Asymmetric Heterogeneous Catalysis: Transfer of Molecular Principles

---

Research Article pubs.acs.org/acscatalysis

Asymmetric Heterogeneous Catalysis: Transfer of Molecular Principles to Nanoparticles by Ligand Functionalization Imke Schrader,† Sarah Neumann,† Anda Šulce,† Fabian Schmidt,† Vladimir Azov,†,‡ and Sebastian Kunz*,† †

Institute of Applied and Physical Chemistry (IAPC), Center for Environmental Research and Sustainable Technology, University of Bremen, Leobener Straße, 28359 Bremen, Germany ‡ N. D. Zelinsky Institute of Organic Chemistry, Russian Academy of Sciences, Leninsky prospect 47, Moscow 119991, Russia S Supporting Information *

ABSTRACT: We demonstrate the functionalization of supported nanoparticles (NPs) with ligands as an approach to bridge the gap between homogeneous and heterogeneous catalysis. The use of chiral ligands is shown to allow for achieving stereoselectivity with supported catalysts. While these materials are considered to be very complex, we introduce a simple ligand−reactant interaction model that can explain the origin of stereoselectivity on a molecular level. Application of general guiding principles from asymmetric homogeneous catalysis to our model allows for enhancing stereoselectivity in a rational manner to an enantiomeric excess of 74%. This highlights the potential of the presented approach for asymmetric heterogeneous catalysis. In addition, we demonstrate the feasibility of combining this asymmetric control with our previously introduced molecular concept to tune chemoselectivity. In this way supported catalysts are transformed to be stereo- as well as chemoselective, which highlights the potential arising from the idea of transferring principles from molecular catalysis to heterogeneous catalysts. KEYWORDS: ligand-functionalized nanoparticles, stereoselective heterogeneous catalysis, asymmetric control, ligand−reactant interaction, stereoselective hydrogenation, supported platinum nanoparticles hydrogenated under catalytic conditions.12 As a result, their structure and mode of action changes, which causes selectivity losses.13 In comparison to asymmetric homogeneous catalysis the modifier approach is very complex. In homogeneous catalysis the stereoselectivity is primarily determined by the ligand−reactant interaction.14 In contrast, the modifier approach is sensitive to parameters (e.g., particle size) that have no relevance in homogeneous catalysis.15 As a result, the knowledge gained from homogeneous catalysis was reported to be of very limited value for the modifier concept.16 A recent approach to manipulate the catalytic properties of supported NPs that may furthermore exhibit the potential to overcome the limitations of chiral modifiers is surface functionalization with ligands. In contrast to modifiers, ligands are insoluble in the reaction medium and are strongly fixated to the particle surface.10 As a result, they remain bound to the surface under catalytic conditions.17 It has been shown that altering the geometric and electronic properties of metal surfaces with ligands can lead to tremendous improvements in e.g. chemoselectivity and activity.18−23 In addition, ligand− reactant interactions can be achieved on NP surfaces.24 This

1. INTRODUCTION In comparison to homogeneous counterparts, heterogeneous catalysts bear significant economic and ecological advantages regarding product/catalyst separation and the possibility of performing continuous processes.1 However, homogeneous catalysts are in many aspects chemically superior. A particularly important example is asymmetric homogeneous catalysis. Highly stereoselective metal-organic catalysts have been developed over the last decades by designing tailored chiral ligands.2 In contrast, supported nanoparticles (NPs) are not stereoselective because bare NPs do not have chiral properties. An established approach of introducing a chiral information to supported catalysts is the use of “modifiers”.3 Modifiers are chiral molecules that are usually used as auxiliaries for stereoselective synthesis.4 They are dissolved in the reaction medium together with the reactant.5 If the modifiers adsorb on the metal surface in a specific mode and interact properly with the reactants, a stereoselective conversion can occur.6 High enantiomeric excesses (ees) have been demonstrated in particular for α- and β-keto esters, and applications on an industrial scale have been reported.7,8 However, the solubility of modifiers is for several reasons a crucial limitation, as recently discussed in more detail, and attempts to prepare immobilized forms of modifiers led to significant stereoselectivity losses.9−11 Further problems arise from the fact that the best modifiers are © 2017 American Chemical Society

Received: February 8, 2017 Revised: April 29, 2017 Published: May 3, 2017 3979

DOI: 10.1021/acscatal.7b00422 ACS Catal. 2017, 7, 3979−3987

Research Article

ACS Catalysis provides a crucial link for transferring concepts from homogeneous catalysis to supported catalysts. By surface functionalization with chiral ligands asymmetric information can be introduced to supported NPs in a way similar to that in homogeneous catalysis. However, until now ees remained low (≤34% ee) and too little was known about the asymmetric control for these materials that would allow achieving high stereoselectivities. To explore supported ligandfunctionalized NPs systematically, we established a rationaldesign approach with independent control over the particle, ligand, and support properties. In contrast to the use of chiral modifiers, which utilizes commercial catalysts, our preparation concept is quite complex. However, it allows for determining selectively the influence of each parameter on the catalytic properties and hence to determine the knowledge needed to understand these novel complex materials.25 Using our preparation concept, we recently showed that, in contrast to the modifier approach, the stereoselectivity of ligand-functionalized Pt NPs does not depend on the particle size.26,27 Instead, it is primarily determined by the ligand−reactant interaction. This is analogous to the basic principles of asymmetric homogeneous catalysis,14 indicating that general concepts from metal-organic catalysis should be applicable to ligandfunctionalized NPs. The present work is focused on (i) developing a ligand− reactant interaction model and (ii) applying general guiding principles from asymmetric homogeneous catalysis to this model. This allowed us to enhance the stereoselectivity in a rational manner to a level that demonstrates the potential of ligands for asymmetric heterogeneous catalysis. Furthermore, we show that this asymmetric control can be combined with our previously introduced molecular concept to render supported catalysts that are highly chemoselective.17 As a result, heterogeneous catalysts are transformed to being highly chemo- and stereoselective by functionalization with ligands.

