V₂O₅-WO₃/TiO₂ catalysts usually contain 0.5%-3% V₂O₅, 8%-10% WO₃ and TiO₂ as the support, and have been widely adopted for the selective catalytic reduction of NO_x by NH₃ (NH₃-SCR) ^{[1, 2, 3, 4, 5, 6]}. Exhaust gas from stationary sources such as power plants always contains alkali oxides and salts which can deactivate the catalyst. Potassium exhibits the most serious poisoning effect on vanadium-based catalysts of the alkali and alkali earth metals ^{[7, 8, 9, 10]}. Among different potassium salts, potassium chloride gives severe deactivation ^{[11, 12, 13]}.

It is generally accepted that the alkali interacts strongly with dispersed vanadia species, neutralizes Brönsted acid sites, decreases the ammonia adsorption capacity and reduces the redox property, leading to a decrease in the SCR activity of catalyst. However, the potassium poisoning mechanism of V₂O₅-WO₃/TiO₂ catalysts has not been well established. The roles of vanadia and especially tungsten oxide in this process have not been clearly clarified. For example, Wan et al. ^[7] suggested that the formation of inactive metavanadate species such as KVO₃ is responsible for the deactivation caused by alkali (earth) metals on V₂O₅-WO₃/TiO₂ catalysts. Nevertheless, they did not give proof of the existence of these inactive species. Bulushev et al. ^[13] identified the vanadia species in K-doped V₂O₅/TiO₂ catalysts by H₂ temperature-programmed reduction and FT-Raman spectroscopy as a function of the K/V ratio. They found that K-doped monomeric species were formed with the addition of a small amount of K, and more amorphous KVO₃ existed on the surface at K/V = 1. In our previous study ^[14], the alkali resistance of V₂O₅-WO₃/TiO₂ catalysts prepared by different impregnation methods was investigated. The wet-impregnated sample with more polymeric vanadia species was found to be less sensitive to potassium than the isolated vanadia species. Kamata et al. ^[15] also confirmed the formation of KVO₃ on K-impregnated V₂O₅-WO₃/TiO₂ catalyst by Raman spectroscopy. They suggested that potassium probably coordinates to the hydroxyl groups on WO₃ but without any experimental evidence support. Thus, it is important to identify the interaction between potassium and the different components from the impregnation of KCl on V₂O₅-WO₃/TiO₂ catalysts, and clarify the roles of tungsten oxide and vanadia in the alkali resistance of the catalyst.

In the present work, V₂O₅-WO₃/TiO₂ catalysts of various compositions were treated with KCl to investigate their alkali resistance. The surface acidity, structural, textural and redox properties of the catalysts were characterized by a series of methods. The specific roles of vanadia and especially tungsten oxide in the chemical deactivation of the catalysts were discussed.

The catalysts were prepared by excess water solution impregnation. Nano TiO₂ powders (Millennium Chemicals DT51, USA, S_BET = 81 m²/g) were mixed with the ammonium paratungstate solution (Beijing Chemical Reagent, China) by magnetic stirring for 1 h. The mixture was dried at 110 °C overnight and calcined at 450 °C for 4 h in air to obtain the WO₃/TiO₂ (WT) powders. The nominal mass loading amount of WO₃ was 10%. V₂O₅/TiO₂ catalysts were prepared by a similar impregnation method using ammonium metavanadate (Beijing Chemical Reagent, China, 99.9%) as the precursor. The catalysts obtained with a nominal V₂O₅ mass loading amount of 1% and 3% were denoted as V1T and V3T, respectively. The V₂O₅-WO₃/ TiO₂ catalysts were also prepared by impregnating ammonium metavanadate on the WT powder. The catalysts obtained were denoted as V1WT and V3WT, respectively.

The potassium poisoned catalysts were prepared by impregnating an aqueous solution of KCl on the fresh samples with a nominal mass ratio 0.83% of K. The impregnated powders were then dried overnight and calcined at 450 °C for 2 h. The samples were denoted as WT-K, V1T-K, V3T-K, V1WT-K and V3WT-K, respectively. According to the TGA study by Salinas et al. ^[16], the phase transformation from anatase to potassium titanate in 20% KNO₃/TiO₂ occurred at 670 °C, which was much higher than the calcination temperature (450 °C). In addition, the K/Ti ratio was also much higher than the K/Ti ratios in the poisoned catalysts. Thus, the formation of potassium titanate was not considered in the present work.

