Mechanistic Insights into the Proline-Directed Enantioselective

Langmuir 2007, 23, 6113-6118

6113

Mechanistic Insights into the Proline-Directed Enantioselective Heterogeneous Hydrogenation of Isophorone Alexander I. McIntosh, David J. Watson, and Richard M. Lambert* Department of Chemistry, UniVersity of Cambridge, Cambridge CB2 1EW, U.K. ReceiVed October 18, 2006. In Final Form: February 20, 2007 The adsorption rates onto a range of platinum single-crystal surfaces of key species involved in the proline-directed heterogeneous enantioselective hydrogenation of isophorone were investigated by electrochemical means. Specifically, the uptakes of the prochiral reactant (isophorone), the chiral hydrogenation product (3,3,5-trimethylcyclohexanone), and the chiral directing agent ((R)- and (S)-proline) were examined. The effects of R,S chiral kink sites on the adsorption of (R,S)-proline were also studied. The reactant adsorbs ∼105 times faster than the chiral modifier so that under conditions of competitive adsorption the latter is entirely excluded from the metal surface. Supplementary displacement and reaction rate measurements carried out with practical Pd/carbon catalysts show that under certain reaction conditions isophorone quickly displaces preadsorbed proline from the metal surface. Thus both kinetics and thermodynamics ensure that the chiral modifier can play no role in any surface-mediated process that leads to enantiodifferentiation. These results are fully consistent with the recent proposal1 that the crucial step leading to enantiodifferentiation occurs in the solution phase and not at the metal surface. In addition, it is found that there is no preferred diastereomeric interaction between (R,S)-proline and R,S step kink sites on Pt{643} and Pt{976}, implying that such sites do not play a role in determining the catalytic behavior of supported metal nanoparticles.

1. Introduction The heterogeneous enantioselective hydrogenation of functionalized alkenes is a little explored yet very important area from the viewpoint of organic synthesis. Nitta et al. investigated palladium catalysts modified by cinchonidine in the enantioselective hydrogenation of the CdC function in R-phenylcinnamic acid derivatives.2-4 Tungler et al. worked on the metal-catalyzed, proline-directed, enantioselective hydrogenation of isophorone to form 3,3,5-trimethylcyclohexanone (hereafter TMCH), where they found relatively modest enantiomeric excesses (ee’s) using a range of metal catalysts.5,6 This latter system is the subject of the present article. Here we use two techniques to examine the adsorption rates and strengths of the relevant molecular species (reactant, product, chiral directing agent) to obtain additional fundamental insight into the reaction mechanism, which, according to our recent proposal1 does not involve the metal surface in the crucial step that leads to enantioselectivity. First, cyclic voltammetry (CV) was used to study adsorption rates onto a number of different platinum single-crystal surfaces. Then, the displacement of one species by another on Pd/C catalyst samples was studied under reaction conditions. (Pt and Pd catalysts exhibit generally similar behavior toward the reaction of interest, exhibiting similar ee’s at equivalent conversion, as would be expected on the basis of the “metal-independent” mechanism that we proposed in ref 1.) Our earlier catalytic measurements1,7 were carried out with Pd/C. However the production of chiral * Corresponding author. E-mail: [email protected]. (1) McIntosh, A. I.; Watson, D. J.; Burton, J. W.; Lambert, R. M. J. Am. Chem. Soc. 2006, 128, 7329-7334. (2) Nitta, Y. Chem. Lett. 1999, 635-636. (3) Nitta, Y.; Shibata, A. Chem. Lett. 1998, 161-162. (4) Sugimura, T.; Watanabe, J.; Okuyama, T.; Nitta, Y. Tetrahedron: Asymmetry 2005, 16, 1573-1575. (5) Tungler, A.; Kajtar, M.; Mathe, T.; Toth, G.; Fogassy, G.; Petro, J. Catal. Today 1989, 5, 159-171. (6) Tungler, A.; Mathe, T.; Petro, J.; Tarnai, T. J. Mol. Catal. 1990, 61, 259267. (7) McIntosh, A. I. Doctoral Thesis, University of Cambridge, Cambridge, U.K., 2006.