microscopy), and the size is maintained in all following preparation steps.17,25 To isolate the “ligand-free” Pt NPs, 50 mL of 1 M HCl (VWR) was added after cooling to ambient temperature. The precipitated particles were separated from the supernatant by centrifugation and washed once with 1 M HCl. For further preparation steps the cleaned particles were redispersed in 100 mL of cyclohexanone (≥99.0%, Sigma-Aldrich). 2.1.2. Synthesis of Ligand-Functionalized Pt Nanoparticles. Ligand solutions with a ligand concentration of 16 mM were prepared. In order to bind the ligands to the NPs, the amino group has to be deprotonated.17 Therefore, the ligand solutions have to be alkaline, which is achieved by adding NaOH. In this way the ligand becomes deprotonated in the aqueous phase. However, if the OH− concentration is too high, OH− starts to adsorb on the particle surface,29 thus competing with the ligands for free binding sites. Therefore, the NaOH concentration used for functionalization has to be adjusted with respect to the pH properties of each ligand. We will specifically address this point in more detail in a subsequent publication. To determine the ideal NaOH concentration for every ligand, we performed functionalizations at different NaOH concentrations, prepared catalysts with the resulting particles, and measured the enantiomeric excess (ee) achieved with these catalysts for the hydrogenation of MAA (see red reaction in Figure S1 in the Supporting Information). The NaOH concentration that led to the highest ee was determined for each ligand and used in the following for the catalyst preparation. The ideal NaOH concentrations for the different ligands are L-proline (≥99%, Sigma-Aldrich, see blue structure in Figure 1) with 30 mM NaOH, (R)-2-(methoxymethyl)pyrrolidine (>99.0%, TCI, see orange structure in Figure 2) with 20 mM NaOH, D-piperidinecarboxylic acid (>98.0%, TCI, see black structure in Figure 2) with 50 mM NaOH, and (R)(−)-3-piperidinecarboxylic acid (97%, Sigma-Aldrich, see pale red structure in Figure 2) with 40 mM NaOH. For Pt NP functionalization 400 mL of the ligand solution was added to the previously prepared 100 mL of Pt NP dispersion (with respect to cyclohexanone; see above). This corresponds to a ligand/Pt ratio of 12.3. The excess of ligand is used to achieve a nearly ligand saturated surface on the NPs.25 The number of ligand-free surface atoms is then determined by the steric demand of the ligand structure.25 The resulting emulsion was vigorously stirred for 45 min. Successful ligand functionalization is indicated by transfer of the NPs from the organic phase into the aqueous alkaline ligand solution. As the colloids are black, the particle transfer is accompanied by a color change of the two phases. The initially black organic dispersion turns clear, and the aqueous solution turns black. The black ligandfunctionalized Pt NP dispersion was separated from the clear organic phase in a separation funnel. 2.1.3. Deposition of “Ligand-Free” and Ligand-Functionalized Pt Nanoparticles. For catalytic investigations the “ligand-free” and ligand-functionalized Pt NPs were deposited onto Al2O3 (Puralox SCCa 150/200; Sasol, grain size 200−500 μm). Therefore, the support material was added to the desired particle dispersion (ligand-free NPs were dispersed in cyclohexanone, ligand-functionalized in the according aqueous ligand solution) in order to give a nominal metal loading of 2 wt %. The solvent was removed using a rotary evaporator (P = 20 mbar; T = 60 °C for “ligand-free” and T = 40 °C for ligand-functionalized Pt NPs), yielding a brown-gray solid.

2. EXPERIMENTAL SECTION 2.1. Material Synthesis. The preparation of supported, ligand-functionalized nanoparticles (NPs) follows a previously established rational-design concept that allows stepwise assembling of the desired catalysts.25 “Ligand-free” Pt NPs are prepared and subsequently functionalized with the desired ligand in a separate step. These colloids can be deposited onto any given support material. As the particle size is maintained in each step, the particle, ligand, and support properties can be controlled individually and their influence on the catalytic properties be evaluated selectively. The experimental details of the different preparation steps are described in the following. 2.1.1. Synthesis of “Ligand-Free” Pt Nanoparticles. For the preparation of “ligand-free“ Pt NPs, a slightly modified procedure of Wang et al. was used.28 A 0.25 g portion of H2PtCl6·H2O (40% Pt, ChemPur) was dissolved in 25 mL of ethylene glycol (EG, 99.8%, Sigma-Aldrich), and 25 mL of a 0.5 M NaOH (98.9%, Fisher Chemical) solution in ethylene glycol was added. The mixture was vigorously stirred and heated to 150 °C using a preheated oil bath. After about 5 min the color changed from yellow to black, which indicated the formation of Pt NPs. To ensure complete reduction of the Pt precursor, the reaction mixture was kept at 150 °C for 1.5 h. As some metal is lost due to the formation of a black residue, the yield of Pt in the form of stable colloids is ∼80%. The resulting particles are 1.2 ± 0.3 nm in size (determined by transmission electron 3980

