1 Introduction

Vanadium-phosphorous oxide catalysts (VPO) are commercial catalysts for the transformation of n-butane into maleic anhydride [1,2,3,4,5], which is used for the manufacture of important materials such as polyurethanes and polyester resins. This process is an example of commercial application of an alkane partial oxidation reaction. Although this catalytic system is efficient for other partial oxidation reactions, such as propane oxidation and ammoxidation [4,5,6,7,8]. Several VPO phases have been described, with different V/P molar ratio and different vanadium oxidation states [6,7,8], resulting in quite a complex catalytic system. In addition, the VPO phases interconvert during reaction [9,10,11], and subsequently the catalytic behavior can also be modified. This demonstrates the importance of monitoring the oxidation state of vanadium species during reaction to understand and to control the catalytical behavior. Operando and in situ methodologies characterize materials exposed to reaction conditions, offering insight into the chemical structure of catalytic materials at work; i.e., the active phases and the reaction mechanisms [12,13,14]. Therefore, due to the extraordinary complexity chemistry of vanadium compounds, and VPO in particular, it is necessary to characterize their catalytically active phases during reaction with techniques that allow to uncover which are their chemical environment, their structure, and their oxidation state. 51V NMR spectroscopy is a powerful technique for the investigation of V-containing catalytic materials since the 51V chemical shift and quadrupolar tensors parameters for different systems can be compiled using original and published 51V isotropic chemical shifts data. This provides important chemical information about complex heterogeneous catalytic systems [15,16,17]. By using 51V NMR spectroscopy, several types of vanadia sites can be identified in supported vanadia catalysts [17, 18], giving insights about the structure and properties of these materials. We have previously applied 51V and 31P NMR to characterize VPO catalysts [18], demonstrating the simultaneous determination of chemical shielding anisotropy and quadrupole tensor parameters, as well as their distributions, delivering an insight of the local environment of vanadium sites in the V–P–O catalysts. Electron spin resonance spectroscopy (ESR) [19, 20] can quantify the relative amount of paramagnetic V4+ ions. V3+, a d2 system, does not produce an ESR signal, not even at ambient nor at liquid nitrogen temperature. Very short electron spin–spin and spin–lattice relaxation times are the most probable reason for this [21], but the amount of V3+ ions could be determined by magnetic susceptibility measurements and by the difference Vtotal − V5+ − V4+. Another characterization method that can be performed under controlled conditions of temperature and atmosphere is XANES [22], which can provide valuable information to elucidate the nature of the chemical bonds at the catalyst surface to better understand reaction mechanisms and intermediates. Thus, the objective of present study is to perform a combination of different characterization techniques (NMR, ESR and XANES) using operando and in situ conditions which may provide complementary information on the active species relative to the oxidation state and the chemical environment to alumina-supported VPO catalysts. The use of supported catalysts maximized the exposed fraction of active sites, allowing the characterization of the outermost layer [9, 23, 24], most relevant to catalysis.

2 Experimental

2.1 Characterization

Solid-state NMR experiments were performed using Bruker MSL-400 (9.4 T) spectrometer at resonance frequency of 105.20 MHz for 51V, 104.2 MHz for 27Al and 161.9 for 31P. Bruker 4.0 mm MAS probe was used for acquisition of static and 15 kHz MAS spectra. The single pulse sequence with radiofrequency pulse duration of 1 μs (less than π/8) and recycling time from 0.1 to 5 s was used for 51V, and 27Al, for 31P pulse duration of 5 μs and recycling time from 1 to 5 s were used. For 51V the chemical shift value referenced to external VOCl3, for 27Al—to external Al(H2O)63+, for 31P—to 85% H3PO4, NMR spectra were performed for hydrated samples (hydration of the samples was achieved by exposing them to ambient humidity). Simulations of 51V, 27Al static and MAS NMR spectra were performed considering second-order quadruple correction using general purposes simulation program NMR5 (modified NMR1) described earlier [17]. The relative content of different species was determined from integral intensities of the spinning sidebands of the central transition and from static spectra.

The ESR spectra were recorded on a Bruker ER200D spectrometer with a microwave frequency of 9.2 GHz (X-band). The samples were placed into a double cavity which permitted the measurement of a sample and a reference under identical conditions, particularly at the same microwave power. A solid solution of Mn2+ in MgO (hyperfine structure with six lines, g = 1.9796) was used as reference material.

