Kinetic Analysis as an Optimization Tool for Catalytic Esterification with a Moisture-Tolerant Zirconium Complex

This work describes the use of kinetics as a tool for rational optimization of an esterification process with down to equimolar ratios of reagents using a recyclable commercially available zirconocene complex in catalytic amounts. In contrast to previously reported group IV metal-catalyzed esterification protocols, the work presented herein circumvents the use of water scavengers and perfluorooctane sulfonate (PFOS) ligands. Insights into the operating mechanism are presented.

* sı Supporting Information ABSTRACT: This work describes the use of kinetics as a tool for rational optimization of an esterification process with down to equimolar ratios of reagents using a recyclable commercially available zirconocene complex in catalytic amounts. In contrast to previously reported group IV metal-catalyzed esterification protocols, the work presented herein circumvents the use of water scavengers and perfluorooctane sulfonate (PFOS) ligands. Insights into the operating mechanism are presented.

■ INTRODUCTION:
Esterification of alcohol and carboxylic acid with the release of water is one of the most fundamental organic reactions. In synthesis, the Brønsted acid-catalyzed Fischer esterification reaction has served the community well since 1895. 1 However, this equilibrium process typically requires high excess of one reagent and/or removal of water to drive the reaction to high yields, and is unsuitable for many applications. As a result, techniques that rely on stoichiometric activation of the carboxylic acid are often utilized, such as Steglich esterification and the Schotten−Baumann reaction. 2 A breakthrough was seen in the early 2000s when Yamamoto and co-workers demonstrated that Lewis acids efficiently catalyze dehydrative esterification using equimolar ratios of the starting materials, 3 and the field expanded rapidly to include homogeneous, heterogeneous, and micellar protocols. 4 The seminal papers by Yamamoto identified chloride and alkoxide complexes based on hafnium and zirconium as the most efficient catalysts in refluxing toluene with azeotropic water removal. The required use of a dehydration technique is a common feature for protocols using early transition metal catalysts with halide and alkoxide ligands, due to their tendency to undergo hydrolytic decomposition. 5 In contrast, analogous fluoroalkyl sulfonate complexes render moisture-tolerant catalysts for a variety of applications. 6,7 Despite the benefits that water-stable catalysts present, particularly in the context of dehydrative reactions, Lewis acidic fluoroalkyl sulfonate metal complexes remain a surprisingly understudied catalyst class with respect to (re)activity and mechanistic action. 8 While the use of the onecarbon trifluoromethane sulfonate (triflate) unit is unrestricted, its longer chain analogue perfluorooctane sulfonate (PFOS) displays bioaccumulative and toxic properties 9 and is regulated by, e.g., the European Chemicals Regulation (REACH EC no. 1907 and the United States Environmental Protection Agency's (EPA) PFOA Stewardship Program. 10 As a result, the use of PFOS in early transition metal catalysis is hampered despite its recently demonstrated efficiency as a ligand in, e.g. zirconium-catalyzed esterification. 11 During the last decades, modern and user-friendly methods for intuitive visual kinetic analysis of organic reactions have been developed, starting with Blackmond's pioneering reaction progress kinetic analysis (RPKA), 12 and recently complemented with Bureś' variable time normalization analysis (VTNA). 13 These methodologies enable facile extraction of kinetic data by the utilization of full reaction profiles for a minimal number of experiments under synthetically relevant conditions to generate an in-depth understanding of a chemical system, and have been successfully employed for the mechanistic elucidation of a wide variety of organic transformations. 14 While valuable from a fundamental perspective, mechanistic insights can also be converted into strategic modifications of reaction conditions to improve yields and selectivities. 15 In this work, integrated use of kinetics enables rational optimization of zirconium-catalyzed esterification to provide more insight compared to traditional screening relying on single data points, typically yield at a specified time. In addition, insight into the operating mechanism is provided based on kinetics and NMR spectroscopy.

■ RESULTS AND DISCUSSION
Based on previous work using zirconocene triflate, 6i we expected that the complex would display catalytic activity for dehydrative ester condensation in tetrahydrofuran (THF). Indeed, a first attempt using an equimolar ratio of benzyl alcohol and benzoic acid with a 2 mol % loading of the zirconium complex resulted in 6% benzyl benzoate after 24 h at 80°C. Taking off from this starting point, an assessment of solvents was carried out (Supporting Information (SI)) that indicated byproduct formation at high reactant concentrations. This prompted a switch to 2-phenylethanol as the new benchmark alcohol and a lower reactant concentration for further screenings. By plotting product concentration as a function of time, it became clear that aromatic hydrocarbons, and benzotrifluoride in particular, 16 were favorable for ester formation, whereas the reaction rate decreased with increasing polarity and coordination capability of the solvent (Figure 1, left). Despite low to moderate yields after 24 h in ethers and sulfolane, continuous build-up of product indicated that decomposition or irreversible inhibition of the catalyst was not taking place, suggesting that these solvents may be used as co-solvents. As expected, the reaction rate was shown to be temperature dependent, with virtually no reaction occurring at room temperature (Figure 1, right). At 60°C, a steady accumulation of product was observed, whereas reactions at 80 and 100°C had increasingly faster rates. An approximate 70% yield of 2-phenylethyl benzoate (3a) for the 0.5 M equimolar reaction of 1a and 2a was reached at different reaction times depending on the temperature, whereas higher yields were not observed at any of the evaluated temperatures even with prolonged reaction times. For further assessments, a reaction temperature of 80°C was chosen to allow for compatibility with a wider range of functionalized substrates.