DOI: 10.1021/acscatal.7b00422 ACS Catal. 2017, 7, 3979−3987

Research Article

ACS Catalysis The supported “ligand-free” catalysts were rinsed twice with acetone and the ligand-functionalized catalysts twice with ethanol to clean the sample from residues and residual ligands that do not bind to the surface. Both catalysts were dried after washing at reduced pressure for 5 min. The actual metal loading of each catalyst was analyzed by atomic absorption spectroscopy (AAS, Carl Zeiss Technology AAS 5 FL) and used to normalize the reaction rates. Samples for AAS were prepared by digesting the supported NPs with freshly prepared aqua regia. 2.2. Catalytic Investigation. 2.2.1. Catalytic Hydrogenation. Catalytic experiments were performed using inhouse-designed stainless steel autoclaves. Five autoclaves were connected to the same H2 (Linde 5.0) gas line and placed on a 5-fold stir plate in order to perform five experiments in parallel under identical conditions. The autoclaves were loaded with 1 mL of reactant and 9 mL of dioxane (≥99.9%, Carl Roth) for activity measurements and with 0.5 mL of reactant and 4.5 mL of dioxane for the enantioselectivity experiments (except for results in Figure 2, EOPP, and Table 1; here 1 mL of reactant and 9 mL of solvent were used). As reactants methylacetoacetate (MAA; 99%, Sigma-Aldrich), 2-heptanone (HEP; >98.0%, TCI), 4-methyl-2pentanone (MPE; >99.5%, TCI), acetylacetone (ACAC; >99.0%, TCI), ethylacetoacetate (EAA; ≥99.0%, SigmaAldrich), propylacetoacetate (PAA; >98.0%, TCI), butylacetoacetate (BAA; >98.0%, TCI), isopropyl acetoacetate (IPA; > 97.0%, TCI), tert-butyl acetoacetate (TBA; >95.0%, TCI), methyl 3-oxovalerate (MOV; >98.0%, TCI), methyl 3oxohexanoate (MOH; >96%, TCI), methyl 3-oxoheptanoate (MOHp; >95.0%. TCI), methyl isobutyrylacetate (MIBA; >97.0%, TCI), methyl 4,4-dimethyl-3-oxovalerate (MDOV; >95.0%, TCI)), and ethyl 3-oxo-3-phenylpropanoate (EOPP; ≥97.0%, Sigma-Aldrich) were used. In each experiment 0.2 g of catalyst was used (except for ACAC, where 0.4 g was used). The autoclaves were flushed three times with hydrogen and filled to a pressure of 20 bar. All hydrogenation reactions were carried out at 23 °C and at a stirring rate of 800 rpm. For the structure−activity and structure−selectivity studies shown in Figures 1 and 2, the conversion was kept below 10% to achieve differential operation conditions.30 Conversiondependent tests revealed no significant influence of the conversion on the stereoselectivity. Enantiomeric excesses (ees) reported for β-keto esters of different steric demands and values reported for the hydrogenation of EOPP (see Table 1) correspond to experiments run to full reactant conversions. The corresponding turnover numbers (TONs) are given in the captions. As previously demonstrated, no indication for metal leaching was found by performing relevant tests.17 Thus, the catalytic reaction truly occurs on supported NPs and is not related to homogeneous catalytic species formed by metal leaching. The experimental errors of the activities and enantioselectivities were determined by estimating the standard deviation from 10 individually performed catalysis experiments. The catalyst of each of these experiments was prepared separately. Thus, the experimental errors contain deviations that arise from the catalyst preparation, performance of catalysis experiments, and product analysis. 2.2.2. Catalyst Recycling Experiments. As previously shown by mass spectrometry, ligands do not desorb under catalytic conditions.27 The catalysts should thus be recyclable. To test the recoverability of supported ligand-functionalized Pt NPs, PRO-Pt NPs were used as catalysts prepared according to

section 2.1 and the hydrogenation of MAA was applied as a test reaction (see section 2.2.1). After full conversion was obtained, the supported catalyst was recovered by filtration, dried under ambient conditions, and subsequently used again as catalyst for the hydrogenation of MAA. Following this protocol, the catalyst was recycled two times. 2.2.3. Product Analysis of Catalytic Experiments. The conversion and enantioselectivity were determined by gas chromatography (PerkinElmer GC-Clarus 580) equipped with a flame ionization detector (FID). He (Linde, 5.0) was used as carrier gas. All samples were diluted in a 1:1 ratio with acetone (99.9%, VWR) for analysis, unless otherwise specified. For the determination of conversions of all reactants, a Zebron (ZB-WAXplus, 30 m length, 0.25 mm inner diameter, 0.25 μm film thickness) column was used (except for EOPP, conversions were here determined using a chiral column, see below). The injector temperature was set to 200 °C. The oven temperature program was as followed (for the method for MPE and ACAC see further below): the oven was held at 110 °C for 11 min, then heated to 160 °C at a rate of 5 °C min−1, and kept for 1 min. After further heating to 180 °C at a rate of 20 °C min−1 the temperature program was stopped after 1 min. For conversion determination of MPE the oven was held at 90 °C for 20 min, then heated to 160 °C at a rate of 5 °C min−1, and kept for 1 min. After further heating to 180 °C at a rate of 20 °C min−1 the temperature program was stopped after 1 min. For the product determination of ACAC the oven was held at 50 °C for 10 min, then heated to 70 °C at a rate of 30 °C min−1, and kept for 10 min. After further heating to 180 °C at a rate of 5 °C min−1 the temperature program was stopped after 10 min. For conversion measurements acetophenone was added after performing the catalytic reactions and used as internal standard. None of the experiments showed any detectable carbon losses. The enantiomeric excesses (ees) were determined using a Lipodex E column (Macherey-Nagel, 27 m length, 0.25 mm inner diameter, 0.25 μm film thickness). Prior to analysis the products were derivatized with 20 μL of N-methylbis(trifluoroacetamide) (97%, abcr) for 12 h (except for MAA, MPE, ACAC, MP, and EOPP). The injector temperature was set to 200 °C. Here the oven program was in general as follows (for the method for HEP, MPE, ACAC, MIBA, and EOPP see further below): the oven was held at 85 °C for 20 min, then heated to 180 °C at a rate of 30 °C min−1, and kept for 10 min. Then the temperature program was stopped. For ee determination of HEP and MPE the oven was held at 50 °C for 20 min. After further heating to 180 °C at a rate of 30 °C min−1 the temperature program was stopped after 10 min. For the ee determination of MIBA and ACAC the oven was held at 50 °C for 1 min, then heated to 100 °C at a rate of 5 °C min−1, and kept for 1 min and then heated to 110 °C at a rate of 1 °C min−1 and kept for 1 min. After further heating to 180 °C at a rate of 30 °C min−1 the temperature program was stopped after 1 min. For conversion and ee determinations of EOPP hydrogenation experiments, the oven was held at 80 °C for 5 min and then heated to 139 °C at a rate of 0.5 °C min−1. After further heating to 180 °C at a rate of 30 °C min−1 the temperature program was stopped after 11 min. EOPP samples were diluted in a 1/1 ratio (sample/acetone) for analyzing conversions and a 1/5 ratio for determination of ee. In order to identify and assign the side products formed by hydrogenation of the aromatic moiety, GC-MS was performed at a Finnigan MAT 95 spectrometer coupled with an Agilent 6890N 3981