V K-edge XANES/EXAFS spectra of the samples were acquired at the BM31 beamline at the European Synchrotron Radiation Facility (ESRF) in Grenoble in the 4500–5600 eV range, with an energy step of 1 eV/s and using a Si (111) monochromator. A vanadium metal foil was used as reference to calibrate the pre-edge absorption energy of the V spectra. We used a reaction cell previously described [25] to obtain the spectra under reaction conditions, which is a modified commercial (Specac) IR cell, with Kapton windows. For each experiment, 50 mg of sample were pressed in a stainless-steel holder. This cell is equipped with a thermocouple for controlling the temperature, and mass flow controllers were used for feeding the gases to the feed. A total flow of 20 ml/min was used.

Nitrogen adsorption isotherms (− 196 °C) were recorded on an automatic Micromeritics ASAP-2000 apparatus. Prior to the adsorption experiments, samples were outgassed at 140 °C for 2 h. BET areas were computed from the adsorption isotherms (0.05 < P/P0 < 0.27), taking a value of 0.164 nm2 for the cross- section of the adsorbed N2 molecule at − 196 °C.

2.2 Preparation of Samples

The synthesis method for preparing VPO supported on g-alumina by incipient wetness co-impregnation with different P/V atomic ratios (0.1–1.1) and loadings (4–9 V + P monolayers), has already been described [9]. The dispersion limit, understood as the maximum surface loading of VOx units that remain dispersed, with no crystalline V2O5, was determined by Raman spectroscopy to be near 9 VOx units per nm2 of alumina support, in accordance with a previous report on another alumina [26]. Table 1 shows the V + P coverage and V/P atomic ratio values of the prepared catalysts, as well as the BET surface area values. The nomenclature used was xVyPAl, where x and y are numbers of atoms of V and P per nm2 of alumina support.

Table 1 BET surface areas and P/V molar ratio values of VPO alumina supported catalysts

2.3 Propane Ammoxidation Tests

Activity measurements were performed at atmospheric pressure, using a conventional micro-reactor with 9 mm diameter size and minimum void volume. The feed stream and effluents of the reactor were analyzed by an on-line gas chromatograph equipped with flame ionization and thermal-conductivity detectors. The accuracy of the analytical determinations was checked for each test by verification that the carbon balance (based on the propane converted) was within the cumulative mean error of the determinations (± 10%). The catalytic tests were made using 0.2 g of powder sample with particle dimensions in the 0.250–0.125 mm range. The axial temperature profile was monitored by a thermocouple sliding inside a quartz tube inserted into the catalytic bed. Tests were made using the following reaction feed composition (% volume): total flow 20 ml/min; 25% O2, 9.8% propane, 8.6% ammonia and rest He. The quantity of catalyst and total flow were determined in order to avoid internal and external diffusion limitations in previous studies [24]. Yields and selectivities in products were determined based on the moles of propane feed and products, considering the number of carbon atoms in each molecule. Space–time yields were calculated as kg of acrylic acid per kg of active phase and hour of reaction.

3 Results

51V NMR spectra of samples with different V + P coverages and V/P molar ratio are shown in Fig. 1. In accordance with previous 51V NMR studies on V catalytic materials [27, 28], the spectra can be interpreted in terms of three components (A, B and C marked in the Fig. 1A) that have been analyzed by more specific NMR experiments with model samples [27,28,29], showing that different chemical environments are identified for V5+ species. Three lines are observed in 51V NMR spectrum of the sample 4VAl: the line A is assigned to decavanadate (δ = − 350 ppm), line B corresponds to weakly bound V, with a maximum at − 550 ppm; and line C is assigned to strongly bound vanadium with a maximum at − 720 ppm [27,28,29]. The presence of P results in vanadium reduction, though the decavanadate (line A) contribution appears unaffected (Fig. 1A line at − 350 ppm). Line B (− 550 ppm) significantly decreases at V/P = 1, while P excess results in full disappearance of line C and significant decrease of line B. Thus, the addition of phosphorous induces the reduction of vanadium species and it is probable that some of the reduced sites have V–O–P bonds, since V–O–P species have been identified by Raman spectroscopy [9] in V–P–O alumina supported samples. Figure 1B shows the effect of V/P addition to the local environment of vanadium centers; as expected, the relative amount of V5+ species decreases as the phosphorous content increases.

