A Comparative Study in Vanadium and Tungsten Leaching from Various Sources of SCR Catalysts with Local Di ﬀ erence

: Direct leaching with NaOH can be an economically acceptable method for vanadium (V) and tungsten (W) recovery from spent selective catalytic reduction (SCR) catalysts. However, di ﬀ erent chemical-physical characteristics of catalysts would a ﬀ ect the V and W leaching. In this paper, the V and W leaching behavior of various sources of SCR catalysts with a local di ﬀ erence (yellow and gray color) were systematically investigated with alkali leaching solution under ambient pressure. Di ﬀ erent leaching e ﬃ ciencies from yellow and gray color areas were correlated with oxidation states and species of V and W on catalyst surfaces, as characterized by X-ray photoelectron spectroscopy (XPS), Raman, Fourier transform infrared spectroscopy (FTIR), and other analytic methods. For the V leaching e ﬃ ciency, the samples from a gray area of catalysts (40.0–51.0%) were lower than that from the yellow area (66.8–69.8%). The higher molar ratio of V 3 + and a lower molar ratio of V 5 + , and the lower total V content on the surface of the samples from the gray area could be the main reasons for the lower V leaching e ﬃ ciency. As for the W leaching e ﬃ ciency, the samples from the gray area (44.6–57.3%) were slightly higher than that from the yellow area (38.0–52.6%) of catalysts. The less total W content of surface species and stronger interaction among V–W–Ti of yellow area samples resulted in the lower leaching e ﬃ ciency. These di ﬀ erential leaching e ﬃ ciencies needed to be taken into consideration for recovering V and W from spent SCR catalysts.


Introduction
The selective catalytic reduction (SCR) technology is the most efficient and industrially widespread technology for reducing NO x emission from stationary sources [1,2]. For decades, as a mature catalyst with a high NO x conversion rate, a wide operating temperature window, and excellent stability, a considerable amount of V 2 O 5 -WO 3 /TiO 2 catalysts have been installed in the SCR system. Final disposal, such as landfill or recovery for other uses, is unavoidable after catalyst operating life [3,4].
As a hazardous solid waste, the landfill disposal cost is relatively high. Thus, various approaches have been developed for recovering valuable metals from spent SCR catalysts. As compared with high-temperature roasting or high-pressure leaching method, which means high energy cost and leads to a high concentration of impurities in leaching solutions, such as silicon, Na 2 TiO 3 , Na 2 Ti 3 O 7 [5][6][7], Table 1. The vanadium (V) and tungsten (W) leaching efficiency of selective catalytic reduction (SCR) catalyst by direct leaching method reported in the literature.

Ref. Reagents Temperature ( • C) Time (min) Particle size (µm) Efficiency (%)
Wu et al. [ This work focused on the different leaching behaviors of V and W between the yellow and gray color areas in various sources of SCR catalysts with local differences. The leaching efficiency of V and W from the yellow and the gray area samples were tested under the same condition based on the previous work [10]. Also, the reasons why the yellow and the gray area samples from the same one catalyst show different leaching behaviors were discussed based on the characterizations.

Sample Preparation
The commercial honeycomb catalyst was obtained from coal-fired power plants. All the selected catalysts had a gray cylindrical region through the axle center (Figure 1a). S1 and S2 were fresh catalysts; S3, S4, and S5 were spent catalysts. The solids deposited on the SCR catalysts were physically removed by compressed air (0.5 MPa), the yellow area and the gray area were carefully separated manually, and the color transition part was discarded. Then, the samples from the yellow area and the gray area were pulverized by a hammer, ground by the vibration mill (LDP-750A; Hongtaiyang, Yongkang, China), and sieved after drying them to the constant weight. Finally, the SCR catalyst samples with particle size smaller than 100 meshes were obtained. The tested samples extracted from a gray area were labeled as "G", which means gray (Figure 1b), and that from a yellow area labeled as Sustainability 2020, 12, 1499 3 of 12 "N", which means normal (Figure 1c). For example, S1G means the sample from the gray area of the S1 catalyst, and S2N means that from the yellow area of the S2 catalyst.
Sustainability 2020, 12, x FOR PEER REVIEW 3 of 12 yellow area labeled as "N", which means normal (Figure 1c). For example, S1G means the sample from the gray area of the S1 catalyst, and S2N means that from the yellow area of the S2 catalyst. Figure 1. Photos of catalysts (a); the tested particle samples extracted from the gray area (b) and yellow area (c) of the catalysts with color variation, respectively.

