Ligand and solvent effects on CO2 insertion into group 10 metal alkyl bonds.

The insertion of carbon dioxide into metal element σ-bonds is an important elementary step in many catalytic reactions for carbon dioxide valorization. Here, the insertion of carbon dioxide into a family of group 10 alkyl complexes of the type (RPBP)M(CH3) (RPBP = B(NCH2PR2)2C6H4 -; R = Cy or t Bu; M = Ni or Pd) to generate κ1-acetate complexes of the form (RPBP)M{OC(O)CH3} is investigated. This involved the preparation and characterization of a number of new complexes supported by the unusual RPBP ligand, which features a central boryl donor that exerts a strong trans-influence, and the identification of a new decomposition pathway that results in C-B bond formation. In contrast to other group 10 methyl complexes supported by pincer ligands, carbon dioxide insertion into (RPBP)M(CH3) is facile and occurs at room temperature because of the high trans-influence of the boryl donor. Given the mild conditions for carbon dioxide insertion, we perform a rare kinetic study on carbon dioxide insertion into a late-transition metal alkyl species using ( t BuPBP)Pd(CH3). These studies demonstrate that the Dimroth-Reichardt parameter for a solvent correlates with the rate of carbon dioxide insertion and that Lewis acids do not promote insertion. DFT calculations indicate that insertion into ( t BuPBP)M(CH3) (M = Ni or Pd) proceeds via an SE2 mechanism and we compare the reaction pathway for carbon dioxide insertion into group 10 methyl complexes with insertion into group 10 hydrides. Overall, this work provides fundamental insight that will be valuable for the development of improved and new catalysts for carbon dioxide utilization. This journal is © The Royal Society of Chemistry.

Introduction

Carbon dioxide (CO2) is an attractive target as a chemical feedstock due to its low cost, non-toxic nature, and abundance.[1] Although there are several chemicals currently industrially prepared from CO2, such as urea and carbonates,[2] only a small fraction of our available CO2 is used as a source of carbon for either commodity[1] or fine[3] chemical synthesis. The full exploitation of CO2 as a feedstock is limited by its kinetic and thermodynamic stability, which poses a scientific challenge for developing synthetic routes to make more valuable chemicals from CO2. Utilizing homogeneous transition metal catalysts is one attractive approach for solving kinetic problems associated with the conversion of CO2 into more valuable chemicals and a variety of systems have been developed for the thermal, electrochemical, and photochemical conversion of CO2.[1,3] An elementary step which is proposed in many of these catalytic reactions is CO2 insertion into an M–E σ-bond (for example E = H, OR, NR2, or CR3) to form a product with a M–O bond (eqn (1)).[4] To date, the majority of catalysts feature late transition metals,[1,3] likely due to the relative weakness of the M–O bonds that are formed,[4] which makes cleavage of the M–O bond and subsequent transformation more facile.

The crucial role of CO2 insertion into late transition metal element σ-bonds in catalysis means that understanding the pathway for insertion is important for rationally designing improved catalytic systems.[4] In particular, understanding the impact of the ancillary ligand, solvent, and additives that are commonly used in catalysis, such as Lewis acids, is valuable for the development of optimized catalysts and reaction conditions. Although the influence of these factors on CO2 insertion into metal hydrides has been studied,[5] there is limited information on how these variables affect CO2 insertion into metal alkyl bonds.[6-13] These reactions are particularly important because they can result in the generation of products containing C–C bonds from CO2, which has been identified as a high priority research target by the United States National Academies of Science.[2] For example, group 10 catalysts have been used for the reductive carboxylation of a variety of alkyl halides and pseudo halides.[3] In these reactions, C–C bonds are proposed to form between CO2 and the alkyl electrophile via the insertion of CO2 into a metal alkyl bond.

A major limitation in studying CO2 insertion into group 10 metal alkyl bonds is the paucity of systems that are stable and react under mild conditions.[14] In fact, the majority of well-defined examples of CO2 insertion reactions into group 10 metal alkyls require elevated temperatures, long reaction times, and do not give quantitative yields of products, which makes it difficult to perform kinetic studies (Fig. 1a). Due to the need for high temperatures, there has been a focus on complexes supported by pincer ligands, which often result in complexes with high thermal stability. For example, Wendt and co-workers demonstrated that treatment of (PCP)Pd(CH3) (PCP = 2,6-C6H3(CH2PBu2)2) with 4 atm of CO2 at 80 °C results in complete conversion to the κ1-acetate complex (PCP)Pd{OC(O)CH3}, but the reaction takes two days.[15] Similarly, we reported that the reaction of 1 atm of CO2 with the related compound (PCP)Ni(CH3) requires heating to 150 °C to form the acetate complex (PCP)Ni{OC(O)CH3}, but only approximately 75% of the acetate product is formed.[16] Changing the pincer ligand from PCP to a system containing an sp3-hybridized carbon donor or replacing one of the phosphine arms of the pincer ligand with a hemi-labile nitrogen donor only leads to small changes in the rate of reaction.[17] A notable example of a group 10 methyl complex that does not contain a pincer ligand and inserts CO2 at room temperature is an unusual anionic palladium dimethyl complex, with a pendant Lewis acid, prepared by Jordan and co-workers.[7] Although this system inserts CO2 rapidly it requires careful handling as it is unstable at room temperature. Overall, in order to gain detailed information on the factors that influence the rate of CO2 insertion there is a need for stable alkyl complexes that cleanly react with CO2 under mild conditions.