step-kinked single-crystal Pd electrodes is nontrivial,8 so the electrochemical measurements were carried out using Pt samples. Several very different models have been proposed to explain the origin of enantioselectivity in metal-catalyzed hydrogenations: all of them have a common feature, namely, that the metal surface is directly involved in the reaction step that leads to enantiodifferentiation. The behavior of those that involve cinchonidine as the chiral agent has been interpreted in terms similar to those used for the enantioselective hydrogenation of R-ketoesters, also directed by cinchonidine. Hypotheses include either (i) the formation of ordered arrays of chiral adsorption sites on the metal surface,9 (ii) a 1:1 docking interaction between randomly adsorbed chiral templates and the prochiral substrate,10,11 or (iii) selective adsorption of the chiral modifier at chiral (R,S) kink sites, thereby converting a racemic catalyst into a chiral one as suggested by Attard et al.12 However, a very different mechanism has been proposed6 for the isophorone/ proline system. According to this view, the surface is not pretemplated in any way by the chiral agent. Instead, the reaction proceeds via a reactive intermediate that is preformed in solution and then undergoes enantioselective hydrogenation after adsorption on the metal surface. Nevertheless, the metal surface is seen as being critically involved in the hydrogenation event that leads to enantioselectivity. Very recently, however,1 we have demonstrated that this view cannot be correct. We showed that the key step that leads to enantiodifferentiation occurs homogeneously in the solution phase: the racemic heterogeneous hydrogenation of isophorone is followed by kinetic resolution that effectively removes one enantiomer of the TMCH product. The crucial point is that the metal surface merely supplies chemisorbed hydrogen (8) Attard, G. Personal Communication. (9) Sutherland, I. M.; Ibbotson, A.; Moyes, R. B.; Wells, P. B. J. Catal. 1990, 125, 77-88. (10) Bonello, J. M.; Williams, F. J.; Lambert, R. M. J. Am. Chem. Soc. 2003, 125, 2723-2729. (11) Simons, K. E.; Meheux, P. A.; Griffiths, S. P.; Sutherland, I. M.; Johnston, P.; Wells, P. B.; Carley, A. F.; Rajumon, M. K.; Roberts, M. W.; Ibbotson, A. Recl. TraV. Chim. Pays-Bas Belg. 1994, 113, 465-474. (12) Attard, G. A. J. Phys. Chem. B 2001, 105, 3158-3167.

10.1021/la063064h CCC: $37.00 © 2007 American Chemical Society Published on Web 04/18/2007

6114 Langmuir, Vol. 23, No. 11, 2007

McIntosh et al.

Figure 1. Cyclic voltammograms of Pt{111}. (a) CV of the clean Pt{111} surface with key features identified. (b) In the presence of 5 × 10-7 M isophorone. (c) In the presence of 5 × 10-4 M TMCH. (d) In the presence of 5 × 10-2 M (S)-proline, 40 cycles at 50 mV s-1 between 0 and 0.8 V, showing every fifth cycle for clarity. In b-d, red curves correspond to the clean Pt{111} surface.

atoms and is not involved in chiral direction of the reaction. This leaves open the question as to why the system behaves in this waysthe surface plays no role in steering the reaction by interacting with either the chiral modifier or any species derived from it. Here, by means of electrochemical measurements, we report the uptake rates from solution of the reactant (isophorone), the product (racemic TMCH), and the chiral modifier ((R)- and (S)proline) on a range of platinum single-crystal surfaces encompassing flat {111}, stepped {211} and {331}, and stepped chiral {643}S, {643}R, {976}S, and {976}R cases. The results provide an answer to the question posed above. 2. Experimental Section Cyclic voltammetry experiments were conducted in a threeelectrode dual-compartment electrochemical cell. Data acquisition was carried out using EChem software (AD Instruments) in conjunction with a PowerLab/200 data acquisition unit (AD Instruments) and a PG 1.0 potentiostat (Ionic Systems). All quoted potentials were measured against a saturated Pd-hydrogen electrode in direct contact with the electrolyte. The electrolyte used was a 0.1 M solution of sulfuric acid (Aldrich 99.999%). A polycrystalline Pt mesh was used as the counter electrode. Solutions were prepared using (R)- or (S)-proline (Aldrich, 99%), 3,3,5-trimethylcyclohexanone ((()-TMCH) (Aldrich 98%), and isophorone (Aldrich 97%), with the latter two being purified by freeze-thaw and then subsequently analyzed by GC. Millipore Milli-Q water (18.2 MΩ cm-1) was used for making up solutions and for cleaning the cell. Prior to use, the working solutions were deoxygenated by bubbling oxygen-free nitrogen through them for 30 min: deoxygenated solutions were preserved during experiments by maintaining an inert