Fig. 1
figure 1

A 51V NMR static spectra of samples 4VAl (green), 16V16PAl (dark blue), 8V8PAl (black), 4V4PAl (red). B 51V NMR spectra of samples 9V8PAl (blue), 8V8PAl (black) and 8V9PAl (red)

Figure 2 shows the 27Al NMR spectra of samples with different V + P coverages and V/P molar ratios. Figure 2A illustrates the spectra of bare alumina and samples with V or P (4VAl and 4PAl); these spectra clearly illustrate that 5-coordinated Al appears in P/Al series and is also practically absent in 4VAl; phosphorous introduction also results in the appearance of a second octahedral Al (− 10 ppm). The 5-coordinated Al sites observed on pure Al2O3 disappeared upon vanadium impregnation and became more pronounced and slightly shifted upon phosphorous impregnation (4PAl sample). Figure 2B illustrates that not only vanadium becomes silent for 51V NMR, but also aluminum turns 27Al NMR silent upon phosphorous impregnation. Significant broadening of the signal from tetrahedral Al sites is apparent in P-containing samples. This broadening can be caused both by paramagnetic effect and by larger values of the quadrupolar constant and its distribution. Figure 2C illustrates the effect of V + P coverage whereas 2D, the effect of V/P. The ratio between penta- and hexa-coordinated aluminum sites depends on vanadium content and on the V/P atomic ratio.

Fig. 2
figure 2

A 27Al MAS NMR spectra of bare Al2O3 (black) and samples 4VAl (red) and 4PAl (blue), spectra normalized in height; B 27Al MAS NMR spectra of bare Al2O3 (black) and samples 4VAl (blue), 8V9PAl (red), 16V16PAl (green), spectra with real intensity; C 27Al NMR spectra of samples 4V4PAl (blue), 8V8PAl (black) and 16V16PAl (red), spectra normalized in height; D 27Al NMR spectra of samples 8V8PAl (black), 9V8PAl (blue) and 8V9PAl (red), spectra normalized in height

31P NMR spectra of samples with different V + P coverages and V/P molar ratio are shown in Fig. 3. The effect of V + P coverage in Fig. 3A shows a broad line for 4PAl sample near − 12 ppm, which shifts to − 25 ppm upon vanadium impregnation. The narrowest line with maximum shift is apparent for 8V8PAl. The effect of the V/P ratio in the spectra is illustrated in Fig. 3B, showing that both P or V excess, broadens and shits the signals to lower field.

Fig. 3
figure 3

A 31P MAS NMR spectra of samples 4PAl (green), 16V16PAl (blue), 8V8PAl (black), 4V4PAl (red). B.31P MAS NMR spectra of samples 8V9PAl (red), 8V8PAl (black) and 9V8PAl (blue)

The ESR spectra, and their relative intensities are shown in Figs. 4, 5. All ESR spectra can be described by the superposition of two lines: line with axial symmetry with hyperfine splitting (hfs) and line similar to that in VPO. The room temperature ESR spectrum in Fig. 4A, can be regarded as representative, and its axial line with hfs is well described by g||= 1.937, and g = 1.976 and the hyperfine coupling constants A = 63.6 × 10–4 and A||= 173 × 10–4 cm−1, as determined in our previous study [30]. For the sample with V/P = 1 content of V4+ depends on Vanadium content (Fig. 4B), but not linearly.

Fig. 4
figure 4

A ESR spectra of samples 9V8PAl (black), 4V4PAl (red), their difference (green and VPOC (blue). B 4VAl (green), 16V16PAl (red), 8V8PAl (light green), 4V4Pal (khaki) and their relative intensities

Fig. 5
figure 5

ESR spectra of samples 9V8PAl (rose), 8V8PAl (green), 8V9PAl (turquoise) and their relative intensities

In order to evaluate the catalytic properties during partial oxidation reaction, the samples were tested during propane ammoxidation reaction, and the results are shown in Table 2. 4V4PAl catalyst presents a quite poor yield to acrylonitrile production, due to the low coverage and subsequently low number of active species on the surface. Previous studies have shown that two nearby sites are required for the propane transformation into acrylonitrile [31, 32]. The yield to acrylonitrile is low for the sample 16V16PAl (Table 2), in this case, the coverage is quite high and quite big crystalline phases are formed, that seems to be not as efficient as site isolated well distributed active phases as in the case of 8V9PAl and 8V8PAl. Based on these results (Table 2), 8V9PAl catalyst was selected for the operando-XANES study, since is presents the highest yield to acrylonitrile formation. Thus, XANES spectra were recorded under propane ammoxidation conditions with the reactor-cell described in the experimental section.

