Recent Developments in the Scope, Practicality, and Mechanistic

Article Cite This: Acc. Chem. Res. XXXX, XXX, XXX−XXX

pubs.acs.org/accounts

Recent Developments in the Scope, Practicality, and Mechanistic Understanding of Enantioselective Hydroformylation Anna C. Brezny† and Clark R. Landis*

Downloaded via KAOHSIUNG MEDICAL UNIV on August 19, 2018 at 02:33:05 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Department of Chemistry, University of WisconsinMadison, 1101 University Avenue, Madison, Wisconsin 53706, United States

CONSPECTUS: In the nearly 80 years since catalytic hydroformylation was first reported, hundreds of billions of pounds of aldehyde have been produced by this atom efficient one-carbon homologation of alkenes in the presence of H2 and CO. Despite the economy and demonstrated scalability of hydroformylation, the enantioselective process (asymmetric hydroformylation, AHF) currently does not contribute significantly to the production of chiral aldehydes and their derivatives. Current impediments to practical application of AHF include low diversity of chiral ligands that provide effective rates and selectivities, limited exploration of substrate scope, few demonstrations of efficient flow reactor processes, and incomplete mechanistic understanding of the factors that control reaction selectivity and rate. This Account summarizes developments in ligand design, substrate scope, reactor technology, and mechanistic understanding that advance AHF toward practical and atom-efficient production of chiral α-stereogenic aldehydes. Initial applications of AHF were limited to activated terminal alkenes such as styrene, but recent developments enable high selectivity for unactivated olefins and more complex substrates such as 1,1′- and 1,2-disubstituted alkenes. Expanded substrate scope primarily results from new chiral phosphine ligands, especially phospholanes and bisdiazaphospholanes (BDPs). These ligands are now more accessible due to improved synthesis and resolution procedures. One of the virtues of diazaphospholanes is the relative ease of derivatization, including attachment to heterogeneous supports. Hydroformylation involves toxic and flammable reactants, a serious concern in pharmaceutical production facilities. Flow reactors offer many process benefits for handling dangerous reagents and for systematically moving from research to production scales. New approaches to achieving good gas−liquid mixing in flow reactors have been demonstrated with BDP-derived catalyst systems and lend assurance that AHF can be practically implemented by the pharmaceutical and fine chemical industries. To date, progress in AHF has been empirically driven, because hydroformylation is a complex, multistep process for which the origins of chemo-, regio-, and enantioselectivity are difficult to elucidate. Mechanistic complexity arises from three concurrent catalytic cycles (linear and two diastereomeric branched paths), significant pooling of catalyst as off-cycle species, and multiple elementary steps that are kinetically competitive. Addressing such complexity requires new approaches to collecting kinetic and extra-kinetic information and analyzing these data. In this Account, we describe our group’s progress toward understanding the complex kinetics and mechanism of AHF as catalyzed by rhodium bis(diazaphospholane) catalysts. Our strategy features both “outside-in” (i.e., monitoring catalytic rates and selectivities as a function of reactant concentration and temperature) and “inside-out” (i.e., building kinetic models based on the rates of component steps of the catalytic reaction) approaches. These studies include isotopic labeling, interception and characterization of catalytic intermediates using NMR techniques, multinuclear high-pressure NMR spectroscopy, and sophisticated kinetic modeling. Such broad-based approaches illuminate the kinetic and mechanistic origins of selectivity and activity of AHF and the elucidation of important principles that apply to all catalytic reactions.

■

INTRODUCTION

Since its discovery in 1938,1 hydroformylation (Scheme 1) has become one of the largest scale, organotransition metal catalyzed processes in the chemical industry. Common © XXXX American Chemical Society

