Babak
Karimi
*a,
Asghar
Zamani
a,
Sedigheh
Abedi
a and
James H.
Clark
b
aDepartment of Chemistry, Institute for Advanced Studies in Basic Sciences (IASBS), PO. Box 45195–1159, Gava Zang, Zanjan, Iran. E-mail: karimi@iasbs.ac.ir; Fax: +98-241-4214949; Tel: +98-241-4153225
bClean Technology Centre, University of York, York, Yorkshire, UK YO10 5DD. E-mail: jhc1@york.ac.uk; Fax: +44 1904 432705; Tel: +44 1904 432559
First published on 14th November 2008
Preparation and characterization of a variety of immobilized palladium catalyst, based on either ligand functionalized amorphous or ordered mesoporous silica, is described. The resulting Pd-loaded materials act as efficient catalyst for the oxidation of a variety of alcohols using molecular oxygen and air. Our studies show that in the case of supported palladium catalyst on hybrid amorphous silica, the nature of ligand and the solvent could effectively control the generation of nanoparticles. Furthermore, we have found that nanoparticles with smaller size and higher activity were generated from the anchored palladium precursor when the aerobic oxidation of alcohols was carried out in α,α,α-trifluorotoluene (TFT) instead of toluene. On the other hand, in the case of aerobic oxidation reactions by using supported palladium catalyst on hybrid SBA-15, the combination of organic ligand and ordered mesoporous channels resulted in an interesting synergistic effect that led to enhanced activity, prevention of Pd nanoparticles agglomeration, and finally generation of a durable catalyst.
![]() | ||
Scheme 1 |
Catalyst 1 was characterized by atomic absorption spectroscopy (AA), thermogravimetric analysis (TGA), and DRIFT-IR spectroscopy.15 From the TGA analysis of L1@SiO2, it was calculated that the loading of the bipyridyl ligand bound to the silica surface was 0.13 mmol.g−1. The loading of palladium in 1 was determined using AA and shows a loading at 0.120 ± 0.001 mmol.g−1. This indicates that more than 90% of the surface-bound ligand were complexed with palladium.†15
The second catalyst (Cat 2, Pd@SiO2-L2) was synthesized by APS (0.33 mmol.g−1) precursor using a known procedure with slight modifications (Scheme 2).16
![]() | ||
Scheme 2 |
TGA analysis was used to determine the amount of ligand incorporated into APS.† Weight loss is mainly divided into three regions: below 100 °C, 150 °C–400 °C and 400–614 °C. Weight loss 30–151 °C was assigned to the loss of adsorbed water (1.8%). The large weight loss between 480–615 °C was owing to the decomposition of covalently bonded organic groups (4.6%). The amount of ligand anchored on the surface of L2@SiO2 was found to be 0.28 mmol.g−1. This data was further confirmed by elemental analysis. The catalyst 2 (Pd@SiO2-L2) was then prepared by stirring a mixture of L2@SiO2 (4 g) and Pd(OAc)2 (0.112 g, 0.5 mmol) in dry acetone (100 mL) at room temperature for 24h. After stirring, the resulting solid was filtered, and washed with acetone until the washings were colourless and dried at 95 °C overnight to afford Pd@SiO2-L2. The amount of palladium anchored on the surface of 2 was found to be 0.08 mmol.g−1 (corresponding to 28% of the ligands available) on the basis of atomic absorption spectroscopy. To determine the thermal stability, TGA analysis of 2 was also conducted in air from room temperature to 700 °C, and typical weight loss curves have been shown.† This sample shows a weight loss of a small amount of loosely bound water (less than 2%) below 200 °C. This is followed by a weight loss of about 9.1% between 350 and 500 °C due to the decomposition of organic ligand bound palladium complex. There after, an additional loss of 3.2% is observed, probably resulting principally decomposition of free anchored imine complex (see supporting information).
