Markus
Schörner
a,
Stefanie
Kämmerle
a,
Dorothea
Wisser
b,
Benjamin
Baier
a,
Martin
Hartmann
b,
Matthias
Thommes
c,
Robert
Franke
de and
Marco
Haumann
*a
aFriedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Lehrstuhl für Chemische Reaktionstechnik (CRT), Egerlandstr. 3, 91058 Erlangen, Germany. E-mail: marco.haumann@fau.de
bFriedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen Center for Interface Research and Catalysis (ECRC), Egerlandstr. 3, 91058 Erlangen, Germany
cFriedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Lehrstuhl für Thermische Verfahrenstechnik (TVT), Egerlandstr. 3, 91058 Erlangen, Germany
dEvonik Operations GmbH, Paul-Baumann-Str. 1, D-45772 Marl, Germany
eRuhr-Universität Bochum, Lehrstuhl für Theoretische Chemie, Universitätsstr. 150, D-44780 Bochum, Germany
First published on 23rd August 2022
In this work, the continuous gas-phase aldol condensation reaction of n-pentanal was investigated on silica supports. Since high aldol formation can lead to deactivation by pore blocking, the reaction is important to improve continuous gas-phase hydroformylation processes. Aldehyde conversion was monitored depending on the surface acidity. The spent catalysts were analyzed by thermogravimetric analysis (TGA) to evaluate the accumulation of substrate and product inside the pores. The pore size was altered using hydrothermal treatment. The obtained supports were analyzed using N2-sorption, mercury pore intrusion/extrusion, point-of-zero-charge measurements, temperature programmed desorption measurements (CO2 and NH3), and 29Si MAS NMR. A variation of reaction duration, pressure, and temperature was carried out. The influence of the silica texture on activity and accumulation was investigated using different particle size fractions and median pore diameters. In larger pores, the total volume-based accumulation was lower compared to the one in smaller pores. At the same time, the aldol was enriched compared to n-pentanal in the condensed liquid inside the pore network.
In order to achieve long-term stability, liquids with little to no vapor pressure must be employed. We became interested in supported ionic liquid phase (SILP) materials more than fifteen years ago for the continuous hydroformylation of small alkenes like propene and but-1-ene.7,8 In such SILP materials, the transition metal complex is dissolved in ionic liquids (see Fig. 2).9 Since these ionic liquids have an extremely low vapor pressure, they do not evaporate during continuous gas-phase processing. Several examples of SILP gas-phase catalysis have been reported over the past years.10 Most of these studies placed the focus on the proof of concept, demonstrating the performance of the SILP catalyst with respect to activity, selectivity and, to a lesser extent, stability. However, the stability is crucial when it comes to the industrial implementation and the catalyst's time on stream (TOS) should typically exceed 8000 h. Here, only few examples of longer runtimes are known in literature.11,12 We are currently developing a SLP hydroformylation process using monolithic support structures coated with a polymeric separation layer as novel membrane reactor within the course of a H2020 project.13,14
Fig. 2 Schematic illustration of pore filling by accumulation due to the aldol condensation reaction of n-pentanal to 2-propyl-2-heptenal on porous support materials. |
For all SLP catalyzed gas-phase hydroformylation reactions, the major challenge for long term stability is to avoid the consecutive aldol condensation reaction (see Scheme 1). The formed aldehydes are highly reactive and can undergo self-condensation, catalyzed by traces of acid or base, to yield products of twice the molecular weight. These compounds would then be prone to liquefy under reaction conditions inside the pores of the SLP catalyst.
As a result, one can expect growing of the liquid film over time (see Fig. 2) that inevitably leads to blocked transport pores and hence deactivated SLP catalysts.
The self-condensation of aldehydes or ketones has been widely studied in the literature. A variety of studies are conducted in the liquid phase using e.g. sulfuric acid,15 aqueous sodium hydroxide,16 functionalized ionic liquids,17 zeolites18 or oxides19,20 as catalysts. Studies investigating the gas-phase kinetics are mainly using zeolites, aluminosilicates or different metal oxides as active species.21–30 While some of these studies already involve observations of accumulation of coke,21 carbonaceous species28 or condensation products,25 the experimental temperature exceeds the typical temperature for SILP-hydroformylation of 373–393 K by far. Thus, to get a better understanding of the kinetics and the accumulation in the gas-phase aldol condensation of n-pentanal, a more detailed study is required. For SILP and other SLP materials the presence of the support renders the situation, since the surface acidity (or basicity) can catalyze the aldol condensation reaction significantly. In a previous study by Kaftan et al., the hydroformylation of but-1-ene was monitored by means of in situ DRIFTS.31 Here, the formed n-pentanal consecutively reacted to 2-propyl-2-heptenal which then accumulated in the pores (see Fig. 2).