Received: July 5, 2018

A

DOI: 10.1021/acs.accounts.8b00335 Acc. Chem. Res. XXXX, XXX, XXX−XXX

Article

Accounts of Chemical Research

Recent progress in AHF includes new ligands that favor production of branched aldehydes with high enantioselectivity, expansion of the substrate scope to include functionalized alkenes, demonstration of effective catalyst immobilization strategies that enable recycling of precious rhodium and expensive ligands, and establishment of flow reactor technologies that reproduce the selectivity and activity of batch processes.5 Hydroformylation is now accessible at the research lab scale (milligrams to tens of grams) because inexpensive pressure bottle reactors can be used in place of expensive autoclaves. The mechanism of AHF is complex, with three isomeric cycles that each comprise multiple elementary steps. Progress in elucidating the features that control activity, regioselectivity, and enantioselectivity in AHF has come from the application of techniques both long-standing (analysis of isotope distributions obtained in deuteroformylation experiments6 and kinetic analysis of catalytic reactions) and more recent, at least for hydroformylation (application of operando NMR and IR spectroscopic techniques7 and the interception and characterization of catalytic intermediates).8

Scheme 1. Hydroformylation of Alkenes

catalysts are based on cobalt or rhodium complexes.2 Hydroformylation constitutes a catalytic, atom-economic, one-step, one-carbon homologation of alkene feedstocks that produces aldehydes. Aldehydes are versatile and valuable synthetic intermediates because their carbon oxidation level is easily increased to the carboxylic acid level or reduced to that of an alcohol and because they participate in a variety of C−C bond-forming reactions. Commodity scale applications of hydroformylation focus on production of achiral (commonly linear) aldehydes that serve as intermediates for the production of solvents, plasticizers, and detergents.3 Enantioselective, or asymmetric, hydroformylation (AHF) dauntingly requires control of both enantio- and regioselectivity to high levels. Nevertheless, chiral α-stereogenic aldehydes are attractive precursors for the synthesis of fine chemicals and pharmaceuticals.2e,4

Table 1. Select Examples of Improvements to Substrate Scope in AHF

B


Article


for the branched aldehyde in high ee (Table 1) with several phosphine ligands.19 This result appears to be unique to 1,1disubstituted alkenes with two small electron-withdrawing substituents (e.g., −CF3 and −OAc); other 1,1-disubstituted alkenes yield linear hydroformylation products with these ligands.20 Our group is currently investigating other substrates with a variety of catalyst systems that can produce the branched, α-tetrasubstituted aldehydes.21

In this Account, we summarize strategies that we and others have used to expand the practical utility and scope of AHF and to better understand the mechanistic and kinetic features of this iconic organotransition metal catalyzed reaction.

■

RECENT ADVANCEMENTS IN THE ACCESSIBILITY OF AHF Advances in the practical application of AHF over the past decade primarily address expansion of the substrate scope, immobilization of catalysts, and development of flow processes. Such developments make AHF more attractive to both academic and industrial researchers and provide a convenient, atom-efficient route to chiral α-stereogenic aldehydes.

Improving Practicality and Ease of Use

Bis(diazaphospholane) (BDP) ligands exhibit high reactivity and selectivity in rhodium-catalyzed asymmetric hydroformylation.22 Unlike phosphine syntheses that utilize phosphide or Grignard reagents, BDP synthesis tolerates the presence of functional groups such as carboxylic acids. However, the original synthesis of resolved BDP ligands was inconvenient due to the number of steps, low yield, and difficult resolution.5 Recently our group, in collaboration with Eli Lilly, reported an improved, scalable synthesis and resolution procedure that avoids the need for chromatography or specialized equipment.23 Separation of precious metal catalysts from product and catalyst recycling are concerns with AHF and many other catalytic processes. Immobilization of catalysts on a solid support may be an attractive remedy, so long as the selectivity and activity of the catalyst is not compromised. The carboxylic acid functional groups of the BDP ligands enable immobilization of the ligands to resins via simple coupling reactions. For example, Rh(BDP) catalysts supported on Tentagel resin exhibit high selectivity and activity, similar to the homogeneous system.24 Vigorous mixing of a solution of substrate, synthesis gas, and immobilized catalyst achieves high turnover numbers and enable separation by filtration (Figure 1).

