1 Introduction

In the wake of the biodiesel boom, there is a large interest in the side-product glycerol as a starting point for fuel and chemical synthesis [1, 2]. The lowest cost quality of glycerol is the so-called crude grade, which is obtained via esterification of triglycerides with methanol, resulting in fatty acid methyl esters and glycerol. The crude grade will contain not only the glycerol, but also the sodium hydroxide used as a catalyst, other salts, some water as well as methanol, mono-, di-, and triglycerides and matter organic non-glycerol (MONG) in different proportions depending on source [2]. In synthesising chemicals and fuels, there is a desire to keep cost down, and at the same time meet product specifications. There is quite some clean-up required before using the glycerol in the production and the state-of-aggregation in the transformation as well as the target product to a large degree determine the strategy for upgrading. One avenue pursued for transforming glycerol is the gas-phase dehydration into acrolein, which has several potential applications. The molecule is already used as an intermediate today, based on propene oxidation traditionally over copper catalysts (supported or non-supported), and due to its two double bonds can react in a multitude of ways [3]. Possible products include allyl alcohol (1-propenol), pyridine, 3-picoline, acrylonitrile and acrylic acid.

The catalysts used for the dehydration reaction are mainly acidic and can be characterised according to the material used. In a previous paper by the authors [4] as well as a few available reviews [5,6,7] gives a good overview of the subject; the classes of catalysts used include: heteropoly acids and other supported inorganic acids [8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26], zeolites [21, 27,28,29,30,31,32,33], mixed oxides [34,35,36,37,38,39,40,41], and sulphated zirconia [42, 43], all of which are solid acid materials. The acids usually contain or are promoted with P, W, S, Mo, and B, while ZrO2, Al2O3, SiO2, and TiO2 are supports of choice for heteropoly acid and mixed oxides. Glycerol conversion is normally high (80–100%) with yields maintained in the 60–80% range at high conversion rates [40]. The reaction is normally performed at moderate pressure and at medium to low temperature. Catalyst stability is an issue and carbon formation limits the time-on-stream.

There are several papers that investigate the deactivation of these kinds of catalysts under this reaction regime and carbon formation appears to be the general consensus. The carbon formation is suggested to depend on oligomer formation with hydroxyacetone as a precursor [31, 35]; indeed recycling of hydroxyacetone very much favours carbon formation [4]. Deleplanque et al. [35] imply that formation of olefins (yielding subsequent aromatics, oligomers and coke) is a possible parallel deactivation mechanism and in a separate paper Suprun et al. derived the deactivation to Brønstead acid sites [15]. The same researchers also found that the higher the temperature the lower the hydrogen-to-carbon quote was in the resulting deposits. Further, aside from the temperature, there is a marked lower carbon formation rate with increased water and oxygen content in the feed [26]. As a curiosity, it has also recently been reported that the formation of carbon on the catalyst increase the selectivity of WO3/TiO2 and heteropoly acid catalyst to acrolein [24, 25].

In previous research, the potential for the pore condensation of glycerol due to physical reasons was proposed as a means of increased deactivation rate of catalysts for glycerol dehydration [4]. This behaviour, higher carbon formation with smaller pores, has been reported in literature [33]. In this paper, this insight is used to thermally modify catalysts and thereby improving their lifetime and yield of acrolein. The morphological changes of the materials are reported as well as the initial activity and selectivity of the modified catalysts. Selected catalysts were also lifetime tested over many 100 s of hours. The target of the catalysts pre-treatment is to achieve a larger pore diameter to avoid pore condensation.

2 Experimental

This section describes the catalyst-testing procedure, along with the methods used for catalyst characterisation.