In order to prepare the third catalyst (Pd@SBA-15-L3, 3), we chose to employ the ordered mesoporous silicate SBA-1517 as a support for a bidentate ligand because it has regular porosity, and high surface area. In addition, it is well known that organic groups inside large pore mesoporous materials are more accessible than those on amorphous silica. Therefore we reasoned that SBA-15, owing to the above-mentioned property, might be better suited as a support for preparing heterogeneous Pd catalysts. SBA-15 was obtained from pluronic P123 (EO20PO70EO20, EO = ethylene oxide, PO = propylene oxide, MAV = 5800, Aldrich) as a lyotropic ligand that was liquid crystal templated, and (EtO)4Si under acidic conditions following the reported literature procedure.17b The resulting SBA-15 was then functionalized with a bipyridylamide ligand followed by complexation with Pd(OAc)2 to give the corresponding immobilized catalyst 3 (Scheme 3).†
![]() | ||
Scheme 3 |
A typical nitrogen adsorption/desorption type IV profile with a sharp hysteresis loop, which is characteristic of the highly ordered mesoporous materials, was obtained for 3 (Fig. 1a).
![]() | ||
Fig. 1 The isotherm plot of a sample of Catalyst 3: (a) before the first reaction cycle and; (2) after the first reaction cycle. |
A BET surface area of 455 m2.g−1 and a total pore volume of 0.76 cm3.g−1 were obtained for the material. Thus values are smaller than those for the starting SBA-15 (864 m2g−1). BJH calculations showed an average pore diameter of 7.6 nm for 3, a value which is in good agreement with the pore diameter estimated from the TEM image (Figs. 4a, b). On the basis of the AA analysis of a solution obtained by washing the catalyst with nitric acid, the amount of palladium loading on 3 was found to be 0.022 ± 0.001 mmol.g−1. We used such a low loading in order to minimize catalyst leaching.
|
|||||||
---|---|---|---|---|---|---|---|
Entry | Catalyst/support | x (mol%) | solvent | time (h) | yield (%)a,b | ||
O2 | air | O2 | air | ||||
a GC yield unless otherwise stated. b Yields in parentheses refer to isolated pure products. c TFT = α,α,α-trifluorotoluene. | |||||||
1 | 1/SiO215 | 4.0 | toluene | 8 | 12 | >99 (85) | >99 |
2 | 1/SiO2 | 4.0 | TFT c | 8 | 12 | >99 | >99 |
3 | 1/SiO2 | 4.0 | toluene | 2 | 2 | 7 | 5 |
4 | 1/SiO2 | 4.0 | TFT | 2 | 2 | 11 | 9 |
5 | 2/SiO2 | 3.2 | toluene | 1.25 | 1.25 | >99 | >99 |
6 | 2/SiO2 | 2.0 | toluene | 3 | 3 | >99 | >99 |
7 | 2/SiO2 | 1.0 | toluene | 7 | 7 | >99 | >99 |
8 | 2/SiO2 | 0.4 | TFT | 1.5 | 2 | >98 | >98 |
9 | 2/SiO2 | 0.2 | TFT | 4 | — | >98 | — |
10 | 2/SiO2 | 0.2 | Toluene | 2 | — | 35 | — |
11 | 2/SiO2 | 0.2 | TFT | 2 | — | 51 | — |
12 | 3/SBA-1512 | 0.4 | toluene | 3.5 | 5.5 | >99 (83) | >99 |
13 | 3/SBA-15 | 0.4 | TFT | 3.5 | 5 | >99 | >99 |
14 | 3/SBA-15 | 0.4 | toluene | 2 | — | 72 | — |
15 | 3/SBA-15 | 0.4 | TFT | 2 | — | 73 | — |
16 | NHC-Pd/IL@SiO223 | 5 | toluene | 12 | — | Nd | — |
17 | NHC-Pd/IL@SiO223 | 5 | TFT | 12 | — | Nd | — |
Although all catalysts were able to furnish the corresponding benzaldehyde in excellent yield within the indicated time obviously, catalyst 2 in TFT (entry 11) and catalyst 3 in both TFT and toluene (entries 14, 15) exhibited higher catalytic activities than that of catalyst 1 (entries 3, 4).
It is worth mentioning that catalyst 2 in TFT (equivalent to 0.4 mol% Pd) is even more effective than the same catalyst sample in toluene (entry 11 vs. 12). Moreover, our preliminary investigations showed that catalyst 2 in TFT and toluene gave consistent activity in 5 and 6 subsequent reactions and exhibited TONTFT = 2500 and TONTol = 187, respectively at 80 °C, with almost quantitative yield of the product. To clarify the above mentioned observations, we have studied the evolution of 2 by means of transmission electron microscopy (TEM) analysis. Interestingly, inspection of the TEM image of a sample of catalyst 2 after recovery from the aerobic oxidation of benzyl alcohol in TFT and toluene clearly indicates, the generation of Pd nanoparticles onto the surface of amorphous SiO2 (Fig. 2).