Fig. 1

(a) Previous examples of pincer-supported group 10 methyl complexes that undergo CO2 insertion to form κ1-acetate complexes. (b) RPBP supported palladium complexes studied in this work, which undergo CO2 insertion under milder conditions.

In this work, we describe the insertion of CO2 into palladium and nickel methyl complexes supported by RPBP (RPBP = B(NCH2PR2)2C6H4−; R = Cy or Bu) pincer ligands, which contain a strongly donating central boryl donor (Fig. 1b). As a result of the strong trans-influence of the boryl donor, these complexes insert CO2 at significantly milder conditions (room temperature) than previously reported pincer systems. This enables us to measure the kinetics of CO2 insertion and study the effect of modulating the pincer ligand and the solvent on CO2 insertion. We demonstrate that the Dimroth–Reichardt ET(30) parameter[18] for a solvent correlates with the rate of CO2 insertion and show that Lewis acids, which are often present in catalysis, do not promote CO2 insertion. Our experimental studies are complemented by computational investigations to understand the pathway for CO2 insertion and identify the structure of key transition states and intermediates. Overall, our work is one of the first quantitative investigations of the impact of the ancillary ligand, solvent, and LAs on CO2 insertion into metal alkyl bonds and will assist in the development of improved and new catalysts for CO2 utilization.

Results and discussion

Synthesis of RPBP supported nickel and palladium methyl complexes and reactivity with CO2

Previous studies on CO2 insertion into pincer supported group 10 metal hydrides demonstrate that the rate of CO2 insertion increases as the trans-influence of the donor opposite to the hydride ligand increases.[19] As a result, nickel and palladium hydride complexes supported by CyPSiP (CyPSiP = Si(Me)(2-PCy2-C6H4)2) ligands, which contain a strongly donating silyl ligand trans to the hydride insert CO2 more rapidly than the corresponding species containing RPCP ligands. We hypothesized that a similar trend may also occur for nickel and palladium methyl species. Unfortunately, the previously described complexes (CyPSiP)M(CH3) (M = Ni or Pd) undergo an intramolecular ligand rearrangement,[20] as does (PSiP)Pd(CH3) (PSiP = Si(Me)(2-PBu2-C6H4)2, see ESI†) which complicates studies of CO2 insertion. Therefore, we prepared nickel and palladium methyl complexes supported by RPBP ligands (RPBP = B(NCH2PR2)2C6H4; R = Cy or Bu) that feature a boryl donor, which has previously been shown to exert a large trans-influence,[21] opposite the methyl ligand.

(CyPBP)Ni(CH3) was synthesized via an analogous route to that previously described for (PBP)Ni(CH3).[21] Initially, NiCl2·DME was treated with CyPBHP to generate (CyPBP)NiCl in 68% yield (Scheme 1). The subsequent reaction of (CyPBP)NiCl with MeMgCl generates (CyPBP)Ni(CH3) in 26% yield. The major cause of the low yield of (CyPBP)Ni(CH3) is the need to perform sequential recrystallizations to remove magnesium salts. Specifically, considerable care is required to remove MgCl2 by-products from the Grignard reagent, which cause the methyl species to convert back to the chloride complex. (CyPBP)Ni(CH3) was characterized in the solid-state using X-ray crystallography (Fig. 2) and has an analogous distorted square planar geometry to (PBP)Ni(CH3).[21] The P(1)–Ni(1)–P(2) bond angle is 158.67(10)°, which is similar to that observed in (PBP)Ni(CH3) (157.58(2)°). The B(1)–Ni(1)–C(1) bond angle is almost linear (179.1(4)°), which presumably results in the boryl ligand exerting the maximum possible trans-influence on the methyl ligand. The Ni(1)–B(1) bond length in (CyPBP)Ni(CH3) is 1.927(9) Å, which is comparable to other species with a Ni–boryl bond.[21] The nickel methyl bond length is 2.019(9) Å, which is significantly shorter than the nickel methyl bond length in (PBP)Ni(CH3), which is 2.059(2) Å. The longer bond length in (PBP)Ni(CH3) is likely due to steric factors, as the larger steric properties of the tert-butyl substituents in (PBP)Ni(CH3) may cause the nickel methyl bond to elongate. The increased steric demand of the PBP ligand was confirmed using buried volume calculations (SambVca 2.1 (ref. 23)) using the crystal structures of (RPBP)Ni(CH3), which show that the buried volume of (PBP)Ni(CH3) is 92.1% compared to 89.3% for (CyPBP)Ni(CH3) (see ESI†).