In order to evaluate the influence of silica, the standard support materials for SILP catalysis in general and for hydroformylation in particular, we carried out experiments with different particle sizes as well as different pore diameters. For SILP catalyzed hydroformylation, we observed a pronounced transport limitation at large particle sizes, which only diminished at particle sizes smaller than 500 μm. This pore diffusion limitation could be overcome by the use of a support material with enlarged pores from a hydrothermal treatment.32 In the present work, we use this hydrothermal treatment to adjust the pore diameter of silica, while keeping the particle size and surface chemistry more or less intact.
Mercury intrusion porosimetry (MIP) was used to determine pore volume, median pore size d50, and specific surface area of macroscopic pore systems. The measurements were carried out on a PASCAL system from Thermo Scientific/Porotec (pressure range 0.1–400 MPa).
Point-of-zero-charge (PZC) analysis and temperature-programmed desorption measurements (CO2 and NH3-TPD) were conducted to reveal the surface acidity/basicity. For the PZC analysis 50 mg of support calcined support (873 K, air) were dispersed in 12 ml pH solution (range from pH 3 to 12, HNO3 used for acidic solution, NaOH for basic solution). The dispersions were shaken at room temperature for 20 h before measuring the pH value. The measured values were plotted against the pH value of the pure solution. The PZC-value was determined at the intersection of the measured values and the bisector. For the TPD measurements, 0.4 g of sample was used. The measurements were carried out on a Thermo Scientific TPDRO 1100 instrument equipped with a TCD. The sample was calcined at 873 K in air and pretreated at 773 K for 30 min in helium before measuring. Ammonia loading was carried out at 393 K for 60 min, while CO2 loading was done at 443 K for 60 min. For the loading pure ammonia and CO2 was used. After that, the gas phase was flushed with helium for 120 min at 313 K. The analysis was carried out with a heating ramp of 10 K min−1 up to 1073 K in helium.
Magic angle spinning (MAS) NMR measurements were conducted on an Agilent 500 WB spectrometer (11.7 T) (resonance frequencies: 499.9 for 1H, 99.3 MHz for 29Si, MAS rate: 15 kHz, rotor: 3.2 mm zirconia, room temperature). The rotors were loaded in the Glovebox in argon atmosphere. The measurements were carried out as described in an earlier work.32
A mass loss Δmi of each stage i (i = I–III) was calculated from the data. Since the density of the herein used hydrocarbons are expected to be in the same range, the density of n-pentanal was used to calculate the mass loss into a volume according to eqn (1).
(1) |
(2) |
(3) |
(4) |
(5) |
The isotherms were analyzed using nonlocal density functional theory (NLDFT) that assume cylindrical silica pores. To obtain the pore volume and pore size distribution, the adsorption branch was used by a dedicated NLDFT kernel that takes the delay caused by metastable pore fluid quantitatively into account.36,37 In contrast to the samples treated at 393 K, the one treated at 453 K did not show a plateau close to the saturation pressure p0. This indicated pores in the macroscopic range which cannot be measured with the N2-sorption. Therefore, this sample was analyzed using mercury intrusion porosimetry (see Fig. S6 in ESI†) which revealed pores up to 120 nm. The resulting sum function graphs are summarized in Fig. 5. An increase in pore size with increasing treatment time or temperature was clearly visible. Based on these hydrothermal treatments the median value of the sum function distribution Q3(d50) was used as descriptor for the following catalytic studies. The median pore size d50 is defined as the pore size where the sum function Q3 is equal to 0.5. Here 50% of the pore volume is spread across pore sizes below the value d50 and the other 50% is spread across pore sizes larger than d50.