The rate dependencies on reactant concentrations were assessed with different excess experiments 12 and determined to be close to zero by comparison with standard conditions (equimolar ratios of starting materials) (Figure 2, top). Furthermore, no reaction took place in the absence of zirconium. The order in [Zr(Cp) 2 (CF 3 SO 3 ) 2 ·THF] was estimated to be 0.75 (Figure 2, middle) for different global reaction concentrations. 13 The less than first order in [Zr] suggests that the zirconium is partitioned between catalytically active species and inactive forms of higher order, similar to what has previously been described for other catalytic systems. 17 The catalyst stability was probed with two same excess experiments (Figure 2, bottom), 12 where the equimolar esterification of 1a and 2a (0.5 M) (circles) was compared to two separate reactions simulating 25% conversion (0.375 M 1a and 2a) in the absence (squares) and presence (triangles) of the corresponding amount of water (0.125 M). As evident, the time-adjusted profiles for the same excess experiments overlay very well with that of standard conditions, indicating that the catalyst does not undergo significant inhibition or decomposition under standard conditions.  The combined kinetic results suggested that the reaction would tolerate an increase in zirconium and reactant concentrations by a decrease in solvent volume for the formation of ester 3a, without the risk of catalyst deactivation. This was indeed found to be the case (SI) and concentrations of 1 M and 0.02 M for substrates and Zr(Cp) 2 (CF 3 SO 3 ) 2 ·THF, respectively, were chosen as the starting points for the substrate evaluation. Under these conditions, the turnover number (TON) for the first 6 h of the reaction time was estimated to be 19.7, corresponding to a turnover frequency (TOF) of around 3.3 h −1 (SI). 18 The catalyst was found to be moisture stable, as assessed by the addition of water at the outset of the reaction (Figure 3, top left). Similar reaction rates compared to standard conditions were observed for reactions with 25 and 50 equiv of water relative to zirconium, whereas the addition of 250 molar equiv of water quenched catalysis. It has previously been demonstrated that hydrolytic decomposition of titanocene triflate and other Lewis acidic trifluoromethane sulfonate complexes can occur under certain conditions to release triflic acid, 19 which can be used as a Brønsted acid catalyst for Fischer esterification. 20 To probe the nature of the active catalyst in our system, a set of experiments was performed using 2 mol % of either zirconocene triflate or triflic acid (Figure 3, top right). The use of the latter resulted in a slightly faster reaction compared to the zirconium complex, hence reaching a higher yield of 3a over 24 h under standard conditions at 0.5 M. The addition of 250 equiv of water relative to the catalyst resulted in the product in 37% yield in the presence of triflic acid after 24 h, whereas only traces of 3a were observed in the presence of the zirconocene complex. Interestingly, the addition of molecular sieves (MS) suppressed the reaction rate of the zirconiumcontaining reaction significantly and almost completely quenched triflic acid catalysis, resulting in only trace amounts of product 3a after 24 h (Figure 3, top right; for additional information, see SI). This decelerating behavior stands in contrast to what is typically observed for dehydrative transformations, where water removal shifts the equilibrium toward product formation. 6i While the origin of catalytic inhibition by molecular sieves was not the subject of further investigations, the different responses from triflic acid and zirconocene triflate suggest that the esterification is indeed zirconium-catalyzed under our conditions. 21 Further support for zirconium catalysis was obtained by 13 C NMR spectroscopy. While a sharp carbonyl peak was observed at 169.3 ppm for 1 3 C(1)-labeled benzoic acid, introduction of Zr-(Cp) 2 (CF 3 SO 3 ) 2 ·THF resulted in a downfield shift to 171.18 ppm and considerable broadening of the carbonyl peak ( Figure  3, bottom), suggesting carbonyl coordination to zirconium in a fashion reminiscent of what has been observed for similar systems. 17a, 22 The line broadening indicates exchange between the free and coordinated carboxylic acid.