Scheme 1

Synthesis of (CyPBP)Ni(CH3).

Fig. 2

Solid-state structure of (CyPBP)Ni(CH3) with thermal ellipsoids at 30% probability. Hydrogen atoms are omitted for clarity. Selected distances (Å) and angles (°): Ni(1)–B(1) 1.927(9), Ni(1)–C(1) 2.019(9), Ni(1)–P(1) 2.148(2), Ni(1)–P(2) 2.138(2), B(1)–Ni(1)–C(1) 179.1(4), B(1)–Ni(1)–P(1) 80.1(3), C(1)–Ni(1)–P(1) 100.6(3), B(1)–Ni(1)–P(2) 78.7(3), C(1)–Ni(1)–P(2) 100.6(3), P(1)–Ni(1)–P(2) 158.67(10).

Palladium chloride complexes supported by RPBP (R = Cy or Bu) ligands were synthesized through the reaction of the protonated RPBHP ligand with commercially available [(η3-allyl)PdCl]2 in yields of 69% and 63%, respectively. This is different to a previous route utilized to metallate the PBP ligand, which used the more expensive Pd(COD)Cl2 as the palladium source.[21] Treatment of (RPBP)PdCl with MeLi results in the formation of the (RPBP)Pd(CH3) in yields of 25% for (CyPBP)Pd(CH3) and 60% for (PBP)Pd(CH3) (Scheme 2). In an analogous fashion to (CyPBP)Ni(CH3), it is important to remove alkali halide by-products as they can cause the palladium methyl species to be converted back to the palladium chloride precursor and the multiple recrystallizations required for (CyPBP)Pd(CH3) lowers the yield. (PBP)Pd(CH3) was characterized using X-ray crystallography (Fig. 3). The geometry around palladium is distorted square planar. The P(1)–Pd(1)–P(2) bond angle is 153.74(3)°, which is a greater deviation from linearity than that observed in a related complex with a PCP ligand. Specifically in (PCP)PdCl P–Pd–P is 165.79(2)°.[24] As a consequence, the fourth coordination site in pincer complexes supported by PBP is likely more sterically accessible than in species supported by PCP ligands. The Pd(1)–B(1) bond length in (PBP)Pd(CH3) is 2.027(4) Å, which is significantly longer than the Pd–B bond length in the only other crystallographically characterized palladium complex with a PBP ligand, (PBP)PdCl (Pd–B is 1.972(4) Å). This is consistent with the methyl ligand exerting a significantly stronger trans-influence than the chloride ligand.

Scheme 2

Synthesis of (RPBP)Pd(CH3) (R = Cy or Bu).

Fig. 3

Solid-state structure of (PBP)Pd(CH3) with thermal ellipsoids at 30% probability. Hydrogen atoms are omitted for clarity. Selected distances (Å) and angles (°): Pd(1)–B(1) 2.027(4), Pd(1)–C(1) 2.185(4), Pd(1)–P(1) 2.3030(8), Pd(1)–P(2) 2.3035(8), B(1)–Pd(1)–C(1) 178.09(15), B(1)–Pd(1)–P(1) 77.36(12), C(1)–Pd(1)–P(1) 103.49(10), B(1)–Pd(1)–P(2) 77.33(12), C(1)–Pd(1)–P(2) 102.02(10), P(1)–Pd(1)–P(2) 153.74(3).