An asymptotic growth of the median diameter d50 from 7 to 26 nm was obtained when hydrothermal treatment at 393 K was carried out for up to 200 h (Fig. 6a). This asymptotic growth was already known for similar treatments32,33 and is in good agreement with the model from Leboda et al.34 This model assumes that the silica skeleton consists of globules with different sizes. Under hydrothermal treatment conditions, the smaller globules fuse to the larger ones according to Ostwald ripening. Depending on the temperature, the skeleton homogenizes at a different maximum size of the globules, since more energy and hence higher temperatures are required to fuse bigger globules. Comparing different treatment temperatures for 24 h, an exponential increase of the median pore size is achieved from 7 to 54 nm for a treatment at 453 K (Fig. 6b) as it would be expected from the temperature dependency of Ostwald ripening.38
An overview of all materials is given in Table 1. No change in pore volume, size or surface area was achieved by sole grinding of the particles (see entries 1–5). The corresponding N2-sorption isotherms are shown in the ESI† (see Fig. S5). When comparing the surface area of samples hydrothermally treated at a constant temperature for different times (entries 1, 6–7), a decrease in surface area was observed with increasing pore size while the pore volume only slightly changed. A similar behavior was found when comparing samples that were hydrothermally treated at different temperatures for the same time (entries 1, 11, 12). Point-of-zero-charge measurements for treated and untreated samples (entries 1–5, 12) gave similar values in the neutral range, as previously published.32
Entry | d particle | d particle,av | t HT | T HT | d 50 | V pore | A | PZC |
---|---|---|---|---|---|---|---|---|
μm | μm | h | K | nm | cm3 g−1 | m2 g−1 | — | |
d particle = particle range from sieved fraction, dparticle,av = averaged particle diameter of sieved fraction, tHT = time of hydrothermal treatment, THT = temperature of hydrothermal treatment, d50 = median pore diameter, Vpore = pore volume, PZC = point of zero charge. a Analyzed with N2-sorption. b Analyzed with mercury intrusion porosimetry. | ||||||||
1 | 63–200 | 132 | — | — | 7a | 0.811a | 503a | 7.1 |
2 | 200–450 | 330 | — | — | 7a | 0.817a | 512a | 7.1 |
3 | 450–630 | 540 | — | — | 7a | 0.820a | 514a | 7.2 |
4 | 630–800 | 720 | — | — | 7a | 0.835a | 519a | 7.1 |
5 | 800–1000 | 900 | — | — | 7a | 0.769a | 523a | 7.1 |
6 | 63–200 | 132 | 9 | 393 | 9a | 0.739a | 320a | — |
7 | 63–200 | 132 | 16 | 393 | 12a | 0.738a | 252a | — |
8 | 63–200 | 132 | 24 | 393 | 16a | 0.720a | 180a | — |
9 | 63–200 | 132 | 100 | 393 | 23a | 0.712a | 120a | — |
10 | 63–200 | 132 | 200 | 393 | 26a | 0.692a | 82.2a | — |
11 | 63–200 | 132 | 24 | 423 | 21b | 0.639b | 136.3b | — |
12 | 63–200 | 132 | 24 | 453 | 54b | 0.760b | 77.1b | 7.2 |
We carried out 29Si cross polarization (CP) MAS NMR studies for the untreated material as well as materials hydrothermally treated at 393 K for 24 and 100 h (see Fig. 7a). Cross polarization probes predominantly the surface of the material, where Q3 groups, indicative of one OH-group per silica, and to a smaller extent also Q2 (two OH groups per silica) and Q4 groups (zero OH groups per silica) were detected. Compared to the untreated sample, a lower concentration of Q2 groups was observed in the samples that underwent hydrothermal treatment. This is to be expected since Ostwald ripening will induce condensation of silanol groups. In all samples, the calcination step after the hydrothermal treatment dominates the final degree of hydroxylation.39,40 This step reduces the signal intensity and hence the possible changes that occur during the hydrothermal treatment, allowing only a qualitative analysis. For a quantitative analysis, NH3-TPD measurements were carried out (Fig. 7b). Again, untreated silica samples as well as samples treated at 393 K for 24 and 100 h were compared. No peaks above 673 K were detected for all samples, thus we assume the absence of strong acid sites on the silica support.41–45 The results in the low temperature range can be explained by weakly bonded ammonia on silanol groups, which is in good agreement also with the 29Si MAS NMR results.42,45–47 The plots were fitted by Gaussian peaks according to Katada et al.42 resulting in two overlapping peaks around 415 K and 525 K.
The calculated amount of acid sites for these two peaks is given in Table 2 for all samples. While the first peak decreased, the second peak increased with increasing treatment time. The total mass-based concentration remained constant. Since the hydrothermal treatment decreased the surface area of the samples, the surface-based concentration of acid sites increased from 0.05 to 0.24 μmol m−2 (see entries 1 to 3). The CO2-TPD measurement only showed a very weak signal at 450 K for all three samples without a visible trend with temperature (see Fig. S19 ESI†). The results obtained by TPD measurements were in agreement with the less sensitive point-of-zero-charge measurement which also indicated an almost neutral surface as both the acid and basic sites were only present at low concentrations.