Table 2 Catalytic properties of catalysts in propane ammoxidation reaction at 500 °C

Figure 6 shows the main edge position energies for 8V9PAl catalyst and for some references phases during operando-XANES propane ammoxidation. V2O5 has a coordination of strongly distorted square pyramids of VO5 [33], and NH4VO3 has distorted tetrahedral coordination [34]. These two oxides, in which vanadium species are V5+, exhibit a high edge energy value. VOSO4 and (VO)2P2O7 references phases, with essentially V4+ and with a distorted VO6 coordination [35], presents lower main edge position energy values. This illustrates the influence of the oxidation state of V species on the main-edge peak position. At ambient conditions, the vanadium edge energy is higher than 5485 eV, indicative that most vanadium species must be V5+. When the reactants flow, even at a low temperature (70 °C, A70) a shift to 5484 eV indicates the reduction of some vanadium to V4+. This edge value remains similar up to 500 °C. However, the edge value approaches that observed at ambient conditions at higher temperatures (550 °C, A550). It should be noted that the optimal acrylonitrile yields are obtained in the 480–500 °C range [9], when the energy is close to 5484 eV.2. Once that the temperature reaction has reached a maximum near 550 °C, the temperature was cooled and the spectrum was also taken at 300 °C. In this case a value pretty close to that previously obtained at the optimal temperature reaction (480–500 °C) was obtained (Fig. 6); this suggests that some vanadium species reduce to V4+ when the temperature decreases; this underlines the dynamic character of vanadium species in this catalystic system.

Fig. 6
figure 6

Main edge position energies for 8V9PAl catalyst and reference samples during propane ammoxidation reaction conditions at different temperatures. Total flow 20 ml/min; feed composition: 25% O2, 9.8% propane, 8.6% ammonia in He

4 Discussion

Several individual compounds may form in the ternary V2O5–P2O5–Al2O3 system, being AlVO4, AlPO4, and VOPO4 the most common, and their NMR spectra are well known. There are three nonequivalent vanadium and aluminum sites in the structure of AlVO4, respectively, being their 51V and 27Al NMR spectra characterized by a superposition of 3 lines in 51V NMR spectra [15], and in 27Al NMR spectra [36]. 51V NMR spectra of (αI, αII, β, γ, δ)-VOPO4 have axial symmetry, a large value of chemical shift anisotropy (800–950 ppm) and an isotropic shift in the range of − 690 to − 780 ppm [18, 37], whereas 31P NMR spectra of these compounds are observed in the range of 40–10 ppm. Furthermore, a reversible transformation between the orthophosphate phases and the vanadyl pyrophosphate is possible; this corresponds to a change of the oxidation state of vanadium atoms between V5+ to the V4+ and, possibly, to V3+ state and vice versa [38, 39], observed in this case under reaction conditions by XANES (Fig. 6). Since V4+ and V3+ species are paramagnetic, they are not directly visible to solid-state 51V NMR. However, static 31P spin-mapping NMR spectroscopy allows the discrimination of signals due to phosphorus atoms in the proximity of V5+ (40–10 ppm), V4+ (very broad signal at ca. 2500 ppm), and V3+ species (very broad signal at ca. 4650 ppm) in VPO compounds [40, 41].

51V NMR spectra rule out the presence of any crystalline vanadia phases in our VPO samples. As indicated above, 51V NMR spectra (Fig. 1) of all the samples represent a superposition of three types of lines: A—from vanadium in decavanadate (manifested in static spectra by δ = − 350 ppm), B—vanadium oxide sites weakly bound to the surface by one or two bonds (line at − 550 ppm) and line C (at − 720 ppm) for strongly bound VOx species (bound to the surface by 3 pods). The position of the line C resembles that of AlVO4, in which three nonequivalent isolated (no V–O–V connection) tetrahedral sites exist. The introduction of phosphorus leads to a partial reduction of vanadium, while decavanadate (signal A) practically does not interact with phosphorous, since the content of decavanadate is similar in all three samples (Fig. 1A). Line B is significantly reduced at 1/1 V/P ratio, excess phosphorus completely depletes line C and as a result V5+ in xVyPAl system remains in isolated or slightly associated surface sites and as decavanadate species. The operando XANES results (Fig. 6) may indicate that these chages in the vanadium environment induced by P species are quite dynamic since some changes can occur under reaction conditions as demonstrated operando experiments.