We treated complexes of the type (RPBP)M(CH3) (M = Ni or Pd, R = Cy or Bu) with 1 atm of CO2 in C6D6 at room temperature (Scheme 3). In the cases of (RPBP)Pd(CH3) (R = Cy or Bu) NMR spectroscopy indicated clean and essentially quantitative CO2 insertion to the corresponding acetate complexes at room temperature. The reactions took 1 and 6 hours to reach completion for (CyPBP)Pd(CH3) and (PBP)Pd(CH3), respectively. These are the most rapid insertions of CO2 into pincer-supported group 10 metal alkyl complexes observed to date. The acetate products were isolated and (PBP)Pd{OC(O)CH3} was characterized using X-ray crystallography (Fig. 4). The solid-state structure of (PBP)Pd{OC(O)CH3} has the same distorted square planar geometry as (PBP)Pd(CH3) and confirms κ1-binding of the acetate. As expected, the Pd(1)–B(1) bond length (1.975(2) Å) in (PBP)Pd{OC(O)CH3} is significantly shorter than the Pd–B bond length in (PBP)Pd(CH3) (2.027(4) Å) due to the weaker trans-influence of the acetate ligand. Further, the increased Pd–O bond length in (PBP)Pd{OC(O)CH3} (2.1861(14) Å) in comparison to (PCP)Pd{OC(O)CH3} (2.1279(19) Å)[15] is consistent with the increased trans-influence of the PBP ligand in comparison to the related PCP ligand. In contrast, the insertion of CO2 into (CyPBP)Ni(CH3) was not clean. Although NMR spectroscopy suggested that (CyPBP)Ni{OC(O)CH3}, which was not isolated, was the major product, free ligand and several unidentified decomposition products were present (see ESI†). Additionally, it took approximately 4 days at room temperature for all of the (CyPBP)Ni(CH3) to disappear. Overall, from these initial qualitative examples of CO2 insertion into (CyPBP)M(CH3) (M = Ni or Pd) and (PBP)Pd(CH3) we can make the following conclusions: (i) the insertion of CO2 into (CyPBP)Pd(CH3) is faster than the corresponding insertion into (CyPBP)Ni(CH3). This is in agreement with results for CO2 insertion into (PCP)M(CH3) (M = Ni or Pd)[15,16] but stands in contrast to computational studies of CO2 insertion into (PCP)MH (M = Ni or Pd), where insertion is proposed to be faster for nickel.[19] (ii) The insertion of CO2 into (PBP)Pd(CH3) is significantly faster than the corresponding insertion into (PCP)Pd(CH3).[15] This suggests that the rate of CO2 insertion can be increased by having a stronger trans-influence donor opposite the methyl, similar to what has been observed for CO2 insertion into group 10 metal hydrides.[19] (iii) The faster insertion of CO2 into (CyPBP)Pd(CH3) compared to (PBP)Pd(CH3) suggests that steric factors are important.

Scheme 3

Generation of (RPBP)M{OC(O)CH3} (M = Ni, R = Cy; M = Pd, R = Cy or Bu) through CO2 insertion into (RPBP)M(CH3).

Fig. 4

Solid-state structure of (PBP)Pd{OC(O)CH3} with thermal ellipsoids at 30% probability. Hydrogen atoms are omitted for clarity. Selected distances (Å) and angles (°): Pd(1)–B(1) 1.975(2), Pd(1)–O(1) 2.1861(14), Pd(1)–P(1) 2.3256(5), Pd(1)–P(2) 2.3336(5), B(1)–Pd(1)–O(1) 174.99(7), B(1)–Pd(1)–P(1) 77.32(7), O(1)–Pd(1)–P(1) 100.17(4), B(1)–Pd(1)–P(2) 79.00(7), O(1)–Pd(1)–P(2) 103.05(4), P(1)–Pd(1)–P(2) 155.805(18).

In contrast to the quantitative or relatively clean insertion of CO2 observed for (CyPBP)Pd(CH3), (PBP)Pd(CH3), and (CyPBP)Ni(CH3), insertion into (PBP)Ni(CH3) was not clean and was significantly slower than for the other methyl compounds. After 10 days at room temperature, approximately 30% of (PBP)Ni(CH3) was still observed, along with 50% of a product we propose to be (PBP)Ni{OC(O)CH3} on the basis of its similar 31P NMR chemical shift to (CyPBP)Ni{OC(O)CH3}, and 20% of unidentified decomposition products (Scheme 4). Attempts to independently synthesize (PBP)Ni{OC(O)CH3} via reactions between silver acetate and (PBP)NiCl or sodium acetate and (PBP)Ni(OTf) were unsuccessful. Although, initially a new resonance is observed in the 31P NMR spectra of these reactions with the same chemical shift as the proposed insertion product with CO2 (84.9 ppm), this resonance disappears over time, indicating that (PBP)Ni{OC(O)CH3} is likely unstable in solution. This decomposition prohibited isolation and full characterization of (PBP)Ni{OC(O)CH3}. Nevertheless, since the rate of CO2 insertion into (PBP)Ni(CH3) is presumably independent of the decomposition of the corresponding insertion product, we can make qualitative conclusions about the rate of insertion. We hypothesize that the slower rate of CO2 insertion is related to the increased steric bulk of (PBP)Ni(CH3), which makes it harder for CO2 to interact with the complex. When (PBP)Ni(CH3) was treated with CO2 at higher temperatures even larger amounts of decomposition products were observed. Control experiments indicate that (PBP)Ni(CH3) is not stable even when heated under an N2 atmosphere. Although we were unable to identify all of the decomposition products of (PBP)Ni(CH3), an unusual dinuclear species (PBMeP)2Ni2(μ-N2) (PBMeP = MeB(NCH2PBu2)2C6H4), containing two nickel(0) centers and a bridging dinitrogen ligand crystallized out of solution (Fig. 5). In this complex, the nickel methyl ligand has undergone reductive coupling with the boryl ligand to form a new C–B bond. The boron atom no longer coordinates to the nickel center and the two phosphorus donors of the new bidentate PBMeP ligands do not coordinate to the same nickel center but instead coordinate to two different nickel atoms. The geometry around the nickel centers is distorted trigonal planar and the dinitrogen ligand coordinates in an end-on fashion. The N–N bond distance of 1.145(8) Å is slightly elongated compared to that of free N2 (1.098 Å), consistent with weak activation of N2 and what has been observed in related nickel dinitrogen complexes.[25] The decomposition of (PBP)Ni(CH3) into (PBMeP)2Ni2(μ-N2) demonstrates a new reductive pathway for decomposition of complexes supported by RPBP ligands. However, this is likely a particular problem for the PBP ligand when it is coordinated to Ni because of steric constraints associated with the bulky tert-butyl substituents and small size of nickel.

