## Introduction

Emission control of NOx (NO and NO2) is an important task for an industrial chemical process that relies on thermal energy produced by the combustion of fossil fuels. For stationary deNOx systems such as those in coal-fired power plants and waste treatment plants, selective catalytic reduction (SCR) has been used with ammonia as a reducing reagent to convert harmful NOx to harmless N2 and H2O1,2,3:

$$4{\mathrm{NO}} + 4{\mathrm{NH}}_3 + {\mathrm{O}}_2 \to 4{\mathrm{N}}_2 + 6{\mathrm{H}}_2{\mathrm{O}}.$$
(1)

Although the stationary deNOx systems have been successfully established, a conventional catalyst needs a high temperature (>300 °C) for NH3-SCR to proceed3,4. In most cases, deNOx catalysts are placed just after the boiler system for a high reaction temperature4. Consequently, deNOx catalysts are deactivated by ash and sulfate generated from the reaction of ammonia and SO2 because the catalysts are exposed to the gas before passing through the dust collection and deSOx system. DeNOx catalysts can be moved to the latter part of the system to avoid the deactivation of the deNOx catalyst works at a low temperature (<150 °C) since the downstream gas is usually cooled down to 100–150 °C.

Vanadium (V) oxide-based catalysts (V2O5/TiO2, V2O5–WO3/TiO2) have been used as industrial catalysts for stationary boiler systems because they show high N2 selectivity, good thermal stability, and low SO2 oxidation activity to SO32. However, their high working temperature limits the latitude of the deNOx process. Therefore, many studies have been carried out to develop low-temperature SCR catalysts such as metal oxide-based materials (V, Mn, Cr, Cu, W, Ce, and Fe)1,5,6,7,8,9,10 and ion-exchanged zeolite (Cu-ZSM-5, Fe-ZSM-5, Cu-CHA, and Cu-SSZ-13)1,11,12,13,14,15,16,17.

We previously showed that vanadium oxide with a bulk crystal structure works as an NH3-SCR catalyst at a low temperature (<150 °C) with high N2 selectivity and less SO2 oxidation activity in contrast to the supported vanadia species18,19. Recent studies provided an insight into the effect of the coordination environment, atomic configuration, and vanadium surface density of the catalyst on SCR activity20,21,22. It was proposed in previous reports that supported vanadia catalysts are composed of VO4 units with O-exposed moieties2,23. On the other hand, bulk V2O5 has a metal (V)-exposed part in addition to the O-exposed part24, and it might show a different catalytic property from that of a supported vanadia catalyst because of their different redox properties, bond strengths, and coordination environments23,25. Bulk crystal of V2O5 is composed of a two-dimensional sheet structure with VO5 units that are connected by a weak van der Waals force24,26,27,28,29. Consequently, the morphology of V2O5 is thermally reconstructed because of its weak interaction in the crystal structure. According to previous studies, the poor stability of V2O5 can be overcome by doping another metal with a three-dimensional coordination environment30,31,32. Bulk tungsten oxide (WO3) is composed of WO6 units that are three-dimensionally connected33,34,35. Furthermore, WO3 is known as a promoter for supported vanadia catalysts that enhance the reactivity of vanadia sites36. Therefore, the incorporation of tungsten sites into bulk V2O5 would facilitate not only the NH3-SCR activity but also the structural stability of bulk vanadium oxide-based catalysts.

The negative effect of water, which is intrinsically included in the exhaust gas (10–30 vol%)2, is critical for low-temperature NH3-SCR, while the effect would be much less in a high-temperature range. The effect of water is thought to be inhibition of the adsorption of the reactant37,38,39,40,41,42,43 and/or the reaction between NO and adsorbed NH344,45,46. The water adsorption behavior was studied for a V2O5–WO3 solid solution system and V2O5 without tungsten47,48,49. Broclawik et al.48 theoretically suggested that dissociative adsorption of water occurs on adjacent V–O–W sites, leading to an increase in the concentration of Brønsted acid sites. On the other hand, water merely adsorbs in the case of V2O5 without tungsten, and a Brønsted acid site is not newly created. Thus, NH3-SCR would efficiently proceed with a vanadium–tungsten complex oxide even in the presence of water vapor owing to the large population of protonic Brønsted acid sites.