Substrate Scope

The most common “benchmark” substrates for asymmetric hydroformylation are functionalized, prochiral terminal alkenes such as vinyl arenes or vinyl acetates.2e Inductively electronwithdrawing substituents commonly direct reaction to the branched product2e and these functional groups are present in many of the substrates that are successful for AHF. Nevertheless, broader application of AHF in organic synthesis requires expansion of the method beyond terminal, activated alkenes. In this section, we highlight a few important advances at the expense of a comprehensive review. Simple alkenes that lack electron-withdrawing groups are a challenge for AHF, as typical ligands favor the linear product. For example, the hydroformylation of unactivated 1-alkenes using Rh(binaphos)9 and Rh(BDP)10 gives branched/linear (b:l) ratios of 1:3.2 and 1:1.5, respectively. Recent work from Clarke and co-workers, which used Rh(acac)(CO)2 with (Sax,S,S)-bobphos catalyst system, achieved significant regioselectivities and enantioselectivities for a variety of simple 1alkenes (Table 1).11 Our group has recently reported the synthesis of reduced backbone BDPs that yield high branched selectivity for difficult allyl substrates (Table 1),12 though the origin of this preference is not clear. In contrast to AHF of monosubstituted terminal alkenes, which necessarily yields a methyl group at the stereogenic center, AHF of 1,2-disubstituted alkenes is attractive because it creates stereogenic centers with greater complexity. However, the increased steric bulk at both positions of the olefin makes it difficult to attain high selectivity for only one position. Both the Tan and Breit groups have used a scaffolding catalyst strategy to impart regiocontrol.13 In addition, both binaphos and BDP have been successfully used for AHF of activated internal alkenes.14 AHF of 1,2-disubstituted Z-enamides and enol esters with BDP ligands yield α-functionalized aldehydes with high regioselectivity and ee values (Table 1).15 In these reactions, the electron-withdrawing group directs placement of the formyl moiety to the α-position, and the high intrinsic activity of BDP-based catalysts enables useful rates and the novel synthesis of sequence-specific and stereocontrolled oligoesters with near complete atom-economy.16 A third challenging class of substrates is 1,1-disubstituted alkenes. Due to increased steric bulk at the α-position, the majority of examples produce the linear product. Although stereogenic centers are created for linear and branched isomers,17 the branched product is more attractive because it possesses a stereogenic, tetrasubstituted center. However, facial discrimination of 1,1-disubstituted alkenes is challenging and often yields minimal ee.18 A notable exception is the AHF of αtrifluoromethyl vinyl acetate, which provides good selectivity

Figure 1. After reaction with immobilized Rh(BDP), the colorless solution indicates low metal leaching (top). Structure of tentagel bead with BDP ligand (bottom).

Larger scale application of hydroformylation in a pharmaceutical environment must address the safety challenges of using flammable and toxic gases. Flow processes represent an attractive approach to scaling intrinsically dangerous processes, but gas−liquid mixing can be problematic for flow reactors. In collaboration with Eli Lilly, we reported a pipes-in-series flow reactor for asymmetric hydroformylation with a homogeneous catalyst, which mitigates the safety hazards by minimizing the quantity of reagent gas (Figure 2).25 Additionally, this design enables good gas/liquid mixing, an important feature because C


Article


insertion to yield 4l or 4b, or is this step reversible? How do temperature and gas pressures affect reversibility? There are two general approaches to elucidating the origins of selectivity in a catalytic mechanism. The first we call the “outside-in” approach: one monitors the inputs and outputs of the catalytic “black-box” in an attempt to infer what happens inside. Measurements may include rates, selectivities, and isotopic label distributions as a function of reaction conditions. The alternative “inside-out” approach focuses on direct observation of the catalyst speciation under catalytic conditions or determining the rates and selectivities of elementary steps that make up the full catalytic cycle. Deep understanding of complex catalytic reaction mechanisms requires both outside-in and inside-out strategies.

regio- and enantioselectivity strongly depend on gas concentrations.26

Outside-in Approach

The pressure of CO affects both the regioselectivity and enantioselectivity of styrene hydroformylation.28,26 For reactions catalyzed with the BDP ligand at 80 °C and constant dihydrogen pressure, common linear/branched and enantiomeric ratios are illustrated in Figure 3. Under these conditions,

Figure 2. Pipes-in-series flow reactor design used by Eli Lilly in collaboration with Landis and co-workers. Adapted with permission from ref 25a. Copyright 2016 American Chemical Society.

