On the selective aerobic oxidation of benzyl alcohol with Pd/Au-nanoparticles in batch and flow

3.1 Preparation and characterization of unsupported nanoparticles

We prepared the monometallic Pd NPs by the polyol process [18]. Therefore, the precursor Na₂PdCl₄ and a threefold excess of PVP were dissolved, and the pH was set to 10.5. Ethylene glycol acts as both the solvent and reducing agent. The bimetallic NPs were synthesized by successive reduction of the metal salts, i.e., the same procedure was employed with a subsequent addition of HAuCl₄•3 H₂O via syringe pump. Thus, the Pd/Au NPs with varying amounts of Au were obtained (0.5, 1, 3, 5, 10, and 20 mol%). All NPs exhibited a diameter of ca. 2 nm and were analyzed by UV/Vis spectroscopy and/or SAXS. Figure 1 compares the scattering curves obtained from SAXS experiments of different sets of Pd NPs prepared under identical reaction conditions. The intensity of the scattered beam is plotted against the scattering vector q, which is inversely proportional to the size of a particle at which the incident X-ray beam is scattered. With the help of these plots, one can easily see that the different sets are equal in size. The TEM image of one set of Pd NPs confirms a narrow size distribution (Figure 1B and C). Thus, the applied preparation method allows synthesizing Pd NPs with similar size distributions. This is also true for the bimetallic Au/Pd NPs with a low Au content (0.5–5%). In contrast, the curves of the different sets of bimetallic NPs with 20% Au, for example (Figure 2A), showed major differences at higher q values indicating differences in the particle size. Figure 2B shows a TEM image of NP set no. 3 in Figure 2A, which was used to determine the particle size, revealing a broader particle size distribution. Usually, compared to Pd NPs, a significantly larger excess of stabilizer is required to prepare pure Au NPs of approximately 2 nm diameter under the applied reaction conditions. As we kept the amount of stabilizer constant for the preparation of all Au/Pd NPs with varying compositions, an uncontrolled growth and/or agglomeration of Au might occur during the successive preparation at larger amounts of Au leading to a broad particle size distribution. Obviously, the applied successive reduction method is not suitable for the reproducible preparation of PVP-stabilized NPs of this composition. However, all the synthesized NPs reported herein were subjected to a standardized catalytic test reaction to ensure the reproducibility of the NP synthesis (at 25°C, with atmospheric pressure of air).

Figure 1

(A) Scattering curves of the different sets of PVP-stabilized Pd NPs. (B) TEM image of a Pd NP sample. Scale bar is 5 nm. (C) Size distribution derived from TEM.

Figure 2

(A) Scattering curves of the different sets of PVP-stabilized 20% Au/Pd NPs. (B) TEM image of a 20% Au/Pd NP sample. Scale bar is 5 nm. (C) Size distribution derived from TEM.

Based on the scattering curves, the SAXS measurements also allow to deduce the mean particle sizes and particle size distributions. Table 1 gives an overview on particle sizes of Pd and Pd/Au NPs with an Au content of 20% and 0.5%, respectively, and it shows clearly that the SAXS and TEM characterization delivered well-correlating results.

3.2 Catalytic testing in batch

First, the catalytic activity of the monometallic Pd NPs was tested in batch at varying temperatures with air at atmospheric pressure (Figure 3). Within the tested parameters, higher conversions were obtained at higher temperatures. Raising the temperature from 25°C to 35°C resulted in a more distinct increase of conversion than the temperature rise from 35°C to 50°C. At 50°C, complete conversion was achieved in <60 min (Figure 3). It has to be mentioned here that selectivities of ≥96% were obtained in all the experiments described in the publication at hand.

Figure 3

Temperature dependence of conversion with Pd NPs at atmospheric pressure of air in batch. Reaction conditions: 1 mmol benzyl alcohol, 25 μmol Pd NP in 10 ml iPrOH/water (ratio 4/1), atmospheric pressure of air.