The reaction mechanisms of NH3-SCR have been widely investigated for a supported vanadium-based oxide catalyst under a dry gas condition using in situ and operando spectroscopic techniques (infrared (IR), ultraviolet–visible (UV–Vis), Raman)20,36,38,46,50,51,52,53,54. The current consensus is that both Lewis acid and Brønsted acid sites participate in the reaction mechanism. The reaction mechanism should be investigated in the presence of water as well as under dry conditions in order to understand catalytic activity in the actual atmosphere. Although Topsøe et al.45 and Song et al. 55 investigated the reaction mechanism of 3–6 wt% V2O5/TiO2 under a 1.7–3.0 vol% water atmosphere at 250–300 °C, the reaction mechanism of the vanadium–WO3 system has not been investigated under wet conditions, especially in a low-temperature range. An insight into the reaction mechanism under wet conditions would provide a milestone for designing the active site of a low-temperature NH3-SCR catalyst under actual conditions.

Herein, we report a bulk tungsten-substituted vanadium oxide catalyst for low-temperature NH3-SCR (<150 °C) under a wet atmosphere. The structural features of bulk tungsten-substituted vanadium oxide are studied. We examine the effect of tungsten substitution on water tolerance to NH3-SCR activity and catalytic stability. The reaction mechanism is investigated by operando spectroscopies under dry and wet conditions to elucidate the role of vanadium and tungsten sites in the catalytic cycle. Finally, we compare the difference between bulk tungsten-substituted vanadium oxide catalyst and a current titania-supported vanadia catalyst.

## Results

### Synthesis of tungsten-substituted vanadium oxide

Tungsten-substituted vanadium oxide was synthesized by the oxalate method using ammonia metavanadate, ammonia metatungstate, and oxalic acid. Water-insoluble ammonia metavanadate was converted to soluble vanadium oxalate in aqueous media, and ammonia metatungstate was then added to the aqueous solution followed by evaporation. The resulting precursor powder was then calcined in air. The synthesized x mol% tungsten-substituted vanadium oxides were denoted as xW–V (x = 0, 1, 3.5, 7, 10, 15, and 40). The precursor powder of the 3.5W–V sample showed a strong exothermic peak assigned to its decomposition at around 297 °C from thermogravimetry/differential thermal analysis measurement (Supplementary Fig. 2). Thus, we calcined the catalyst twice at 300 °C for 4 h each time to obtain V2O5-based catalysts with large surface areas. The specific surface areas of the xW–V (x = 0–40) catalysts were measured to be 32–41 m2 g−1 (Supplementary Table 2), which are almost in the same range and comparably large.

### Characterization of tungsten-substituted vanadium oxide

Atomic-resolution high-angle annular dark-field imaging-scanning transmission electron microscopy (HAADF-STEM) images were measured to directly determine the incorporation of tungsten into vanadium oxide. The image of 0W–V (V2O5 without tungsten) showed a lattice fringe based on the (101) plane (Fig. 1a). No bright spots of tungsten atoms were observed for 0W–V. On the other hand, we confirmed atomic bright spots on the vanadium oxide lattice for 3.5W–V (Fig. 1b), indicating that tungsten atoms were atomically dispersed. An enlarged view of the HAADF-STEM image for 3.5W–V (Fig. 1d) showed that lattice vanadium atoms were partially substituted by tungsten atoms (Fig. 1e). Aggregated tungsten sites were observed on the surfaces of catalysts when 7 mol% tungsten was added to vanadium oxide (7W–V; Supplementary Fig. 3). Adjacent and aggregated tungsten moieties were more clearly observed when 40 mol% tungsten was added to vanadium oxide (40W–V; Fig. 1c), indicating that a WO3 phase partially forms with an increase in the amount of tungsten. Thus, 0, 3.5, and 40W–V can be regarded as catalysts without tungsten, with a moderate amount of tungsten and with an excess amount of tungsten, respectively.

