The sol-gel procedure resulted in an immediate gelation process and an orange gel was observed. At this point, stirring using a magnetic stir bar was not possible due to the high viscosity of the solution. After three hours of aging, the mixture turned into a bright yellow slurry. The surface area of the slurry following atmospheric drying is 56 m²/g. The DRIFTS spectrum of this sample is shown in Figure 1 and the comparison with other reference materials is shown in Table 1. All of the vibrations of VOPO₄·2H₂O are present [23]. However, some of the bands assigned to this compound are missing, and some other weak ones (2993 cm⁻¹, 2893 cm⁻¹, 1460 cm⁻¹ and 1364 cm⁻¹) are not related to any of the vanadium phosphates. These bands are attributed to intercalated compounds as discussed below.

The X-ray diffraction pattern of AD, Figure 2, shows the presence of many of typical reflections of α-VOPO₄ as well as some of the reflections of VOPO₄·2H₂O [15], Table 1 compares the spacing and intensity of these referenced compounds with the AD. However, there are other very intense reflections at 8.71 Å, 5.03 Å and 2.91 Å that are not characteristic of any of the most well known VPO phases. These reflections are attributed to the intercalation of either solvent or isopropyl alcohol. Okuhara and coworkers recently explored the phenomena of exfoliation by means of intercalating alcohols into the layers of VOPO₄·2H₂O. The process was carried out for three hours at relatively low temperatures (30–70 °C) and the subsequent reduction of the exfoliated VPO by primary alcohols yielded the precursor [7,30]. Previous works employed alkylphosphonic acid that reacted with V₂O₅ to intercalate the corresponding alkyl groups and intercalated pyridine into VOPO₄·2H₂O by reflux [31,32].

Thus, the DRIFTS and XRD patterns in Figure 1 and Figure 2 show VOPO₄ intercalated either with isopropyl alcohol or THF. This possibility is also supported by the presence of weak bands in the DRIFTS spectrum at 2993 cm⁻¹ and 2893 cm⁻¹ that correspond to C-H stretches of THF or isopropyl alcohol and the remaining weak bands that can be ascribed to these two compounds, Table 1. To substantiate that the sample was an intercalated VOPO₄, the XRD powder pattern was indexed to obtain the cell parameters. This calculation was done with CELREF software, version 3. A tetragonal cell symmetry was considered, which is the known symmetry of VOPO4·2H₂O and other intercalated compounds. This calculation gave cell parameter values of a = 6.158 Å and b = 8.868 Å. The main (hkl) indices are shown in Figure 2. The hard sphere diameter for THF and 2-propanol was estimated according to the methodology reported by Misawa [33] and both are ~4.8 Å. This hard sphere diameter is approximately half of the b parameter for the powder pattern shown in Figure 2 (8.86 Å), which can be interpreted as an association of the intercalated molecules with each of the oxide layers that bound the intercalation space. This situation has been denominated as noninterdigitated intercalation, where double layers of the alcohol or THF are included into a VOPO₄ lamellar structure [34]. Even though THF was not considered as an exfoliation agent, THF or isopropyl alcohol could act as donor reagents, intercalating the solid layers by donating electrons to the vanadium (V). At this time, we are unable to prove whether THF or isopropyl alcohol is the intercalated molecule.

The pressure in the autoclave was measured for three levels of solvent (100 mL, 50 mL and 20 mL) as a function of temperature and is shown in Figure 3. For comparison, a theoretical curve of the liquid-vapor equilibrium of THF is also plotted [35]. When high and medium amount of solvent are added to the slurry before the drying step, the system develops a behavior similar to the vapor-liquid equilibrium of pure solvent and the liquid vaporizes very close to the critical point. In contrast, when a low amount of solvent is added, the liquid completely vaporizes and never reaches the critical point.

The morphology of the particles is strongly affected by the type of drying. Figure 4 shows that high pressure drying yields platelets while atmospheric pressure drying yields spherical particles (the atmospheric drying is in agreement with literature [36]). Among the high pressure treatments, the low amount of solvent (LS) produced larger particles and there is not a significant difference between the medium amount (MS) and the high amount (HS). The morphology of LS is typical of the precursor with phosphorous/vanadium ratios close to one [14]. Thus, in the autoclave drying, the amount of added solvent influences the conditions where it is completely vaporized [24] and influences the final size and morphology. An additional and interesting aspect is that the observed morphology of the particles after the drying step is very similar to those obtained by Okuhara and coworkers [7], but have a smaller particle size. Their claims include the formation of platelets of the reduced compounds with particle sizes of 500–6000 nm in length and 69–156 nm in thickness (calculated from surface area measurements). The reduced compounds shown in Figure 4 b,c are approximately 400 nm in length and the length of crystallites in Figure 4 a is 500 nm. For LS, MS and HS precursors the thickness of crystallites was not calculated from surface area data due to the impurity (non-complete reduction) of samples.