Leaching
A chosen condition based on the previous work [10] (temperature 100 °C, liquid/solid (L/S) 15, stirring speed 900 rpm, NaOH concentration 1.5 mol/L, leaching time 120 min) was utilized to evaluate the different leaching behaviors among all the various samples. The leaching tests were performed in a 50 mL Teflon tube immersed in an oil bath with magnetic stirring at a required temperature (±0.5 °C). After leaching, the slurry was sampled and separated by vacuum filtration, and then the clear liquid was filtered with a 0.22 μm membrane filter. The cake was washed sufficiently with deionized water by vacuum pump until the pH of the filtrate was lower than 8. Then, the cake was dried at 105 °C for 24 h.

Analysis
The concentration of V and W in the leaching solution was measured by an inductively coupled plasma optical emission spectrometer (ICP-OES, Optima 7000DV, Waltham, MA, USA). The elemental analysis of catalysts and residuals was conducted on an X-ray fluorescence spectrometer (XRF, Axios mAX, PANalytical B.V., Almelo, Holland). As considering the deviation of XRF, the V and W content of original catalysts were corrected by a wet method. A 0.1 g catalyst was added to 6 mL digestive solution composed of hydrochloric acid, nitric acid, and hydrofluoric acid with ratio 3:2:1 in a closed digestion vessel and digested in a graphite digestion instrument. Then, the solution was analyzed by ICP-OES, and the results were used to calculate the leaching efficiency by using Equation (1): where Ci is the concentration of V and W in the leaching solution; V is the volume of the leaching solution; m is the mass of the initially added spent SCR catalyst; Wi is the content of V, and W is the initially added spent SCR catalyst. Photos of catalysts (a); the tested particle samples extracted from the gray area (b) and yellow area (c) of the catalysts with color variation, respectively.

Leaching
A chosen condition based on the previous work [10] (temperature 100 • C, liquid/solid (L/S) 15, stirring speed 900 rpm, NaOH concentration 1.5 mol/L, leaching time 120 min) was utilized to evaluate the different leaching behaviors among all the various samples. The leaching tests were performed in a 50 mL Teflon tube immersed in an oil bath with magnetic stirring at a required temperature (±0.5 • C). After leaching, the slurry was sampled and separated by vacuum filtration, and then the clear liquid was filtered with a 0.22 µm membrane filter. The cake was washed sufficiently with deionized water by vacuum pump until the pH of the filtrate was lower than 8. Then, the cake was dried at 105 • C for 24 h.

Analysis
The concentration of V and W in the leaching solution was measured by an inductively coupled plasma optical emission spectrometer (ICP-OES, Optima 7000DV, Waltham, MA, USA). The elemental analysis of catalysts and residuals was conducted on an X-ray fluorescence spectrometer (XRF, Axios mAX, PANalytical B.V., Almelo, Holland). As considering the deviation of XRF, the V and W content of original catalysts were corrected by a wet method. A 0.1 g catalyst was added to 6 mL digestive solution composed of hydrochloric acid, nitric acid, and hydrofluoric acid with ratio 3:2:1 in a closed digestion vessel and digested in a graphite digestion instrument. Then, the solution was analyzed by ICP-OES, and the results were used to calculate the leaching efficiency by using Equation (1): where C i is the concentration of V and W in the leaching solution; V is the volume of the leaching solution; m is the mass of the initially added spent SCR catalyst; W i is the content of V, and W is the initially added spent SCR catalyst.