DOI: 10.1021/acscatal.7b00422 ACS Catal. 2017, 7, 3979−3987

Research Article

ACS Catalysis

strongly altered by the ligand. Interestingly, the presence of PRO enhances the hydrogenation rate of MAA by almost a factor of 2. In this context one has to consider that the rates shown in Figure 1 are normalized to the total number of surface atoms. However, surface atoms that bind ligands are expected to be blocked.34 A nearly PRO saturated Pt NP surface exhibits a ligand coverage of 0.85.17 If only ligand-free surface atoms are considered to be catalytically active, it is seen that the presence of PRO enhances the hydrogenation rate for MAA by a factor of 10 (see Figure S2 in the Supporting Information). We have previously shown an activity enhancement for the hydrogenation of acetophenone induced by PRO. This finding we related to a ligand acceleration effect that is known from homogeneous catalysis as the N−H effect.17,35 A similar activation of ketones has also been discussed for cinchonidine modifiers.36 Upon binding to late d metals the amine-bound hydrogen becomes acidic.37 This proton can activate the CO group for hydrogenation, opening an alternative reaction pathway with a higher reaction rate in comparison to that for the purely metal catalyzed reaction.38 However, since only MAA shows a pronounced activity enhancement (see Figure S2), we conclude that the ester group plays a major role in the activation of MAA by PRO. Similar activity enhancements can be achieved with phenyl substituents.17 We thus assume that not an ester group but an additional substituents being close to the CO group (in α or β position) and suitable for interaction with the PRO ligand is a prerequisite for a strong promotion of the CO activation. In contrast to ligand-free Pt NPs, the PRO-functionalized Pt NPs show stereoselectivities for all reactants (ee values shown in Figure 1 at the top of the bars). As previously shown, the stereoselectivity of ligand-functionalized NPs is determined by the ligand−reactant interaction.27 HEP and MPE both do not exhibit any additional functional groups that may result in specific electronic interactions with PRO. The different stereoselectivities for these two reactants (15% ee for HEP, 21% ee for MPE) are thus concluded to arise from sterically induced changes in the ligand−reactant interaction. The ee for MAA is significantly higher (39% ee), even though its steric demand does not vary strongly from that of HEP and MPE. Thus, we conclude that the ester group plays a specific role for the asymmetric bias that is not related to a steric effect. Instead, direct intermolecular interactions between the ester group of MAA and the ligand should be considered for an explanation of the enhanced stereoselectivity. PRO exhibits two functional groups to interact electronically with the ester moiety of MAA: (i) the amine proton and (ii) the carboxyl group. To shed light onto the MAA−PRO interaction, we investigated the influence of different amine ligands on the activity and stereoselectivity (see Figure 2). Often ligand-functionalized NPs cannot be cleaned sufficiently to determine e.g. conclusive ligand coverages.39,40 As this was the case for some of the ligands shown in Figure 2, only the catalytic rates normalized to the total number of surface atoms are discussed and not rates normalized to the number of ligand-free surface atoms. The reaction rates shown in Figure 2 reveal that irrespective of the ligand structure a rate enhancement in comparison to the ligand-free Pt NPs is observed for all ligand-functionalized NPs. Thus, we can conclude that neither the position nor the presence of the carboxyl group is decisive for the rate enhancement, only the presence of the amine proton.

chromatograph. For product separation a CP-Sil 8 CB column (27 m length, 0.25 mm inner diameter, and 0.25 μm film thickness) was used. After sample injection (at 250 °C), the temperature was held at 110 °C for 4 min. Then the temperature was increased to 120 °C at a rate of 4 °C min−1, kept for 10 min at 120 °C, and then raised to 290 °C at a rate of 10 °C min−1 and finally kept at this temperature for 5 min.

3. RESULTS AND DISCUSSION We started our investigation by focusing on the hydrogenation of methylacetoacetate (MAA; Figure 1 and Figure S1 in the

Figure 1. Hydrogenation rates (normalized to the total number of surface atoms) of different ketones over “ligand-free” (a) and PROfunctionalized Pt NPs (b; see the ligand structure at the top). The hydrogenation of MAA shows a significantly enhanced rate over PROPt NPs in comparison to “ligand free” NPs, whereas for HEP and MPE the reaction rates are reduced. The highest ee (see values on top of columns) is obtained with MAA. The experimental error for the ee is ±1%.

Supporting Information) as a reactant and L-proline (PRO) as a ligand (please see ref 17 for detailed characterization of PRO-Pt NPs). As previously shown by NMR spectroscopy, PRO does bind via the amine group but not by interaction between the carboxyl group and Pt.17 The lack of any strong binding between COOH and “surfactant-free” Pt NPs has further been demonstrated experimentally by Wang et al.31 By optimizing the catalyst preparation procedure and the reaction conditions, we could achieve an enantiomeric excess (ee) of 39% (highest ee achieved so far 34%).27 In homogeneous catalysis stereoselectivity is mainly determined by steric and electronic interactions (e.g., intermolecular hydrogen bonds) between the reactant and the chiral ligands.14,32,33 MAA is the simplest β-keto ester with the least steric demand. In addition to the reactive keto group that is supposed to be catalytically hydrogenated, MAA contains an ester group, expected to interact sterically and/or electronically with the PRO ligand. To clarify the relevance of these two interactions for the asymmetric bias, we tested two model reactants (HEP and MPE; Figure S1). These two ketones exhibit a similar steric demand in comparison to MAA but no additional functional groups. The results for ligand-free and PRO-functionalized NPs (see Figure 1) reveal that the catalytic properties of the NPs are 3982

DOI: 10.1021/acscatal.7b00422 ACS Catal. 2017, 7, 3979−3987

Research Article

ACS Catalysis

shown in Figure 3a (and Figure S3a) is more suitable to describe the interaction of MAA with Pt-bound PRO.

Figure 3. Two-point-binding (a) and one-point binding (b) interaction models for β-keto esters with Pt-bound PRO. The resulting alcohols with the corresponding absolute stereochemistry are shown as insets in the top right corners.

