Acrylate Formation from CO2 and Ethylene ... - ACS Publications

---

Article pubs.acs.org/Organometallics

Acrylate Formation from CO2 and Ethylene Mediated by Nickel Complexes: A Theoretical Study Philipp N. Plessow,†,‡ Ansgar Schaf̈ er,† Michael Limbach,‡,§ and Peter Hofmann*,‡,∥ †

Quantum Chemistry, BASF SE, GVM/M-B009, Carl-Bosch-Strasse 38, D-67056 Ludwigshafen, Germany CaRLa (Catalysis Research Laboratory), Ruprecht-Karls-Universität Heidelberg, Im Neuenheimer Feld 584, D-69120 Heidelberg, Germany § Synthesis & Homogeneous Catalysis, BASF SE, GCS/C-M313, Carl-Bosch-Strasse 38, D-67056 Ludwigshafen, Germany ∥ Organisch-Chemisches Institut, Ruprecht-Karls-Universität Heidelberg, Im Neuenheimer Feld 270, D-69120 Heidelberg, Germany ‡

S Supporting Information *

ABSTRACT: The first step of the nickel-mediated synthesis of acrylates from carbon dioxide (CO2) and ethylene is the formation of nickelalactones. Nickelalactones could then react to give complexes containing acrylic acid moieties. Such a spontaneous reaction, however, is usually not observed and most progress has recently been made by cleaving the lactones with auxiliaries such as Lewis acids and strong bases. Here we investigate in detail the coupling of CO2 and ethylene and further reactivity in absence of auxiliaries mediated by nickel complexes bearing bidentate ligands. We have found a new mechanism for lactone formation which, for some bidentate ligands, is favored over the mechanism described in the literature. Furthermore, we found that β-H elimination leading to ring contraction of the lactone is almost feasible at room temperature. Importantly, however, all investigated mechanisms for formation of acrylic acid complexes require higher activation free energies of ca. 130 kJ/mol and are not accessible at ambient or slightly elevated temperatures.

■

INTRODUCTION

monodentate DBU ligand (1,8-diazabicyclo[5.4.0]undec-7-ene) has shown2 that productive formation of acrylic acid complexes is not feasible at ambient or slightly elevated temperatures (ΔG⧧ = 164 kJ/mol). It has also been pointed out that the formation of acrylic acid from CO2 and ethylene is endergonic and therefore cannot be utilized for a productive thermal catalytic process. Recent experimental work has focused on the use of bidentate ligands, such as bis-phosphines. Two main problems have emerged: (1) nickelalactones are formed from CO2 and ethylene only for a minority of bis-phosphine ligands and (2) nickelalactones either are stable or react in unwanted side reactions.5,6,43 In no case do they yield the desired π complex of acrylic acid (or any other form of acrylic acid that could be used in a catalytic process). Productive cleavage of nickelalactones has been achieved by adding methyl iodide,3 which leads to stoichiometric formation of methyl acrylate, and by the addition of strong bases, which leads to catalytic formation of sodium acrylate.4 In both processes it is clear a priori that the auxiliaries are required to allow liberation of acrylates thermodynamically. Our computational studies have furthermore shown that MeI4 and strong bases5 kinetically facilitate cleavage of nickelalactones with activation barriers that are feasible at room temperature. The most recent experimental work has revealed that the addition of neutral strong Lewis acids induces β-H elimination, which leads to ring contraction of the five-membered lactone to the four-

The synthesis of acrylates from ethylene and CO2 and related chemistry has already been studied in the 1980s by Hoberg and co-workers.1 One of the important results has been the isolation of the coupling products of ethylene and CO2, nickelalactones (see Scheme 1). A computational study by Graham et al. based on Hoberg’s experimental findings with the Scheme 1. Introduction of the Simplified Representation Used for dtbpe Throughout This Work and of the Most Important Organometallic Speciesa

a

Note that the formation of acrylic acid is endergonic under typical reaction conditions. © 2014 American Chemical Society

Received: February 11, 2014 Published: April 16, 2014 3657

dx.doi.org/10.1021/om500151h | Organometallics 2014, 33, 3657−3668

Organometallics

Article

membered lactone.6 Furthermore, the reverse reaction of the desired transformation, i.e. the reaction of an acrylic acid π complex back to a nickelalactone, results from the addition of Lewis acids. The bidentate phosphine ligand dtbpe (1,2-bis(di-tertbutylphosphino)ethane; see Scheme 1) is one of the few ligands which enable the direct synthesis of lactones from CO2 and ethylene and for which probably most mechanistic information is available.4,5 We limited the broad investigation of reaction mechanisms to dtbpe and describe results for other ligands only for selected steps. In this study we want to revisit the thermodynamics of acrylic acid formation as well as the lactone formation and its cleavage mainly on the basis of the bidentate phosphine dtbpe but also monodentate ligands such as PMe3. Among the various routes for cleavage the effect of a strong Lewis acid (BF3) is interesting to us in comparison to mechanisms for deprotonation with strong bases, which we have described previously in less detail and at other levels of theory.5