Characterization
The crystalline phase of samples was detected by X-ray diffraction (XRD, X'Pert Pro, PANalytical B.V., Almelo, Holland) at 40 kV and 40 mA with Cu Kα ray. The elemental valence was detected by X-ray photoelectron spectroscopy (XPS) analysis, which was collected on an ESCALAB 250Xi photoelectron spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) with monochromatic Al Kα radiation (E = 1486.2 eV) using the binding energy (284.8 eV) of amorphous carbon C 1s as a reference point. The further analysis of the XPS spectra of samples was conducted by "Avantage" software supplied by Thermo Fisher Scientific Corporation, and uses the "Scofield factor" as a sensitive factor. The specific surface areas and pore size distributions were determined by N 2 adsorption/desorption isotherms at 77K with Brunanuer-Emmett-Teller (BET) and Barrett-Joyner-Halenda (BJH) methods (Quantachrome Corp., Boynton Beach, FL, USA). Raman analysis was conducted by a Laser Raman spectrometer (LabRAM Aramis, Horiba Jobin Yvon, Paris, France). The Fourier transform infrared spectra (FTIR) of the catalysts were recorded on Bruker Vertex 70 infrared spectrometer (Bruker, Karlsruhe, Germany). Figure 2 shows the V leaching efficiency of yellow and gray areas from five different kinds of catalysts. The sequence of leaching efficiency of V from yellow samples was S1N (69.8%) ≈ S2N (69.7%) > S4N (68.8%) > S5N (67.7%) > S3N (66.8%), and this was similar to the result of our previous work in which the V leaching efficiency from catalyst without color difference was about 66.3%. As for gray samples, the leaching efficiency was much lower and in the following sequence: S1G (51.0%) > S2G (49.1%) ≈ S5G (49.0%) > S3G (42.1%) > S4G (40.0%). Thus, it could be concluded that the V leaching efficiency from the yellow area was higher (67-70%) than that from the gray area (40-51%).

Vanadium and Tungsten Leaching Behavior
Sustainability 2020, 12, x FOR PEER REVIEW 4 of 12

Characterization
The crystalline phase of samples was detected by X-ray diffraction (XRD, X'Pert Pro, PANalytical B.V., Almelo, Holland) at 40 kV and 40 mA with Cu Kα ray. The elemental valence was detected by X-ray photoelectron spectroscopy (XPS) analysis, which was collected on an ESCALAB 250Xi photoelectron spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) with monochromatic Al Kα radiation (E = 1486.2 eV) using the binding energy (284.8 eV) of amorphous carbon C 1s as a reference point. The further analysis of the XPS spectra of samples was conducted by "Avantage" software supplied by Thermo Fisher Scientific Corporation, and uses the "Scofield factor" as a sensitive factor. The specific surface areas and pore size distributions were determined by N2 adsorption/desorption isotherms at 77K with Brunanuer-Emmett-Teller (BET) and Barrett-Joyner-Halenda (BJH) methods (Quantachrome Corp., Boynton Beach, FL, USA). Raman analysis was conducted by a Laser Raman spectrometer (LabRAM Aramis, Horiba Jobin Yvon, Paris, France). The Fourier transform infrared spectra (FTIR) of the catalysts were recorded on Bruker Vertex 70 infrared spectrometer (Bruker, Karlsruhe, Germany). Figure 2 shows the V leaching efficiency of yellow and gray areas from five different kinds of catalysts. The sequence of leaching efficiency of V from yellow samples was S1N (69.8%) ≈ S2N (69.7%) > S4N (68.8%) > S5N (67.7%) > S3N (66.8%), and this was similar to the result of our previous work in which the V leaching efficiency from catalyst without color difference was about 66.3%. As for gray samples, the leaching efficiency was much lower and in the following sequence: S1G (51.0%) > S2G (49.1%) ≈ S5G (49.0%) > S3G (42.1%) > S4G (40.0%). Thus, it could be concluded that the V leaching efficiency from the yellow area was higher (67-70%) than that from the gray area (40-51%). The W leaching efficiency of yellow and gray areas from five different kinds of catalysts are shown in Figure 3. For yellow area samples, the sequence of leaching efficiency of W was: S4N (52.6%) > S5N (44.7%) > S1N (44.2%) > S2N (41.4%) > S3N (38.0%), and this was similar to the result of our previous work in which the W leaching efficiency from catalyst without color difference was about 48.5%. The gray area samples had similar sequence: S4G (57.3%) > S1G (54.2%) > S5G (53.2%) > Figure 2. Vanadium leaching efficiency of the yellow and the gray area samples from fresh (S1 and S2) and spent (S3, S4, and S5) catalysts.