The NH₃-SCR activity measurement was carried out in a fixed-bed reactor with 200 mg of catalysts (diluted to give 1 mL with silica pellets). The reaction gas mixture consisted of 5×10^-4 NO, 500×10^-4 NH₃, 5% O₂ and N₂ balance. Prior to the measurement, the catalysts were treated in a 20% O₂-80% N₂ flow at 450 °C for 30 min. The measurement was performed at fixed temperatures from 150 to 450 °C with 50 °C intervalsand kept for 20 min at each point. The gas hourly space velocity (GHSV) was 30000 h^-1. The feed and product concentrations were measured by a Nicolet 380 apparatus (ThermoFisher, USA) at 120 °C. The NO_x (NO, NO₂ and N₂O) conversions were calculated as NO_x conversion (%) = (NO_in - NO_out - NO_2out - 2N₂O_out)/NO_in x 100.

The inductively coupled plasma (ICP) analysis was carried out using an ICP emission spectrometer (Vista-MPX, Varian, USA).The specific surface areas of the samples were measured using N₂ adsorption at -196 °C by the four-point Brunauer-Emmett-Teller (BET) method using an automatic surface analyzer (F-Sorb 3400, Gold APP, China). The samples were degassed in vacuum at 200 °C for 2 h before the measurement. The Raman spectra of the samples were obtained at ambient condition on a confocal micro-Raman apparatus (IDSpec Aurora, China) using Ar⁺ laser with a CCD detector. Pure powder supported on a sheet glass was used without any pretreatment and the wavelength of laser was 632.8 nm and the exposure time was 20 s while the spectra were recorded.

H₂ temperature-programmed reduction (H₂-TPR) was conducted on a chemical adsorption instrument AutoChem II 2920 (Micromeritics, USA). Prior to the test, 100 mg catalyst was purged in O₂ (50 mL/min) at 500 °C for 30 min and then cooled down to room temperature. The reactor temperature was raised to 900 °C at a heating rate of 10 °C/min in 10% H₂-90% Ar (50 mL/min).

The in situ infrared (IR) spectra of ammonia adsorption were recorded on a Nicolet 6700 FT-IR spectrometer (ThermoFisher, USA) equipped with an MCT detector. The catalyst was pretreated in 20% O₂-80% N₂ flow (100 mL/min) at 500 °C for 30 min, and then saturated with ammonia (500×10^-4 NH₃ in N₂, 100 mL/min) for 30 min. After purging in N₂ for 30 min at different temperatures (from 50 to 450 °C at intervals of 50 °C), the spectra were collected by accumulating 32 scans at a resolution of 4 cm^-1.

Figure 1 shows the NH₃-SCR activity of the fresh and poisoned catalysts. As shown in Fig. 1(a), the two fresh V₂O₅/TiO₂ catalysts showed much higher deNO_x activity than WO₃/TiO₂, indicating that vanadia species are the active sites for the SCR reaction. Increasing the vanadia loading enhances the catalytic activity of the catalyst ^{[17, 18]}. Meanwhile, the addition of tungsten oxide improved the deNO_x efficiency of the V₂O₅/TiO₂ catalysts significantly, as has been reported ^[1]. The V3WT catalyst achieved above 90% NO_x conversion in a wide temperature window 200-450 °C. Almost no NO₂ or N₂O was detected in the downstream gas in the whole temperature region, indicating that a high selectivity to N₂ was achieved for all the fresh catalysts.

Fig. 1. NH3-SCR activity of (a) fresh and (b) poisoned catalysts. Reaction conditions: NO = NH3 = 5×10-4, O2 = 5%, N2 in balance, GHSV = 30000 h-1.