![]() | ||
Fig. 2 TEM images of the recovered catalyst 2. (a) After the first reaction cycle in toluene (particle size in the range of 6–17 nm and an average particle size ≈10 nm); (b) After the first reaction cycle in TFT (particle size in the range 3–14 nm and an average particle size ≈6.7). |
Furthermore, the images show that the oxidation reaction using catalyst 2 in TFT clearly favors production of smaller metal nanoparticles than those obtained in toluene. It is well-known that the catalyst performance can be very sensitive to particle size because the surface structure and electronic properties can change greatly within the nano-size range.18 Moreover, it has been demonstrated that the catalytic activity of Pd nanoparticles increases with decreasing their average size.19 Therefore, it would be logical to speculate that the increase in the size of the nanoparticles onto the surface of 2 in toluene is the reason why the process is slower than in TFT.
This observation might be due to the fact that TFT more efficiently caps many of the free metallic surface sites of the nanoparticles and provides fewer sites for the Ostwald ripening process.20 On the other hand, the lower catalytic activity (higher catalyst loading) of 2 in toluene is probably due to generation of a smaller amount of nanoparticles of the bigger size throughout the silica surface due to larger nanoparticle aggregations. We may expect that in the case of 1 (Pd@SiO2-L1) and 3 (Pd@SBA-15-L3), TFT will result in rapid nucleation of a great number of smaller Pd nanoparticles compared to toluene and therefore cause a similar solvent effect. However, as can be seen from the data in Table 1 neither 1 nor 3 showed this effect (entries 3 vs. 4 and 14 vs. 15).
Therefore it seems that in the case of 1 and 3 other factors should also be taken into account. On the basis of our previous studies, we also found that catalyst 1 requires high catalyst concentration (up to 4 mol%, 80 °C, 8 h) and it also suffers from the disadvantage of prolonged reaction time and a significant reduction in its catalytic activity after three reaction cycles.15 At first glance, one might conclude that this catalytic deactivation is due to Pd leaching from the solid to the solution. However, both individual AA analysis of the solution of each three reaction cycles and hot-filtration test indicates no Pd species that leached into the solution.15 To gain better insight into the structural changes, the recovered sample of 1 has also been studied by means of TEM to try to find a reason for its rapid deactivation. Fig. 3 shows a typical TEM (in dark-field mode) image of 1 after the third cycle of aerobic oxidation of benzyl alcohol. It can be seen, this material shows extensive agglomeration of palladium with an irregular size distribution above 100 nm.
![]() | ||
Fig. 3 Dark-field TEM image of catalyst 1 after the third reaction cycle. |
This result clearly shows that Pd-agglomeration is the major reason for deactivation of catalyst 1 just after the third cycle. One reason the process is so much more prominent in the case of both 2 and 3 than in 1 is presumably owing to smooth hydrolysis of un-complexed imino group in 1 during the oxidation reaction using the water by-product. Unfortunately, we have not yet been able to obtain any evidence for the hydrolyzed bipyridyl ketone that leached from catalyst 1. However, TGA analysis of the recovered catalyst 1 showed a slight decrease in the weight loss as compared with the fresh catalyst, in support of the above proposal.
The lack of binding provides a means of rapid sintering of small palladium particles throughout the surface of silica and produce very large (much less reactive) palladium clusters. This may also explain why the catalyst 1 shows similar behaviour in both TFT and toluene, i.e. in the absence of a well-bonded ligand, Pd agglomeration would rapidly result in catalyst deactivation regardless to the small difference in the stabilizing ability of TFT compared to toluene. This observation also highlights the crucial role of anchored ligands in generating and stabilizing nanoparticles during a typical catalytic process.