The crystal structures of xW–V were confirmed from X-ray powder diffraction (XRD) measurements to determine the effect of tungsten doping (Fig. 2a). For 0, 1, and 3.5W–V, all of the diffraction peaks were assigned as vanadium oxide phase (Fig. 2a, b), and no WO3 crystalline phase was confirmed. Given that the HAADF-STEM images showed tungsten dispersed atomically, the XRD patterns suggested that tungsten atoms were atomically incorporated into the vanadium oxide lattice and a W–V oxide solid solution was formed. We confirmed broad XRD peaks of WO3 (Fig. 2a, c) at 2θ = 23° for 7, 10, 15, and 40W–V in addition to the XRD patterns of vanadium oxide. These results indicate that tiny WO3 crystalline particles form when the tungsten content is higher than 7 mol% as shown by HAADF-STEM images. The peak positions of (010) reflection at around 2θ = 20.2° slightly increased with an increase in the amount of tungsten (Fig. 2a, inset), indicating that the lattice spacing decreased along the b-axis. The lattice parameters were calculated by Rietveld analysis. The lattice parameters of the b-axis decreased with an increase in the amount of tungsten (Supplementary Fig. 4b). These results show that WO6 units connect the vanadium oxide layers by bond formation (e.g., V = O → V–O–W). The lattice parameters along the a-axis and c-axis were expanded because of the incorporation of tungsten atoms with a larger ionic radius (Supplementary Fig. 4a, c).

We also conducted IR measurements for 0, 3.5, and 40W–V to determine the effect of tungsten substitution on the structure (Supplementary Fig. 5). Although an obvious difference in the infrared (IR) spectra was not found for 3.5 and 0W–V, the IR peak of 40W–V assigned to V = O (1020 cm−1) became weak compared to those of 0 and 3.5W–V, results similar to those reported by Satsuma et al.56. The number of V = O bonds would decrease to form V–O–W bonds by the addition of tungsten into the vanadium oxide lattice.

### NH3-SCR activity of tungsten-substituted vanadium oxide

First, we investigated the effects of different amounts of tungsten on the NH3-SCR activity of xW–V catalysts under dry and wet (10 vol% water) atmospheres at 150 °C (Fig. 3a). The 0W–V catalyst (V2O5 without tungsten) showed NO conversions of 82% (dry) and 47% (wet). The NO conversion increased with an increase in the amount of tungsten up to 3.5 mol%. The NO conversions of 3.5W–V were >99% (dry) and 94% (wet). The negative effect of water was suppressed by the incorporation of tungsten into vanadium oxide. The 3.5W–V catalyst showed the best NH3-SCR activity among the synthesized catalysts, while the NO conversion decreased with a further increase in the amount of tungsten. The 40W–V catalyst showed NO conversions of 29% (dry) and 19% (wet). Thus, an excess amount of tungsten decreased the NH3-SCR activity.

The dependence of NO conversion on reaction temperature was examined for 0W–V, 3.5W–V, and 1 wt% V2O5–5 wt% WO3/TiO2 (V–W/TiO2, a model of an industrial catalyst; Fig. 3b). The 3.5W–V catalyst showed NH3-SCR activity at a low temperature (<150 °C) under both dry and wet conditions in which the conventional V–W/TiO2 catalyst did not show sufficient NH3-SCR activity. The 0W–V catalyst (V2O5 without tungsten) also showed low-temperature NH3-SCR activity under a dry condition, but the negative effect of water was critical at a low temperature. The temperatures for 80% NO conversion (T80) were 110 °C (3.5W–V), 157 °C (0W–V), and 225 °C (V–W/TiO2) under a dry condition and 135 °C (3.5W–V), 178 °C (0W–V), and 236 °C (V–W/TiO2) under a wet (10 vol% water) condition. N2 selectivity was >99% during the reaction, and undesirable N2O production was not observed for any of the catalysts (Supplementary Fig. 6).