After impregnation with KCl, the catalysts were deactivated to different degrees as shown in Fig. 1(b). The NO_x conversion of WO₃/TiO₂-K was always close to zero within the whole temperature range measured, and that of V1T-K was generally below 10%. The maximum NO_x conversion of V1WT-K and V3T-K catalysts were 41% and 50% at 350 °C, respectively. V3WT-K exhibited relatively high activity at low temperatures, with the maximum NO_x conversion of 60% was achieved at 250 °C. Notably, the NO_x conversion turned negative (-45%) at 450 °C due to the over-oxidation of ammonia to NO_x at high temperatures. Again, almost no NO₂ or N₂O was detected for these poisoned catalysts, implying that the main product of ammonia oxidation was NO.

In order to determine the contents of vanadium, tungsten and potassium, ICP measurements were applied. The results are listed in Table 1. It is clear that the poisoned catalysts have similar contents of V₂O₅ and WO₃ to the corresponding fresh catalysts. No significant loss of vanadia occurred because of the alkali poisoning treatment adopted. It has been reported in our previous study that due to its high volatility, the residual content of chloride on V₂O₅-WO₃/TiO₂ was much lower than that of potassium ^[11]. A similar phenomenon is observed in this work. We also found that the catalyst was hardly deactivated by such low contents of Cl ^[11], and therefore the influence of chloride is not discussed here. V3WT-K and V3T-K with a high vanadia loading maintained a similar K content to the nominalmass value (0.83%), while a little bit of potassium were leached out of the V1WT-K, V1T-K and WT-K catalysts. The fresh catalysts have similar BET surface areas of 80-83 m²/g, which were close to that of the titania support (81 m²/g). The poisoned catalysts experienced a small loss in the surface area due to pore blocking by the potassium salt. Based on the above results, the physical properties including the potassium content and surface area were not responsible for the different deactivation degree of the catalysts.

Table 1 Compositions and BET surface area of the catalysts a. Catalyst Element mass contents b (%) SBET/ (m2/g) V2O5 WO3 K Cl V1WT(-K) 0.93 (0.88) 9.91 (9.72) — (0.75) — (0.0154) 83 (80) V3WT(-K) 2.55 (2.46) 9.99 (9.50) — (0.82) — (0.0152) 80 (76) V1T(-K) 0.86 (0.90) — — (0.76) — (0.0159) 82 (78) V3T(-K) 2.54 (2.63) — — (0.82) — (0.0157) 85 (78) WT(-K) — 9.76 (9.35) — (0.74) — (0.0116) 83 (80) a The data of the fresh and poisoned catalysts are listed outside of and in brackets, respectively. b Measured by ICP.

Table 1 Compositions and BET surface area of the catalysts a.

The states of the tungsten oxide and especially vanadia species on the catalysts are important in determining the catalytic behavior of catalyst. Fig. 2 shows the Raman spectra under ambient conditions. All the catalysts exhibited a broad band at 796 cm^-1 corresponding to the Ti-O vibration of anatase TiO₂. V3T and V3WT contained both isolated monomeric vanadyl species (1005 or 1012 cm^-1) and polymeric (metavanadate-like) species (917 cm^-1) ^{[19, 20, 21]}. These bands became much weaker in the Raman spectra of V1T and V1WT due to the low vanadia loading. The isolated vanadyl band was shifted slightly towards higher frequency (1012 cm^-1) and increased in intensity for V3WT compared with V3T. These indicated that the addition of tungsten oxide facilitated the distribution of V^x+ species in the form of isolated vanadyl. For the WO₃/TiO₂ catalyst, a band at 978 cm^-1 was assigned to the symmetric W=O groups. It was shifted slightly to higher frequency (982 cm^-1) in the Raman spectrum of V1WT, which may be caused by the overlapping of the isolated vanadyl band ^[22].

Fig. 2. Raman spectra of (a) fresh and (b) poisoned catalysts.