2.1 Catalyst Testing

The catalysts were investigated in two different test set-ups. The first one was used for the activity and selectivity measurements and the second one for the long-term deactivation experiments. In the activity and selectivity experimental setup, a carrier gas (hydrogen) was mixed with liquid 20% glycerol (99.5% Karlshamn Tefac AB) by weight in water before being vaporised in a pre-heater. The vaporised liquid/gas mixture was then passed through a stainless steel reactor with an inner diameter of 15 mm and an inlet temperature of 553 K, on to a condenser and gas separation step, and finally collected in a vessel. The stream of carrier gas (hydrogen) was regulated by a mass-flow controller to 1.67 cm3 s−1 and the stream of liquid to 0.005 g s−1 by a high-pressure liquid chromatography pump, resulting in a gas hourly space velocity of 500 h−1. The gas and liquid were mixed in a manifold before entering the pre-heater. The reactor and pre-heater were situated in two tube furnaces, controlled individually so that the inlet temperature of the catalyst bed was the preferred one. The catalyst bed temperature was monitored using two thermocouples, situated at the inlet and approximately 10 mm before the end of the catalyst bed. The reactor outlet pressure was set to 0.5 MPa absolute. The long-term testing rig was similar to the one described above and used the same conditions with respect to catalyst amount, flowrate and composition. The pressure drop over the reactor was monitored continuously. The reactor inner diameter was 17 mm for both systems.

A gas and a liquid sample were taken from time to time over the entire test period. The gas was analysed using a Varian CP-4900 two-channel micro gas chromatograph (equipped with one Mol Sieve 5A PLOT column and one PoraPlot column, with double TCD detectors). The liquid sample was analysed using a Varian CP-3800 gas chromatograph equipped with a 1177 split/splitless injector, a CP-wax 58/FFAP column, and a flame ionisation detector.

2.2 Material Preparation

The catalyst used was a commercially available catalyst that is 10% WO3 by weight supported on ZrO2 in the form of beads 1–2 mm in diameter (Daiichi KKK), pre-treated in different ways. The pretreatment performed was a heat treatment at different temperatures (500, 600, 700, 800, 850, 900 and 1000 °C) for 4 h.

2.3 Catalyst Characterisation

Several catalyst characterisation methods have been used to understand the changes in the catalyst morphology. These methods, presented here, are surface area measurement using nitrogen adsorption, mercury intrusion porosimetry, Raman spectroscopy as well as temperature programmed ammonia desorption. All catalysts were analysed for changes in their BET surface areas measured by the adsorption of nitrogen at liquid nitrogen temperature using a Micromeritics ASAP 2,400 instrument after degassing for 16 h at 473 K. Pore volume analysis was performed using the BJH method [44]. For calculating the adsorbed layer thickness (t value), The Halsey formula [45] was used. In the analysis, the desorption-isotherm was used.

Fourier transform (FT)-Raman spectra were recorded on a Bruker IFS 66 FT-IR spectrometer fitted with the Bruker FRA 106 FT-Raman add-on. A low-power, diode-pumped Nd:YAG laser, model DPY 301 (1064 nm) from Adlas, and a liquid-nitrogen-cooled, high-sensitivity model 403 germanium diode detector from Applied Detector Corp. were used. In the analysis, the laser power was 100 mW and the resolution was 4 cm−1. Every spectrum reported is an average of 800 scans. Mercury intrusion porosimetry (MIP) was performed using a Micromeretics AutoPore IV instrument at a maximum pressure of 414 MPa, using an assumed 130 degrees in contact or wetting angle, a surface tension of 485 mN m−1 and a density of 13 541.3 kg m−3 for the mercury.

The acidity of the catalyst was measured for the investigated catalysts using ammonia-TPD on a Micromeretics 2900. The catalyst sample was placed in a glass tube and degassed at 673 K for 1 h in helium atmosphere. Post degassing, the temperature was lowered to 323 K and the catalysts were saturated with ammonia using a gas mixture of 5% NH3 in He. Physisorbed ammonia was desorbed at 323 K by switching back to He before initiating the temperature ramp. The temperature ramp was set to 973 K with 10°/min ramp-rate.