Vanadium and Tungsten Leaching Behavior
The W leaching efficiency of yellow and gray areas from five different kinds of catalysts are shown in Figure 3. For yellow area samples, the sequence of leaching efficiency of W was: S4N (52.6%) > S5N (44.7%) > S1N (44.2%) > S2N (41.4%) > S3N (38.0%), and this was similar to the result of our previous work in which the W leaching efficiency from catalyst without color difference was about 48.5%. The gray area samples had similar sequence: S4G (57.3%) > S1G (54.2%) > S5G (53.2%) > S2G (52.0%) > S3G (44.6%). S3N showed the lowest leaching efficiency of about 38.0%, and S4G had the highest leaching efficiency (57.3%). Opposite to the leaching behaviors of V, the W leaching efficiency of the yellow area samples (38.0-52.6%) was slightly lower than that of the gray area samples (44.6-57.3%).

Chemical and Textural Characterizations
XRD was used to analyze the composition and phase structure of the ten samples. As shown in Figure 4, all the diffraction peaks could be indexed to anatase TiO2 only, and neither V2O5 nor WO3 was detected among all the samples. The absence of these bulk phases in the XRD patterns implied that V and W species were present in either non-crystalline state or lower content than the detection limit of XRD.
As seen in Table 2, the elemental composition of all the samples included TiO2, V2O5, WO3, SiO2, CaO, and Al2O3, which was totally higher than 95%. These main components of two samples from yellow and gray areas from one catalyst were similar. Moreover, there were apparent differences in V and W contents among various sources catalysts as a result of manufacture formulations or the V and W loss to varying extents after the serving time.

Chemical and Textural Characterizations
XRD was used to analyze the composition and phase structure of the ten samples. As shown in Figure 4, all the diffraction peaks could be indexed to anatase TiO 2 only, and neither V 2 O 5 nor WO 3 was detected among all the samples. The absence of these bulk phases in the XRD patterns implied that V and W species were present in either non-crystalline state or lower content than the detection limit of XRD.

Figure 3.
Tungsten leaching efficiency of the yellow and the gray area samples from fresh (S1 and S2) and spent (S3, S4, and S5) catalysts.

Chemical and Textural Characterizations
XRD was used to analyze the composition and phase structure of the ten samples. As shown in Figure 4, all the diffraction peaks could be indexed to anatase TiO2 only, and neither V2O5 nor WO3 was detected among all the samples. The absence of these bulk phases in the XRD patterns implied that V and W species were present in either non-crystalline state or lower content than the detection limit of XRD.
As seen in Table 2, the elemental composition of all the samples included TiO2, V2O5, WO3, SiO2, CaO, and Al2O3, which was totally higher than 95%. These main components of two samples from yellow and gray areas from one catalyst were similar. Moreover, there were apparent differences in V and W contents among various sources catalysts as a result of manufacture formulations or the V and W loss to varying extents after the serving time. As seen in Table 2, the elemental composition of all the samples included TiO 2 , V 2 O 5 , WO 3 , SiO 2 , CaO, and Al 2 O 3 , which was totally higher than 95%. These main components of two samples from yellow and gray areas from one catalyst were similar. Moreover, there were apparent differences in V and W contents among various sources catalysts as a result of manufacture formulations or the V and W loss to varying extents after the serving time.  Table 3 lists the textural properties of catalysts, including surface area, pore-volume, and pore diameter. There were slight differences between gray and yellow area samples among all the catalysts, and the range of specific surface area from 40.3 m 2 /g to 56.7 m 2 /g was mainly caused by different formulations for sampled catalysts. Dong et al. prepared SCR catalysts under different oxygen content atmospheres, and the color of samples changed from yellow to gray when the oxygen content decreased [19]. Also, the lower specific surface area of the yellow sample than the gray sample might be due to the sufficient oxidation of V and W.