The effects of water concentration in flow gas on NH3-SCR were also investigated for 0W–V and 3.5W–V at 150 °C (Fig. 4a). Although NO conversion of 3.5W–V was slightly decreased by the addition of 5% water into reaction gas, ~93% NO conversion was maintained in the presence of 5–20 vol% of water. On the other hand, NO conversion was drastically decreased from 82% to 35% in the case of 0W–V when water vapor was introduced to the reaction gas. Although NO conversion decreased with the addition of water, the values were almost the same regardless of the concentration of water (2–20%). Previous studies showed that a further inhibitory effect of water on the reaction rate did not occur at a high water concentration (>5%)2,39,57. Water vapor would affect a catalytic cycle more strongly at a low temperature such as 150 °C. Thus, we found that tungsten-doped vanadium oxide (3.5W–V) shows a low-temperature NH3-SCR activity in the presence of a high concentration of water.

A catalytic stability test (150 °C) was conducted for 0W–V and 3.5W–V to determine the effect of tungsten on the stability (Fig. 4b). In the presence of 10 vol% water, the NO conversion was decreased for both 0W–V and 3.5W–V. After turning off the water addition, the NO conversion of 3.5W–V recovered to the original value (>99%), but that of 0W–V decreased from 78 to 59%. The specific surface areas of 0 and 3.5W–V were measured before and after the catalytic stability test (Fig. 4c). The specific surface area of 0W–V decreased after the stability test (41 → 19 m2 g−1), while 3.5W–V had the same value (39 m2 g−1). Although a porous structure was confirmed by scanning electron microscope measurement for 0W–V before the activity test, it disappeared and the surface became smooth after the test (Supplementary Fig. 7). Obvious morphological changes were not confirmed for 3.5W–V. There were no changes in the XRD patterns of 0 and 3.5W–V after the stability test (Supplementary Fig. 8). We assume that the bulk tungsten sites, which have a three-dimensional coordination environment (WO6 units), retain the structure by incorporation into the layered vanadium oxide lattice, leading to the retention of surface area.

### Reaction kinetics of tungsten-substituted vanadium oxide catalysts

Reaction rate, apparent activation energy (Ea), and reaction order were investigated to understand the reaction kinetics for low-temperature NH3-SCR. Reaction rates for the kinetics were calculated by adjusting the weights of the catalysts to control the conversion to <20%. Reaction rates per surface area were 3.2 × 10−9 (0W–V), 5.2 × 10−9 (3.5W–V), and 0.2 × 10−9 molNO m−2 s−1 (V–W/TiO2) in a dry condition (Fig. 5a and Supplementary Table 3), indicating that the reaction site of 3.5W–V was favorable for NH3-SCR to proceed compared to the reaction sites of 0W–V and V–W/TiO2. The reaction rate of 0W–V considerably decreased to 1.4 × 10−9 molNO m−2 s−1 in a wet condition and that of 3.5W–V was 4.2 × 10−9 molNO m−2 s−1. Thus, the inhibitory effect of water on the reaction was less for the tungsten-substituted vanadium oxide catalyst.