A tentative catalytic cycle is depicted in Figure 4. The positive rate dependence on [Zr] and close to zero order in [benzoic acid] and [2-phenylethanol] suggest that the turnover-limiting step is found late in the catalytic cycle. Since the same excess experiments indicated insignificant product inhibition, the release of ester and water is likely not turnover-limiting under the examined conditions. Hence, our data suggest that the slow step in the catalytic cycle is the collapse of the tetrahedral intermediate resulting from nucleophilic attack by the alcohol on the coordinated carboxylic acid. This corresponds to a barrier of approximately 16.7 ± 0.5 kcal/mol (SI) and is of the same order of magnitude as what has previously been estimated by density functional theory (DFT) calculations for collapse of the tetrahedral intermediate in zirconium-catalyzed amidation. 17a Using the optimized conditions, a range of carboxylic acids and alcohols were evaluated as substrates. Our model product, 2-phenylethyl benzoate (3a), was formed in a 78% yield, whereas excess alcohol or carboxylic acid resulted in increased yields, as expected for an equilibrium process ( Figure 5). Benzoic acid derivatives with electron-withdrawing or electrondonating groups resulted in good to moderate yields (3b−i), tolerating ketone and aldehyde substituents. Heteroaromatic carboxylic acids delivered the expected products 3j and 3k in moderate yields, whereas carboxylic acids with pyridine, imidazole, and indazole backbones failed to form esters (vide infra). Aliphatic carboxylic acids were smoothly converted into their corresponding esters 3l−u in good to excellent yields using equimolar amounts of alcohol, including fatty acids 3m and 3n, sterically congested substrates 3o and 3p as well as diacids 3q and 3r. The anti-inflammatory drug indomethacin was converted into its respective ester 3s in a good yield, and we were pleased to see that the corresponding 2-phenethyl ester 3t of Boc-protected L-alanine retained >99% enantiomeric excess (ee), similar to what has previously been described for zirconium-catalyzed amidation. 23 In contrast to the corresponding PFOS complex, 11 zirconocene triflate in catalytic amounts preferentially mediates esterifications over transesterifications (SI). Gratifyingly, this differentiation in carbonyl activation allowed for selective synthesis of the unsymmetric diester 3u  from the corresponding methyl ester-substituted phenylacetic acid without any transesterification product observed. The catalyst can be recycled and used for at least four consecutive cycles with negligible loss of activity for the formation of ester 3l (SI).
A range of aliphatic alcohols proved to be suitable esterification coupling partners ( Figure 6), including oleyl alcohol and allyl alcohol, which smoothly yielded esters 3v and 3w with both aryl and alkyl carboxylic acids. The electron-poor benzylic alcohol and the heteroaromatic 2-thiophene ethanol were acylated to form 3x and 3y in high yields. Trifluoroethanol formed the corresponding phenylacetate 3z in moderate yield, as did the sterically hindered adamantyl alcohol and cholesterol, rendering 3aa and 3ab, respectively. The corresponding benzoic and phenylacetic esters of 1-phenyl-2-propanol formed in different yields, giving benzoic ester 3ac in a moderate yield, whereas the use of (R)-(−)-1-phenyl-2-propanol (>98% ee) furnished phenylacetic ester 3ad in an excellent yield and retained enantiomeric excess.
Aromatic alcohols and N-heterocyclic carboxylic acids and alcohols with a basic nitrogen failed to form esters, and the starting materials could be recovered in near-quantitative amounts. To probe the origin of this observation, nicotinic acid and phenol were added separately to standard reactions after 3 h of reaction time (Figure 7). Interestingly, whereas the effect of phenol addition on the formation rate of 3a lies within the variability for the standard reaction (SI), the addition of nicotinic acid completely quenched the catalysis, indicating that the inability of aromatic alcohols and basic heteroaromatic compounds to form esters has different origins. The inhibiting effect of nicotinic acid, occurring already at a nearly 1:1 ratio to zirconium (SI), may be explained by the formation of catalytically inactive zirconium species that could form by Ncoordination of the pyridine or by coordination of a negatively charged carboxylate species after deprotonation by pyridine. On the contrary, 2-phenylethyl benzoate 3a continued to form at a similar rate after the addition of 50 equiv of phenol relative to zirconium with only traces of phenyl ester formation (see SI).  The Journal of Organic Chemistry pubs.acs.org/joc Article As secondary and tertiary alcohols perform well as coupling partners (Figure 6), the low reactivity of the phenol is likely not a function of steric hindrance; rather, its poor nucleophilicity under non-basic conditions is expected to be the main reason for the sluggish performance. As suggested from Figure 7, the low reactivity of phenols under standard conditions would allow for esterification of substrates substituted with unprotected aromatic alcohols. Indeed, acylation of 2-(4-hydroxyphenyl)ethanol proceeded with full selectivity for the aliphatic over the aromatic alcohol to form ester 3ae (Scheme 1).