atmosphere of nitrogen in the cell. Experiments were conducted at room temperature using Pt {111}, {211}, {331}, {643}S, {643}R, {976}S and {976}R single-crystal electrodes, prepared and cleaned by the technique developed by Clavilier et al.13 CV sweeps were run at 50 mV s-1 between 0 and 0.8 V (i.e., within the hydrogen UPD and double-layer regions). The time dependence of the uptake of proline, isophorone, and racemic TMCH (hereafter all references relate to racemic TMCH) were investigated by repeatedly running CVs in 0.1 M sulfuric acid solutions containing varying concentrations of each individual species. The working electrode was first cleaned in a gas-air flame for ∼60 s, cooled in a H2-rich environment to maintain good 1 × 1 order of the surface,13 and then quenched in ultrapure water. The electrode was protected from contamination during transfer to the electrochemical cell by leaving a droplet of ultrapure water adhering to the surface. A meniscus contact was then made between the electrode and electrolyte such that only the well-defined singlecrystal face was in contact with the electrolyte. The first CV was recorded immediately after contact, with additional CVs being recorded repeatedly until no further changes were detectable. Displacement experiments were conducted using standard Quickfit glassware with a Pd/C catalyst supplied by Alfa Aesar, and the resulting solutions were examined using Electrospray ionization mass spectrometry (ESI-MS).

3. Results and Discussion 3.1. Relative Uptake Rates of Isophorone, (S)-Proline, and TMCH. The rate of uptake of each adsorbate was followed by measuring the attenuation of the HUPD peak area as described (13) Clavilier, J.; Faure, R.; Guinet, G.; Durand, R. J. Electroanal. Chem. 1980, 107, 205-209.

Mechanistic Insights into Isophorone Hydrogenation

Langmuir, Vol. 23, No. 11, 2007 6115 Scheme 1 . Reaction Scheme Proposed for the Enantioenrichment by Kinetic Resolution of (S)-TMCH, (S)-3, in the Pd/C Catalyzed Hydrogenation of (S)-proline, (S)-2, and Isophorone, 1

Figure 2. Uptakes of (a) 5 × 10-7 M isophorone, (b) 5 × 10-4 M TMCH, and (c) 5 × 10-2 M (S)-proline onto Pt{111}.

Figure 3. Hydrogen uptake for the Pd/C-catalyzed hydrogenation of (a) 1:1 proline/isophorone added simultaneously (2 mmol of isophorone and 2 mmol of (S)-proline in 4 mL of methanol with 10 mg of Pd/C) and (b) isophorone added to proline-premodified Pd/C (2 mmol of isophorone in 4 mL of methanol with 10 mg of Pd/C).

previously.14 This method also provides information about the discrimination (if any) shown by the adsorbate toward different types of surface sites. The clean-surface CV enables the identification of key features, as indicated in Figure 1a. The broad peak between 0 and 0.2 V is attributed15 to the underpotential deposition of hydrogen on (111) terraces of platinum. The sharper peak superimposed on this at 0.21 V is due to a small number of (110) defect sites on the surface.15 The second broad peak (centered around 0.4 V) results from the specific adsorption of sulfate anions on the (111) terraces, and the sharp peak at ∼0.47 V is diagnostic of good long-range order on the (111) terraces.15 The concentrations of (S)-proline, TMCH, and isophorone were adjusted so as to give similar rates of uptake that could be conveniently observed at a sweep rate of 50 mVs-1 between 0 and 0.8 V. The resulting CVs for the adsorption of these species onto Pt {111} are shown in Figure 1b-d. In every case, it is apparent that with increasing elapsed time the area under the CV curves decreased as the number of sites available (14) Stephenson, M. J.; Lambert, R. M. J. Phys. Chem. B 2001, 105, 1283212838. (15) Clavilier, J. Interfacial Electrochemistry; Wieckowski, A., Ed.; Marcel Dekker: New York, 1999.