There are, however, cases where the structure in solution differs significantly from the gas-phase structure. In these cases, the solvation free energy has been determined using reoptimized structures (BP86/def2-SV(P)/COSMO) and free energies obtained in this way are denoted Gs = ERPA + ΔGstat + ΔGsolv,reopt. In the most extreme case, a gas-phase structure corresponding to the structure in solution does not exist. The free energy, Gss, was then obtained by comparison of the free energy obtained with COSMO-RS at the BP86/def-TZVP/ COSMO//def2-SV(P)/COSMO level in solution, Gsolution, with a suitable reference free energy where a gas-phase structure exists, so that Gss = Gsolution + Gref − Gref,solution. This approach was necessary only in a few cases: we chose C1 as reference point for TS-B1-C2 and LC1 for TS-LC1-B1 and TS-LB1-LC2. We followed this slightly complicated hierarchical protocol, because (1) we have no general access to precise free energies in solution without computing precise gas-phase total energies as with the RPA and (2) it is not feasible to reoptimize every transition state in solution. The connectivity of transition states has been confirmed by a displacement along the transition vector followed by steepest descent optimization. In order to decide how relevant a mechanistic scenario is, we consider the half-life (t1/2) of a first-order reaction with the given activation barrier at the given temperature. We consider a reaction relevant if t1/2 < 12 h. This means for T = 25 °C that ΔG⧧ < 100 kJ/mol and for T = 80 °C that ΔG⧧ < 119 kJ/mol. Care must be taken, since many of the studied reactions are not first order, so that this procedure is not applicable for arbitrary concentrations of the involved species. However, most of the reactions are only first order in metal complex so that typically pseudo-first-order kinetics apply in the presence of reactants that have a much higher concentration. Thermodynamics of Catalytic Acrylate Formation. As has already been pointed out, the formation of acrylic acid from carbon dioxide and ethylene is endergonic under typical experimental conditions.2 While the introduction of auxiliaries was mainly motivated as a means of cleaving the kinetically inert nickelalactone, they are also essential to make the overall reaction exergonic and to therefore allow catalysis. Although this is certainly not the way how these reactions work, the thermodynamics of auxiliary-assisted reactions are understood best as a subsequent reaction of acrylic acid (see Scheme 2).

■

RESULTS AND DISCUSSION Computational Details. Molecular structures were optimized at the BP86/def2-SV(P)7−13 level of theory. Calculations were carried out using the resolution of the identity approximation (RI) and appropriate auxiliary basis functions14 with TURBOMOLE.15 Statistical corrections to the gas-phase Gibbs free energies, ΔGstat, were obtained within the usual rigid-rotator, harmonic-oscillator approximation. Absolute electronic energies were obtained using the random phase approximation (RPA) with PBE/def2-QZVPP16−18 KS orbitals with TURBOMOLE.19,20 We have decided to use this method because of its good performance in benchmark calculations which are available in the Supporting Information. For benchmark calculations, single-point energies have been obtained at the coupled cluster level of theory, CCSD(T). Furthermore, structures have also been optimized at the MP2/ def2-TZVP level of theory, where indicated. RPA, MP2, and CCSD(T) calculations were carried out within the RI approximation, corresponding auxiliary basis sets,21−23and the frozen core approximation (only orbitals with eigenvalues >−3Eh were correlated). For reactions involving larger molecules, namely the reaction of (dtbpe)nickelalactone C1 with NaOMe, MeOH, or acrylic acid, the smaller def2-TZVPP basis was employed for RPA calculations. The differences in reaction energies obtained with the RPA between the def2TZVPP and def2-QZVPP basis sets werewhere it could be checkednever large, typically smaller than 5 kJ/mol. Free energies in solution (THF) were obtained by adding solvation free energies calculated with COSMO-RS24,25 and the parametrization for BP86/def-TZVP, which requires calculation of BP86/def-TZVP/COSMO and gas-phase energies. In general, it is sufficient to compute the solvation free energies using the gas-phase optimized structures. This is the standard way in which the Gibbs free energies, G = ERPA + ΔGstat + ΔGsolv,SP, described here were obtained. The free energies are given in kJ/ mol for a reference state with a molar fraction of χ = 0.1, which corresponds roughly to c ≈ 1 mol/L in THF. If not explicitly stated otherwise, all free energies will be given at this level of theory. We will always discuss free energies, G, relative to a common reference point (G = 0) defined in the context. Free energy differences which do not refer to this reference point will be denoted ΔG.