3 Results

This section presents the results of the catalyst characterisation, along with the results of the catalyst testing. As a point of reference, to determine whether pore condensation are in place during the used test conditions, the Kelvin equation has been used (Eq. 1) in which R is the gas law constant, T the absolute temperature, P 0 the bulk saturated vapour pressure, P C the pressure at which pore condensation occurs, σ VL the gas–liquid surface tension derived from Gallant [46], θ the contact angle (12° used in the assessment) [47], n L and n G the bulk molar densities of the gas and liquid, respectively, and H the pore diameter. Using the equation with these assumptions for the testing conditions described above gives pore condensation in pores smaller than 45 Å (4.5 nm).

$$RTln\left( {\frac{{{P_C}}}{{{P_0}}}} \right)=\frac{{ - 2{\sigma _{VL}}\cos \theta }}{{\left( {{n_L} - {n_G}} \right)H}}$$
(1)

3.1 Characterisation

The aim of the thermal pre-treatment is to eliminate the pores below 45 Å, to avoid any pore condensation. To investigate the changes to the pore system, MIP was used. The results of these analyses may be viewed in Fig. 1. It is evident that during the first thermal treatments (400–800 °C) there is not much change to the catalyst. At 850 °C there is some change initiated, which is further continued with the treatment at 900 and 1000 °C. At 850 °C, the first peak in intrusion volume (small diameter pores) is moved upwards from the 100–120 Å range to 150–160 Å. Continued heat treatment at 900 °C shifts this peak to the 360–370 Å range and at 1000 °C to the 2000 Å range. For the 850 and 900 °C samples, there is also an increase in pore volume in the 500–700 Å range; for the 900 °C sample, there is also a significant shift upwards in pore size for this second peak compared to the fresh and low temperature treated samples. In the case of the 1000 °C, there are no other pores than the ones in the 2000 Å range.

Fig. 1
figure 1

The incremental pore volume of the heat-treated catalysts

Full size image

The behaviour with a shift in the pore volume above 800 °C is confirmed by the BET analysis, Fig. 2. There is not much change in the first number of samples (400–700 °C), whilst there is a slight decrease in surface area (11%) at 800 °C. Thereafter the change in surface area is more pronounced, at 850 °C the decrease is almost 40%, at 900 °C it is 66% and at 1000 °C heat pre-treatment it is 93%.

Fig. 2
figure 2

The results from BET and TPR analysis

Full size image

It is interesting to note that even though there is little change to the catalyst BET surface area for the first four catalyst samples, there are other changes. As evident from Fig. 2 the overall acidity increases with heat treatment up to the 600 °C sample, thereafter the specific acidity (mmol/g) as measured by ammonia adsorption is decreasing rather linearly with decreasing surface area.

In Fig. 3, the specific acid density (mmol/m2) and desorption maxima are reported for the catalysts investigated. It can be seen that initially, the specific acidity increases, but it then more or less levels out at about 2.05 × 10−6 mmol of NH3 per square meter of catalyst surface area. The final value for the 1000 °C sample is for obvious reasons less precise (lower values, higher uncertainty). At the same time, the temperature for maximum ammonia desorption is shifted downwards in temperature for the last three samples, above 800 °C. This implies that even though the specific acidity is constant or indeed slightly increasing for the samples above 800 °C, the acid strength of these sites decrease.

Fig. 3
figure 3

Specific acid density (mmol/m2) of the catalyst and desorption maxima for ammonia

Full size image

As may be viewed in Fig. 4, the Raman spectrum for the fresh catalyst and for the catalysts heat treated from 400 to 850 °C catalyst is consistent with that of monoclinic zirconia with strong intensity bands at 180, 188, 222, 305, 333, 345, 378, 475, 534, 569, 618, and 634 cm−1 [48]. The small shoulder at 850 cm−1 may be attributed to the W–O–W bond and the unresolved band at 900–990 cm−1 indicates W=O terminal bonds [8]. The presence of intensity bands at about 810 and 720 cm−1 are consistent with the formation of triclinic WO3 (tetragonal) for the catalyst samples heat-treated above 850 °C [49].

Fig. 4
figure 4

Raman-spectra of selected dehydration catalyst samples

Full size image

3.2 Activity Measurements

The sintered catalysts were evaluated by activity tests in a micro-reactor, as a comparison the result from the fresh catalyst (not sintered), during a 20 h run, is given in Fig. 5.

Fig. 5
figure 5

Fresh non-sintered dehydration catalyst. Run at 553 K, 0.5 MPa and GHSV 500 h−1

Full size image

Figures 6, 7 and 8 shows the results from the runs with the prepared catalysts sintered at 850, 900 and 1000 °C. The effect on the pore size distribution of the treatment can be seen in Fig. 1 above.