Next, we set out to test the bimetallic Pd/Au NPs of varying compositions (0.5–20 mol%). It is known from literature, that the addition of a second metal to NPs can enhance the catalytic activity [3]. We kept the total amount of metal constant to bear up the number of potential active species. The experiments were performed at room temperature in order to see the differences in conversion more clearly (Figure 4).

Figure 4

Time conversion plots of Pd NPs with different amounts of Au in the oxidation of benzyl alcohol at atmospheric pressure of air in batch at 25°C (A), and 50°C (B). Reaction conditions: 1 mmol benzyl alcohol, 25 μmol Pd NP in 10 ml iPrOH/water (ratio 4/1), atmospheric pressure of air.

From Figure 4A at 25°C, two general findings can be deduced in the range of tested parameters: NPs with an Au content of up to 3% delivered (1) higher conversions compared to pure Pd NPs, and (2) considerably higher conversions compared to NPs with 10% and 20% Au, respectively. At 50°C, the Au amount has less effect on the benzyl alcohol conversion (Figure 4B). Deviations of the corresponding conversions are within the limit of error. However, NPs with an Au amount of 0.5% tend to deliver the best conversion. This is especially true for reaction times up to 15 min. A maximum conversion (>99% and 96% selectivity) was achieved in 40 min. As already stated, all the experiments delivered selectivities >96%. A tendency of the Au amount effect on the selectivity cannot be deduced as all the selectivity values were in the error limits. It was already stated above that SAXS profiles of different sets of NPs with 20% Au showed significant deviations indicating varying NP sizes. Moreover, the TEM image revealed a larger particle size distribution, i.e., more NPs of diameters larger than 2 nm. Thus, the lower catalytic activity of 20% Au/Pd NPs could be explained assuming that larger NPs exhibit less catalytic activity. This is also reflected in the results obtained with the different sets of these NPs. The conversions varied strongly from 30% to 75% after 2 h. Figure 4A shows the average of six runs. This also true for NPs with 10% Au amount. We did not prepare as many different sets as for the 20% Au NPs but expect similar differences in size distributions.

Table 2 compares the results obtained in this work with selected results on the oxidation of benzyl alcohol to benzaldehyde reported in literature. High conversion and selectivity was achieved at comparably mild reaction conditions.

3.3 Catalytic testing in a microfluidic capillary reactor

Owing to the higher activity and selectivity of the bimetallic NPs with 0.5%–5% Au, we decided to use these in a microfluidic reactor under continuous conditions. Our aim was to exploit the provision of high interfacial areas in a gaseous/liquid system in microfluidic devices. Additionally, parameters such as temperature, residence time, or substrate ratios can be adjusted more precisely. We tested several types of reactors. Capillary reactors made of polytetrafluoroethylene (PTFE) or fluorinated ethylene propylene (FEP) turned out to be the most suitable. In glass reactors, the NPs were retained in the reactor due to interactions of the NP solution in iPrOH/water with the reactor wall, which was rather problematic.

The setup for the continuous processing is depicted in Figure 5. The NP solution and a solution of benzyl alcohol in iPrOH/H₂O (4/1) were delivered via syringe pumps and heated up to the desired temperature prior to mixing. Air was added with a mass flow controller to adjust the volumetric flow and to generate a reproducible slug flow in the capillary. The whole setup was kept in a thermostat. PTFE and FEP capillaries with the IDs of 1000 and 1600 μm were used. The mean residence time τ was measured with a stopwatch and ranged from 7.5 to 15 min.

Figure 5

Schematic setup of the continuous aerobic oxidation of benzyl alcohol.