Scheme 2. Thermodynamics of Acrylic Acid Formation along with Potential Consecutive Reactions

We will now discuss contributions to the reaction free energy for the formation of acrylic acid. The gas-phase reaction energy at the RPA(PBE/def2-QZVPP)//BP86/def2-SV(P) level is ΔE = 22 kJ/mol and at the CCSD(T)/def2-QZVPP level is ΔE = 11 kJ/mol for the same geometry. Our most accurate electronic reaction energy (CCSD(T)/cc-pV5Z//MP2/def2-TZVP)26,27 is ΔE = 10 kJ/mol. From statistical thermodynamics (harmonic 3658

dx.doi.org/10.1021/om500151h | Organometallics 2014, 33, 3657−3668

Organometallics

Article

for a model complex with two PH3 ligands are ΔEerr = 20 kJ/ mol relative to CCSD(T)/def2-QZVPP benchmark calculations. While CCSD(T) is certainly not as reliable for the nickel complexes (D1-diagnostic ≈ 0.3)31 as it is for organic molecules, these results nevertheless indicate that reaction energies for the incorporation of CO2 into the ethylene complex are underestimated by about 20 kJ/mol. A comparable error for dtbpe would explain the observed reactivity. The formation of nickelalactones has been studied in the literature for two coordinated, monodentate DBU ligands,2 for bipyridine (bipy),32 and for chelating bis-phosphine ligands.5 It is widely accepted that the reaction starts from the ethylene complex, since it is found to be more stable in calculations and its formation prior to lactone formation is observed experimentally. For the more oxophilic molybdenum, lactone formation occurs via prior formation of a stable CO2−ethylene complex by ligand exchange and by subsequent coupling of CO2 and ethylene to give a lactone.33,34 For nickel(0) complexes with bidentate ligands this is less likely due to the lack of a free coordination site. Therefore, the insertion is believed to occur in one step, where the CO2 molecule concertedly coordinates to nickel and inserts into a Ni−C bond (G⧧ = 124 kJ/mol). For reasons discussed below, we will refer to this mechanism as an “inner-sphere” mechanism. The same mechanism has been proposed for the reaction with DBU as a ligand, where two DBU ligands according to calculations stay coordinated.2 Depending on the ligand and the level of theory, we found a weak associative precoordination of CO2, which however certainly does not indicate a stable compound of any significance (see the Supporting Information for details). For bipy, such a structure has been described as a flat area on the potential energy surface (PES), but not as a minimum.32 We have studied a mechanism with precoordination of CO2 to a vacant coordination side created by the dissociation of one arm of the bidentate ligand, but even for a strained four-membered chelate formed by bis(di-tert-butylphosphino)methane (dtbpm), this turned out to be difficult with a transition state ΔΔG⧧ = 15 kJ/mol higher in free energy than that of the innersphere mechanism (see Schemes 3 and 4). We have furthermore studied a mechanism where CO2 attacks the coordinated ethylene in the “outer sphere” without precoordination of CO2 (G⧧ = 110 kJ/mol). This reaction is perhaps most easily understood as the reverse reaction of the decarboxylation of the β-H agostic, formally zwitterionic intermediate B1 (see Scheme 4 and Figure 1). A similar mechanism for attack of CO2 at amide and hydroxide groups of nickel compounds with subsequent insertion of CO2 into the Ni−N/Ni−O bonds via zwitterionic intermediates has been computed recently.35 Carboxylation of Ni(II) and Pd(II) allyl complexes has also been proposed to proceed via zwitterionic intermediates resulting from the attack of CO2 at the terminal carbon of the η 1 -coordinated allyl group. 36−41 B1 is isoelectronic with the thermodynamic insertion product in cationic olefin polymerization such as in the Brookhart systems.42 Intermediate B1 is fairly stable, if solvation is taken into account. This means that a second, lower barrier (Gs⧧ = 82 kJ/mol) needs to be overcome for recoordination of the carboxylate unit to nickel. In the gas phase, B1 is not stable but corresponds to a very flat area on the PES. The transition state for the outer-sphere attack of CO2 in the gas phase therefore leads directly to subsequent recoordination of the carboxylate, from A to C1. Still, there is a gas-phase structure corresponding to B1 (upon optimization with COSMO it