The postulated interaction mode (see Figure 3a) leads to a so-called two-point binding of the reactant with the ligand. This phenomenon is known from homogeneous catalysis, and it has also been proposed for Raney Ni modified with tartaric acid and β-keto esters as reactants.41,42 A two-point binding reduces the conformational flexibility of the reactant, which in turn increases the asymmetric bias. In the present model, the interaction between the alkoxy and the COOH group is found to play a specific role in the asymmetric bias (see section 2 in the Supporting Information). This is in contrast with the model postulated for the Raney Ni−tartaric acid system that is not suitable to explain our experimental findings. In the Raney Ni− tartaric acid binding model the alkoxy group does not take part in the binding.42 Furthermore, while in our model the CO moiety of the ester group interacts with the same group as the reactive CO group (see Figure 3a), both CO groups interact with different hydroxyl groups of the tartaric acid in the modified Raney Ni system. Further evidence for the proposed interaction model can be found by analyzing the product configuration. Hydrogenation proceeds via hydride transfer from the metal surface to the C atom and proton transfer from the amine to the O atom.27 The CO group of MAA exhibits two enantiotopic faces (see Figure S4 in the Supporting Information) at which the hydride can attack the C atom. Due to the two-point binding, one of these faces is oriented toward the surface (see Figure S5a in the Supporting Information) and the other one toward the reaction medium (see Figure S5b). The hydride will preferentially attack the reactant at the face that points to the surface. As a result, the alcohol with R configuration should be preferentially formed. This prediction agrees with the product configurations we determined experimentally.27 Our interaction model is hence able to correctly predict the stereochemistry of the product. Using the interaction model and the knowledge established in homogeneous catalysis, we next demonstrate how the ee of the reaction can be enhanced in a rational manner. The origin of stereoselectivity lies in the interaction of a prochiral reactant with a chiral ligand. This interaction leads to two diastereotopic reaction pathways with transition states of different free energies. The two-point binding mode discussed above reflects rather the adsorption complex from which the reactant will react, but not the transition state (see Figure S6 in the Supporting Information for illustration of energy profile). However, hydrogenations of ketones are highly exergonic.43

Figure 2. Hydrogenation rates normalized to the total number of surface atoms for MAA over different ligand-functionalized Pt NPs (catalyst structures are shown at the top). All ligand-functionalized Pt NPs show enhanced reaction rates in comparison to the “ligand-free” NPs. The activity enhancement of these amine ligands does not depend on the presence or position of a carboxyl group. However, significant stereoselectivities are only obtained with a carboxyl group in an α position with respect to the amine group. The experimental error for the ee is ±1%.

The NPs are functionalized in an excess of ligands (ligand/Pt = 12.3) to ensure that a nearly ligand saturated particle surface is achieved. The coverage is then determined by the steric demand of the ligand.25 As the two five-membered-ring ligands exhibit similar sizes, the ligand coverage and thus the activity are expected to be similar. The same should hold for the sixmembered-ring ligands. On the basis of this idea, the higher activity obtained for six-membered-ring ligands in comparison to five-membered-ring ligands can be explained as well. On nearly ligand saturated surfaces the reaction is expected to be limited by the availability of ligand-free surface atoms, required for the adsorption and activation of H2.27 A six-membered-ring ligand is slightly larger in size than a five-membered ring. Hence, it may lead to a lower ligand coverage and a higher number of ligand-free sites. As a result, the activity of Pt NPs functionalized with six-membered-ring ligands is slightly higher than that obtained for the five-membered-ring structures. The activity enhancement was found to depend neither on the position nor on the presence of the ligand’s carboxyl group (see Figure 2). However, an ester group within the reactant is needed (see Figure 1). Hence, the activity enhancement must be related to an interaction of the reactant’s ester group with the ligand’s N−H proton. In contrast, the ee (see Figure 2) is sensitive to both the presence and position of the carboxyl group within the ligand. These findings lead to the conclusion that the interaction of the ester group with PRO must be 2-fold. First, it interacts with the amine proton of PRO, leading to the pronounced activity increase. Second, it interacts with the carboxyl group of PRO enhancing the asymmetric bias. On the basis of these conclusions two ligand−reactant interaction modes can be postulated (see Figure S3 in the Supporting Information). Further tests (see section 2 in the Supporting Information) using ACAC as a reactant reveal that the model 3983

DOI: 10.1021/acscatal.7b00422 ACS Catal. 2017, 7, 3979−3987

Research Article

ACS Catalysis

Figure 4. Stereoselectivities for the hydrogenation of β-keto esters of different steric demand over PRO-Pt NPs (see box in the middle for reaction equation). By tuning the steric demand of the β-keto ester substituents in accordance with the interaction model illustrated in Figure 3, the ee can be enhanced to 73% (see MDOV). The experimental error for the ee is ±1. All ees correspond to experiments run to full conversion (=850 turnovers per ligand).

Changes on the ester side (R′′, highlighted in green in Figures 3 and 4) should destabilize the desired two-point binding, as the oxygen of the alkoxy group becomes less accessible to interact with the carboxyl group (see Figure 3a). Thus, the ΔΔG⧧ value and the ee are expected to decrease. Extending the chain length on the ester side (R′′) shows a minor increase of the ee from 39% (MAA) to 42% (BAA). Linear alkyl substituents are flexible structures that can adopt various conformations to minimize repulsive steric interactions.45 In the present case the alkyl substituents may rearrange in a way that they point away from the surface and toward the reaction medium as in an ordered SAM (self-assembled monolayer).46 This diminishes their influence on the transition state stability. In contrast, when rigid bulky groups (isopropyl and tert-butyl) are introduced that are not able to rearrange, the ee decreases to 35% (TBA) in accordance with our proposal derived from the postulated interaction model. The steric properties of the reactant affect not only the stereoselectivity but also the activity (see Figure S9 in the Supporting Information). The active center is located at the surface. Hence, the reactant has to pass the ligand shell to reach the active center in order to become activated and catalytically converted. When the size of the reactant is increased by introducing bulkier groups such as isopropyl and tert-butyl, the probability for the CO group to reach the active center decreases. As a result, the catalytic activity decreases with increasing bulkiness of the reactant’s substituents. A significant advantage of heterogeneous over homogeneous catalysts is the possibility of recycling the catalyst by filtration, because heterogeneous catalysts are not dissolved in the reaction medium. For PRO-functionalized NPs we have previously shown that recycling does not significantly alter the chemoselectivity but the stereoselectivity was found to decrease.17 As mentioned above, we have optimized the catalyst preparation for this work to maximize the stereoselectivity. Recycling experiments performed with PRO-functionalized Pt NPs prepared by the new protocol (see section 2.1) reveals that the catalyst can be recovered at least two times without any significant loss of stereoselectivity (see Table S1 in the Supporting Information). We thus conclude that catalysts prepared by the new protocol are more stable and that now recycling can be performed without inducing any significant change within the ligand shell. Finally, we demonstrate that the knowledge presented above about the asymmetric control can be combined with our