In case of the capillaries with an ID of 1600 μm, an additional back pressure was necessary to get a steady slug flow. Therefore, the liquid was collected in a pressurized bucket at the end of the capillary. Only the air was expanded to prevent pressure fluctuations. Despite these efforts, it was hardly possible to create a stable flow regime. In contrast, back pressure regulation was not necessary to form a reproducible slug flow in a capillary with an ID of 1000 μm. Here, the pressure decrease was about 0.3 bar over the entire capillary length. This capillary was, therefore, used for all the continuous experiments discussed herein. Figure 6A summarizes the results obtained with bimetallic NPs of varying compositions under standard conditions at 50°C. The liquid phase was delivered at an overall flow rate of 0.4 ml/min (0.1 mmol/ml), and air was supplied in a nearly 1 to 1 ratio with 4.25 ml/min. At these flow rates, the mean residence time τ was determined to be 15 min.

Figure 6

Oxidation of benzyl alcohol in a PTFE capillary. (A) Influence of the Au content of the nanoparticles on the benzyl alcohol conversion. (B) Influence of the molar oxygen/alcohol ratio on the benzyl alcohol conversion for 0.5% Au/Pd NP: 1/1 (21%); 2.9/1 (50%); 4.7/1 (100%). Reaction conditions: 50°C, 20 m capillary length, 1000-μm ID, 0.4 ml/min overall liquid flow rate, 4.25 ml/min air/gas flow rate, τ=15 min.

Clearly, under the given reaction conditions (50°C, atmospheric pressure of air) NPs with an Au amount of 0.5% delivered the highest conversion (91%) and a selectivity of 97% within approximately 15 min (Figure 6A, Table 3). Doubling the amount of the catalyst caused only a slight increase in the conversion of 2%. In further experiments, we investigated the influence of varying substrate/oxygen ratios. The mixtures of air and pure oxygen were generated in order to keep the gas flow rate constant at different oxygen contents. Therefore, a second flow controller was installed, and the two gas flows were mixed before entering the reaction capillary, whereas all the other parameters were kept constant. Figure 6B shows the conversions obtained with an oxygen content of 21%, 50%, and 100%, respectively. The oxygen/substrate ratio varied from 1/1 to 4.7/1. Surprisingly, the conversion of the alcohol decreased with increasing amount of oxygen. This unexpected behavior is discussed in more detail in the next paragraph. However, to summarize and evaluate our results, they are put in context of selected results reported in literature (Table 3).

3.4 Deactivation and recyclability of the unsupported nanoparticles

So far, the experiments in flow revealed that a lower conversion of alcohol was obtained with an increasing oxygen/substrate ratio. In batch, we observed the same tendency at 50°C and varying air pressures (1 to 5 bar (a), Figure 7). Lower conversions were obtained at higher pressures. This was true for the monometallic Pd NPs as well as the bimetallic Au/Pd-NPs, which is exemplified by the 20% Au/Pd NP in Figure 7B. Generally, the higher the air pressure, the faster the maximum conversion was reached.

Figure 7

Conversion vs. time plots for the benzyl alcohol oxidation in batch at different oxygen pressures (A) with monometallic Pd NPs and (B) with 20% Au/Pd NPs. Reaction conditions: 1 mmol benzyl alcohol, 25 μmol Pd NP in 10 ml iPrOH/water (ratio 4/1), 50°C.

This indicated an inhibition of the Pd NPs by oxygen. Therefore, the Pd NPs were dissolved in the typical solvent mixture iPrOH/H₂O (4/1) and treated with 4 bar of air prior to the usual reaction protocol. After this treatment, the NPs exhibited no significant catalytic activity. Oxidation of PVP stabilized Pd NPs was already described in literature. Here, the irreversible formation of an oxide coating was reported [22]. We, therefore, did further investigation with the pure Pd NPs.