oscillator−rigid rotator approximation) a contribution of 48 kJ/ mol arises at both BP86/def2-SV(P) and MP2/def2-TZVP levels of theory. The total gas-phase reaction free energy under standard conditions is therefore around ΔGgas = 58 kJ/mol at the single-point coupled-cluster level of theory. Unfortunately, we could not find experimental gas-phase free energies of formation for acrylic acid and the enthalpies of formation are scattered over a wide range. The computed ΔH = 18 kJ/mol (CCSD(T)/cc-pV5Z//MP2/def2-TZVP) is at the edge of the range of experimental values ΔHexp = 4−16 kJ/mol which can be obtained from different experimental heats of formation of acrylic acid.28 Calculating the solvation free energy on the basis of the COSMO-RS model with single-point energies on the BP86/ def2-SV(P) geometries gives ΔGsolv = −36 kJ/mol in THF. The reaction free energy in THF is therefore around ΔG = 21 kJ/ mol. In order to be able to synthesize acrylates from CO2 and ethylene with the help of auxiliaries, a subsequent reaction has to gain at least this amount of energy. This means for quantitative deprotonation (ΔG ≤ −10 kJ/mol) that the base should have a conjugate acid that is less acidic than acrylic acid by at least 6 units on the pKa scale. In water, this would mean that typical amine bases are just capable of making the overall reaction quantitative.29 Taking into account concentration effects (high concentrations of base and ethylene and CO2, low concentration of acrylate) will make the reaction even more exergonic. In aprotic solvents, where acrylic acid is much less acidic relative to neutral bases, things are less clear. In DMSO, protonated DBU is only about ΔpKa = 1.6 less acidic than acetic acid (in water, acetic acid is ΔpKa = 0.5 less acidic than acrylic acid).29,30 The strongest neutral base that has been tested for the deprotonation of nickelalactones in ref 5 was tertbutyliminotris(pyrrolidino)phosphorane, the corresponding acid of which should then be approximately 5−6 pKa units less acidic then acetic acid in DMSO.30 According to these results such a strong neutral base could be just capable of making the formation of acrylate from ethylene and CO2 exergonic in aprotic solvents. This would mean that the observed complete lack of reactivity is due to a kinetically unfavorable reaction in the absence of weak Lewis acids, such as Na+.5 Nickelalactone Formation. We will study the formation of the nickelalactone C1 from the ethylene complex A (see Scheme 4). This means that we do not address the formation of the nickel ethylene complex from a precursor. As has been discussed previously, 5 phenyl-substituted bis-phosphines (dppm, dppe, dppp) readily form unproductive bis-chelate complexes, for example Ni(dppe)2, if Ni(COD)2 is used as a precursor. We will not discuss such problems here, because we think that this is more of a practical issue since other precursors are available. Nevertheless, one should keep in mind that the ethylene complexes are not always trivial to make. The computed reaction free energy for lactone formation in general is endergonic (ΔG = 15 kJ/mol for dtbpe), while reaction energies are exothermic. This is in contrast to experimental findings, where the dtbpe-based nickelalactone C1 is formed readily from CO2 and ethylene. One source of error for the reaction free energy in solution is the computed solvation free energy, which partially cancels the very endergonic statistical gas-phase free energy contribution. The other source of error is of course the absolute energy. We have seen above that the error for acrylic acid formation is about ΔEerr = 11 kJ/mol. Similarly, errors for lactone formation 3659

dx.doi.org/10.1021/om500151h | Organometallics 2014, 33, 3657−3668

Organometallics

Article

probably things are most easily understood for the series dtbpm, dtbpe, dtbpp: increasing the bite angle leads to larger steric bulk and the inner-sphere barrier5 systematically increases while the outer-sphere barrier is much less affected. The deviation of the calculated activation barriers from the benchmark (PH3 ligands; CCSD(T)/def2-QZVPP) is for outer-sphere (inner-sphere) mechanism ΔEerr = 17 (22) kJ/ mol. Assuming a similar error for dtbpe gives G⧧ = 93(102) kJ/ mol, in reasonable agreement with experiment and previous results. We will not attempt to systematically correct our results, since the relative comparison of ligands remains unaffected, as do the relative energies of CO2-containing complexes studied later on. β-H Elimination of Nickelalactones. So far there is no direct evidence for the reaction of nickelalactones with bidentate chelate ligands leading to acrylic acid compounds. In the literature such a reaction is expected to proceed by β-H elimination and a subsequent transfer of the hydrogen from the metal to the oxygen of the acrylate moiety.3,43 β-H elimination itself has been found for Lewis acid activated nickelalactones where either a neutral boron acid or a methyl cation is bound to the oxygen.4,6 In both cases ring contraction of the fivemembered to a four-membered nickelalactone is observed. This is expected to occur via β-H elimination, rotation of the acrylate moiety, and reinsertion, as has been confirmed by calculations for the methyl case.4 So far, it is not clear if the lack of reactivity of C1 and the activation by Lewis acids is solely a kinetic or thermodynamic effect or if both apply. We have found three pathways for β-H elimination of neutral nickelalactones that connect five-membered (C1) and fourmembered (C2) species (cf. Scheme 5). One involves decoordination of the oxygen of the lactone to B1 followed by rotation of the acrylate moiety (TS-B1-C2: Gss⧧ = 165 kJ/ mol) and is very similar to that described for methylated nickelalactones.4 This path corresponds to the reaction responsible for branching in olefin polymerization42,44 and is strongly disfavored if solvation effects are not taken into account. The other possibility is a path where the oxygen stays coordinated and the trigonal-bipyramidal intermediate B2 is formed. The hydride in B2 is trans to the oxygen (see Figure 2). From this intermediate, both carbons can insert into the Ni−H bond, to give either C2 or B1, where the latter barrier is higher (TS-B1-B2: G⧧ = 113 kJ/mol). The third path involves formation of the κ1O-coordinated hydride complex B3. Again, the acrylate can insert into the Ni−H bond to give either C2 or B1, the barrier now being slightly higher for C2 (TS-B3-C2: G⧧ = 109 kJ/mol). Overall, the rearrangements where oxygen stays coordinated to nickel (via either intermediate B2 or B3) are similar in activation barrier and are clearly preferred over the direct, zwitterionic pathway. Both of these processes are