Two additional Raman bands at 934 and 904 cm^-1 appeared in the spectra of the poisoned catalysts V1T-K, V3T-K and V1WT-K, which were due to the symmetric O-V-O stretching mode of “amorphous” KVO₃ in the form of small crystalline particles or a thin amorphous layer. This was confirmed by the characteristic bands of standard KVO₃. Furthermore, the red shift of the high frequency band indicated a lengthening of the V=O bond due to the presence of potassium and an electrostatic interaction ^[23]. Thus, the bands at 982-970 cm^-1can be assigned to K-doped monomeric species such as K-O-V=O sites. The main Raman frequency of K₂WO₄ was located at 922 cm^-1. The Raman spectra of TiO₂ supported K₂WO₄ could not be analyzed because of the very weak signals detected, as reported by Erdöhelyi et al. ^[24]. The red shift of the band from 978 to 965 cm^-1in the spectrum of WT-K may also be caused by K-doped species such as K-O-W=O sites. Compared with V3T-K, V3WT-K exhibited the band at a slightly higher frequency (982 cm^-1). This was related to the overlapping of a shoulder assigned to some residual isolated monomeric species, which was confirmed by the maintaining of the band (917 cm^-1) assigned to polymeric metavanadate species ^[15].

It has been established that the surface acidity of V₂O₅-WO₃/TiO₂ catalyst plays an important role in the selective reduction of NO_x with NH₃. To investigate the influences of the catalyst composition and alkali poisoning on the acid sites of the catalyst, the IR spectra obtained after NH₃ adsorption at 50 °C are shown in Fig. 3. As shown in Fig. 3(a), the bands at 1672 and 1447 cm^-1 were assigned to the symmetric and asymmetric bending vibrations of NH⁴⁺ coordinated to Brönsted acid sites, respectively. The bands at 1594 and 1280-1100 cm^-1 were assigned to the asymmetric and symmetric bending vibrations of NH₃ coordinated to Lewis acid sites ^{[25, 26]}. The corresponding stretching modes were found at 3391 and 3224 cm^-1 for coordinated NH₃, and at 3022 and 2812 cm^-1 for ammonium ions. It was obvious that the NH₃ adsorbed on Brönsted acid sites was more dominant than that on Lewis acid sites, which is common for vanadium-based catalysts ^{[22, 27, 28]}.

Fig. 3. FT-IR spectra of (a) fresh and (b) poisoned catalysts after NH3 adsorption at 50 °C for 30 min. (1) V1WT(-K); (2) V3WT(-K); (3) V1T(-K); (4) V3T(-K); (5) WT(-K).

The IR spectra of the NH₃ adsorbed at 50 °C over the potassium poisoned catalysts are shown in Fig. 3(b). Both the ammonium ions (NH₄⁺) coordinated to Brönsted acid sites (3022, 2784, 1672 and 1443 cm^-1) and NH₃ coordinated to Lewis acid sites (3398, 3260, 1601, 1213 and 1116 cm^-1) were observed for all the poisoned catalysts. Nevertheless, NH₃ adsorption was severely suppressed on the poisoned catalysts with bands ofmuch lower intensity.The V3WT-K and V3T-K catalysts exhibited a stronger band at 1447-1443 cm^-1 than V1WT and V1T. This indicated that there were more Brönsted acid sites on the poisoned catalysts at a high vanadia loading.

The addition of potassium not only suppressed NH₃ chemisorption significantly at 50 °C, but also reduced the stability of adsorbed NH₃ at high temperatures. The amount of Brönsted acid sites was calculated by integrating the area of the band at 1447-1443 cm^-1 as a function of temperature. The results are shown in Fig. 4. It is clear that less vanadia in the catalysts led to weaker stability of the ammonium ions adsorbed on Brönsted acid sites. Among the fresh catalysts, the ammonium ions adsorbed on WT showed the highest thermal stability in Fig. 4(a). It is interesting that V1WT presented less thermally stable Brönsted acid sites than WT and V1T. It has been reported that as the temperature goes above 150 °C, the formation of V-O-W bonds, which causes the Brönsted acid site to be unstable, explains a sharp decrease of the surface acidity of the ternary catalyst ^[29]. Similarly, V3WT exhibited a larger amount of Brönsted acid sites at temperatures lower than 200 °C and then a similar amount to WT. As shown in Fig. 4(b), the amount of ammonium ions decreased more quickly with temperature on the poisoned catalysts. The surface acidity was quite low when the temperature increased to 250 °C. The advantage of WT for the thermal stability of the acid sites disappeared with the formation of potassium tungstate. With the addition of tungsten oxide as the promoter, V1WT-K and V3WT-K retained more Brönsted acid sites in the whole temperature range than V1T-K and V3T-K, respectively.