XPS Analysis
As shown in Figure 5a, the binding energy of Ti 2P 3/2 (458.7 eV) was attributed to the typical characteristic peak of Ti 4+ ion. Chen et al. and Zhang et al. concluded that the Ti 4+ would shift to higher binding energy after adding WO 3 [20,21]. The lower binding energy of Ti 4+ (458.4 eV) was found on both S1G and S2G, indicating weak interaction between WO 3 and TiO 2 . A similar phenomenon was found on O1s ( Figure 5b) and W 4f (Figure 5c), for S1G and S2G; the O1s binding energy shifted from 530 to 529.6 eV, and the W 4f shifted to lower binding energylevels by 0.3 eV. It is well known that the existing form of V in the commercial SCR catalyst is rather intricate. It could exist as dispersed monolayer isolated vanadyl [12], polymeric surface VOx species [20,22], V species in oligomeric structure [23]. Also, V can enter into the TiO2 crystal lattice due to the very close ionic radius of V 4+ to that of Ti 4+ [24].
XPS was used specifically to analyze the surface compositions and oxidation states of all the catalysts. The V 2p3/2 spectra were resolved into three peaks [25] (Figure 6), and the molar ratios of V 5+ , V 4+ , and V 3+ are shown in Table 4. The bulk (measured by XRF and ICP) and the surface (performed by XPS) molar compositions of all the samples are listed in Table 4. As shown in Table 4, the bulk amounts of V and W were similar between gray and yellow area samples, and lower than surface amounts. It implied that the V and W species migrated from internal onto the surface during the calcine or reaction phase [11,20].
For V species, the surface amounts of gray samples were lower than that of yellow samples for all catalysts. Nevertheless, for W species, that was the opposite. According to the well-known shrinking core models [9,26], the higher surface amounts of V on the yellow samples and W on the gray samples could be one of the reasons that the yellow samples showed higher V leaching efficiency but lower W leaching efficiency. It is well known that the existing form of V in the commercial SCR catalyst is rather intricate. It could exist as dispersed monolayer isolated vanadyl [12], polymeric surface VO x species [20,22], V species in oligomeric structure [23]. Also, V can enter into the TiO 2 crystal lattice due to the very close ionic radius of V 4+ to that of Ti 4+ [24].
XPS was used specifically to analyze the surface compositions and oxidation states of all the catalysts. The V 2p 3/2 spectra were resolved into three peaks [25] (Figure 6), and the molar ratios of V 5+ , V 4+ , and V 3+ are shown in Table 4. The bulk (measured by XRF and ICP) and the surface (performed by XPS) molar compositions of all the samples are listed in Table 4. As shown in Table 4, the bulk amounts of V and W were similar between gray and yellow area samples, and lower than surface amounts. It implied that the V and W species migrated from internal onto the surface during the calcine or reaction phase [11,20].
For V species, the surface amounts of gray samples were lower than that of yellow samples for all catalysts. Nevertheless, for W species, that was the opposite. According to the well-known shrinking core models [9,26], the higher surface amounts of V on the yellow samples and W on the gray samples could be one of the reasons that the yellow samples showed higher V leaching efficiency but lower W leaching efficiency.
For all catalysts, the ratios of the V 3+ valence state of the gray area samples were higher than that of yellow area samples, and the ratios of the V 5+ valence state of the gray area samples were lower than that of yellow samples. It is known that the colors of vanadium oxides vary with the change in the valence of V. Yaws et al. [27] mentioned that V 2 O 3 is black color, VO 2 is blue-black color, and V 2 O 5 shows yellow-brown color. As shown in Figure 1, all the selected catalysts had locally different colors, and the samples from gray areas contained higher V 3+ or V 4+ species (V 2 O 3 and VO 2 ) than that from yellow areas; this might be caused by inadequate oxygen supply, resulting in insufficient oxidation of V in the gray area during the sintering process [11,19]. As compared with amphiprotic oxide VO 2 Sustainability 2020, 12, 1499 8 of 12 and V 2 O 5 , V 2 O 3 is hardly solved in alkaline or acid, and then additional oxidation is usually used to enhance vanadium leaching [18,[28][29][30][31]. Thus, the higher molar ratios of V 3+ and lower molar ratios of V 5+ valence could be another reason for the lower V leaching efficiency of gray samples.  For all catalysts, the ratios of the V 3+ valence state of the gray area samples were higher than that of yellow area samples, and the ratios of the V 5+ valence state of the gray area samples were lower than that of yellow samples. It is known that the colors of vanadium oxides vary with the change in the valence of V. Yaws et al. [27] mentioned that V2O3 is black color, VO2 is blue-black color, and V2O5 shows yellow-brown color. As shown in Figure 1, all the selected catalysts had locally different colors, and the samples from gray areas contained higher V 3+ or V 4+ species (V2O3 and VO2) than that from yellow areas; this might be caused by inadequate oxygen supply, resulting in insufficient oxidation of V in the gray area during the sintering process [11,19]. As compared with amphiprotic oxide VO2 and V2O5, V2O3 is hardly solved in alkaline or acid, and then additional oxidation is usually used to enhance vanadium leaching [18,[28][29][30][31]. Thus, the higher molar ratios of V 3+ and lower molar ratios of V 5+ valence could be another reason for the lower V leaching efficiency of gray samples.  Also, the XPS spectra results of Ti 2p (Figure 7a) and O 1s (Figure 7b) before and after leaching are displayed. Except for S1G and S2G with lower Ti2p and O1s binding energy before leaching, other catalysts obviously shifted to lower binding energy after leaching. Ti 2p binding energy shifted from 458.7 to 458.4 eV, and O 1s binding energy shifted from 530.0 to 529.6 eV, indicating that the surface W species was leached. This was consistent with the phenomenon found by Chen et al. and Zhang et al. that the modification with WO 3 results in the Ti 4+ shift to higher binding energy [20,21].