Scheme 3. Competing Mechanism for Lactone Formation for the dtbpm Ligand via Opening of the dtbpm Chelatea

a

The molecular structure of the transition state is also depicted. Bond lengths are given in pm. The hydrogen atoms on the dtbpm ligand are omitted, with the exception of the tert-butyl group that has a H−Ni agostic interaction.

rearranges to B1), where the agostic hydrogen is more distant from the carbon and more hydridic and where the carboxylate group is coordinated. We therefore decided to associate this structure with B1 and the same for TS-B1-C1 and we labeled the gas-phase “outer sphere” TS TS-A-B1 as well. For other ligands such as dmpm, dmpe, bipy, and dtbpm (see Table 1 for acronyms), B1 is not stable and the outer-sphere transitions state leads directly to C1. In general, in the gas phase the inner-sphere mechanism is favored, but this changes for typical organic solvents such as THF. The vibrational entropic contribution to the difference in the activation free energies of inner- and outer-sphere mechanisms is small, with ΔΔG⧧ = 4 kJ/mol in favor of the outer-sphere mechanism. The outer-sphere mechanism is significantly more favorable for sterically hindered ligands, and the activation barriers for ligands such as bipy (see Table1) become essentially identical for both mechanisms. More electron rich ligands lead in general to lower barriers for both mechanisms. However, while steric hindrance does not affect the outer-sphere mechanism, it leads to much higher barriers for the inner-sphere mechanism. Although it is in general not completely clear which mechanism is favored for what reason,

Scheme 4. Competing Mechanisms for Lactone Formation for the dtbpe Liganda

a

Reference point: G(A) = 0. 3660

dx.doi.org/10.1021/om500151h | Organometallics 2014, 33, 3657−3668

Organometallics

Article

Figure 1. Structures of the transition states involved in lactone formation in solution (optimized at the BP86/def2-SV(P) level of theory with COSMO). From left to right: TS-C1-B1, TS-B1-A, and TS-C1-A. Bond lengths are given in pm. Hydrogen atoms on the dtbpe ligand are omitted for clarity.

Table 1. Activation Free Energies for the Nickelalactone Formation for Different Ligands and Inner-Sphere (TS-A-B1) and Outer-Sphere (TS-A-C1) Mechanismsa

a

The activation barrier for recoordination (TS-B1-C1) of the carboxylate and the free energy of the nickelalactone (B1) are also given. The values have been obtained from COSMO-RS BP86/def-TZVP/COSMO//BP86/def2-SV(P)/COSMO free energies by applying four different shifts, δG, for the four different reactions so that the values described above are exactly reproduced for dtbpe. Where the exact values are available (values in boldface), the agreement is in general good; at least the relative order for each ligand is right. NHCP/C refers to the isomer where CO2 inserts trans to the NHC carbon and the other way around for NHCP/P (reference point: G(A) = 0). Asterisks in the body of the table indicate that no transition state could be obtained. TS-A-B1 connects in these cases A and C1 directly and the region where TS-B1-C1 was expected is a very flat area on the PES. Attempts to optimize a minimum B1 led in these cases to C1 (see the Supporting Information for more details).

In conclusion, β-H elimination giving rise to equilibria between neutral four- and five-membered lactones C1 and C2 is almost feasible at room temperature or might even operate, taking into account errors of the employed methods. We note that very recently the mentioned isomerization between C1 and C2 has indeed been observed for the ligand bis(dicyclohexylphosphino)ethane (dcpe) at 55 °C after 1 day by Bernskoetter and co-workers.45 In the same work, a mechanism for isomerization has been proposed that corresponds to the direct transformation in Scheme 5 via rotation of the olefin moiety in the zwitterionic transition state. As discussed above, our results suggest that this path is certainly less likely than the other two paths (via B2 or B3). This is in fact not in contradiction with the computational results in ref 43, since therein no transition state was reported and a zwitterionic intermediate was computed with a free energy of 118 kJ/mol, which is already slightly higher than the expected barrier, on the basis of the experimental findings. The pathways investigated for formation of the coordinated acrylic acid complex are not feasible even at elevated temperatures. According to these results, the lack of reactivity between acrylic acid complexes and lactones observed for bidentate nickel complexes, which in the literature has been attributed to a high kinetic barrier for β-H elimination, is more precisely due to a missing low-energy path for H transfer from either the carbon or the nickel atom to oxygen. We took a tentative look at the simplest TS for the reaction of lactone to acrylic acid compound, TS-B1-C3, and computed it for the