Thus, the structure of the transition state will closely resemble that of the adsorbed state (Hammond’s postulate). The diastereotopic transition states should be similar to the interaction mode shown in Figure 3a. Its diastereotopic counterpart that leads to the product with opposite configuration (S) is obtained by rotating the reactant 180° along the N−H bond (see Figure 3b). If the reactant adsorption is quasi-equilibrated and additional side reactions are neglected, the ee is determined by the difference in free activation energy of the two diastereomeric transition states (ΔΔG⧧, see Figure S6).44 According to the Curtin−Hammett principle, the larger the ΔΔG⧧, the higher the ee. In the present case the transition state leading to the product with R configuration must be that with lower free energy: that is, the state with the two-point binding (see Figure 3a). This can be explained by the additional interaction between the ester and the carboxyl group that does not occur for the other interaction mode (see Figure 3b). To increase the ee, the desired two-point binding transition state has to be further stabilized (see section 4 and Figure S7 in the Supporting Information) or the undesired one-point binding transition state has to be destabilized (see Figure S8 in the Supporting Information). In homogeneous catalysis this aim is achieved by tuning the steric and electronic interactions between ligand and reactant. In the present case we already take advantage of specific electronic effects (two-point binding vs one-point binding). Therefore, steric interactions remain as a parameter to further tune the ligand−reactant interactions and the resulting asymmetric bias. For this purpose we varied the steric demand of the reactant’s substituents and determined the influence on the ee (see Figure 4). Introducing sterically more demanding groups on the carbonyl side (R′, highlighted in red in Figure 4) should destabilize the undesired one-point binding (see Figure S8 in the Supporting Information) for steric reasons, as indicated in Figure 3b. Thus, the ee is expected to increase. Extending the chain length leads to an increase in the ee from 39% (MAA) to 64% (MOHp). By introducing bulkier groups such as isopropyl and tert-butyl the ee is even enhanced to 73% (MDOV). This vast improvement that we achieved by understanding the ligand−reactant interaction on a molecular level clearly demonstrates the enormous potential for applying molecular principles in heterogeneous catalysis to tune the catalytic properties of supported catalysts. 3984

DOI: 10.1021/acscatal.7b00422 ACS Catal. 2017, 7, 3979−3987

Research Article

ACS Catalysis

Figure 5. Hydrogenation of EOPP (see black structure labeled 1) over ligand-free (a) and PRO-functionalized Pt NPs (b). While over ligand-free NPs hydrogenation of the phenyl substituent occurs, PRO-Pt NPs are highly chemoselective toward the unsaturated alcohol. Furthermore, PRO-Pt NPs are also highly stereoselective as indicated by the different sizes of 2.1 and 2.2 (see Table 1 for ee).

previous findings on the molecular control of chemoselectivity by PRO.17 We have identified β-keto esters as privileged for a strong asymmetric bias and varied the steric demand of the substituents in order to enhance the stereoselectivity. Thereby, we focused on alkyl substituents which are not reactive under the applied reaction conditions. However, most organic molecules that are used as chiral building blocks for pharmaceuticals and fine chemicals contain reactive moieties: in particular, phenyl substituents.47 This means that a catalyst has to be stereo- and chemoselective. To demonstrate that this problem can be tackled by ligand functionalization, we investigated the hydrogenation of EOPP (see black structure labeled 1 in Figure 5). The desired reaction is the chemoselective hydrogenation of the CO group to obtain the unsaturated alcohol (see green structures 2.1 and 2.2 in Figure 5). This product is of interest, for example, as a chiral building block for fine chemicals.48 Over ligand-free NPs substantial hydrogenation of the phenyl group is obtained (see Table 1), leading to the formation of the saturated ketone (see red structure 3 in Figure 5a) and the saturated alcohol (see gray structure 4 in Figure 5a). If the reaction is run to full reactant conversion, the chemoselectivity toward the desired unsaturated alcohol is only 63% (see Table 1) with no ee, due to the lack of any chiral information. In contrast, by functionalization with PRO a chemoselectivity of more than 99% is obtained at full conversion and no side products are obtained, even when the reaction time is prolonged (see Figure S10b in the Supporting Information). Previously, we have shown that the functionalization of Pt NPs with PRO allows for achievement of such high chemoselectivity for the hydrogenation of acetophenone.17 Thereby, the effect of the ligands is 2-fold. First, it dilutes large ensembles of adjacent surface atoms.49 As large ensembles are needed for the adsorption and hydrogenation of aromatic moieties,50 dilution by ligands inhibits this undesired reaction pathway.51 Second, due to the ligand’s N−H proton an alternative pathway for the hydrogenation of the CO group is opened that exhibits a higher reaction rate than the purely metal-catalyzed reaction.35

On the basis of our molecular model introduced above for controlling the asymmetric bias, EOPP should be well suited for obtaining a high ee, as it exhibits a bulky phenyl substituent at the position that causes destabilization of the undesired transition state (see Figure 3b). In fact, we obtained an ee of 74% (see Table 1), which is similar to the value obtained for Table 1. Ratio and Stereoselectivity of Products Formed by Hydrogenation of EOPP (See Figure 5) over “Ligand-Free” and PRO-Functionalized Pt NPsa product (%) ligand-free Pt NPs PRO-Pt NPs

2.1 + 2.2

3

4

ee for 2.1 (%)

63 >99

3

34

0 74 ± 1

a

The values were taken from measurements run to full reactant conversion. This corresponds to 1700 turnovers per ligand for PRO-Pt NPs. The formation of the different products over time is shown in Figure S10 in the Supporting Information.