With regard to material 3 we have reported in an earlier communication12 that at 0.4 mol% it is an efficient and durable heterogeneous catalyst for the aerobic oxidation of a wide range of alcohols including benzyl alcohol itself. In this study, we have found that the catalytic activity and durability of 3 was not significantly altered by changing the solvent from toluene to TFT (Table 1, entries 14 vs. 15). Moreover, after the first oxidation cycle using 3 (Table 1, entry 8) to afford benzaldehyde in 83% isolated yield (>99% conversion), the recovered catalyst exhibited consistent catalytic activity in 12 consecutive reactions (total TON ≅ 3000 at 80 °C),12 which is much higher than those observed in the case of both 1 and 2. To find a reason for this high reactivity and more importantly, high durability of 3 at a molecular level, TEM analysis has also been performed on a sample of 3 before and after catalysis. Fig. 4 shows representative TEM images of 3 before the first cycle of aerobic oxidation both perpendicular (Fig. 4a) and across (Fig. 4b) to the hexagonally uniformed channels of ≈7 nm in size. On the other hand, Fig. 4c shows a representative TEM image of the same sample of 3 after recovery from the first cycle of aerobic oxidation of benzyl alcohol. By comparing these two set of TEM images before and after the first reaction cycle, we can see that Pd nanoparticles with regular size (≤ 7 nm) were mostly generated inside the highly ordered channels and that the nanoarchitecture of the catalyst (SBA-15 channels) largely survived.21
![]() | ||
Fig. 4 TEM images of fresh catalyst 3. (a) Perpendicular to the channels; (b) across the channels; (c) TEM image of the recovered catalyst 3 across the channels, well-dispersed palladium nanoparticles with a relatively regular size can clearly be seen inside the channels. |
Moreover, the N2adsorption-desorption analysis of the recovered catalyst (Fig. 1b) showed a very similar isotherm to those of the fresh catalyst 3 (Fig. 1a) with relatively sharp adsorption and desorption branches in the P/P0 range of 0.5–0.8. This strongly indicates a relatively narrow mesopore size distribution even in the recovered catalyst, although the total pore volume decreases from 0.76 to 0.57 cm3 g−1. This result also suggests that most of the nanometre-scaled void space and channels of the host SBA-15 remain open, although a small portion of the channels may be blocked by Pd nanoparticles (Fig. 2b). In order to better clarify the role of bipyridyl ligands in our protocol we set up two sets of controlled experiments. First, we prepared a new catalyst in which SBA-15 lacking organic ligands was loaded with Pd(OAc)2 with the same Pd loading as 3. The oxidation of benzyl alcohol was then conducted under the same reaction condition using this catalyst. Interestingly, we found that the corresponding benzaldehyde was produced in >99% conversion after 5 hours in the first experiment. However, when the oxidation of benzyl alcohol was repeated for two subsequent runs with the same catalyst sample, the catalyst activity was dramatically decreased. The significant deactivation of the material along with a color change to dark grayish is presumably due to the formation of large palladium clusters (palladium black) onto the outer surface of SBA-15.12 In the second experiment, the SBA-15 with 3-cyanopropyl group was loaded with Pd(OAc)2 and the resulting pale yellow solid was tested for catalytic activity in the same process. In this case, the solid catalyst showed a high degree of leaching and also the corresponding benzaldehyde was produced in low (less than 25%) yield after 5 h due to the rapid formation of Pd-black. Therefore, we believe that the bipyrydyl ligands in catalyst 3 provide a means of uniform distribution of mononuclear palladium center throughout the solid support, to ensure controlled nanoparticle formation mostly inside the ordered mesoporous channels of SBA-15.12 We can also explain why the catalyst 3 shows high and the same catalytic activity in both TFT and toluene. The cooperation of functionalized organic ligand (the bipyridine ligand) inside the ordered mesoporous channels and size restriction imposed by meso-channels of the parent SBA-15 resulted in an interesting synergistic effect enhancing the activity, preventing Pdnanoparticles agglomeration, and producing an inherently durable catalyst independent of the choice of TFT or toluene. Much to our surprise, our recently developed silica supported N-heterocyclic carbene palladium/ionic liquid catalysts that have been shown to be highly efficient and recyclable catalysts for the Heck reaction, were ineffective in oxidizing benzyl alcohol under similar reaction conditions to those of catalysts 1–3 (Table 1, entries 16, 17).22 The reasons for this latter observation are under investigation.