The surface areas of all autoclave-prepared precursors are shown in Table 2 and the influence of the amount of added solvent in this parameter is evident. The values are also higher than the results reported by Ennaciri for atmospheric pressure drying in air [36]. The analysis of variance for the experiment with the three levels of solvent results in the conclusion that there is no difference between MS and HS, but LS is different from the average HS and MS. The Bonfferoni t test was used to compare the means [37]. These results indicate that vaporization of the solvent during drying before the critical point leads to a dramatic loss of surface area.

Infrared spectra and X-ray diffraction patterns were obtained for the materials dried with different amounts of solvent. These spectra, along with reference spectra, are shown in Figure 5a,e are DRIFTS fingerprints of VOPO₄ compounds shown in Figure 1 and the precursor respectively [23,38] Figure 5b–d show the evolution of spectra from VOPO₄ towards the precursor when the amount of added solvent is decreased. Bands at 1054 cm⁻¹, 977 cm⁻¹ and 1196 cm⁻¹ appear in Figure 5b, then 1104 cm⁻¹ and 932 cm⁻¹ are added in Figure 5c and finally 1133 cm⁻¹ band shows up in Figure 5d to complete the six characteristic bands of the fingerprint spectra of the precursor. These spectra are very similar to those reported by Kamiya et al. for exfoliated-reduced VOPO₄·2H₂O [7]. The amount of added solvent in the three cases influences the chemical composition of the final materials. Similarly, Figure 6a,e are X-ray diffraction patterns of the VOPO₄ phase shown in Figure 2 and the precursor respectively [15]. In Figure 6b the reflections at 5.67 Å and 2.91 Å appear as the first sign of the presence of the hemihydrate. Figure 6c is mostly equal to the a), however the peak at 4.5 Å can be also observed and is associated with the main reflection of the hemihydrate (4.53 Å). Figure 6d is the closest pattern to Figure 6e since it contains additional reflections at 3.64 Å, 3.08 Å, 2.77 Å, 2.59 Å, 2.39 Å, 1.89 Å and 1.84 Å. Thus, the crystallography of the resultant materials evidences the transformation of the yellow slurry into a compound that can be identified as the precursor.

These results show that in addition to the physical changes that occur during autoclave drying, chemical transformations also occur. Interestingly, while HS and MS had nearly the same surface area, they have quite different phase compositions. This suggests that the differences in surface area for these materials are not just due to differences in phase composition: the autoclave drying procedure has a substantial role in producing high surface-area vanadium phosphates.

As discussed in the previous sections, when the material was dried inside the autoclave, chemical transformations occurred, which yielded products containing reduced-vanadium phases. The reduction process can be attributed to the alcohol generated as byproduct during the hydrolysis of the vanadium alkoxides. It is known that alcohols are good reducing agents to prepare the precursor [39]. Reduction of VOPO₄ with isopropyl alcohol and other alcohols has been reported to yield non-agglomerated particles with platelet crystalline morphology as a result of the reduction of VOPO₄ [14]; this is in agreement with the trend seen in Figure 4a–c. As more solvent was added to the slurry, the concentration of alcohol decreased and less reduced products were obtained. In contrast, low amount of added solvent resulted in more reduced phases. As measured through titration, the average oxidation state of the samples decreased with the amount of solvent.

The materials prepared on this study are compared with a control synthesized through the traditional methodology [14]. Conversion, selectivity and intrinsic activity are shown in Table 3. Traditionally prepared catalysts behaved in similar way as reported by Horowitz [14], even though their study activated the precursors under an oxidizing atmosphere and our activation (dehydration process) was under anaerobic conditions. The slight difference might be due to the formation of non-selective V³⁺ phases during activation under nitrogen; intercalated benzyl alcohol might have acted as a reducing agent [40]. Despite the possible formation of V³⁺ we chose this activation procedure to avoid the over-oxidation of the precursors with an aerobic atmosphere as suggested by the same authors [38]. Thus, this catalyst is an appropriate baseline to compare with the sol-gel prepared materials.

The conversion levels of the four catalysts were not significantly different, as seen by the overlapping of the confidence intervals in Table 3. Selectivity shows the tendency to increase when the amount of solvent used during the precursor’s drying decreases. The confidence intervals for selectivity of traditional and low solvent treatment do not overlap meaning that the low solvent treatment is more selective. The confidence interval for the medium solvent treatment overlaps with the traditionally prepared material, suggesting similar selectivity, while the high solvent treatment has a lower selectivity than the traditional treatment. These results are attributed to the fact that the materials prepared with low amounts of the solvent have higher concentrations of the precursor and therefore, higher concentration of the active phase after activation. The activated catalysts are confirmed to be (VO)₂P₂O₇ by the XRD patterns in Figure 7 [15].