Scheme 4

CO2 insertion into (PBP)Ni(CH3).

Fig. 5

a) Decomposition of (PBP)Ni(CH3) under N2 at 60 °C. (b) Solid-state structure of (PBMeP)2Ni2(μ-N2) with thermal ellipsoids at 30% probability. The model is near the crystallographic inversion center; half of the structure is generated by symmetry. Hydrogen atoms are omitted for clarity. Selected distances (Å) and angles (°): Ni(1)–N(1) 1.854(4), Ni(1)–P(1) 2.293(11), Ni(1)–P(2) 2.164(11), N(1)–N(1A) 1.145(8), P(1)–Ni(1)–P(2) 122.0(5), N(1)–Ni(1)–P(1) 118.9, N(1)–Ni(1)–P(2) 110.8(3), Ni(1)–N(1)–N(1A) 178.7(5).

Kinetic studies of CO2 insertion into (PBP)Pd(CH3)

It has traditionally been challenging to perform kinetics experiments on CO2 insertion into group 10 alkyl complexes either because of the instability of the complexes[14] or the harsh conditions required for insertion.[16,17] In contrast, the insertion of CO2 into (PBP)Pd(CH3) under mild conditions enables kinetics experiments to be performed. Initially, 1H or 31P NMR spectroscopy was used to monitor the reaction of (PBP)Pd(CH3) with CO2 as a function of time (Fig. 6a). It was possible to measure the concentrations of both the (PBP)Pd(CH3) starting material and the (PBP)Pd{OC(O)CH3} product, which enabled us to confirm the reaction is clean and quantitative. An excess of CO2 was used to ensure that the reactions were performed under pseudo-first order conditions with the concentration of CO2 measured using quantitative 13C NMR spectroscopy (see SI).[26] Initial experiments in C6D6 indicate that the reaction is first order in both [(PBP)Pd(CH3)] and [CO2], demonstrating that the overall rate law is k1[(PBP)Pd(CH3)][CO2] (Fig. 6b and 7). The kobs values obtained from a plot of ln([(PBP)Pd(CH3)]) versus time were divided by the concentration of CO2 to obtain values of k1 (see ESI†). Monitoring the rate of the reaction in C6D6 at various temperatures allowed for the activation parameters to be determined through Eyring analysis (see ESI†). The enthalpy of activation, ΔH‡, 10.9 ± 1.1 kcal mol−1, for insertion into (PBP)Pd(CH3) is significantly lower than that for (PCP)Pd(CH3) (ΔH‡ = 17.4 ± 1.7 kcal mol−1),[26] which is the most closely related system for which an experimental activation energy for CO2 insertion has been measured. This is consistent with the hypothesis that the strong trans-influence of the PBP ligand weakens the palladium methyl bond and destabilizes the ground state. The entropies of activation, ΔS‡, for (PBP)Pd(CH3) and (PCP)Pd(CH3) are within error of each other at −32.2 ± 3.2 and −28.2 ± 4.54 cal mol−1 K−1, respectively. The similar negative values suggest that the nature of the rate-limiting transition state in both reactions are analogous and is in agreement with two molecules combining to form one molecule in the transition state (vide infra). The overall difference in ΔG‡ at 298 K between (PBP)Pd(CH3) (20.5 ± 2.1 kcal mol−1) and (PCP)Pd(CH3) (25.8 ± 2.6 kcal mol−1) is consistent with the much faster rate of insertion observed for the PBP system.

Fig. 6

Representative traces for the insertion of CO2 into (PBP)Pd(CH3) at 40 °C in C6D6 with 1 atm of CO2 showing (a) the concentrations of (PBP)Pd(CH3) and (PBP)Pd{OC(O)CH3} as a function of time and (b) the ln of the concentration of (PBP)Pd(CH3) as a function of time.

Fig. 7

Reaction order in [CO2] of the insertion of CO2 into (PBP)Pd(CH3) in C6D6 at 40 °C.