computed to have activation barriers which are slightly too high to be accessible at room temperature. Finally, the question arises how an acrylic acid π complex can be formed. Obviously β-H elimination with ring contraction is not productive, since the hydrogen is still bound to a carbon atom. Possibly, the investigated mechanisms are understood best if we study the reverse reaction, starting from the acrylic acid π complex C3. The π-bound acrylic acid can protonate the nickel atom to generate (intermediate) β-H agostic complexes that rearrange to the five-membered (TS-B1-C3: G⧧ = 131 kJ/ mol) and four-membered (TS-C2−C3-I: Gs⧧ = 138 kJ/mol) lactones. Furthermore, acrylic acid can directly protonate its own carbon atoms, leading againvia β-H agostic intermediatesto the four- and five-membered nickelalactones. This is not unreasonable for protonation of the β-carbon (leading to the four-membered ring C2), as the transition state for proton transfer is a five-membered ring (TS-C2-C3-II: G⧧ = 154 kJ/ mol). The third variant for hydrogen transfer to the oxygen is described best as an internal deprotonation of intermediate B3 where the carboxylate deprotonates the hydride to give the highly unstable intermediate B4. This is similar to the mechanism proposed for complexes bearing one DBU ligand.2 The transition state TS-B4-B5 (G⧧ = 133 kJ/mol) refers to a rotation of the κ1O-coordinated acrylic acid moiety to give the η2-C,O-coordinated acrylic acid complex B5, which easily rearranges to the favored η2-C,C binding mode. 3661

dx.doi.org/10.1021/om500151h | Organometallics 2014, 33, 3657−3668

Organometallics

Article

Scheme 5. Intramolecular Reaction Paths Connecting Five- and Four-Membered-Ring Lactones C1 and C2 and the Acrylic Acid π Complex C3a

a

Reference point: G(C1) = 0.

Effect of Lewis Acids. If the mechanisms discussed apply, Lewis acids should stabilize structures involving a dangling carboxylate (RCO2−) and consequently the barrier for rotation of the acrylate moiety TS-B1-C2 should decrease. Furthermore, we would expect that the acidity of coordinated acrylic acid increases, which lowers the barriers for the protonation at nickel in the reverse reaction to the nickelalactones. The basicity of the dangling carboxylate should decrease and disfavor intramolecular deprotonation. Indeed, this has been observed experimentally for a 1,1-bis(diphenylphosphino)ferrocene dppf-ligated lactone:6 Upon addition of the neutral Lewis acid tris(pentafluorophenyl)borane (BArF3) to the fivemembered-ring lactone (analogous to C1), ring contraction took place at room temperature within hours. Addition of the Lewis acid to the acrylic acid π complex led to fast rearrangement to the more stable lactones, in a mixture which then equilibrated over time.6 We have investigated computationally the effect of BF3 on the dtbpe-ligated nickelalactone. This can of course not reproduce the results obtained for dppf/BArF3 but will allow us to study the effect of a neutral strong Lewis acid on the reaction paths investigated before. In contrast to the discussion

Figure 2. Structures of the transition state for the formation of the acrylic acid π complex C3 from intermediate B1 and hydride intermediate B2 (both optimized at the BP86/def2-SV(P) level of theory). Bond lengths are given in pm. Hydrogen atoms on the dtbpe ligand are omitted for clarity.

NHCP and the dmpe ligand. The barriers relative to the corresponding lactone are even higher than for dtbpe (NHCP, 144 kJ/mol; dmpe, 138 kJ/mol). 3662

dx.doi.org/10.1021/om500151h | Organometallics 2014, 33, 3657−3668

Organometallics

Article

Scheme 6. Intramolecular Reaction Paths Connecting Five- and Four-Membered-Ring Lactones and the Acrylic Acid Compound Activated by BF3a

a

Reference point: G(LC3) = 0.

species (methanol for water and other carboxylic acids or the Ni−acrylic acid complex instead of acrylic acid).