for hydrogen atom adsorption/desorption was reduced. The “spike” at ∼0.47 V was rapidly attenuated, almost entirely vanishing after the first cycle, indicating that the long-range order on the terraces was immediately disrupted by all three adsorbates. These observations suggest that the adsorbates were immobile; otherwise, one would expect the defect sites (characterized by the peak at ∼ 0.21 V) to be blocked before the terrace sites were populated. The data shown in Figure 1 were used to construct uptake curves for each of the adsorbates as follows. The time dependence of the area under each curve (between 0.0 and 0.6 V and excluding the double-layer region) was evaluated. For a one-electron process such as HUPD, these areas provide a measure of the total charge on the electrode. By taking a ratio of these values with the total charge for the clean Pt{111} surface (240 µC cm-2), one obtains a value for the fraction of the surface that remains uncovered (θfree). Therefore, the fraction of surface covered by the adsorbate is (1 - θfree) ) θ, and the time dependence of this quantity for each of the three species is shown in Figure 2. Keeping in mind the very different concentrations used for the three adsorbates, the data clearly show that uptake rates for the three species are very different. The initial rate of isophorone adsorption is ∼ 103 times greater than that of TMCH, which is itself adsorbed ∼102 times faster than (S)-proline. Essentially the same behavior (not shown) was exhibited by the {211}, {331}, {643}R/S, and {976}R/S surfaces, and in every case, there was no detectable tendency for the adsorbates to preferentially populate one type of surface site first. Note that in every case some HUPD sites remained vacant at saturation coverage of the organic adsorbate. This may be understood as follows. The

6116 Langmuir, Vol. 23, No. 11, 2007

McIntosh et al.

Figure 4. Cyclic voltammograms of Pt{643}R and Pt{643}S in 0.1 M H2SO4 with 0.05 M (a) (S)-proline and {{643}R, (b) (R)-proline and {643}R, (c) (S)-proline and {643}S, and (d) (R)-proline and {643}S at 50 mV s-1 between 0 and 0.8 V, showing every fifth cycle for clarity.

relatively large organic molecules adsorb randomly and are immobile: this will always leave a number of bare Pt ensembles “trapped” because they cannot accommodate the organic adsorbate. However, the very small H adatoms can get into these nooks and crannies, hence the observed effect. The very different uptake rates of the chiral agent and the prochiral substrate strongly suggest that in the proline/isophorone system premodification of the catalyst surface by (S)-proline can be ruled out. The reactant, isophorone, adsorbs ∼105 times faster than the chiral directing agent so that the latter is totally excluded from the surface, which it cannot therefore modify before hydrogenation commences at the isophorone saturated surface. This stands in sharp contrast to other prototypical systems, such as cinchonidine-2-4,16-23 and tartaric acid-directed24,25 ketoester hydrogenations. Parenthetically, we note that the adsorption rate of TMCH is ∼100 times slower than that of isophorone. The implication is that as long as significant quantities of reactant remain, poisoning of the surface by adsorption of the product is unimportant. (16) Orito, Y.; Imai, S.; Niwa, S. J. Chem. Soc. Jpn. 1979, 1118. (17) Orito, Y.; Imai, S.; Niwa, S. J. Chem. Soc. Jpn. 1980, 670. (18) Orito, Y.; Imai, S.; Niwa, S. J. Chem. Soc. Jpn. 1980, 137. (19) Orito, Y.; Imai, S.; Niwa, S.; Hung, N. G. J. Synth. Org. Chem. Jpn. 1979, 37, 173. (20) Nitta, Y. Bull. Chem. Soc. Jpn. 2001, 74, 1971-1972. (21) Nitta, Y.; Kobiro, K. Chem. Lett. 1996, 897-898. (22) Nitta, Y.; Kubota, T.; Okamoto, Y. Bull. Chem. Soc. Jpn. 2000, 73, 26352641. (23) Nitta, Y.; Kubota, T.; Okamoto, Y. Bull. Chem. Soc. Jpn. 2001, 74, 21612165. (24) Izumi, Y. AdV. Catal. 1983, 32, 215-271. (25) Webb, G.; Wells, P. B. Catal. Today 1992, 12, 319-337.