the bulky tert-butyl substituent. This highlights the potential of transferring concepts from homogeneous catalysis to supported catalysts because it allows, for example, the achievement of high chemo- and stereoselectivities. Similar to the case in homogeneous catalysis, the stereoselectivity of ligand-functionalized NPs is the result of steric and electronic ligand−reactant interactions. Therefore, it is expected that it is not possible to design a generally applicable catalyst for generating enantioenriched alcohols with high ees; the ligand has to be adjusted with respect to the reactant. The ee of 74% achieved here is still below the high stereoselectivities achieved with tailored homogeneous catalysts and chiral modifiers.42,52 However, it is already above the limit proposed as “high” in the asymmetric catalysis literature (ee ≥65% ee)53 and close to the benchmark reported by industrial chemists for any suitable applicability of an asymmetric catalyst (80% ee).54 The ability to perform simultaneously chemo- and stereoselective catalysis paves the way to studying reactants more complex than those shown in Figure 4. This should allow for crossing the 80% ee limit by 3985

DOI: 10.1021/acscatal.7b00422 ACS Catal. 2017, 7, 3979−3987

ACS Catalysis utilizing even stronger steric constraints and studying the catalytic preparation for chemically more demanding structures. While our work focuses on hydrogenation catalysis, a further interesting reaction class for which ligand-functionalized NPs have already shown potential is the formation of C−C bonds.55,56 In such cases two reactants that both exhibit steric demand are connected. This provides further possibilities to use steric constraints in order to achieve even higher stereoselectivities. The study presented here mainly focused on PRO, but our model suggests that α-amino acids in general are wellsuited to induce a strong asymmetric bias to β-keto esters. We will therefore explore if high stereoselectivities for reactants with relevance as chiral building blocks (e.g., EOPP) can be achieved by varying the ligand based on the model introduced here. In this way it may eventually be possible to achieve stereoselectivities similar to those of tailored homogeneous catalysts.

ACKNOWLEDGMENTS

■

REFERENCES

(1) Heitbaum, M.; Glorius, F.; Escher, I. Angew. Chem., Int. Ed. 2006, 45, 4732−4762. (2) Knowles, W. S. Adv. Synth. Catal. 2003, 345, 3−13. (3) Orito, Y.; Imai, S.; Niwa, S.; Hung, N. G. Yuki Gosei Kagaku Kyokaishi 1979, 37, 173−174. (4) Blaser, H. U. Chem. Rev. 1992, 92, 935−952. (5) Hess, R.; Vargas, A.; Mallat, T.; Bürgi, T.; Baiker, A. J. Catal. 2004, 222, 117−128. (6) Ferri, D.; Burgi, T. J. Am. Chem. Soc. 2001, 123, 12074−12084. (7) Blaser, H.-U.; Pugin, B. In Handbook of asymmetric heterogeneous catalysis; Ding, K., Uozumi, Y., Eds.; Wiley: Hoboken, NJ, 2008; pp 413−437. (8) Blaser, H.-U.; Pugin, B.; Studer, M. In Enantioselective heterogeneous catalysis: academic and industrial challenges; De Vos, D., Vankelecom, I. F. J., Jacobs, P. A., Eds.; Wiley-VCH: Weinheim, Germany, 2000; pp 1−17. (9) Weng, Z.; Zaera, F. J. Phys. Chem. C 2014, 118, 3672−3679. (10) Kunz, S. Top. Catal. 2016, 59, 1671−1685. (11) Huang, Y.; Xu, S.; Lin, V. S. Y. ChemCatChem 2011, 3, 690− 694. (12) Blaser, H. U.; Jalett, H. P.; Monti, D. M.; Baiker, A.; Wehrli, J. T. Stud. Surf. Sci. Catal. 1991, 67, 147−155. (13) Blaser, H. U.; Jalett, H. P.; Lottenbach, W.; Studer, M. J. Am. Chem. Soc. 2000, 122, 12675−12682. (14) Noyori, R.; Ohkuma, T. Angew. Chem., Int. Ed. 2001, 40, 40−73. (15) Wehrli, J. T.; Baiker, A.; Monti, D. M.; Blaser, H. U. J. Mol. Catal. 1989, 49, 195−203. (16) Mallat, T.; Orglmeister, E.; Baiker, A. Chem. Rev. 2007, 107, 4863−4890. (17) Schrader, I.; Warneke, J.; Backenköhler, J.; Kunz, S. J. Am. Chem. Soc. 2015, 137, 905−912. (18) Marshall, S. T.; O’Brien, M.; Oetter, B.; Corpuz, A.; Richards, R. M.; Schwartz, D. K.; Medlin, J. W. Nat. Mater. 2010, 9, 853−858. (19) McKenna, F.-M.; Wells, R. P. K.; Anderson, J. A. Chem. Commun. 2011, 47, 2351−2353. (20) Vilé, G.; Almora-Barrios, N.; Mitchell, S.; López, N.; PérezRamírez, J. Chem. - Eur. J. 2014, 20, 5926−5937. (21) Chung, Y.-H.; Chung, D. Y.; Jung, N.; Sung, Y.-E. J. Phys. Chem. Lett. 2013, 4, 1304−1309. (22) Zhou, Z.-Y.; Kang, X.; Song, Y.; Chen, S. J. Phys. Chem. C 2012, 116, 10592−10598. (23) Jeong, H.; Kim, C.; Yang, S.; Lee, H. J. Catal. 2016, 344, 609− 615. (24) Kahsar, K. R.; Schwartz, D. K.; Medlin, J. W. J. Am. Chem. Soc. 2014, 136, 520−526. (25) Kunz, S.; Schreiber, P.; Ludwig, M.; Maturi, M. M.; Ackermann, O.; Tschurl, M.; Heiz, U. Phys. Chem. Chem. Phys. 2013, 15, 19253− 19261. (26) Baiker, A. J. Mol. Catal. A: Chem. 1997, 115, 473−493. (27) Schrader, I.; Neumann, S.; Himstedt, R.; Zana, A.; Warneke, J.; Kunz, S. Chem. Commun. 2015, 51, 16221−16224. (28) Wang, Y.; Ren, J. W.; Deng, K.; Gui, L. L.; Tang, Y. Q. Chem. Mater. 2000, 12, 1622−1627. (29) Neumann, S.; Grotheer, S.; Tielke, J.; Schrader, I.; Quinson, J.; Zana, A.; Oezaslan, M.; Arenz, M.; Kunz, S. J. Mater. Chem. A 2017, 5, 6140−6145. (30) Perego, C.; Peratello, S. Catal. Today 1999, 52, 133−145.