The advances summarized above make AHF a more attractive and practical process. Progress in AHF has been driven by empiricism; there are few general rules to guide a new ligand design or match a given substrate with optimal catalyst structure or reaction conditions. A more rational design of AHF processes has been inhibited by the complexity of the reaction mechanism, which involves three reactants and a catalyst precursor that traverse three isomeric catalytic cycles, each comprising multiple elementary steps. This complexity has driven us to more deeply examine the mechanistic intricacies of AHF.

Figure 3. Effect of CO pressure on selectivity in styrene hydroformylation by Rh(S,S,S)-BDP.

■

MECHANISTIC STUDIES OF RH(BIS(DIAZAPHOSPHOLANE)) CATALYSTS The generally accepted mechanism for hydroformylation, depicted in Scheme 2, was first proposed by Breslow and Heck in 1961.27 This overall framework is vague concerning the origins of regio- and enantioselectivity: is regioselectivity fixed when the rhodium hydride alkene complex (3) undergoes

the rate law is first-order in catalyst concentration and independent of H2 concentration but exhibits complexity with respect to the order in CO concentration. The rates of formation of the minor S-enantiomer and linear products are inhibited by CO, whereas the CO-dependence of the rate for the major R-enantiomer changes from independent to inhibitory as CO concentration is increased (Figure 4).

Scheme 2. Accepted Mechanism for Hydroformylation

D


Article


concentration lowers all rates because more catalyst is pushed off-cycle to 1; (2) increasing CO concentration increases the rate of b(R) product formation by favoring partitioning of 4b(R) in the forward direction relative to β-hydride elimination back to 3(R). Mathematically, this leads to steady-state expressions shown below (eqs 1 and eq 2u). This is a common trapping scenario, in this case, CO traps the alkyl species competitively with reversion to the hydridoalkene complex. Interestingly, the origin of increased ee and regioselectivity with increased CO pressure lies in decreased rates of the minor products rather than increased rates of the major product, 4b(R).

Figure 4. Effect of CO pressure on pseudo-first-order (in [styrene]) rate constants for linear, branched (R), and branched (S) product formation at 80 °C (solid red lines represent best fit to [CO]−1 dependence and the solid blue line represents best fit to a [CO]0 dependence). Arrows indicate against which x-axis the data are plotted.

As the CO concentration is increased even more, one approaches the limit of complete trapping efficiency (k4[CO] ≫ k−3), causing the rates of production for all products to become inhibited by CO concentration and the selectivities to asymptotically approach maximum values. At high pressures of CO, hydride migration to all alkyl intermediates is irreversible and the relative transition states for alkene insertion control reaction selectivity. This kinetic scheme, like all models based solely on outsidein approaches, does not uniquely accommodate the observed data. Therefore, we looked for additional support for this mechanistic hypothesis. Hydroformylation of α-Deuterostyrene. In order to provide evidence that the linear alkyl 4l was produced from both the Rh−alkene complexes (3(R) and 3(S)), the products from hydroformylation of α-deuterostyrene were examined.26 The stereochemistry of the linear product was used to determine the face of the alkene to which the migrating hydride added. Both enantiomers of the linear deuterated product are observed, thus indicating that both 3(R) and 3(S) could lead to the linear alkyl complex.29 The kinetic model reflects these results, which demonstrates that the linear aldehyde can arise from hydride addition to either face of styrene.

Watkins and Landis proposed the kinetic model for AHF of styrene that is outlined in Scheme 3.26 Critical attributes of this model include (1) the catalyst pools in the form of the hydrido dicarbonyl (1) resting state (verified by operando IR measurements), (2) insertion of the alkene into the Rh−H bond to make linear alkyl 4l and the minor branched alkyl 4b(S), which is effectively irreversible because it is rapidly trapped by CO to go on to the corresponding acyls, 6l and 6b(S), and (3) in contrast, the major branched alkyl 4b(R), which is proposed to form quasi-reversibly, meaning that the partitioning between reversion back to 3(R) (and into the pool comprising 1, 2, and 3(S)) and forward trapping by CO to give b(R) depends on the CO concentration. This model addresses the central question, why does CO pressure affect both regioselectivity and enantioselectivity? One might anticipate that formation of all products should be inhibited because the hydrido dicarbonyl (1) is the resting state. However, if CO is required to competitively trap the major enantiomer alkyl 4b(R) so that it can go on to product, the effect of CO concentration is 2-fold: (1) increasing CO Scheme 3. Kinetic Model for AHF of Styrene