Fig. 6
figure 6

Catalyst sintered at 850 °C. Run at 553 K, 0.5 MPa and GHSV 500 h−1

Full size image
Fig. 7
figure 7

Catalyst sintered at 900 °C. Run at 553 K, 0.5 MPa and GHSV 500 h−1

Full size image
Fig. 8
figure 8

Catalyst sintered 1000 °C. Run at 553 K, 0.5 MPa and GHSV 500 h−1

Full size image

To evaluate the results from the treated samples, some statistical evaluations have been done. Below in Fig. 9, the measured yields of acrolein as function of time on stream for the four samples, fresh and sinterd at 850, 900 and 1000 °C, are shown. The data are fitted with linear regression.

Fig. 9
figure 9

Linear regressions to the yield of acrolein as function of time

Full size image

Straight line equation \({y}=k\, \times \,x+a\), where k is the slope and a is the intercept.

The linear fits to the yield of acrolein are shown in Fig. 9. The calculated regression coefficients R were tested for validity on the 95% level according to Eq. 2.

$$t=R\sqrt {\frac{{n - 2}}{{1 - {R^2}}}} .$$
(2)

With the null hypothesis H0: No correlation exists, Table 1.

Table 1 Test of correlation
Full size table

In Table 1, one can compare the test quantity t and the t-table value. In the case of the fresh sample (non-sintered), the test quantity t is smaller than the t-table value. This means that there is no significant correlation on the 95% level. In this case it indicates that the acrolein yield is a constant, it does not vary. The mean value is then 88.37 ± 0.41% for 95% confidence interval (CI).

In all other cases (the sintered samples) the test quantity is greater than the table value. This means that the null hypothesis “no correlation exists” can be discarded. Correlation exists in those cases, Table 2.

Table 2 CI of the fitted slopes
Full size table

The decreases observed are probably not linear with time over a larger timespan. Catalysts usually have an initial deactivation phase and then the rate of deactivation levels off and the activity becomes constant. Therefore long-term tests were performed with selected catalysts. The catalyst sintered at 850 °C, were evaluated in a long-term activity test and compared to the fresh sample. In Fig. 10, the gas composition during 650 h is shown and it is obvious that the activity and selectivity are constant during the 650 h run.

Fig. 10
figure 10

Long term testing 20 wt% glycerol to acrolein catalyst heat-treated at 850 °C. Run at 553 K, 0.5 MPa and GHSV 500 h−1

Full size image

In Fig. 11, the result from the sintered catalyst at 850 °C and for the un-sintered catalyst is shown. Not surprisingly, there is no correlation of yield to time in the statistic test of the regression coefficient. Meaning the measured values are constant over the timespan. The mean values are 75.0 ± 0.7% for the un-sintered catalyst and 87.6 ± 0.9% for the catalyst sintered at 850 °C. With those mean values and CI it is obvious that the values significantly differ.

Fig. 11
figure 11

Long term testing comparison of un-sintered catalyst and catalyst sintered at 850 °C. Run at 553 K, 0.5 MPa and GHSV 500 h−1

Full size image

The limiting factor for the reaction is the coke formation in the reactor. This is manifested as an increased pressure drop over the reactor. In Fig. 12, the measured pressure drop (dP) over the reactor during the long-term test is shown.

Fig. 12
figure 12

Pressure drop as function of time for the heat-treated (850 °C) and fresh catalyst

Full size image

As a comparison, the recorded pressure drop from the long-term test with the fresh (unsintred) catalyst (from Fig. 12) is shown in the same figure. It takes longer time for the sintered catalyst to reach the same pressure drop as the unsintered catalyst. This means that the time on stream before regeneration has been prolonged for the treated catalyst. This is probably due to the higher selectivity to acrolein obtained by the sintering. If the selectivity towards acrolein is improved, the formation of by-products is decreased. This leads to a longer availability on stream.