Fig. 4. Amount of Brönsted acid sites on (a) fresh and (b) poisoned catalysts calculated from the DRIFTS bands centered at 1447-1443 cm−1 as a function of temperature.

The redox property is one of important characteristics of SCRcatalysts. Thus the H₂-TPR tests were performed. The results are shown in Fig. 5. As shown in Fig. 5(a), the fresh WT catalyst showed two peaks at 430 and 755 °C, corresponding to the reduction of highly dispersed WO_x species and WO₃ crystallites, W⁶⁺→W⁴⁺, respectively ^[27]. The V1T catalyst showed a distinct peak at 416 °C, which was ascribed to the reduction of V⁵⁺→V^{3+ [3, 30]}. The reduction peak shifted to 256 °C with a shoulder at 303 °C for the V3T catalyst at a high vanadia loading, which can be associated with the reduction of surface isolated and polymeric VO_x species and bulk V₂O₅, respectively ^[26]. These reduction peaks were shifted towards lower temperatures (220 and 264 °C) for V3WT. Similarly, the broad peaks at 350 and 426 °C were ascribed to the reduction of vanadia and tungsten oxide over V1WT, respectively. These results confirmed that the increased vanadia loading and the promoted dispersion of vanadia by the addition of tungsten oxide facilitated the redox property of the catalyst ^{[17, 31, 32]}.

Fig. 5. H2-TPR curves of (a) fresh and (b) K-poisoned catalysts. (1) V1WT(-K); (2) V3WT(-K); (3) V1T(-K); (4) V3T(-K); (5) WT(-K).

After the potassium poisoning, the reduction peaks were shifted to 500 and 842 °C for WT-K due to the formation of finely dispersed and bulk K₂WO₄, which correlated with the reduction behavior of K₂WO₄ on the titania support ^[24]. As indicated by the ICP results in Table 1, the molar content of tungsten was two times that of potassium for WT-K. This implied that at least half the tungsten oxide can be maintained as evidenced by the preservation of the peak at 733 °C associated with the reduction of W⁶⁺→W⁴⁺ in WO₃ crystallites. For V1T-K, the distinct reduction peak shift from 416 to 494 °C was due to the formation of KVO₃ ^[13]. The position of the reduction peak of KVO₃ depended not only on the crystallinity of the salt but also on its interaction with the support. The peaks at 380 and 442 °C over V3T-K can be ascribed to the reduction of bulk V₂O₅ and vanadate KVO₃, respectively.Compared with the fresh sample, the shift of the reduction peak of bulk V₂O₅ on the poisoned catalyst to higher temperatures may be caused by oxide sintering and the weakened interaction with the support. Again, this implied the preservation of some vanadia on the poisoned catalyst at a high vanadia loading. The measured K/V ratio (0.73) was less than 1 for V3T-K. The peaks at 510-525, 740 and 820-830 °C for V1WT-K and V3WT-K can be ascribed to the reduction of highly dispersed K₂WO₄, bulk WO₃ and bulk K₂WO₄, respectively. This indicated that tungsten oxide was also combined with potassium besides the formation of KVO₃ over the ternary catalysts. The peaks at 420 and 472 °C for V1WT-K and V3WT-K were likely caused by the overlapping of the reduction of bulk vanadia and potassium vanadate.The reduction peak at a lower temperature (420 °C) demonstrated that there were more bulk vanadia species on V3WT-K than on V1WT-K.

Although the calcination temperature of the poisoning treatment was 450 °C, the volatilization of potassium cannot be completely avoided. This depended on the ease of the generation of the potassium salts and their stability. It is seen in Table 1 that almost no loss of potassium occurred on the catalysts at a high vanadia loading, implying a strong interaction between potassium and vanadia to form potassium vanadate during the poisoning treatment. On the other hand, the loss of potassium was small on the catalysts with a low vanadia loading or without vanadia, indicating that potassium also reacted with tungsten oxide strongly to form potassium tungstate, which was evidenced by the Raman and H₂-TPR results.