E


Article

Accounts of Chemical Research Deuteroformylation of Styrene. Based on earlier work of Nozaki and co-workers,28 Watkins and Landis used deuteroformylation to probe the reversibility of styrene insertion.26 In the limit of irreversible deuteride migration, one would expect to observe deuterium only at the aldehyde and β-positions of the product. Scrambling of the label indicates reversible formation of the alkyl species (Figure 5).

enough to keep up with very high catalytic rates. The strategy used by Tonks and Landis was to bubble a mixture of CO/H2/ N2 through a reaction at high rates such that subambient concentrations of CO and H2 could be maintained at constant values. Hydroformylation is rapid at low pressures (∼1 turnover/s at 14 psia syn gas). Furthermore, Tonks and Landis reported that the hydroformylation of styrene at very low CO pressures leads to inversion of the regioselectivity and complete loss of enantiomeric excess (Figure 7).31

Figure 5. Observed products from deuteroformylation of styrene at 80 °C catalyzed by Rh((S,S,S)-BDP).

Watkins and Landis found that the deuteroformylation of styrene at 80 °C with Rh((S,S,S)-BDP) yielded a significant amount of both the β-deuterostyrene and d1-aldehydes (92% of transformed styrene molecules) (Figure 5). The high proportion of d1-aldehydes was surprising because it implies a higher concentration of Rh−H than Rh−D during the reaction. In the presence of D2 gas, a hydride can only be formed by reversion of an alkyl complex. Additionally, βdeuterostyrene is formed from the reversion of 4b. Because no significant amount of α-deuterostyrene was observed in these experiments, it suggests that when the linear Rh−alkyl isomer (4l) is formed, it does not revert to the Rh−alkene complex.30 Deuterium Exchange As a Function of CO Concentration. A critical test of the kinetic model proposed by Watkins and Landis concerns the influence of CO concentration on the exchange of deuterium into styrene. If CO traps the alkyl species competitively with reversion, then the phenomenological rate of exchange of deuterium into styrene must be inhibited by increased CO concentration. Indeed, the observed rate constant for formation of β-deuterostyrene decreases with increasing pressure of CO (Figure 6).

Figure 7. Regio- and enantioselectivity in the AHF of styrene at varied CO pressures ([styrene]0 = 2.2 M, [Rh(BDP)] = 5 × 10−4, 80 °C, PH2 = PCO). Note the log scale on the x-axis.

The change in regioselectivity is consistent with Watkins and Landis’s model in which the kinetically favored branched alkyl 4b is formed reversibly. Under lower CO pressures, this intermediate can revert to the alkene complex and reinsert to form the thermodynamic distribution of alkyl species. It is unclear why the enantiomeric excess erodes to zero. Further insight required an approach by which the kinetic and thermodynamic preferences of catalytic intermediates could be assessed directly, that is, an inside-out strategy. Inside-out Approach

Because of the several NMR-active nuclei (31P, 1H, 13C) present and the abundant structural information provided by coupling to 109Rh, NMR spectroscopy is the method of choice for inside-out studies. The tetraphenyl-BDP (Figure 8) was used for these studies due to increased solubility and improved line shape in the 31P NMR spectrum compared to (S,S,S)-BDP as used above. Hydroformylation is a rapid catalytic process; in order to intercept intermediates along the cycle, hydrido dicarbonyl 1 was reacted with styrene at low temperatures in the absence of dihydrogen and in the presence of ca. 1 atm of CO. The absence of dihydrogen prevents aldehyde product formation and enables interception and detailed characterization of

Figure 6. Influence of CO pressure on observed rate for formation of β-deuterostyrene.