Entry | t HT | Weak acidic sitesb | Weak basic sitesc | |||
---|---|---|---|---|---|---|
T peak ≤ 415 K | T peak ≤ 525 K | Total | Total | Total | ||
h | μmol g−1 | μmol g−1 | μmol g−1 | μmol m−2 | μmol g−1 | |
Sample treated for 0 h represents untreated material. All samples were calcined at 873 K before analysis. a Time of hydrothermal treatment at 393 K. b Analyzed with NH3-TPD. c Analyzed with CO2-TPD. | ||||||
1 | 0 | 16 | 8 | 25 | 0.05 | ≤2 |
2 | 24 | 7 | 18 | 25 | 0.14 | ≤2 |
3 | 100 | 6 | 23 | 29 | 0.24 | ≤1 |
T/K | 353 | 373 | 393 | 413 | 433 | 453 | 473 | 493 |
X eq/% | 88.9 | 87.2 | 85.6 | 83.9 | 82.3 | 80.7 | 79.2 | 77.8 |
In a first series of experiments, the time on stream (TOS) behavior of the reaction was monitored. Experiments were carried out for 1, 12, 24, 60, 100 and 130 h, during which the effluent gases were monitored with an on-line GC. After the experimental run, the samples were analyzed ex situ via thermogravimetric analysis (TGA). The results are shown in Fig. 8, where the bottom half shows the pore filling degree of hydrocarbons, calculated from eqn (3), and the top half shows the conversion of n-pentanal over time (from GC data). After an induction period the conversion started to decline over time, while the total pore filling degree showed the opposite trend and continuously increased from 0.18 mlhydrocarbons mlpore−1 after 1 h to 0.44 mlhydrocarbons mlpore−1 after 130 h. This increase can be attributed to accumulation of 2-propyl-2-heptenal in the pore system, as shown by TGA of post-run samples. This pore filling increased from 0.07 mlhydrocarbons mlpore−1 after 1 h to 0.30 mlhydrocarbons mlpore−1 after 130 h. The amount of higher hydrocarbons (not shown, 0.6 mlhydrocarbons mlpore−1) and n-pentanal (0.8 mlhydrocarbons mlpore−1) on the other hand remained at a constant level over the observed time. A similar trend was observed by Kaftan et al. using operando DRIFTS (diffuse reflectance IR spectroscopy) measurements in the Rh-catalyzed gas-phase hydroformylation of but-1-ene on silica support.31 The high accumulation after only 1 h TOS can be explained by an adsorber-like behavior right after starting the experiment. A decrease to almost zero of the n-pentanal signal is detected in the online GC for the first 0.25 h. This observation suggests, that all substrate within this time is accumulated in the pores as this behavior was not observed in an empty reactor (see ESI† Fig. S21). A similar stepwise accumulation was observed by Rekoske et al. after exposure of TiO2 with acetaldehyde.49 In contrary to the accumulation, the conversion decreased after an induction period of 5 h continuously from 3 to 2%. A similar deactivation pattern was observed in the gas-phase aldol condensation of acetone on zeolites by Herrmann et al. as a result of the blockage of active sites by reaction compounds.22,23
Increasing the reaction temperature from 373 to 461 K led to a significant increase in conversion from 3.1 to 22.6% after 5 h. However, at higher temperatures a more pronounced deactivation was observed (see Fig. 9a). A reverse trend was observed regarding the post-run pore filling degree. Here, a decrease of total pore filling was measured up to 437 K, which leveled off above that temperature (see Fig. 9b). While higher hydrocarbons remained more or less unaffected by temperature at a value around 0.05 mlhydrocarbons mlpore−1 (not shown), accumulation of n-pentanal (decrease from 0.11 mlhydrocarbons mlpore−1 at 373 K to 0.04 mlhydrocarbons mlpore−1 above 417 K) and 2-propyl-2-heptenal (decrease from 0.16 to 0.10 mlhydrocarbons mlpore−1 above 437 K) became less prominent above 437 K. The fraction of the low-boiling substrate in the accumulating species decreased from 34 to 22% with increasing temperature. Given the large difference in boiling point of 2 and 5 (376 K to 482 K), this result can be expected.