Raman Analysis
As shown in Figure 8a, the vibrations at 196 cm −1 , 397 cm −1 , 518 cm −1 , and 636 cm −1 were fingerprint peaks of anatase TiO2. To get more details, the Raman active modes in the range of 750~1200 cm −1 were also investigated and shown in Figure 8b. All the catalysts showed a peak at 798 cm −1 , which could be assigned to the Ti-O vibration of TiO2 or crystal tungsten oxide [19,20]. The broadband between 925 cm −1 and 1000 cm −1 could be indexed to the overlap of contributions from polymeric vanadate species and both tetrahedrally and octahedrally coordinated polymeric WOx [14,19,32]. In the FTIR spectra (Figure 9), all the samples had a significant vibration band at 1050 cm −1 , which was related to W=O stretches [17]. It also could be found that there was a peak at 1016 cm −1 in the samples with high sulfur content according to XRF and XPS, which might be assigned to the vibration of sulfate [14], and this was consistent with FTIR results showing that there was a vibration band at 1121 cm −1 (Figure 9) [17]. As compared with gray samples, the peak between 925 and 1000 cm −1 of yellow samples was broader and higher, meaning that there is much stronger interaction among W, V, and Ti species [20]. The stronger interaction between W species and TiO2 anatase lattice could result in lower W leaching efficiency [5]. Thus, the stronger interaction among V-W-Ti and the less content of surface W species of yellow area samples might result in their lower W leaching efficiency.

Raman Analysis
As shown in Figure 8a, the vibrations at 196 cm −1 , 397 cm −1 , 518 cm −1 , and 636 cm −1 were fingerprint peaks of anatase TiO 2 . To get more details, the Raman active modes in the range of 750~1200 cm −1 were also investigated and shown in Figure 8b. All the catalysts showed a peak at 798 cm −1 , which could be assigned to the Ti-O vibration of TiO 2 or crystal tungsten oxide [19,20]. The broadband between 925 cm −1 and 1000 cm −1 could be indexed to the overlap of contributions from polymeric vanadate species and both tetrahedrally and octahedrally coordinated polymeric WO x [14,19,32]. In the FTIR spectra (Figure 9), all the samples had a significant vibration band at 1050 cm −1 , which was related to W=O stretches [17]. It also could be found that there was a peak at 1016 cm −1 in the samples with high sulfur content according to XRF and XPS, which might be assigned to the vibration of sulfate [14], and this was consistent with FTIR results showing that there was a vibration band at 1121 cm −1 (Figure 9) [17]. As compared with gray samples, the peak between 925 and 1000 cm −1 of yellow samples was broader and higher, meaning that there is much stronger interaction among W, V, and Ti species [20]. The stronger interaction between W species and TiO 2 anatase lattice could result in lower W leaching efficiency [5]. Thus, the stronger interaction among V-W-Ti and the less content of surface W species of yellow area samples might result in their lower W leaching efficiency.