Ultralow-Pressure Hydroformylation. These data indicate that CO trapping of the alkyl intermediates is fast for the linear and b(S) products at ambient and higher pressures of gas, but what happens at very low pressure? Will selectivities invert because all alkyl intermediates may equilibrate? Running hydroformylation at subambient pressures creates experimental problems such as achieving gas−liquid transport that is fast

Figure 8. Tetraphenyl-BDP (racemic) used in NMR spectroscopy studies for improved solubility and line shape. F


Article


Figure 9. Generalized catalytic cycle for rhodium-catalyzed hydroformylation of styrene, with observed intermediates boxed.

several on- and off-cycle catalyst species by 1H and 31P NMR spectroscopy (Figure 9).32 At early reaction times, reaction of 1 and styrene exhibits strong kinetic preference for the formation of the branched isomer (7b), which is consistent with the assumption in Watkins and Landis’s kinetic model (vide supra).26,31 At longer reaction times, the linear acyl dicarbonyl (7l) becomes the dominant species, indicating that it is the thermodynamically preferred acyl species. This study clearly demonstrates the countervailing kinetic and thermodynamic preferences of the acyl dicarbonyl complexes. In these simple NMR tube experiments, slow, passive gas− liquid mixing during the reaction causes the solution to become starved of CO over time (the formation of acyl 7 from 1 consumes one equivalent of CO). Under CO-starved conditions, small amounts of the four-coordinate branched alkyl monocarbonyl 4b and the five-coordinate linear alkyl dicarbonyl 5l were observed. The different number of terminal CO ligands for the observed alkyl species, 4b and 5l, and the unknown concentrations of CO during the course of the reaction prevent quantitative analysis of the kinetics of branched and linear alkyl formation. Nonetheless, this work provided the first direct observation of alkyl-catalytic intermediates for any rhodium AHF system. Furthermore, these studies demonstrate a kinetic preference for formation of the branched acyl dicarbonyl species (7b), the reversibility of its formation, and a thermodynamic preference for the linear isomer (7l). A limitation of these inside-out studies is that the reaction conditions are very different than normal catalytic reactions.

How can the intermediates be studied under catalytic conditions? The Wisconsin High Pressure NMR Reactor (WiHPNMRR). The WiHP-NMRR33 developed by Landis and coworkers was designed to explore catalyst speciation and reaction kinetics under catalytic conditions (Figure 10). Key to the reactor design is active gas−liquid mixing.

Figure 10. Wisconsin High Pressure NMR Reactor (WiHP-NMRR) developed by Landis and co-workers33 for studying gas-fed reactions. G


Article


Figure 11. Time course data for catalytic hydroformylation reactions collected using the WiHP-NMRR. For each set of conditions, disappearance of styrene, appearance of product (top), and catalyst speciation (bottom) are presented.

Figure 12. Noncatalytic hydrogenolysis experiments in the WiHP-NMRR at different pressures.

Operando Study of Catalytic Hydroformylation. Recently, we used WiHP-NMRR to explore the role of offcycle acyl dicarbonyl species (7) during catalysis.34 The data from catalytic hydroformylation reactions collected by operando spectroscopy enable monitoring of both starting material and product concentrations as a function of time (Figure 11); we found a zeroth order dependence of the rate on styrene concentration. This contrasts with the previous work from Watkins and Landis at higher temperatures, for which a first-order dependence on styrene concentration and a rhodium hydride (1) resting state was reported.26 In the lower temperature (ca. 313 K) studies using WiHP-NMRR, the acyl complexes (7l and one diastereomer of 7b) were observed as the resting state. Hydroformylation rate profiles revealed by WiHP-NMRR studies are surprising. At constant H2 pressure and high CO

pressures (200 psia and 115 psia, panels A and B of Figure 11), the reaction rate is inhibited by CO concentration. This is consistent with the need for CO dissociation from the acyl species (7) before reaction with dihydrogen. However, lowering the pressure of CO to 20 psia (panel C of Figure 11) results in a large rate decrease. The decrease in rate correlates with the change in catalyst speciation from a 3:1 ratio of 7b/7l (at high CO pressure) to an approximately thermodynamic mixture of acyl dicarbonyls (

Recent Developments in the Scope, Practicality, and Mechanistic

Recommend Documents