Advantages of polymer-supported multivalent organocatalysts for the Baylis-Hillman reaction over their soluble analogues

Introduction

Heterogeneous catalysts bear a number of advantages, as compared to their soluble analogues [1]. Among these benefits are the ease of separation and purification of the products, as well as the potential for reuse. However, there are also a number of significant drawbacks, which are usually associated with the conversion of homogeneous catalysts into their heterogenized counterparts [2]. These shortcomings include a multitude of different catalytic sites, difficulties in the catalyst characterization and in the understanding of their mechanistic behavior. The most practically detrimental disadvantage, which is usually associated with the heterogenization of homogeneous catalysts, is the decrease in the activity and selectivity (e.g., chemoselectivity, enantioselectivity) of the catalytic systems. While one can fight the decrease in the activity by increasing the reaction time or the catalyst loading, the reduced selectivity is irreparable.

During the past decade we demonstrated that dendronization of an insoluble polymer support could remedy the compromised activity and/or selectivity of heterogenized organometallic [3–6], as well as organic catalysts [7–11]. In particular, we recently showed that supported dendritic systems decorated with N-alkylated imidazoles exhibit an outstanding multivalency effect in the catalysis of the Baylis-Hillman reaction (Scheme 1) [10, 11]. A remarkable increase in the yield of the Baylis-Hillman adducts followed the increase in the dendritic generation of the spacer tethering the imidazole units to the insoluble matrix. Various modes of connecting N-alkyl imidazole units to the spacer were examined and this, with a secondary amine as a connecting functionality, was found to be the most active. The positive multivalency effect was observed for a number of combinations of substrates (Table 1). Moreover, in all catalytic experiments, carried with N-alkyl imidazoles tethered to the support through monovalent as well as multivalent spacers, the chemoselectivity was outstanding, favoring the conversion of the aldehyde (the limiting reactant) to the Baylis-Hillman product in ca. 95%. In a typical ¹H NMR of the crude extract of the filtrate of the reaction mixture, one can see the Baylis-Hillman product, traces of the starting aldehyde (when the conversion is not complete), traces of solvents (when not evaporated thoroughly) and very little of anything else (see, for instance, Fig. 1).

Scheme 1

Dendritic effect of the polymer-supported catalysts in the model reaction.

Table 1

Dendritic effect in the catalysis of Baylis-Hillman reaction with other substrates.^a

Entry	Catalyst	Ar	R1, R2	Yieldb (%)
1	PS-G0(A-Im)	Ph	Me, H	<1
2	PS-G1(A-Im)	Ph	Me, H	7
3	PS-G2(A-Im)	Ph	Me, H	82
4	PS-G0(A-Im)	4-MeOC6H4	Me, H	<1
5	PS-G1(A-Im)	4-MeOC6H4	Me, H	4
6	PS-G2(A-Im)	4-MeOC6H4	Me, H	38
7c	PS-G0(A-Im)	4-NO2C6H4	-(CH2)2-	2
8c	PS-G1(A-Im)	4-NO2C6H4	-(CH2)2-	8
9c	PS-G2(A-Im)	4-NO2C6H4	-(CH2)2-	90
10c	PS-G0(A-Im)	4-NO2C6H4	-(CH2)3-	<1
11c	PS-G1(A-Im)	4-NO2C6H4	-(CH2)3-	9
12c	PS-G2(A-Im)	4-NO2C6H4	-(CH2)3-	75
13c	PS-G0(A-Im)	4-NO2C6H4	MeO, H	<1
14c	PS-G1(A-Im)	4-NO2C6H4	MeO, H	5
15c	PS-G2(A-Im)	4-NO2C6H4	MeO, H	27

^aReaction conditions: 1 mmol of aldehyde, 3 mmol of electron-deficient olefin and 10 mol% (0.1 mmol) of the catalytic units in 1 mL DMF:H₂O 1:1 v/v mixture, 20 h, at 25 °C.

^bDetermined by NMR.

^cReaction time 24 h.

Fig. 1

A typical ¹H NMR of the crude filtrate of the reaction catalyzed by a polymer-supported catalyst.