The scanning electron micrographs in Figure 8 illustrate the morphology of these materials. Small non agglomerated crystallites of about 600 nm in length are observed as activated catalysts from LS precursor. A similar shape but larger size, 1000 nm, is observed for active phase from MS and HS precursors. The crystallite thickness was estimated from both the Scherer’s equation [29] and the surface area [17]. The XRD-based calculation yielded 14 nm for those having 600 nm in length and 11 nm for those of 1000 nm; the estimation from the surface area measurements gave values of 4 and 7 nm respectively. This suggests that it cannot be assumed that the materials have a single morphology, so the surface area-based calculation is likely in error. Selective parts of the activated catalysts are suspected to be the very small particles present in Figure 8b produced by breaking down the corresponding precursor Figure 4a. The existence of the non selective low surface area domains is the reason for their low selectivity at the evaluated level of conversion (compared to those reported in the literature for small crystallites of the activated catalyst).

As a reference material for the catalytic evaluation, the precursor was prepared according to one of the traditional methods [14]. The procedure comprises the reduction of V₂O₅ with a mixture of alcohols (isobutyl and benzyl alcohol) and the addition of the phosphoric acid, 85% (Fisher). The sol-gel procedure was based on the reported methodology by Ennacciri et al. [30]. Two 1M solutions were prepared: (i) vanadium (V) triisopropoxide oxide (the alkoxide) (Alfa-Aesar) in tetrahydrofuran (THF) (Fisher) and (ii) orthophosphoric acid, anhydrous solid (Fluka), in THF (Fisher). Equal volumes (10 mL) of each solution were mixed. The phosphoric acid solution (ii) was added to the alkoxide solution (i) while stirring. The gel was aged for three hours and the final slurry was dried according to the procedure described below.

As a control procedure, the yellow slurry was dried under nitrogen at atmospheric pressure for four hours. This material will be referred to as AD. To dry the materials under high pressure, a reported procedure was followed [41]. The solvent and slurry was poured into a glass-lined 600-mL Parr reactor. The reactor was pressurized with nitrogen to 7.9 bars and heated from 25 °C to 265 °C at a rate of 1 °C/min. After ten minutes at the maximum temperature, the reactor was vented. In order to reach the critical point and avoid the collapse of the fragile gel, solvent was added before the slurry was dried in the autoclave [24]. To determine the influence of the amount of added solvent in the final product an experiment with a completely randomized design [37] was carried out with three levels or treatments: adding 20 mL, 50 mL or 100 mL (low, medium and high level, respectively) of THF. The resultant materials were referred to as LS, MS and HS respectively. Three replications were performed for each level.

Precursors LS, MS and HS were evaluated as catalysts for the partial oxidation of butane to maleic anhydride. W/F (weight of catalysts per molar flow of n-butane) was adjusted to 60 g h mol⁻¹ for MS and HS (having bulk density up to 0.1 g/mL) and 150 g h mol⁻¹ for LS and traditionally prepared precursor (bulk density of 0.18 g/mL and above). This allowed us to keep gas hourly space velocities (GHSV) between 1700 h⁻¹ and 2200 h⁻¹. The precursors were initially treated under nitrogen from 298 K to 673 K and were kept at this temperature for 2 h. Next, a reacting mixture of 1.7% butane in air was flowed through the catalytic bed. The reaction was carried out for 72 h. The products were analyzed with gas chromatography employing a 5 Å molecular sieve and a Porapaq QS column. The separated products were analyzed with a thermal conductivity detector (TCD) and a flame ionization detector (FID). The activated materials were referred to as CLS, CMS and CHS.

Diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was done with a Thermo Nicolet Nexus™ 670 FT-IR spectrophotometer equipped with a Smart Collector. The samples were diluted to 1–10 weight % in potassium bromide infrared grade (Acros). X-ray diffraction (XRD) was carried out with a Bruker axs D8 advance diffractometer that was set at 40 kV and 40 mA. Scans were from 5° to 70° (2θ) and with a step size of 0.05°. The diffractometer radiation was copper Kα (λ = 1.54 Å). The sample was put on the sample holder and exposed to the atmosphere at room temperature during the analysis. The scanning electron micrographs (SEM) were taken with a Hitachi S-3500N. The surface areas of the samples were determined by an Altamira instruments AMI-200 according to the BET model. The oxidation state of the vanadium in the samples was determined with the volumetric method reported in the literature [42].