In CO2 insertion reactions into metal hydrides, it has been demonstrated that the solvent can have a major influence on the rate of insertion.[5] To date there are no studies exploring how the rate of CO2 insertion into metal alkyls varies as the solvent is changed. We determined the rate of CO2 insertion (k1) into (PBP)Pd(CH3) in toluene-d8, benzene-d6, 1,4-dioxane, THF, and glyme (Table 1). There was only a small solvent effect with approximately a 1.9-fold difference in k1 between the slowest (toluene-d8) and fastest (glyme) solvent (Table 1). This compares with a 30-fold increase in the rate of CO2 insertion into an iridium hydride in a similar set of solvents.[5] We propose that the smaller impact of solvent in this case is because there is less build-up of charge in the rate-determining transition state for CO2 insertion (vide infra) compared to the corresponding transition state for metal hydrides. To determine if the variation in the rate of insertion correlated with any solvent parameters, ln(k1) was plotted against the dielectric constant, Gutmann Acceptor Number (AN),[27] or Dimroth–Reichardt ET(30) parameter[18] of the solvent. No significant correlation was observed between the rate of CO2 insertion and the dielectric constant or AN of a solvent (see ESI†). In contrast, an excellent correlation was observed between the ET(30) value of a solvent and the rate of CO2 insertion (Fig. 8), suggesting that ET(30) is a good predictor of the rate of CO2 insertion. A similar relationship was observed for CO2 insertion into metal hydrides and hydroxides,[5] raising the possibility that this correlation is general across a variety of distinct metal-element σ-bonds. It is likely that ET(30) is a better predictor of CO2 insertion rates because it explicitly considers specific solute/solvent interactions, which are not accounted for when the solvent is modelled using either dielectric constant or AN.[5] Unfortunately, we were unable to extend the correlation between the rate of CO2 insertion into (PBP)Pd(CH3) and ET(30) to more polar solvents. This is because these solvents led to the decomposition of (PBP)Pd(CH3) or resulted in rates that were too fast to measure using NMR spectroscopy. For example, CO2 insertion into (PBP)Pd(CH3) is complete in less than 10 minutes at room temperature in pyridine and benzonitrile, which is too fast to obtain reliable kinetics by NMR spectroscopy. Nevertheless, given the slow rates of CO2 insertion previously observed into pincer-supported group 10 alkyls, the completion of reactions in less than 10 minutes at room temperature is unprecedented.

Effect of solvent on rate of CO2 insertion into (PBP)Pd(CH3) at 40 °C and 1 atm of CO2

Entry	Solvent	k 1 a (M−1 s−1 × 10−2)	Rate enhancement
1	Toluene-d8	1.1	1
2	Benzene-d6	1.3	1.2
3	1,4-Dioxane	1.5	1.4
4	Tetrahydrofuran	1.8	1.6
5	Glyme	2.1	1.9

These values are the average of two trials and the errors are ±10%.

Fig. 8

Plot of ln(k1) for the insertion of CO2 into (PBP)Pd(CH3) versus the Dimroth–Reichardt ET(30) solvent parameter for various solvents at 40 °C and 1 atm of CO2.

Lewis acids have been proposed to increase the rate of CO2 insertion into a variety of different metal-element σ-bonds, but these effects are often dependent on the pathway of insertion and the reaction conditions.[5,7,8,9] As a result, it is difficult to predict when a Lewis acid will increase the rate of a CO2 insertion reaction. We evaluated the impact of a variety of different Lewis acids on the rate of CO2 insertion into (PCP)Pd(CH3) using THF as the solvent (Table 2).[29] Although there is an increase in rate when the reaction was performed in the presence of 20 equivalents of a variety of Lewis acids, a control experiment using 20 equivalents of (Bu4)PF6, which does not contain a Lewis acid, gives the same enhancement. This indicates that there is no increase in rate caused by Lewis acids and instead it is likely the case that increasing the ionic strength of the solution results in a slight enhancement in the rate of CO2 insertion. Consistent with this hypothesis, all of the different Lewis acids evaluated give approximately the same increase in the rate of CO2 insertion. At this stage, it is unclear why no Lewis acid effect is observed in the insertion of CO2 into (PBP)Pd(CH3), given that Li+ has been demonstrated to increase the rate of CO2 insertion into (CO)5W(CH3).[8] However, in the next section, DFT calculations are performed to understand the exact nature of the transition states involved in CO2 insertion into (PBP)Pd(CH3).

Effect of Lewis acids on rate of CO2 insertion into (PBP)Pd(CH3) in THF at 40 °C and 1 atm of CO2

Entry	Additive (20 equiv.)	k obs b (s−1 × 10−3)	Rate enhancement
1	None	1.6	1
2	(NBu4)PF6	2.3	1.4
3	LiNTf2a	2.2	1.4
4	NaNTf2a	2.1	1.3
5	KNTf2a	2.3	1.4
6	LiPF6	2.4	1.5
7	LiBPh4·3DME	2.5	1.6
8	LiOTf	2.1	1.3

NTf = triflimide.