above, we will shift our reference point (G = 0) to the Lewis acid coordinated acrylic acid π complex LC3 (see Scheme 6). While C1 and C3 differ by only 4 kJ/mol, the Lewis acid coordinated nickelalactone LC1 is 27 kJ/mol more stable than LC3. This can be understood in terms of a more favorable coordination of BF3 to the more basic oxygen. The barriers connecting LC1 and LC3 are lowered relative to LC3. Formation of the four-membered ring via transition state TSLC2-LC3-II (Gs⧧ = 114 kJ/mol) and the five-membered ring via transition state TS-LB1-LC3 (Gs⧧ = 109 kJ/mol) becomes almost feasible at room temperature. However, it is important to note that barriers for the reverse reaction (lactone to acrylic acid complex) are not lowered. Lewis acids are therefore expected to thermodynamically and kinetically favor the transformation from acrylic acid complexes to lactones and to thermodynamically disfavor the reverse, desired reaction. As expected, the barrier for ring contraction via rotation of the acrylate moiety is reduced in comparison to the rotation where the carboxylate group is uncoordinated (TS-LB1-LC2, ΔG⧧ = 112 kJ/mol relative to LC1). Importantly, the barrier is not lower than the other barriers which allow equilibration for the neutral lactone (via B2 or B3), so that one would not expect faster equilibration. Therefore, it cannot be excluded that ring contraction of C1 is simply not observed because of the thermodynamics of the reaction. However, as stated above, the BF3/dtbpe system does not relate to any experiment conducted so far but clearly shows the trend of stabilizing one particular reaction channel for ring contraction. Stabilization of comparable, otherwise zwitterionic intermediates upon coordination of Na+ has also been computed in ref 45. Potential Catalytic Transformation from Lactone to Acrylic Acid Complex. Since the rate-limiting step in the direct transformation of nickelalactones to acrylic acid species is the transfer of a hydrogen atom from nickel/carbon to oxygen, we wondered if the process could be catalyzed in a protonshuttling way by protic species. The transition state for water via a six-membered ring is clearly not feasible (G⧧ = 187 kJ/ mol), but proton shuttling by acrylic acid in an eight-membered ring (G⧧ = 120 kJ/mol) is slightly lower in free activation barrier than the intramolecular paths discussed above (cf. Scheme 7). We expect that these results also apply to related

Scheme 7. Equilibrium between Lactone and Acrylic Acid Complex Catalyzed by Proton Shuttlinga

a

Reference point: G(C1) = 0.

The equilibrium could also be catalyzed via an intermediate ring opening of the lactone (cf. Scheme 8): The barrier for ring opening by acrylic acid is quite low (G⧧ = 65 kJ/mol). However, the subsequent deprotonation step yielding the acrylic acid complex and the catalytically active free acrylic acid is again as high in energy as the transition state for proton shuttling. Ring opening by methanol (Scheme 9) is much higher in energy (G⧧ = 104 kJ/mol). Acrylic acid is formed by subsequent β-H elimination followed by reductive elimination of methanol. These two steps are too high in energy to occur even at elevated temperatures. Deprotonation of Nickelalactones. In order to allow direct comparison with the results described above, we have recomputed the mechanism for direct deprotonation of nickelalactone C1 with NaOMe, which we have reported previously at a different level of theory5 (cf. Scheme 10). Initially, NaOMe coordinates most likely to C1 in an exergonic reaction (26 kJ/mol). Alternatively to the direct deprotonation mechanism, C1 could be ring-opened by NaOMe followed by elimination of MeOH. However, the reductive elimination of MeOH is too high in energy to be a realistic option at room temperature and is less likely than the direct deprotonation, which requires ΔG⧧ = 81 kJ/mol relative to N1 (see Figure 3 for a structure of the transition state). As noted in ref 5, a NaOMe tetramer is much more stable (73 kJ/mol per 3663

dx.doi.org/10.1021/om500151h | Organometallics 2014, 33, 3657−3668

Organometallics

Article

Scheme 8. Equilibrium between Lactone and Acrylic Acid Complex Catalyzed by Acrylic Acid through Ring Opening and Subsequent Intramolecular Deprotonation and Dissociation of the Catalytic Acrylic Acida

a

Reference point: G(C1) = 0.

alactone is the only mechanism which explains the fast transformation of lactone to acrylate complex in the presence of base observed experimentally. PMe3 as a Monodentate Ligand. There are numerous reported nickelalactones which are only coordinated by one additional monodentate ligand and are usually stabilized by oligomerization.46 Most progress in the productive coupling of ethylene and carbon dioxide (direct synthesis and lactone cleavage), however, has been made with bidentate ligands. While those ligands are usually tightly bound and stay in their bidentate coordination mode, we were interested to see how nickelalactones with one monodentate ligand could react, this scenario also functioning as a model for hemilabile ligands. As can be seen in Scheme 11, intramolecular equilibration of acrylic acid complex P4 and lactone P1 is not more favorable for PMe3 as a ligand than for complexes with the bidentate bisphosphine dtbpe. The activation barrier required for proton shuttling from P1 to P4 is similar to that of the dtbpe system. However, as P4 is 29 kJ/mol less stable than P1, the reverse reaction is already a reasonable option at room temperature. The nickelalactone isomer P9 (lactone carbon trans to PMe3) is ca. 55 kJ/mol less stable than the other isomer P1. The isomerization barrier between P1 and P9 is negligible (formally negative due to solvation and statistical corrections). One could imagine that different transition states for β-H elimination connect the two lactone isomers P1 and P9 with the hydride isomers P2 and P3. However, all transition state optimizations converged to TS-P1-P2. This can be explained by the negligible isomerization barrier between P1 and P9. Also, TS-P1-P2 “includes” this isomerization concertedly, since lactone carbon oxygen and carbon change positions relative to PMe3. The missing direct transition state to P3 can perhaps be understood in terms of the structural similarity that it would have with the isomerization path TS-P2-P3, which is more favorable. In terms of oligomerization, we have limited our studies to dimerization. The oxygen-bridged lactone dimer Q1 is by far the most stable structure. Indeed, such dimers have been described for nickelalactones bearing one NHC ligand.46 PMe3ligated nickelalactones form tetramers in the solid state and are