In order to have a closer look at the nearby environment of the palladium during the reaction, we applied the XANES and EXAFS measurements. X-absorption near the edge structure (XANES) is able to reveal the oxidation state of X-absorbing metal centers by determination of the energy of the onset of X-ray absorption, the so-called absorption edge [23]. In the case of palladium, the first resonances after the edge do also allow insights into the electronic structure of the X-ray-absorbing center [24]. We prepared different samples: solutions of stored and freshly prepared Pd NPs (named Pd7 and Pd13, respectively) were measured directly in ethylene glycol. For the in situ measurements, the Pd13 NP sample was utilized and transferred to iPrOH/H₂O. Owing to the analysis requirements, a solution of higher concentration was prepared [c(Pd)=12.5 mmol/ml (fivefold), c(alc)=0.2 mmol/ml (twofold)]. Figure 8

Figure 8

XANES spectra of the samples Pd7 (black), Pd13 in ethylene glycol (blue), Pd13 in isopropanol/water (green), Pd13 in isopropanol/water at 50°C (gray), Pd13 in isopropanol/water at 50°C with 3 bar synthetic air (red). The edge area and first resonances after the edge are shown enlarged.

depicts the XANES spectra of the investigated samples. The edge position and the first three resonances are shown enlarged. The absorption transition of the first and second absorption maxima are assigned to 5p- and 4f-type final states, respectively [25].

In Figure 8, it is obvious that the edge position is virtually identical for all the samples, consistent with the same oxidation state of Pd(0). Within the Pd13 samples, a trend in the first and second XANES feature is observed. Changing the solvent of the original Pd13 sample from ethylene glycol to isopropanol/water causes a slight increase in both the signals at around 24.378 and 24.4 keV, respectively. Raising the temperature to 50°C causes a further increase in both the signals, whereas the addition of synthetic air (3 bar) to the reaction mixture does not induce any further changes. Although these changes are rather small, they indicate alterations in the electronic structure, which could be due to the size effects or redox processes. In order to gain further insights in particle size and the local structure around the palladium metal centers, EXAFS analysis of the samples was carried out [23]. The results are summarized in Table 4, the according EXAFS spectra k³ (k) are given in Figure 9.

Figure 9

EXAFS spectra of the samples Pd7 (A), Pd13 in ethylene glycol (B), Pd13 in isopropanol/water (C), Pd13 in isopropanol/water at 50°C (D), Pd13 in isopropanol/water at 50°C with 3 bar synthetic air (E). The experiment is shown in black, the fit in green.

EXAFS spectroscopy allows to determine particle sizes in situ, due to the size dependency of the Pd-Pd coordination numbers [26–28]. The coordination number of the nearest neighbor shell in metal particles is nearly independent of the particle shape and follows an analytical form which is displayed in Figure 10 [27]. According to this function, a Pd-Pd coordination number of six is attributed to a cluster size of around 20 atoms. Following the work by Jentys [27], this number, in turn, can be correlated to a particle diameter of 1.5 nm if a spherical shape is assumed. Only a small variation of the Pd-Pd coordination number can be found under the described environmental and conditional variations. In fact, they are identical within the error bar. However, a trend can be observed. Changing the solvent from ethylene glycol to isopropanol/water slightly increases the Pd-Pd contribution from 6.1 to 6.5, which can be interpreted by a growth of the cluster size to around 24 atoms. Heating to 50°C causes a further increase of the coordination number to 7.1, according to a new cluster size of 32 atoms. Addition of 3 bar synthetic air leads to a slight reduction of the Pd-Pd coordination number again. It has to be mentioned that only for the latter two samples a Pd-O shell is adaptable, both with a coordination number of 0.3, i.e., only 0.3 oxygen atoms per palladium center can be rationalized by EXAFS spectroscopy. However, inclusion of this shell did not improve the quality of fit to a significant extent. Despite the missing statistical significance, this oxygen shell can explain the small changes in the XANES. However, the results clearly show that the formation of a Pd oxide shell does not occur during the oxidation reaction. The 0.3 oxygen atoms may be due to adsorption processes, which is not comparable to the irreversible formation of Pd oxide as reported by Tylus et al. [22].

Figure 10

Correlation between the particle size (number of atoms) and the nearest neighbor coordination number.