The decrease in the intensity of 51V NMR signal upon phosphorus progressive addition is most likely due to the reduction of part of vanadium V5+ species to V4+ and V3+. In the ESR spectra (Figs. 4, 5), V4+ is indeed observed, but there is no a linear relationship between the decrease of V5+ and appearance of V4+. There are also no lines in the 31P NMR spectra (Fig. 3) from phosphorous connected to V4+ or V3+ (there are no signals in the 2500 and 4600 regions). The latter can be explained by both the small content of V4+ and V3+ sites and the large linewidth in 31P connected with paramagnetic centers. The presence of reduced V4+ species is relevant since they are involved in the catalytic cycle (Fig. 6) and (VO)2P2O7 has been described as the active phase [42], being vanadium as V4+, although the presence of V3+ is probably also involved in the redox cycles during reaction. This is in line with theoretical studies [31, 32] that indicated that two different V structures are required for the propane transformation into acrylonitrile, one more oxidized (5+) and the other more reduced (+4). This is in line with the operando XANES study (Fig. 6) and also with the NMR and ESR results. These results are also in line with previous operando Raman studies [9] with similar catalysts, that indicated the presence of V4+ species in vanadyl pyrophosphate as well as V5+ species as dispersed VOx structures along with VPO dispersed structures. By this manner it is clear that the active catalyst it is not a well-defined crystalline structure, instead, it is flexible, and it is able to harbor V species at different oxidation states. By this manner, the catalysts are able to suffer a redox cycle during reaction, according to Mars-van-Krevelen mechanism [14, 43].

The 27Al NMR studies of AlPO4 obtained through different synthesis methods were performed in Ref. [44]. When the compound was made by the sol–gel method with subsequent temperature treatment, and calcined at 500 °C, two signals are observed in the 27Al NMR spectra due to tetrahedral and octahedral positions and two 31P NMR signals with shifts of − 7.5 and − 14 ppm [44]. When it was obtained at high pressure (for example, two phases synthesized at a pressure of 4–5 GPa and 6–7 GPa) two nonequivalent positions of aluminum at 40 and 43 ppm and two positions of phosphorus at − 27 ppm and − 24 ppm were revealed in 27Al and 31P NMR spectra [45, 46]. According to the 27Al NMR spectra (Fig. 2), the formation of AlPO4 in our VPO samples can be excluded in this series, but some changes are induced upon introduction of phosphorus, since not all Al remains visible in 27Al NMR spectra (Fig. 2B) and new signals from five-coordinated aluminum (at 40 ppm) and from the second octahedron (− 10 ppm) (Fig. 2A–C) become apparent. The ratio between the new signals of 5- and 6-coordinated aluminum sites depends on the ratio between vanadium and phosphorous. The number of 5-coordinated species increases with vanadium content (up to 8 V), then decreases (16 V), being maximum at V/P = 1 (Fig. 2C, D). 5-coordinated aluminum at Al2O3 surface disappears upon vanadium deposition on Al2O3 (Fig. 2A). The appearance of a new 5-coordinated aluminum at 40 ppm (Fig. 2A, B) upon phosphorus addition may be associated with the formation of some Al–O–P bonds on Al2O3 surface.

The presented 51V, 27Al, 31P NMR and ESR data show that varying the vanadium content and the V/P ratio generates a wide variety of spectra and properties of the samples. With these data, it has been possible to identify the optimal formulation in the 8V9PAl catalysts. It has the minimum concentration of V5+ (Fig. 1B), a complete disappearance of line C (isolated strongly bound VO4 species) (Fig. 1B, red), and a high content of V4+ and V3+ (ESR and XANES data, Figs. 5, 6). Presence of V3+ in this sample follows from the comparison of V5+ (which is minimum) and V4+ (which is not maximum), thus the rest part is V3+. This sample has distorted P sites (Fig. 3B, blue trace) and mostly distorted tetracoordinated Al sites (27Al NMR, Fig. 2D, red trace) and the lines from both 4- and 5-Al coordinated sites are strongly broadened, which can be due to paramagnetic effects, although it is also possible that this broadening is caused by larger value of quadrupole constant (CQ) and its distribution. Operando-XANES experiments during propane ammoxidation demonstrate that this catalyst undergoes changes during catalytic cycles at 400–500 °C temperature range, which are related with the reduction of some vanadium species. These data show the importance of the environment of V and P sites during the catalytic cycle, and its effect on the coordination and environment of the Al support sites.

5 Conclusions

Operando XANES, in combination with NMR and ESR spectroscopies, has proven to be an effective characterization technique for identifying the oxidation states and chemical environments of complex oxide systems such as the VPO catalytic materials. These studies have allowed to identify the characteristics of the optimized catalytic formulation (8V9PAl sample), and their synthesis parameters, that is the one that possess a high contribution of reduced V4+ species and distorted tetrahedral Al sites.