These values are the average of two trials and the errors are ±10%.

Computational studies of CO2 insertion into (RPBP)Pd(CH3) and related species

Previous studies investigating CO2 insertion into PCP-supported palladium and nickel methyl complexes have proposed two plausible pathways for insertion: (i) a 1,2-insertion (inner sphere) or (ii) an SE2 pathway (outer sphere).[16,17] In the 1,2-insertion pathway, the O-bound acetate product is formed in one step and CO2 interacts with the metal in the key transition state (Fig. 9a). This means that both the C–C and M–O bonds are formed at the same time. In contrast, the SE2 pathway involves two steps and starts with electrophilic attack of CO2 on the coordinated methyl, without any direct interaction between CO2 and the metal (Fig. 9b). This results in the formation of the new C–C bond. In the second step, the intermediate containing an acetate-metal ion-pair rearranges via TS2 to form the O-bound acetate adduct. Here, we used DFT methods to understand the pathway for CO2 insertion into (PBP)M(CH3) (M = Ni or Pd). We also independently calculated CO2 insertion into (PCP)Pd(CH3) in order to directly compare it with the PBP systems.

Fig. 9

Two plausible mechanisms for CO2 insertion into pincer-supported group 10 methyls: (a) 1,2-insertion and (b) SE2.

Our calculations indicate that CO2 insertion into (PBP)Pd(CH3) proceeds via an SE2 pathway (Fig. 10, S43, S44† & Table 3). The rate-limiting step is the initial attack of CO2 on the palladium methyl group (TS1_Out), which has a computed barrier of 19.0 kcal mol−1 (298 K). This is within error of the activation barrier measured experimentally (vide supra). Although there is no interaction between CO2 and the metal in TS1_Out, we note that there are two C–H…OCO interactions between the tert-butyl substituents of the PBP ligand and CO2 (2.43 and 2.74 Å, respectively), which provide charge stabilization. This may explain why the addition of Lewis acids to the system does not result in a rate enhancement (vide supra). The second step of the SE2 pathway has a very small barrier and involves rotation of the acetate-palladium ion pair to form the O-bound acetate product (Fig. 10). In contrast, the barrier for the alternative 1,2-insertion is significantly higher at 25.1 kcal mol−1 (TS1_Inn, Fig. 10 and S45†). The high barrier is likely caused by the bulky tert-butyl substituents of the PBP ligand, which make it difficult for CO2 to approach palladium without steric repulsion. In order to investigate this hypothesis, we calculated CO2 insertion into (MePBP)Pd(CH3), where the tert-butyl substituents on the PBP ligand have been replaced with methyl substituents. For CO2 insertion into (MePBP)Pd(CH3) the 1,2-insertion pathway is preferred by 3.1 kcal mol−1 over the SE2 mechanism (Fig. 10, S46 and S47†). Thus, the size of the pincer ligand has a decisive role in promoting either a 1,2-insertion (small phosphine substituents) or an SE2 pathway (large phosphine substituents). Similar ligand effects were previously observed for NHC–Cu complexes.[30]

Fig. 10

Top: computed free energy profile (B3LYP-D3/def2-TZVPPD[CPCM]//B3LYP-D3/def2-SVP[CPCM], 298 K) for CO2 insertion into (PBP)Pd(CH3), bottom: optimized transition state structures for CO2 insertion into (RPBP)Pd(CH3), with R = Bu or Me. All energies are in kcal mol−1.

Calculated free energy barriers for inner and outer sphere C–CO2 bond formation with different pincer-supported palladium and nickel complexes (kcal mol−1, B3LYP-D3/def2-TZVPPD[CPCM]//B3LYP-D3/def2-SVP[CPCM], 298 K)a

Complex	TS1_Out (kcal mol−1)	TS1_Inn (kcal mol−1)
(tBuPBP)Pd(CH3)	19.0 (23.9)	25.1 (30.6)
(MePBP)Pd(CH3)	22.5 (23.4)	19.4 (21.0)
(tBuPBP)Ni(CH3)	24.5 (27.8)	31.9 (33.7)
(tBuPCP)Pd(CH3)	24.2 (29.2)	36.7 (41.1)
(MePCP)Pd(CH3)	26.8 (30.0)	25.0 (27.8)

For comparison, results for the ωB97XD functional are given in parenthesis, showing the same mechanistic trends.