An approach developed by Wolf et al. was used to account for the deactivation and to calculate the effective reaction rates reff.50 The deactivation rate constant kd(T) was expressed as a function of a reference temperature T0 according to eqn (6) and (7). As shown later, no dependency of substrate or product partial pressure on the deactivation was observed (see Fig. S23 ESI†). Hence it is not included in eqn (6).
reff(t) = r0exp−kd(T)t | (6) |
(7) |
The deactivation could be sufficiently fitted with a value for kd,0 (423 K) of 0.0167 h−1 and a temperature dependency of EA,kd of 19.7 kJ mol−1. The low temperature dependency of the deactivation suggests that no bond formation or cleavage is present in the deactivation mechanism but rather blocking of active sites by physisorbed compounds.
Next, a partial pressure variation of the substrate n-pentanal was carried out. Between 5 and 50 kPa, the effective reaction rate increased linearly (see Fig. S23†) from 0.13 × 10−3 to 1.1 × 10−3 mmol h−1 m−2. In this work an effective reaction order of 0.9 for n-pentanal was obtained (see Fig. 11a). A similar value was measured for n-butanal by Shylesh et al. on a silica-supported amine catalyst in the gas-phase.27 Herrmann and Iglesia showed, that the theoretical reaction order present at saturation coverage is 1 for the heterogeneous acid-catalyzed aldol condensation of acetone.22 Regarding the measured accumulation, saturation coverage can be also assumed in the present work. No dependency of the deactivation on the substrate or product partial pressure was observed in our study. After 24 h time on stream the runs were deliberately stopped and the catalysts were again analyzed by TGA (see Fig. 11b). A minor increase in pore filling degree due to higher 2-propyl-2-heptenal accumulation was detected up to 20 kPa. Increasing the partial pressure to 30 kPa resulted in a steep increase in n-pentanal condensation from approx. 0.1 to 0.56 mlhydrocarbons mlpore−1. As a result, the total pore filling degree almost reached 1, indicating completely filled pores. A further increase of the pressure did not have significant effect on the composition of accumulated species until a partial pressure of 70 kPa was used. Here, a significant change in the composition was found with 80% n-pentanal (from 50%) and 14% 2-propyl-2-heptenal (from 33%). At the same time, the measurement in the GC started to fluctuate strongly (see ESI† Fig. S24), which made a proper evaluation impossible. Using the Kelvin-equation for cylindrical pores, a value of p = 66 kPa was calculated for pore condensation of pure n-pentanal at 373 K in 7 nm pores (for calculation details see ESI†). The presence of pore condensation would lead to the increased amount of n-pentanal in the pores as well as the fluctuating GC-values. Although this calculated pressure for pore condensation is in good agreement with the observations at 70 kPa, it cannot serve as good explanation for the steep increase of n-pentanal condensation at 30 kPa.
In a next set of experiments the support characteristics, e.g. the influence of the particle and the pore size, were investigated. Table 4 summarizes the results for different particle size fractions at 393 K. No clear trend can be found with respect to the initial reaction rate or the final conversion level. This again hints for the absence of internal diffusion limitations, hence all n-pentanal can reach the active sites faster than it is converted at these sites. Comparing the total pore filling degrees after a reaction time of 24 h revealed no influence of the total filling degrees with particle size, which ranged between 0.33 and 0.37. However, 2-propyl-2-heptenal accumulated stronger while the n-pentanal fraction declined with increasing particle size. It can be assumed that the longer diffusion times inside the larger supports allowed more aldehyde to be converted into aldols. In addition, these aldols now have a smaller effective diffusion coefficient and a higher tendency for pore condensation.
Entry | d particle | r 0 | α tot | x n-pentanal | x aldol | x higher hydrocarbons |
---|---|---|---|---|---|---|
μm | 10−3 mmol h−1 m−2 | ml mlpore−1 | % | % | % | |
d particle = particle range from sieved fraction, r0 = initial reaction rate, αtot = total pore filling degree, xi = composition of liquid phase in pores. | ||||||
1 | 63–200 | 0.21 | 0.33 | 34 | 48 | 18 |
2 | 200–450 | 0.13 | 0.35 | 28 | 54 | 18 |
3 | 450–630 | 0.14 | 0.35 | 24 | 58 | 18 |
4 | 630–800 | 0.20 | 0.34 | 23 | 60 | 17 |
5 | 800–1000 | 0.15 | 0.37 | 24 | 58 | 18 |
To evaluate the condensation behavior in more detail, we varied the median pore diameter for the smallest particle fraction. The measured pore filling degrees (total, n-pentanal, 2-propyl-2-heptenal) are shown in Fig. 12. Next to the measured volume-based pore filling degrees, Fig. 12a shows the decline in surface area with increasing median pore size as discussed earlier (see Table 1). All filling degrees declined exponentially with increasing pore size. This result can be expected assuming reduced condensation inside wider pores (see Kelvin equation in ESI†). When relating the pore filling to the surface area (in mlhydrocarbons Apore−1) and plotting it over the median pore size (see Fig. 12b), a more or less constant overall filling degree was obtained independent of the median pore size. Interestingly, larger pores led to an increased fraction of high-boilers over n-pentanal in the liquid phase. The fraction of 2-propyl-2-heptenal increased from 48% at 7 nm pores to 68% at 54 nm, while the opposite trend was visible for the low boiling n-pentanal (from 34 to 10%). The latter will have less tendency to condensate in wider pores while 2-propyl-2-heptenal is formed to a larger extent given the increased acidic site density (see Table 2).