In summary, this work demonstrates the use of kinetics as an integrated tool in the optimization of dehydrative esterification using a moisture-tolerant zirconium complex in catalytic amounts. The insights from the kinetic assessment of reaction parameters allowed for rational tuning of conditions and enabled an understanding of why certain substrate classes fail to form products. Furthermore, kinetics and spectroscopy were used to assess catalyst properties and provide support for the proposed mechanism. The present work adds to the general understanding of the reactivity of the understudied moisturetolerant group (IV) metal complexes, a highly interesting compound class for future use in dehydrative catalyzed reactions.
■ EXPERIMENTAL SECTION General Information. All reagents were purchased from commercial suppliers and used without further purification. Reactions were carried out in 4 mL screw neck glass vials furnished with screw caps equipped with poly(tetrafluoroethylene) (PTFE)/rubber septa, and stir bars under ambient atmosphere unless otherwise noted. Silica gel 60 Å (40−60 μm, 230−400 mesh) was used for column chromatography. All NMR spectra were recorded in CDCl 3 using a Bruker AVANCE II 400 MHz or Bruker Avance 500 MHz. Chemical shifts are given in ppm relative to the residual solvent peak ( 1 H NMR: CDCl 3 δ 7.26, 13 C NMR: CDCl 3 δ 77.16) with multiplicity (br = broad, s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet), coupling constants (in hertz), and integration. Kinetic data was analyzed by Agilent 1260 Infinity Quaternary LC (Eclipse Plus 18C column, 3.5 μm, 4.6 × 100 mm 2 ; UV detector, 265 nm) with a gradient of acetonitrile and 0.1% formic acid in Milli-Q water at a flow rate of 1 mL/min. The analytes were calibrated using a five-point calibration curve with threefold dilution between each sample in the series. HPLC with a chiral stationary phase was performed on an Agilent 1100 series instrument. High-resolution mass spectrometry analyses were performed by Thermo Scientific Q Exactive HF Hybrid Quadrupole-Orbitrap HESI or Bruker microTOF ESI, and low-resolution mass analyses by Bruker Daltonics amaZon speed no 06052 ESI.
General Esterification Procedure A for Kinetic Analysis. Recycling Experiments. Recycling experiments were carried out as follows. The reaction was started in accordance with general esterification procedure B on a 0.5 mmol scale. After 24 h, the vial was removed from the heated oil bath and the solvent was evaporated. The reaction mixture was thereafter extracted with 1 mL of petroleum ether (40−65°C bp) and decanted. The opaque solution was injected into an Eppendorf tube (2 mL) and subjected to centrifugation (3000 rpm, 3 min), after which the yellow solution was removed from the black catalyst residue. This procedure was repeated for a total of three extractions. The combined product/substrate fractions were evaporated, weighed, and subjected to 1 H NMR analysis using MeOD-d 4 as the solvent. The black catalyst residue was dissolved in a minimal amount of dichloromethane and added to the original reaction vial and the solvent was evaporated. To the dried catalyst residue, phenylacetic acid, 2-phenylethanol, and benzotrifluoride were then added in accordance with general esterification procedure B and the reaction was stirred at 80°C for another 24 h, after which the recycling procedure was repeated.
General Esterification Procedure B for Product Isolation. For 1 M reaction mixture, Zr(Cp) 2 (CF 3 SO 3 ) 2 ·THF (0.02 mmol, 11.8 mg), carboxylic acid (1 mmol), benzotrifluoride (1 mL, nondried), and alcohol (1 mmol) were added into the reaction vessel under an air atmosphere. The screw cap was tightened and the vial was placed in an oil bath at 80°C (or the indicated temperature). After 24 h, the reaction mixture was brought to room temperature and purified by column chromatography (silica gel 60, 2−10% EtOAc/petroleum ether) unless otherwise stated.
4-Hydroxyphenethyl 2-Phenylacetate (3ae). 3ae was synthesized according to the general esterification procedure B on a 1 mmol scale. The product was isolated as a colorless oil in 43% yield (0.434 mmol, 111 mg). Analytical data matches with the reported literature. 39 3ae: The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.joc.0c00235. 1 H NMR and 13 C{ 1 H} NMR spectra of all new compounds; all known compounds made by a new route not reported in previous studies; 13 C{ 1 H} NMR analysis of Zr(Cp) 2 (CF 3 SO 3 ) 2 ·THF; HPLC chromatograms for esters 3t and 3ad; additional data for the kinetic evaluation of solvents, reproducibility, VTNA analysis, concentration effects, addition of water and molecular sieves, transesterification with Zr- (