Even though isophorone adsorption is overwhelmingly faster than proline adsorption, the possibility remains that preadsorbed proline might survive on the Pt surface to play a chirally directing role if the reactants (hydrogen, isophorone) were to be subsequently added. We therefore examined this possibility as follows by carrying out measurement with a Pd/C catalyst. A catalyst sample (100 mg) was stirred in a methanolic solution of proline overnight (0.5 M), filtered, washed (methanol) and dried. Ten milligrams of this was then added to a methanolic solution of isophorone (0.5 M) onlysno proline was present in solution at this point. This mixture was stirred for 4 h, the time taken by a typical run for the complete hydrogenation of isophorone under these conditions.1 Electrospray ionization mass spectrometry (ESI-MS) was then used to analyze the resulting solution. The only species detected corresponded to the proline/isophorone condensation product that results from the homogeneous reaction between proline and isophorone1 (m/z found: (M + H)+, 236.1642 C14H22NO2 requires M, 236.1645 and (M + Na)+, 258.1454 C14H21NO2Na requires M, 258.1465). The clear implication of this result is that initially adsorbed proline was displaced from the catalyst surface into the solution phase by (more strongly adsorbed) isophorone, as would be expected, for example, on the basis of simple competitive Langmuir adsorption. Finally, we measured the time dependence of hydrogen uptake under catalytic conditions for two different conditions. Figure 3a shows the behavior when 2 mmol of proline and 2 mmol of isophorone were added simultaneously to 10 mg of Pd/C catalyst in methanol in the presence of 1 bar H2. Figure 3b shows hydrogen uptake when a Pd/C catalyst, premodified with proline and then

Mechanistic Insights into Isophorone Hydrogenation

Langmuir, Vol. 23, No. 11, 2007 6117

Figure 5. Cyclic voltammograms of Pt{976}R and Pt{976}S in 0.1 M H2SO4 with 0.05 M (a) (S)-proline and {976}R, (b) (R)-proline and {976}R, (c) (S)-proline and {976}S, and (d) (R)-proline and {976}S for 40 cycles at 50 mV s-1 between 0 and 0.8 V, showing every fifth cycle for clarity.

washed as described above, was added to a methanolic solution of isophorone under 1 bar H2: the TMCH produced after a has an ee of (S)-TMCH of 49% whereas b produced racemic TMCH. Thus pretreatment of the Pd/C catalyst has no effect on the ee of TMCH produced. This, in conjunction with the electrochemical measurements, shows that proline is completely excluded from the catalyst surface. Therefore, under conditions of catalytic hydrogenation the crucial event that leads to an enantiomeric excess of the TMCH product must occur homogeneously, in the solution phase, and not at the metal surface. We have shown elsewhere1 that this event is homogeneous kinetic resolution of racemic TMCH produced by the rapid (racemic) hydrogenation of isophorone. The reaction scheme proposed in ref 1 and verified with respect to every step of the homogeneous chemistry is shown in Scheme 1 (see earlier). The present results provide very strong support for this scheme in that they demonstrate unequivocally that the heterogeneous chemistry cannot involve an enantiodifferentiating event: the metal surface is always saturated with the reactant (isophorone) that undergoes racemic hydrogenation to TMCH. With hindsight, given the functionalities of isophorone and proline, one might have guessed that in competitive adsorption the alkene would win against the amino acid, but it would not have been possible to predict that the effect would be as overwhelming as it actually is. 3.2. Exclusion of a Possible Alternative Mechanism: Uptake of (R)- and (S)-Proline by Chiral Pt Surfaces. A potentially