In order to better understand the factors behind this remarkable effect, we decided to examine soluble N-alkylimidazoles, and particularly soluble analogues of our supported catalysts, in the model reaction of methyl vinyl ketone (MVK) with p-nitrobenzaldehyde.

Results and discussion

Simple N-methylimidazole, as a catalyst, led to rapid consumption of the nitrobenzaldehyde limiting reactant, but promoted a formation of an indecipherable multiproduct mixture (Fig. 2a). N-(3-aminopropyl)imidazole, the commercially available building block, which we used for the construction of the abovementioned supported catalysts, catalyzed the conversion of the nitrobenzaldehyde into the Baylis-Hillman adduct, but also to additional products, as can be seen in the ¹H NMR spectrum of the crude reaction filtrate (Fig. 2b). Following these results, we decided to prepare more precise analogues of the supported non-dendritic as well as dendritic catalysts. In solution, contrary to the synthetic route on solid support, the reaction of the N-(3- aminopropyl)imidazole with the suitable benzyl chlorides is compromised by overalkylation [12].

Fig. 2

A typical ¹H NMR of the crude filtrate of the reaction catalyzed by (a) N-methylimidazole; (b) N-(3-aminopropyl)imidazole.

Accordingly the catalysts were prepared from the corresponding aldehydes via a two-step reductive amination (Scheme 2) [13]. The non-dendritic and the first generation dendritic analogues were prepared in the simpler version (G0(A-Im) and G1(A-Im)), as well as in the more elaborate version with a benzyloxy substituent imitating the Wang linker of the polymer-supported catalysts (BnO-G0(A-Im) and BnO-G1(A-Im) respectively). Since the model catalytic experiments did not reveal any substantial difference between the simpler and the benzyloxy-carrying catalysts of the same generations, most of the subsequent studies were carried out with the more readily attainable simpler versions of the catalysts.

Scheme 2

Synthesis of the soluble analogues of the polymer-supported catalysts. Reagents and conditions: (i) N-(3-aminopropyl)imidazole, toluene, reflux; (ii) NaBH₄, MeOH, rt; (iii) dimethyl-5-hydroxyisophthalate, K₂CO₃, acetone, reflux; (iv) LiAlH₄, THF, reflux; (v) MnO₂, THF, reflux.

The model reaction between methyl vinyl ketone and p-nitrobenzaldehyde, tested with the five potential catalysts, again produced a mixture of products (2 h reaction time). However, these mixtures were less complex then those described in Fig. 2, with the Baylis-Hillman adduct being the dominant component (beyond the starting material). Since in the ¹H NMR spectrum of the mixture at least some of the signals of this adduct are well-resolved and can be integrated separately, calculating the chemoselectivity of these catalysts for the Baylis-Hillman reaction became possible.

The results of these experiments, summarized in Table 2, point to a number of trends that are principally different from those observed with the heterogeneous catalysts. The conversion of the limiting reactant decreases with the increase in the generation of the catalyst. This is in contrast to the case of the supported analogues, which showed an increase in the conversion and yield (practically the same) as the catalyst generation increased. On the other hand, while the catalyst generation did not affect the chemoselectivity of the polymer-supported catalysts (which was almost perfect regardless of the dendritic generation), there is a profound improvement in the chemoselectivity exhibited by the catalysts in solution upon the increase in the dendritic generation.

Following these results, we wondered whether the chemoselectivity is conversion-dependent or catalyst-dependent. In other words, is it a reduced conversion that leads to a better chemoselectivity, or is it truly the generation of the catalyst that does so? In order to examine this question we conducted two experiments. In the first experiment, a series of equivalent catalytic runs were quenched after varying reaction times. An analysis of the crude reaction mixtures demonstrated, as expected, that as the reaction time increases the conversion of the aldehyde into various products increases as well (Fig. 3; Table 3, entries 1–4). It is noteworthy that, concurrent with the increase in the conversion, the chemoselectivity does not deteriorate and even slightly improves. This finding contradicts the conversion-dependent selectivity hypothesis. In another experiment, which proves that the chemoselectivity parameter of this type of catalysts is dependent on the dendritic generation of the catalyst rather than on the reaction conversion, G0(A-Im) and G1(A-Im) catalysts were used in the model reactions of 1 h and 3.5 h, respectively. In this manner conversion in these experiments was practically identical (56%). However, the chemoselectivity and, consequently, the yield obtained with the first generation catalyst were noticeably better than those achieved with its non-dendritic analogue (Table 3, entries 5 and 6).