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acscatal.7b00422. All investigated reactions, a modified version of Figure 1 with the reaction rates being normalized to the number of ligand-free surface atoms, the two suitable interaction models in accordance with the catalytic investigations, illustration of the prochiral faces of the reactant, a model that illustrates how the reaction may proceed on the surface, energy profiles illustrating the effect of stabilizing and destabilizing transition states, and time-dependent product formation for the hydrogenation of EOPP over “ligand-free” and PRO-Pt NPs (PDF)

■

■

I.S. and S.K. gratefully acknowledge the “Fonds der Chemischen Industrie” (FCI) for financial support through a Liebig Research Grant. The authors further acknowledge Dr. Moskaleva and Dr. Schäfer for fruitful suggestions and the DFG for financial support (KU 3152/4-1). The authors thank Dr. Dülcks for performing GC-MS measurements.

4. SUMMARY A model to describe the ligand−reactant interaction on ligandfunctionalized NPs has been developed, and its validity has been tested successfully. Our model explains the origin of stereoselectivity for ligand-functionalized NPs on a molecular level. The application of general guiding principles from homogeneous catalysis enabled us to raise the ee by a rational choice of reactants to a level that demonstrates the potential of ligands for asymmetric heterogeneous catalysis. Our results highlight the possibility of transferring basic principles from homogeneous to heterogeneous catalysis by ligand functionalization and the enormous potential arising from this idea. Furthermore, this work is believed to pave the way for the design of a novel type of asymmetric heterogeneous catalyst that may eventually become competitive with homogeneous catalysts and chiral modifiers used for supported catalysts.

■

Research Article

AUTHOR INFORMATION

Corresponding Author

*S.K.: tel, +49-421-218 63187; e-mail, SebKunz@uni-bremen. de. ORCID

Sebastian Kunz: 0000-0002-2512-9316 Notes

The authors declare no competing financial interest. 3986

DOI: 10.1021/acscatal.7b00422 ACS Catal. 2017, 7, 3979−3987

Research Article

ACS Catalysis (31) Wang, X. D.; Stover, J.; Zielasek, V.; Altmann, L.; Thiel, K.; AlShamery, K.; Baumer, M.; Borchert, H.; Parisi, J.; Kolny-Olesiak, J. Langmuir 2011, 27, 11052−11061. (32) Borner, A. Eur. J. Inorg. Chem. 2001, 2001, 327−337. (33) Sawamura, M.; Ito, Y. Chem. Rev. 1992, 92, 857−871. (34) Park, J. Y.; Aliaga, C.; Renzas, J. R.; Lee, H.; Somorjai, G. A. Catal. Lett. 2009, 129, 1−6. (35) Ohkuma, T.; Ooka, H.; Ikariya, T.; Noyori, R. J. Am. Chem. Soc. 1995, 117, 10417−10418. (36) Blaser, H. U.; Studer, M. Acc. Chem. Res. 2007, 40, 1348−1356. (37) Abdur-Rashid, K.; Faatz, M.; Lough, A. J.; Morris, R. H. J. Am. Chem. Soc. 2001, 123, 7473−7474. (38) Clapham, S. E.; Hadzovic, A.; Morris, R. H. Coord. Chem. Rev. 2004, 248, 2201−2237. (39) Kohlmann, O.; Steinmetz, W. E.; Mao, X. A.; Wuelfing, W. P.; Templeton, A. C.; Murray, R. W.; Johnson, C. S. J. Phys. Chem. B 2001, 105, 8801−8809. (40) Hens, Z.; Martins, J. C. Chem. Mater. 2013, 25, 1211−1221. (41) Shibata, N.; Kohno, J.; Takai, K.; Ishimaru, T.; Nakamura, S.; Toru, T.; Kanemasa, S. Angew. Chem., Int. Ed. 2005, 44, 4204−4207. (42) Osawa, T.; Harada, T.; Takayasu, O. Top. Catal. 2000, 13, 155− 168. (43) Wiberg, K. B.; Crocker, L. S.; Morgan, K. M. J. Am. Chem. Soc. 1991, 113, 3447−3450. (44) Walsh, P. J.; Kozlowski, M. C. Fundamentals of Asymmetric Catalysis; University Science Books: Mill Valley, CA, 2009. (45) Love, J. C.; Wolfe, D. B.; Haasch, R.; Chabinyc, M. L.; Paul, K. E.; Whitesides, G. M.; Nuzzo, R. G. J. Am. Chem. Soc. 2003, 125, 2597−2609. (46) Love, J. C.; Estroff, L. A.; Kriebel, J. K.; Nuzzo, R. G.; Whitesides, G. M. Chem. Rev. 2005, 105, 1103−1169. (47) Blaser, H. U.; Spindler, F.; Studer, A. Appl. Catal., A 2001, 221, 119−143. (48) Xie, J. H.; Liu, X. Y.; Yang, X. H.; Xie, J. B.; Wang, L. X.; Zhou, Q. L. Angew. Chem., Int. Ed. 2012, 51, 201−203. (49) Altmann, L.; Kunz, S.; Bäumer, M. J. Phys. Chem. C 2014, 118, 8925−8932. (50) Puddu, S.; Ponec, V. Recl. Trav. Chim. Pays-Bas 1976, 95, 255− 257. (51) Pang, S. H.; Schoenbaum, C. A.; Schwartz, D. K.; Medlin, J. W. ACS Catal. 2014, 4, 3123−3131. (52) Hu, A.; Ngo, H. L.; Lin, W. Angew. Chem., Int. Ed. 2004, 43, 2501−2504. (53) Lavoie, S.; Laliberté, M.-A.; Temprano, I.; McBreen, P. H. J. Am. Chem. Soc. 2006, 128, 7588−7593. (54) Pugin, B.; Blaser, H. U. Top. Catal. 2010, 53, 953−962. (55) Yasukawa, T.; Miyamura, H.; Kobayashi, S. Chem. Soc. Rev. 2014, 43, 1450−1461. (56) Yasukawa, T.; Miyamura, H.; Kobayashi, S. ACS Catal. 2016, 6, 7979−7988.

3987

DOI: 10.1021/acscatal.7b00422 ACS Catal. 2017, 7, 3979−3987