Scheme 9. Equilibrium between Lactone and Acrylic Acid Complex Catalyzed by Methanol through Ring Opening, Subsequent Β−H Elimination, and Reductive Elimination of Methanola

a

Reference point: G(C1) = 0.

monomer). However, the direct deprotonation reaction can occur similarly, with the difference that now the edge of the tetramer acts instead of the NaOMe monomer. The activation barrier relative to tetramer N6 is similar to that for the monomeric path (ΔG⧧ = 96 kJ/mol). Deprotonation with subsequent (not concerted) rearrangement is slightly more favorable than the direct, concerted deprotonation mechanism.5 For a NaOMe monomer, formation of the deprotonated lactone N5 is endergonic (ΔG = 39 kJ/mol relative to N1) and occurs without barrier. The rearrangement of the enolate to the sodium acrylate complex N4 requires an additional barrier of ca. 10 kJ/mol. In the reaction path involving the tetramer, deprotonation requires a barrier of ΔG⧧ = 73 kJ/mol relative to N6, which is probably a good approximation to the overall barrier for that mechanism, since we do not expect the transition states TS-N6-N7 and TS-N8-N9, which were not explicitly obtained to be substantially higher.5 Once again, also at the level of theory used here and in direct comparison to mechanisms investigated for the formation of acrylic acid complexes, we find that direct deprotonation of the nickel3664

dx.doi.org/10.1021/om500151h | Organometallics 2014, 33, 3657−3668

Organometallics

Article

Scheme 10. Investigated Reaction Paths for the Reaction of Nickelalactone C1 with NaOMea

a

Reference point: G(C1) = 0.

subsequent protonation steps where one of the monomers protonates off the olefin carbon from the other acrylic acid complex. The highest barrier in this process is still below 100 kJ/mol. Oxidative addition of the carboxy group of acrylic acid to nickel yields the most stable monomeric acrylate complex, P8 (cf. Schemes 11 and 12). The computed isomerization path toward the π-bound acrylic acid complex P4 is not very favorable (ΔG⧧ = 116 kJ/mol relative to P8). Intermolecular insertion of the acrylic acid olefin moiety into the hydride of another metal complex is expected to be a much faster process (ΔG⧧ = 33 kJ/mol). Rearrangement can then take place to form a nickelalactone. Further reactions could be oligomerization by repeated intermolecular insertion into nickel hydrides (not investigated) and insertion of the second, dangling acrylate into the remaining hydride. This leads with low barriers to the stable nickelalactone dimer Q1. This last step is extremely exergonic (ΔG = −139 kJ/mol) and therefore irreversible. In conclusion, the reaction of nickel acrylic acid complexes coordinated by only one monodentate ligand to the lactone appears feasible via either proton shuttling or dimerization but not in an intramolecular fashion. The reverse reaction is thermodynamically and kinetically very unfavorable. The dimer of the lactone and potentially also its higher oligomers are thermodynamic sinks. Acrylic Acid Complexes: Ligand Exchange and Deprotonation. We have computed the activation barrier

Figure 3. Structure of the transition state for the direct deprotonation of the nickelalactone by NaOMe (TS-N1-N4) giving directly the sodium acrylate π complex. Also shown is part of the competing stepwise deprotonation mechanism, where the deprotonated nickelalactone rearranges to the sodium acrylate π complex (TS-N5-N4) (both optimized at the BP86/def2-SV(P) level). Bond lengths are given in pm. Hydrogen atoms on the dtbpe ligand are omitted for clarity.

bridged by the Ni-bound oxygen,47 Starting from monomeric acrylic acid complex P4, dimerization gives the most stable dimeric acrylic acid complex Q4. The structure of Q4 can be understood as two stacked square-planar complexes that are bridged by the carboxylic oxygen of acrylic acid. Reaction toward the stable lactone dimer Q1 can now proceed via two 3665

dx.doi.org/10.1021/om500151h | Organometallics 2014, 33, 3657−3668

Organometallics

Article

Scheme 11. Reaction Paths for Equilibration of Acrylic Acid Complex and Lactone Coordinated by a Single PMe3 Liganda

a

Reference point: G(P1) = 0.

lactone with respect to substituents in positions α and β to Ni (cis/trans) (see Scheme 4 for an illustration). We have shown that the intramolecular transformation of nickelalactones to acrylic acid π complexes is somewhat more favorable than that reported in the literature for other ligands (ΔG⧧ = 130 vs 164 kJ/mol). However, these activation barriers would still be too high for efficient catalysis at temperatures