4 Discussion

The goal of this work has been to remove pores with diameter <45 Å in an attempt to improve the performance in the dehydration reaction of glycerol into acrolein. It has been shown in this work, that it is possible to change the pore structure of the WO3/ZrO2 catalyst by thermal sintering in a controlled way (Fig. 1), without changing the morphology of the catalyst in a radical way. Up to a temperature of 850 °C, the WO3 phase remains in the monoclinic phase, but at 900 °C it changes into the triclinic phase (Fig. 4). The specific amount of acidity (mmol/g), measured by NH3 adsorption, decreases when the BET-surface of the catalyst decreases (Fig. 4). However, per m2 of the catalyst, the acidity of the thermally treated catalyst increases compared with the fresh one. Above 850 °C, the acidic strength, measured as peak desorption temperature of ammonia, starts to decrease (Fig. 3).

After the thermal treatment, the initial yields (up to 20 h) differs between the samples 900 °C > 850 °C > Fresh > 1000 °C. The reason for this is an improved selectivity for acrolein and deceased selectivity for mainly hydroxyacetone for the samples treated at 850 and 900 °C. However, when the catalyst is sintered at too high temperature, at 1000 °C, this effect is lost. While the fresh catalyst seems to have an initial stable and constant conversion, or at least the rate-of-deactivation is below the standard deviation in the measurements. All of the thermally treated samples have a significantly decreasing yield in the short time tests (0–20 h). It is not uncommon that fresh catalyst has an initial deactivation period and then the activity level out and become stable.

So the initial deactivation is introduced by the thermal treatment. If water desorption from the catalyst surface is a rate determine step in the reaction, a dehydrated surface will show a higher reaction rate initially than a fully hydrated one. When the water coverage increases, the rate will drop until it reaches steady state and explain the initial behaviour of the catalysts. In the long-term tests, up to 600 h, run in the fully automated long-term test rig, the starting yield was somewhat lower than in the short term run, 75.0 ± 0.7% for the fresh and 87.6 ± 0.9% for the sample sintered at 850 °C (Fig. 11). This is most likely due to the small differences in the set-ups, possible variations in temperature profiles etc. However, no significant drop in the yield could be detected for either of the samples during the run for up to 600 h. The performance of the thermally treated samples, below 1000 °C, is significantly improved with respect to the acrolein yield. The increase in acrolein yield with increasing coking, as previously reported [24, 25], cannot be confirmed in this study, at least not on this timescale; perhaps this phenomenon is more of an initial change.

From the characterization of the samples, we know that the BET surface, the specific acidic amount and the acidic strength decreases due to the thermal treatment. However, the number of acidic sites per m2 of the catalyst increase. If more than one site is required for the dehydration reaction of glycerol, a higher density of acidic sites on the surface could be favourable, or it may be the decrease in acidic-strength that is positive for the glycerol to acrolein reaction.

Decreasing the hydroxyacetone selectivity should give less coke formation in the catalyst, validated in earlier studies by co-feeding hydroxyacetone with the glycerol [4]. To investigate the carbon formation, the pressure drops over the catalytic beds was measured during the long-term test run. It was difficult to obtain stable readings, and other factors than coking could affect the result for instance packing of bed during the run or fragmenting of the catalyst, but the results in Fig. 12 indicates an improved time-on-stream with the catalyst heat-treated at 850 °C compared to the unsintered catalyst.

5 Conclusions

In conclusion, the pre-treatment of tungsta/zirconia catalysts by heat treatment changes the performance of the catalysts. The heat treatment changes the initial deactivation behaviour and the acrolein selectivity is increased initially, most likely due to a difference in surface-water coverage. The heat treatment also influences the catalyst morphology such that heat treatment (at temperatures above 800 °C) result in a catalyst with lower specific surface area, more open pore structure (larger pores), lower absolute acidity and acid strength, but higher specific acid density (mmol/m2). There is also a recrystallization of the tungsta-phase from monoclinic to triclinic between 850 and 900 °C. The catalyst with larger pore system, with no pores at 45 Å or below, also performs better in long-term testing compared to the non-heat treated sample. This indicate that pore condensation may be a contributing factor in catalyst deactivation, but other factors such as acid strength and specific acidity may also influence the behaviour.