Additionally, the semi-online SAXS measurements were performed during the oxidation reaction. Figure 11A shows the set of scattering curves of samples taken directly from the reaction solution at different reaction times. The overall curve shape is similar over the entire reaction process indicating similar NP size and shape. A minor increase in intensity of the scattered X-rays may result from little evaporation of the solvent. TEM images of the Pd NPs before and after the reaction (Figure 11B and D, respectively) confirm only minor changes in NP size and shape. Before the reaction, the NPs exhibit a spherical shape with a mean diameter of around 2 nm. After the reaction, some slightly larger rods are observed and a slightly broader particle size distribution with an approximate mean diameter of 2.5 nm (compare Figure 1C and Figure 11C).

Figure 11

(A) Semi-online SAXS curves of the Pd NP during the reaction at room temperature. TEM images of the utilized Pd NPs (B) before, (D) after the oxidation reaction of benzyl alcohol, and (F) after the first reactivation. All scale bars are 5 nm. Size distribution of the Pd NPs (C) after the reaction and (E) after the first reactivation.

Figure 11F shows a TEM image of NPs, which were reactivated after one run. This required a special treatment of the reaction mixture after the first run and allowed to utilize the NPs in further runs. Therefore, a threefold excess of toluene was added to the reaction mixture. Phase separation occurred, leaving residual benzyl alcohol, product, iPrOH, and toluene in the organic phase, whereas the PVP-stabilized NPs remained in the aqueous phase. After centrifugation and decanting the organic phase, an excess of acetone was added to the aqueous phase to precipitate the NPs. The mixture was centrifuged again, the liquid was decanted, and the black precipitate was redispersed with reaction solution for the next run. Figure 11E and F clearly show that there are only marginal changes of the Pd NP after the first reactivation. This observation is also confirmed by the SAXS measurements shown in Figure 12A. A particle growth in the sub-nanometer range can be concluded from the blue curve that represents the reactivated Pd NPs, which were used for the recycling experiments.

Figure 12

(A) SAXS curves of the Pd NPs in reaction solution before the first run (black), after the first run at quantitative conversion (red), and after the first reactivation prior to the second run (blue). (B) Conversion time plots for successive runs with recycled Pd NPs. Reaction conditions: 1 mmol benzyl alcohol, 25 μmol Pd NP in 10 ml iPrOH/water (ratio 4/1), 50°C, NPs were reactivated after each run.

Figure 12B shows the conversion vs. time plot for three successive runs utilizing the reactivation procedure. Compared to the first run, only slightly lower conversions were achieved in the second run, which may result from minor agglomeration of the Pd NPs as seen in Figure 11F. In the third run, however, a significant drop in the conversion was observed. So far, three effects have to be considered for the NP deactivation: (1) deactivation by oxygen during the reaction, which is at this time attributed to the strong Pd-O₂ interactions in the adsorption processes rather than to the irreversible formation of Pd oxide species (compare the in situ EXAFS measurements). The more oxygen is provided, the faster the active Pd sites seem to be blocked for substrate adsorption and, hence, the lower the catalytic activity. This reversible short-term effect can obviously be overcome by a “washing step” for catalyst recycling. The NP deactivation is further due to (2) particle growth and/or agglomeration or aggregation, and (3) alterations in the stabilizing polymer PVP. The latter two are irreversible long-term effects leading to the decrease of the catalytic activity over several runs (compare Figure 12B). SAXS and TEM revealed only minor changes in the particle size and shape after the first run and reactivation (Figure 11). However, a considerable decrease in the NP solubility was observed in the successive runs. In the third run, the NPs were hardly soluble in the reaction medium after the reactivation procedure, which might be the reason for the lower catalytic activity. In this regard, further investigations are required and intended in order to examine the possible alterations of the stabilizing agent PVP under the applied reaction conditions. Here, solid state-NMR should give more insight in the change of the stabilizer.