Fig. 3

Catalysis parameters as a function of reaction time.

The striking difference in the behavior of the imidazole based catalysts, when free in DMF:H₂O solution vs. when bound to the polystyrene support, forced us to speculate that the hydrophobic environment around the catalytic sites may play an important role. In our previous communication [10], we already demonstrated that in the later case the hydrophobic polystyrene-divinylbenzene matrix of the Wang resin extracts the reactants from the highly polar DMF:H₂O solution [14, 15]. In the homogeneous case, however, it seems that the reaction takes place in the polar solvent bulk. The isolated organic/hydrophobic environment of the catalytic sites that affects substrate specificity and reaction selectivity is frequently found in enzymes [16, 17], and was recently imitated in artificial macromolecular and supramolecular systems [18–22].

In a very primitive attempt to imitate the hydrophobic environment of the supported systems, toluene was added to the reaction mixture, and the reaction was conducted in the biphasic, emulsion-like environment, generated by rapid stirring of the mixture (Table 4). For equivalent reaction times, the reaction conducted with toluene shows lower conversion, but somewhat better chemoselectivity, compared to the reaction without toluene (entry 2 vs. entry 1). For the reaction with toluene that was conducted for longer time, in order to reach the same conversion as in the reaction without toluene, the improvement in the chemoselectivity is more significant (entry 3 vs. entry 1). The impact of the toluene addition on the conversion and selectivity follows, therefore, the same trends imparted by the immobilization of the catalysts on polystyrene, though to a lesser extent.

Conducting the model reaction in a mixture with higher water content (Table 4, entry 4) led to almost quantitative conversion in only 2 h, but caused a three-fold decrease in the chemoselectivity. On the other hand attempts to improve the chemoselectivity by reducing the proportion of water in the mixture (entry 5) or the addition of the non-functionalized Wang polystyrene beads to the mixture (entry 6) did not improve the chemoselectivity. It is possible that longer reaction times in DMF:H₂O 19:1 or even in pure DMF will lead to a better selectivity (since it slightly improves with the increase in the conversion), but it seems that the improvement will not be significant. In the case of the Wang polystyrene additive, it is likely that polar diamino-diimidazole G1(A-Im) catalyst prefers to reside in solution rather than inside the microporous polystyrene pores, unless bound there covalently as in the case of the supported catalysts.

The results described above for the reaction of p-nitrobenzaldehyde with methyl vinyl ketone, were confirmed in two additional Baylis-Hillman reactions (Scheme 3, Table 5), namely that of p-nitrobenzaldehyde with cyclopentenone (entries 1–3) and that of p-chlorobenzaldehyde with methyl vinyl ketone (entries 4–9). In both cases the conversion of the aldehyde (the limiting reactant) declines with the generation increase, but this decrease is compensated by the significant improvement in the chemoselectivity. In this way, the best yield is almost always obtained with the highest generation catalyst. Since these reactions involved substrates that are less reactive than those of the original model reaction, 1:1 DMF:H₂O solution was used in order to accelerate the reactions (entries 1–6) [23]. This change led to low chemoselectivities. For comparison, the reaction of p-chlorobenzaldehyde with methyl vinyl ketone in a 9:1 DMF:H₂O mixture exhibited significantly higher selectivities at the expense of the conversion (entries 7–9).

Scheme 3

Additional Baylis-Hillman reactions.