The PCP-supported complex, (PCP)Pd(CH3), also shows a preference for the SE2 pathway, with a computed barrier of 24.2 kcal mol−1 (TS_Out, Fig. S48 and S49†). Consistent with previous results,[26] the 1,2-insertion mechanism is unlikely for (PCP)Pd(CH3) due to the high barrier of 36.7 kcal mol−1 (TS_Inn, Fig. S50†).[31] Thus, both (PBP)Pd(CH3) and (PCP)Pd(CH3) show a clear preference for an SE2 mechanism. In the rate limiting transition state (TS_Out), (PBP)Pd(CH3) displays a longer C–CO2 bond (by 0.09 Å) and a smaller CO2 angle (by 1.7°) compared to (PCP)Pd(CH3), which indicates an earlier TS. The lower barrier for (PBP)Pd(CH3) (by 5.2 kcal mol−1) is in agreement with our kinetic studies, showing that CO2 insertion is much faster for (PBP)Pd(CH3). The decrease in the barrier for CO2 insertion is also matched by the greater thermodynamic driving force for CO2 insertion into (PBP)Pd(CH3) compared with (PCP)Pd(CH3) (−23.3 kcal mol−1versus −15.8 kcal mol−1). These results highlight how a donor atom with strong trans-influence opposite to the methyl can lower the barrier for insertion and suggest that the enhancement is due to destabilization of the ground state as opposed to a kinetic effect associated with stabilization of the transition state.

The slower qualitative rate of CO2 insertion into (PBP)Ni(CH3) compared to (PBP)Pd(CH3) (vide supra) is surprising given that CO2 insertion into nickel hydrides is typically faster than palladium hydrides.[19] Calculations on CO2 insertion into (PBP)Ni(CH3) confirm our experimental results as the calculated barrier is 24.5 kcal mol−1 at 298 K, more than 5 kcal mol−1 higher than for (PBP)Pd(CH3) (B3LYP-D3, Table 3). The lowest energy pathway for CO2 insertion into (PBP)Ni(CH3) also follows an SE2 mechanism (Fig. S51 and 52†). Geometrical analysis of the rate-determining transition states for CO2 insertion into (PBP)Ni(CH3) and (PBP)Pd(CH3) shows that the Ni–P bonds are shorter by around 0.15 Å compared with the corresponding Pd–P bonds due to the smaller atomic radius of nickel. We propose that the increased steric congestion around nickel results in the higher barrier for CO2 insertion relative to the palladium system.[32] This is in agreement with previous results showing that CO2 insertion reactions into nickel hydrides are highly sensitive to steric factors.[5] It is likely, however, that there is no inherent electronic reason for CO2 insertion into nickel methyl complexes to be slower than palladium methyl complexes, and if a nickel complex could be prepared where steric factors were not relevant,[33] we expect to see comparable rates of insertion.

Conclusions

In this work, we have demonstrated that nickel and palladium methyl complexes supported by RPBP ligands undergo CO2 insertion under significantly milder conditions than any previously reported pincer-ligated group 10 alkyl species. This is because of the strong trans-influence of the central boryl donor in RPBP type ligands, which destabilizes the M–CH3 bond. Qualitatively, we show that CO2 insertion into (RPBP)Pd(CH3) is faster than CO2 insertion into (RPBP)Ni(CH3), which is likely due to steric factors. In contrast, we propose that steric factors are less significant in CO2 insertion into group 10 hydrides, and as a result nickel hydrides typically undergo faster insertion than palladium hydrides.[19] Given the rapid rate of insertion into RPBP-ligated complexes, we were able to complete a rare kinetic study on CO2 insertion into a metal alkyl complex. Our results demonstrate that the solvent parameter that is best for predicting the rate of CO2 insertion is the Dimroth–Reichardt ET(30) parameter. This is analogous to what we have observed for transition metal hydrides,[5] but the dependence of the insertion rate on the solvent is significantly less pronounced in the case of the metal alkyl complex. We hypothesize that this is because there is less charge build-up in the transition state for carbon dioxide insertion into a metal alkyl complex compared to related metal hydrides. Consistent with there being only a small build-up of charge in the transition state, we did not observe any enhancement from Lewis acids in CO2 insertion reactions into (PBP)Pd(CH3). DFT calculations indicate that the proposed mechanism for insertion into (PBP)M(CH3) (M = Ni or Pd) follows an SE2 pathway, but this is dependent on steric factors and compounds with less bulky substituents on the phosphine donors are more likely to undergo insertion via a 1,2-insertion mechanism. Overall, our results indicate that the fastest systems for CO2 insertion into a group 10 methyl will have a strong trans-influence donor opposite the methyl, be able to undergo insertion in solvents with high ET(30) values, and not have sterically bulky ancillary ligands that prevent the approach of CO2. Future research in our group will aim to extend these principles to other metal-element σ-bonds and apply them to improve catalytic reactions.

Data availability

Additional information about selected experiments, NMR spectra, and other details are availabe via the internet.

Author contributions

Project conception: AJD, MRE, DJC, NH; experimental work: AJD, MRE, DJC; X-ray crystallography: BQM; computational analysis: LjP, KHH. All authors contributed to writing the manuscript.

Conflicts of interest

The authors declare no competing financial interests.