The activity data for different median pore sizes is summarized in Table 5. The measured conversion over time and the calculated reff over time graphs are given in the ESI† in Fig. S26 and S27. Both, the initial reaction rate (r0) and the deactivation rate constant (kd) scaled linearly with increasing median pore size. According to Knudsen diffusion, the mass transport is proportional to the pore size (DiK ∼ d). As mass transport limitation could be excluded by the temperature variation for T < 403 K (see Fig. 10) and the particle size variation for particles up to 1 mm (see Table 4), a strong influence of the enhanced mass transport is unlikely. The reason for the higher activity most probably stems from the increased density of acidic sites and not from the improved internal diffusion in the enlarged pores. This assumption is supported by similar findings from Zhao et al. and Ordomsky et al. in related studies.19,24 As the concentration of active sites increased, blocking of such sites seemed more likely. As a result, a faster deactivation was observed. In conclusion, the pore size does not affect the activity or the deactivation significantly in the present study.
In a last set of experiments, the accumulation inside the macro-porous support was observed for up to 100 h TOS and compared to the meso-porous one. Inside the meso-pores a significantly higher amount of hydrocarbons accumulated after 24 h compare 0.33 to 0.05 mlhydrocarbons mlpore−1. Within the error margins, no further change was observed for both the macro-porous silica even at long time on streams up to 100 h (Fig. 13).
In addition, the composition of the liquid phase also remained constant for the macro-porous silica (10% n-pentanal, 70% 2-propyl-2-heptenal, rest higher hydrocarbons). It can be assumed, that the accumulation in larger pores also follows an asymptotic pattern as it was shown in Fig. 8 for a meso-porous support. Depending on the pore size, the maximum value of pore filling as well as the time to reach this value decrease with increasing pore size. Hence, larger pores should have a positive effect on the long-term stability in supported liquid phase catalysis.
Here, an effective activation energy of 43.4 kJ mol−1 was found, which is in good agreement with literature data for liquid-phase condensation reactions. At temperatures higher than 413 K, this apparent activation energy dropped to 22 kJ mol−1, indicating pronounced influence of internal diffusion now. An effective first order dependency was measured in a partial pressure variation of n-pentanal from 5 to 50 kPa. Interestingly, when increasing the pressure from 20 to 30 kPa, a drastic increase of accumulated n-pentanal occurred, which resulted in completely filled pores. Between 50 to 70 kPa, further n-pentanal accumulation occurred, while at the same time the aldol accumulation declined. This result is in line with the estimated pressure for pentanal condensation based on the Kelvin equation. However, liquefied pentanal should have a higher tendency for aldol condensation. It seems that the blocking of acidic site access is a key parameter for the condensation behavior. This was also exemplified by the stronger deactivation in large pore systems, which had a higher density of acidic surface sites. Increasing the pore size from 7 to 54 nm did not affect the activity or deactivation significantly, but the volumetric pore filling degree decreased while the surface area-based accumulation remained constant. Nonetheless, an enrichment of the high-boiling aldol compared to n-pentanal was measured with larger pores. The same trend was observed when increasing the particle size from 200 μm to 1000 μm. A variation of the reaction time with the macro-porous support showed that no change in accumulation was measured between 24 and 100 h TOS. This behavior differed significantly from the meso-porous support and therefore a change to larger pores in supported liquid phase catalysis should lead to an improved long-term stability as pore flooding can be suppressed.
Footnote |
† Electronic supplementary information (ESI) available: Hydrothermal treatment procedure, reactor details, detailed texture data, detailed activity data, TGA data. See DOI: https://doi.org/10.1039/d2re00143h |
This journal is © The Royal Society of Chemistry 2022 |