During the potassium poisoning treatment, tungsten oxide plays an important role in the preservation of vanadia by the formation of potassium tungstate instead of vanadate on the V1WT-K and V3WT-K catalysts. Again, both the Raman and H₂-TPR results confirmed the co-existence of KVO₃ and K₂VO₄ on the ternary catalysts. In this way, more vanadia species remained as active sites and acid sites on V3WT-K, resulting in a relatively high SCR activity of the poisoned catalyst as shown in Scheme 1.

Scheme 1. Structure of the K-poisoned V2O5-WO3/TiO2 SCR catalyst.

The NH₃-SCR activity of the catalyst has been reported to depend on both the redox property and surface acidity. In this work, the NO_x conversion at 250 °C represented the SCR activity of catalyst. The redox property of catalyst was expressed in the form of the first reduction temperature in the H₂-TPR curve. The surface acidity of catalyst was shown by the integrated area of the Brönsted acid site band at 1447-1443 cm^-1 at 250 °C. In this way, the effects of the surface acidity and redox property on the NH₃-SCR activity of catalysts are shown in Fig. 6. It is seen in Fig. 6(a) that for the fresh catalysts with plenty of surface acidity, the SCR activity was determined by the redox property of catalyst although the NO₂-assisted fast SCR reaction at low temperature was not important for the V₂O₅-WO₃/TiO₂ catalysts due to their low activity for NO oxidation. It has been reported that the existence of Brönsted acidity is not a requisite for the SCR activity at low temperatures ^{[3, 33]}. This explains the low SCR activity of the WO₃/TiO₂ catalyst with high Brönsted acidity but poor redox property. The SCR reaction is a redox process that occurs with a redox Mars-van Krevelen mechanism over vanadia-based catalysts ^{[3, 25]}. The slow step of the SCR reaction can be associated with the catalyst oxidation step by oxygen. The increase in catalyst activity observed in Fig. 6(a) by either increasing the V or W loading is ascribable to the increased redox property of the catalyst.

Fig. 6. Effects of surface acidity and redox property on NH3-SCR activity at 250 °C. (a) Fresh catalysts; (b) Poisoned catalysts. The columns represent the relative surface acidity calculated from the IR band of NH3 adsorption centered at 1447-1443 cm−1.

It was shown in Fig. 6(a) that when NH₃ existed at high coverage, there was a weak dependence of the SCR activity on the NH₃ surface concentration. However, when NH₃ coverage was below a threshold value, which was influenced by the redox property of the catalyst, the rate of the deNO_x reaction depended on the NH₃ surface coverage ^[34]. This was exactly the case with the poisoned catalysts as shown in Fig. 6(b). It can be seen that the SCR activity depended not only on the activity of the active sites but also on the surface acidity for these poisoned catalysts. The relatively lower surface acidity of V3T-K was an important factor responsible for its lower SCR activity than V3WT-K, although it has a relatively better redox property. It is worth noting that the broad peak centered at 420 °C for V3WT-K in Fig. 5(b) was caused by the overlapping of the reduction of both vanadia and vanadate, while that at 380 °C for V3T-K was assigned to the reduction of bulk vanadia. Furthermore, there were more active vanadia species that survived on V3WT-K than on V3T-K according to the Raman spectra in Fig. 2. Thus, the first H₂-TPR peak temperature may not be a good indicator of the redox property of the catalyst. On the other hand, the WT-K catalyst was almost totally deactivated with no vanadia as active sites although it maintained some amount of surface acid sites.

Based on the similar catalyst surface area and potassium loading, after potassium poisoning, the survival of vanadia active sites instead of forming inert vanadate and tungstate is critical to the NH₃-SCR activity of V₂O₅-WO₃/TiO₂ catalysts. Increasing the vanadia loading and adding tungsten oxide contributed to the redox property and surface acidity of the catalyst. With sufficient surface acidity, the reactivity of the vanadia species decides the catalytic activity. After potassium impregnation, tungsten oxide acted not only to disperse vanadia but also as a sacrificial agent so that the V₂O₅-WO₃/TiO₂ catalysts maintained their active vanadia species. Among the catalysts investigated, the V3WT catalyst showed higher alkali resistance with 60% NO_x conversion at 250 °C after impregnation of a nominal mass amount of 1% K₂O.