important alternative mechanism needs to be considered. Attard et al. have argued convincingly that in the case of the Pt-catalyzed enantioselective hydrogenation of ethyl pyruvate directed by cinchona alkaloids12,26 diastereomeric interaction of the chiral agent with the R step-kink sites at the surface of a metal nanoparticle can result in an excess of vacant S step-kink sites, thus converting a previously racemic catalyst into a chiral one. If (R,S)-proline were to adsorb strongly enough at (R,S) stepkink sites to compete with isophorone and hydrogenation rates at the vacant step-kink sites were intrinsically much faster than at achiral terrace sites, then this type of mechanism could introduce a heterogeneous component into the overall enantioselectivity of the system. Accordingly, we studied the uptake of (R)- and (S)-proline by Pt {643}R, {643}S, {976}R, and {976}S surfaces in order to determine whether or not there was any preferred diastereomeric interaction between a given enantiomer of proline and R,S stepkinked sites, as has indeed been found for other systems.12,27-29 The relevant results for the uptake of (R)- and (S)-proline by Pt {643}R and {643}S are shown in Figure 4; the corresponding data for Pt {976}R and {976}S are shown in Figure 5. The step(26) Attard, G. A.; Gillies, J. E.; Harris, C. A.; Jenkins, D. J.; Johnston, P.; Price, M. A.; Watson, D. J.; Wells, P. B. Appl. Catal. A 2001, 222, 393-405. (27) Gellman, A. J.; Horvath, J. D.; Buelow, M. T. J. Mol. Catal. A: Chem. 2001, 167, 3-11. (28) Horvath, J. D.; Gellman, A. J. J. Am. Chem. Soc. 2001, 123, 7953-7954. (29) Ahmadi, A.; Attard, G.; Feliu, J.; Rodes, A. Langmuir 1999, 15, 24202424.

6118 Langmuir, Vol. 23, No. 11, 2007

Figure 6. Uptakes of 0.05 M (S)-proline and (R)-proline onto Pt{643}R and Pt{643}S for (a) (S)-proline and {643}R, (b) (R)-proline and {643}R, (c) (S)-proline and {643}S, and (d) (R)-proline and {643}S.

kink geometries of the {976} and {643} surfaces are identical, but the {976} surface contains (111) terraces that are twice the width of those on the {643} surface. The data shown in Figures 4 and 5 were used to construct Figures 6 and 7, which show the uptake behavior of (S) and (R)-proline onto the four different chiral platinum surfaces. It is apparent that the rate of adsorption of (R)- and (S)-proline onto the chiral {643} and {976} surfaces is independent of the chirality of the adsorbate, showing that there is no detectable diastereomeric response in the adsorption of proline by chiral Pt surfaces. Moreover, this lack of discrimination is independent of terrace width. These findings provide strong evidence that in the proline/isophorone system such effects do not act to endow practical dispersed catalysts with chiral catalytic behavior. In particular, they appear to exclude the hypothetical possibility that on practical catalysts consisting of Pt nanoparticles a strongly adsorbed proline enantiomer, preferentially bound at R stepkink sites, could be involved in steering the enantioselective hydrogenation of isophorone. In summary, the proline-directed enantioselective heterogeneous catalytic hydrogenation of isophorone must occur via a

McIntosh et al.

Figure 7. Uptakes of 0.05 M (S)-proline and (R)-proline onto Pt{976}R and Pt{976}S for (a) (S)-proline and {976}R, (b) (R)-proline and {976}R, (c) (S)-proline and {976}S, and (d) (R)-proline and {976}S.

mechanism that does not involve the metal surface in the enantiodifferentiating step. This is so because both kinetics and thermodynamics ensure that the reactant is adsorbed to the complete exclusion of the chiral agent. Moreover, the absence of a preferred diastereomeric interaction between the chiral modifier and chiral adsorption sites appears to exclude the possibility that the uncapped step-kink sites play a role in determining the overall enantioselectivity. The key enantiodifferentiating step must occur homogeneously and not heterogeneously. A strongly surface-tethered chiral agent is a minimum necessary requirement for a heterogeneous enantioselective catalytic process; relevant work is in progress. Acknowledgment. AIM acknowledges the award of a research studentship by the UK Engineering and Physical Sciences Research Council. This work was supported by EPSRC Research Grant EP/D000866/1. LA063064H