Raman Analysis
As shown in Figure 8a, the vibrations at 196 cm −1 , 397 cm −1 , 518 cm −1 , and 636 cm −1 were fingerprint peaks of anatase TiO2. To get more details, the Raman active modes in the range of 750~1200 cm −1 were also investigated and shown in Figure 8b. All the catalysts showed a peak at 798 cm −1 , which could be assigned to the Ti-O vibration of TiO2 or crystal tungsten oxide [19,20]. The broadband between 925 cm −1 and 1000 cm −1 could be indexed to the overlap of contributions from polymeric vanadate species and both tetrahedrally and octahedrally coordinated polymeric WOx [14,19,32]. In the FTIR spectra (Figure 9), all the samples had a significant vibration band at 1050 cm −1 , which was related to W=O stretches [17]. It also could be found that there was a peak at 1016 cm −1 in the samples with high sulfur content according to XRF and XPS, which might be assigned to the vibration of sulfate [14], and this was consistent with FTIR results showing that there was a vibration band at 1121 cm −1 (Figure 9) [17]. As compared with gray samples, the peak between 925 and 1000 cm −1 of yellow samples was broader and higher, meaning that there is much stronger interaction among W, V, and Ti species [20]. The stronger interaction between W species and TiO2 anatase lattice could result in lower W leaching efficiency [5]. Thus, the stronger interaction among V-W-Ti and the less content of surface W species of yellow area samples might result in their lower W leaching efficiency.

Conclusions
The leaching behaviors of V and W for five catalysts with the local color difference were systematically investigated under a specific condition (100 °C, NaOH concentration 1.5 mol/L, L/S 15, stirring speed 900 rpm, time 240 min, and atmosphere pressure). For V, the leaching efficiency of samples from the gray part (40.0-51.0%) was lower than that from the yellow part (66.8-69.8%). On the contrary, the W leaching efficiency of samples from the gray part (44.6-57.3%) was higher than that from the yellow part (38.0-52.6%). According to surface characterization, the higher molar ratio of V 3+ and a lower molar ratio of V 5+ valence, and the lower total V content on the surface of samples could be the main reasons for that gray area samples to had lower V leaching efficiency than yellow area samples. For W leaching, as compared with the gray area samples, the stronger interaction among V-W-Ti and the less W content of surface species of yellow area samples might result in lower leaching efficiency. Therefore, the spent catalysts with local color differences require special attention when they are leached to recycle V and W. The color sorting technology or developing efficient leaching method might be considered based on the economy.

Conclusions
The leaching behaviors of V and W for five catalysts with the local color difference were systematically investigated under a specific condition (100 • C, NaOH concentration 1.5 mol/L, L/S 15, stirring speed 900 rpm, time 240 min, and atmosphere pressure). For V, the leaching efficiency of samples from the gray part (40.0-51.0%) was lower than that from the yellow part (66.8-69.8%). On the contrary, the W leaching efficiency of samples from the gray part (44.6-57.3%) was higher than that from the yellow part (38.0-52.6%). According to surface characterization, the higher molar ratio of V 3+ and a lower molar ratio of V 5+ valence, and the lower total V content on the surface of samples could be the main reasons for that gray area samples to had lower V leaching efficiency than yellow area samples. For W leaching, as compared with the gray area samples, the stronger interaction among V-W-Ti and the less W content of surface species of yellow area samples might result in lower leaching efficiency. Therefore, the spent catalysts with local color differences require special attention when they are leached to recycle V and W. The color sorting technology or developing efficient leaching method might be considered based on the economy.