Entry | Substrate | Product | Pd | time (h) | yield (%)a | ||
---|---|---|---|---|---|---|---|
(mol%) | O2 | air | O2 | air | |||
a GC yield based on an internal standard method unless otherwise stated. b A trace amount of the corresponding esters (≈ 4%) and carboxylic acid (≈ 11%) were formed. | |||||||
1 |
![]() |
![]() |
0.4 | 1.5 | 2 | 98 | 98 |
2 | 0.2 | 4 | — | 97 | — | ||
3 |
![]() |
![]() |
0.1 | 3 | 3 | 100 | 100 |
4 |
![]() |
![]() |
0.2 | 3 | 1.5 | 95 | 95 |
5 |
![]() |
![]() |
0.7 | 10 | — | 91 | — |
6 |
![]() |
![]() |
0.3 | 8 | 15 | 93 | 95 |
7 |
![]() |
![]() |
0.2 | 24 | — | 78 | — |
9 |
![]() |
![]() |
0.2 | 2 | — | 92 | — |
10 |
![]() |
![]() |
0.4 | 2.5 | 4 | >99 | 100 |
0.2 | 8.5 | — | >99 | — | |||
11 |
![]() |
![]() |
0.4 | 2.5 | 4 | 97 | 95 |
12 |
![]() |
![]() |
0.4 | 10 | — | 100 | — |
13 |
![]() |
![]() |
0.4 | 4 | 7 | 100 | 97 |
14 |
![]() |
![]() |
2 | 5 | 8 | 94 | 96 |
15 |
![]() |
![]() |
2 | 7 | 10 | 100 | 94 |
16 |
![]() |
![]() |
2 | 5 | 8 | 100 | 100 |
17 |
![]() |
![]() |
2 | 7 | 10 | 98 | 97 |
18 |
![]() |
![]() |
3 | 24 | — | 13b | — |
19 |
![]() |
![]() |
3 | 24 | — | 21 | — |
20 |
![]() |
![]() |
3 | 24 | — | 18 | — |
In general, many of the previously reported homogeneous transition metal complexes are unable to catalyze the oxidation of alcohols that can chelate Pd(II) catalyst, as a starting material or product, because the strong coordination to metal centers deactivates the catalyst. However, the successful oxidation of benzoin and furfuryl alcohol as model substrates shows a superior capability of 2 in oxidizing similar substrates (Table 2, entries 9, 13). Catalyst 2 in particular showed excellent reactivity for the selective oxidation of various types of allylic alcohols yielding the corresponding α,β-unsaturated carbonyl compounds in excellent yields (Table 2, entries 14–17). It is worth mentioning that in the oxidation of allylic alcohols, CC double bonds remained intact without an intramolecular hydrogen transfer. Secondary benzylic alcohols were also efficiently oxidized into the corresponding ketone (Table 2, entries 10–13). Even the reaction under air (balloon filled) instead of oxygen proceeded well, thus indicating that the reaction is not markedly retarded by the concentration of dissolved oxygen in the solvent. Unfortunately, using 2, aliphatic alcohols such as 1-octanol and 2-octanol and 4-phenyl cyclohexanol proved to be poor substrates, yielding the corresponding carbonyl compound in 13%, 21%, and 18% yields, respectively (Table 2, entries 18–20).
To determine whether the catalyst 2 is functioning in a heterogeneous manner, or whether it is merely a reservoir for more active soluble forms of Pd, various heterogeneity tests were performed. First, the reaction of benzyl alcohol was conducted in the presence of catalyst 2 for 1 hour until a conversion of 55% was reached. Then the solid was hot-filtered and transferred to another Schlenk flask containing K2CO3 in TFT at 80 °C under O2 atmosphere. The catalyst free solution was then allowed to continue to react, but no further reaction took place even after 12 h. Furthermore, our primarily investigation using AA analysis indicates that no Pd species leached into solution within the detection limit. Nevertheless, it is difficult at this stage to exactly attribute the actual catalytic activity solely to the ligand-bound Pd or to Pd nanoparticles. It would not be also surprising if the supported Pd nanoparticles serve as a reservoir for a trace of non-detectable Pd particles which react via a homogeneous pathway.22 To better clarify this issue, we have also conducted a three-phase-test with a heterogeneous alcohol (Scheme 4).
![]() | ||
Scheme 4 |
Unfortunately, this substrate was decomposed during the oxidation reaction so that we have detected the decomposed alcoholic part of the solid in the reaction solution. We are now searching to obtain a more suitable solid alcohol for this purpose and we will present the results in due course.
Footnote |
† Electronic supplementary information (ESI) available: experimental procedures; characterization of the catalysts; schemes and figures. See DOI: 10.1039/b805824e |
This journal is © The Royal Society of Chemistry 2009 |