hydrodemetalation of phpy, i.e. 2-phenylpyridine. The formation of cation 3 was detected only in the case of the carboxylato complexes 1d–f.

The introduction of carbanionic X ligands was also attempted with Grignard reagents such as PhMgBr or CH2=CHMgBr. Quite intriguing was the reaction of 1a with PhMgBr, which produced a large amount of 2-biphenyl-2-ylpyridine (2g)[13] and minute amounts of 1g after about 15 hours of reaction. Attempts to isolate 1g by flash chromatography gave tiny amounts of a material that was characterized by 1H and 13C NMR and electrospray (positive-mode) MS analyses. Monitoring a solution of 1g in CD2Cl2 kept at room temperature for over one month did not reveal any sign of spontaneous conversion into 2g, thereby ruling out mere decomposition of 1g by RE as the origin of 2g. Therefore the formation of 2g in situ necessarily implies the interference of an occasional oxidant: two hypotheses may be put forward: either (1) unreacted 1a somehow acts as an oxidant for 1g and/or (2) in the conditions of the reaction, a Schlenk equilibrium,[14] as established for PhMgBr in THF,[15] produces an electrolyte[16] capable of oxidizing 1g as it forms to give 2g. These conjectures, which require further investigation, are reminiscent of reports by the Nakamura group[13] [17] and by Ackermann, Neidig, and their co-workers[18] on Fe(II)-mediated C–H bond arylations with organomagnesium and -zinc reagents, which all require an oxidative step. It is worth noting that mixing 1a with CH2=CHMgBr (even in slight excess) gave no reaction. Suspecting a difficult nucleophilic displacement of the Co-bound iodo ligand by CH2=CHMgBr, we carried out the same experiment in the presence of one equivalent of Ag[PF6]. This gave a quantitative yield of 2-(2-vinylphenyl)pyridine (2h),[19] suggesting that the replacement of the iodo ligand at 1a and the presence of a one-electron oxidant are both required for the formation of 2h.

Zoom Image
Scheme 2 Experimental assessment of the theoretical predictive approach for X = NHC(O)Me, OAc, OC(O)CF3, OC(O)Ph, Ph, and CH=CH2.

The experimental results disclosed here match the theoretical features shown in Figure [1c]. Indeed, [I]+ systems close to the asymptote, with IBSI(C2′–X) > 0.05 and, therefore, ΔG ‡(RE) < 10 kcal/mol (1b, 1g, and 1h) do actually yield the RE product upon oxidation. At the other end of the graph, [I]+ systems with IBSI(C2′–X) < 0.03 and, therefore, ΔG ‡(RE) > 20 kcal/mol (1a, 1d–f, and 1j) do not undergo RE upon oxidation but, instead, tend to undergo the hydrodemetalation of the phpy ligand and, to a lesser extent CC, as confirmed by experiment. What might happen to systems for which 0.03 < IBSI(C2′–X) < 0.05 remains to be addressed.

In summary, this study has identified the existence of an exponential correlation between RE activation barriers and IBSI, which was validated by a preliminary experimental assessment for a limited number of cases of complexes with X ligands expected to give either an effective RE process with a low activation barrier in the Co(IV) state or no RE process at all. This correlation, although qualitatively meaningful and of fundamental importance, remains to be fully rationalized and validated with different molecular systems from which reactant complexes related to RE can be readily made. The C2′–X IBSI emerges as a reliable reactivity descriptor of the propensity of Co(IV) [CoCp*(phpy)X]+ systems [I]+ to undergo RE: this trend remains to be assessed for cases in which structural changes in the peripheral ligands of Co are considered. This study also confirmed that Co(III)-to-Co(IV) oxidation lowers the barrier of activation toward the RE pathway.[20] It appears that when the RE pathway is not favored with [I]+, other processes requiring lower activation barriers might take over, impacting catalytic efficiency: the most evident of these are the hydrodemetalation of the phpy ligand and a CC pathway that causes an irreversible collapse of the catalytically relevant Cp*Co motif.[9b]


#

Conflict of Interest

The authors declare no conflict of interest.

Acknowledgment

Professor Eric Hénon (University of Reims) is thanked for uplifting discussions on IGM and IBSI.

Supporting Information

    Supporting information for this article is available online at https://doi-org.accesdistant.sorbonne-universite.fr/10.1055/a-1937-9296.
  • Supporting Information
  • XYZ File

Corresponding Authors

Yann Cornaton
Laboratoire de Chimie et Systémique Organométalliques, UMR 7177 CNRS, Université de Strasbourg
4 rue Blaise Pascal, 67000 Strasbourg
France   

Jean-Pierre Djukic
Laboratoire de Chimie et Systémique Organométalliques, UMR 7177 CNRS, Université de Strasbourg
4 rue Blaise Pascal, 67000 Strasbourg
France   

Publication History

Received: 13 July 2022

Accepted after revision: 06 September 2022

Accepted Manuscript online:
06 September 2022

Article published online:
11 October 2022

© 2022. Thieme. All rights reserved

Georg Thieme Verlag KG
Rüdigerstraße 14, 70469 Stuttgart, Germany


   Permissions and Reprints All articles of this category
Zoom Image
Scheme 1 Proposed C–H bond-functionalization catalytic cycle involving Cp*Co complexes with a cyclocondensation (CC) quenching step, the extent of which depends on the nature of X. Note that for X = Cl in 1i, cation 3 is also the dominant product, similarly to 1a.[9b] Cp* = η5-pentamethylcyclopentadienyl.
Zoom Image
Figure 1 (a) and (b) are IGM-Δg inter isosurface plots (cutoff: Δg inter = 0.01 a0 –4) of the interaction between the phpy and X ligands: (a) X = I, (b) X = CH=CH2. (color code: blue = attractive interaction, green = nonbonding interactions, red = repulsive interactions). (c) Correlation plot between the RE Gibbs free energy of activation (ΔG ‡) and the C2′–X intrinsic bond strength index (IBSI) for a series of different X ligands in [I]+ species. The blue dashed line represents a trend curve of the equation ΔG ‡ = a.exp(–b.IBSI) (a = 69 kcal/mol; b =42; r² = 0.94). The red and green dots outline the cases for which experimental results are reported in this work (X = NHAc, OAc, OC(O)Ph, OC(O)CF3, Ph, CH=CH2) or in the literature (X = I, Cl, SCF3), respectively.[9]
Zoom Image
Scheme 2 Experimental assessment of the theoretical predictive approach for X = NHC(O)Me, OAc, OC(O)CF3, OC(O)Ph, Ph, and CH=CH2.