Table 5

Additional reactions with the soluble catalysts.^a

Entry	Catalyst	Ar	R1, R2	Conversionb (%)	Yieldb (%)	Chemoselectivity (%)
1	G0(A-Im)	4-NO2C6H4	-(CH2)2-	90	30	33
2	G1(A-Im)	4-NO2C6H4	-(CH2)2-	73	28	38
3	G2(A-Im)	4-NO2C6H4	-(CH2)2-	69	31	45
4c	G0(A-Im)	4-ClC6H4	Me, H	93	9	10
5c	G1(A-Im)	4-ClC6H4	Me, H	87	13	15
6c	G2(A-Im)	4-ClC6H4	Me, H	87	17	20
7d	G0(A-Im)	4-ClC6H4	Me, H	25	5	20
8d	G1(A-Im)	4-ClC6H4	Me, H	14	6	43
9d	G2(A-Im)	4-ClC6H4	Me, H	12	6	50

^aReaction conditions: 1 mmol of aldehyde, 3 mmol of unsaturated ketone and 10 mol% (0.1 mmol) of the catalytic units in 1 mL DMF:H₂O 1:1 v/v mixture, 17 h, at 25 °C.

^bDetermined by NMR.

^cReaction time 7 h.

^dReaction in 1 mL DMF:H₂O 9:1 v/v mixture.

The question that remained open for a while is the nature of the byproducts, that their formation lowers the chemoselectivity. In an attempt to address this issue, a crude mixture formed in the reaction of p-nitrobenzaldehyde with methyl vinyl ketone was separated on a chromatographic column. Beyond the regular Baylis-Hillman product, the only compound that was isolated in a reasonably pure form in the fraction that followed that of the Baylis-Hillman adduct, was the product obtained by consecutive Baylis-Hillman and aldol reactions (or vice versa, Scheme 4) [24, 25]. Subsequent fractions were obtained as inseparable mixtures. However, valuable information was provided by MS analyses of these fractions. These MS spectra exhibited a number of molecular peaks with m/z, matching the outcome of the formula 70x +151y + 23, where 70 is the molecular mass of methyl vinyl ketone, 151 is the mass of the aldehyde, 23 is the mass of the sodium ion (required for converting the molecules into the positively charged ions), while x and y are integers. Thus, larger peaks were observed for m/z of 616, 686 and 767, which correspond to the formula with x=2/y=3, x=3/y=3 and x=2/y=4 respectively. A number of smaller molecular peaks of higher m/z were observed as well, among them 988, 1209, 1360 and 1511, which correspond to the formula with x=3/y=5, x=4/y=6, x=4/y=7 and x=4/y=8, respectively. These findings point to the formation of multiple molecular adducts, which incorporate several methyl vinyl ketone fragments and several nitrobenzaldehyde fragments into a single molecular structure. These byproducts, which can presumably be formed via combinations of Baylis-Hillman, aldol and Rauhut-Currier reactions [26], as well as via intramolecular cyclizations (in sequences of various order), are likely to be responsible for the observed decline in the chemoselectivity.

Scheme 4

A byproduct obtained via consecutive Baylis-Hillman and aldol reactions.

Conclusions

In conclusion, we discovered that the trends observed in the Baylis-Hillman reaction within the series of soluble imidazole-decorated dendritic catalysts are markedly different from those of their polystyrene-bound analogues. There is a negative dendritic effect on the general catalyst activity. However, there is a significant positive effect on the chemoselectivity. The combination of the two causes superiority of the higher generation dendrons over the lower generation ones or over their monovalent analogues, since the increased reaction time can compensate for the reduction in activity, but the reduced chemoselectivity is irreparable. Furthermore, some of our experiments suggest that the significantly superior chemoselectivity (and the lower activity, as well) of the supported catalysts is imparted by the hydrophobic envelopment of the reactive sites in the polymer. Further studies aimed at supporting this hypothesis, as well as at exploiting it for the design of more selective homogeneous catalysts of the Baylis-Hillman reaction, are underway. Although, the abovementioned effects were not observed in the case of the aldol addition (another organocatalyzed reaction that was explored in our group with dendritic catalysts [7–9]) in spite of the possible byproduct formation [27], the data is not yet sufficient to decide whether the phenomenon is limited to the specific reaction described in this manuscript, or is of a more general character.

Experimental section