The apparent activation energy (Ea) values calculated from Arrhenius plots were 39 kJ  mol−1 (0W–V) and 36 kJ mol−1 (3.5W–V) under a dry condition (Fig. 5b and Supplementary Table 3), indicating that the reaction mechanisms were similar under a dry condition. Ea of 3.5W–V under a wet condition was 40 kJ mol−1, which was close to that under a dry condition, but that of 0W–V was 22 kJ mol−1 and preexponential factor (intercept of the line) decreased in the presence of water vapor. The change in Ea of 0W–V reflects the different reaction mechanism of vanadium oxide without tungsten. Although the Ea of 0W–V became smaller, the number of active sites (preexponential factor) would decrease because of its blockage by water, leading to a decrease in the activity. Ea values of V–W/TiO2 were 43 (dry) and 49 kJ mol−1 (wet). Therefore, the effect of tungsten on the NH3-SCR cycle would be similar to that in the bulk V–W oxide system.

The NH3-SCR of a vanadium-based catalyst proceeds as follows: (1) adsorption of NH3 on an acid site, (2) reaction of NO with NH3 to produce N2 + H2O via a nitrosamide (NH2NO*) intermediate with simultaneous reduction of the vanadia surface, and (3) re-oxidation of the partially reduced vanadia surface by oxygen molecules1,52,53. Therefore, we measured the reaction orders for NH3, NO, and O2 at 150 °C (Supplementary Fig. 9 and Supplementary Table 3). Ranges of gas concentration were 125–500 ppm for NH3, 125–500 ppm for NO, and 2–8% for O2. Reaction orders of 0W–V and 3.5W–V were 0.2–0.3 (NH3), 0.9–1.1 (NO), and 0.3–0.4 (O2) under both dry and wet atmospheres. The small reaction order for NH3 indicates that the adsorption of NH3 to the catalyst surface is strong. The reaction order for NO was high compared to those for the other substrates, indicating that the reaction rate depended on NO concentration in actual conditions. The reaction order for O2 was 0.3–0.4 at 150 °C, although zero-order dependence was reported at O2 concentrations >1% at a higher temperature (>250 °C) for a supported vanadia catalyst2,38,53. Oxygen molecules contribute to the re-oxidation step of the partially reduced vanadia surface (V4+ → V5+) in the reaction mechanism of NH3-SCR. However, the re-oxidation rate can be slower for a low temperature and the surface of the catalyst is not fully oxidized in a low-temperature range2,38,53,54.

### Roles of tungsten substitution in the NH3-SCR cycle

To investigate the redox properties of the catalysts, we also conducted operando UV–Vis measurements (Fig. 7b and Supplementary Figs. 11 and 12). We monitored the change in the pseudo-absorbance (Kubelka–Munk unit) of UV–Vis spectra at λ = 700 nm (ΔKM700), which is assigned to the dd transition of V4+, to observe the change in the valence of vanadium sites (V5+ ↔ V4+). The production of N2 was recorded simultaneously by an MS. The monitoring was performed under NO + NH3 flowing followed by O2 flowing to evaluate the respective reduction and oxidation half-cycles. The correlations between W content, amount of V5+ reduced by NO + NH3 (ΔKM700 under NO + NH3 flowing), and MS intensity for N2 are shown in Fig. 7b. The 3.5W–V showed the largest ΔKM700 and MS intensity of N2 and had superior reduction capability for the production of N2. On the other hand, ΔKM700 considerably decreased when 40 mol% of tungsten was doped, indicating that the catalyst with an excess amount of tungsten was difficult to be reduced. The 3.5W–V also showed the largest ΔKM700 under O2 flowing (Supplementary Figs. 11b, d, f and 12b). Thus, 3.5W–V with moderate tungsten loading turned out to be the most redox-active catalyst, leading to the largest N2 production. Combining with the results of operando IR measurements, tungsten sites serve as acid sites for the adsorption of ammonia, which is stronger than those pre-existing in vanadium oxide without tungsten, and the vanadium (V5+) sites work as redox sites for NH3-SCR. The highest activity for 3.5W–V can be ascribed to the increase in the number of acid sites and the high redox capability of the surface V species. W–O–V units would be needed for the progression of the reaction, and the number of adjacent tungsten sites such as W–O–W units becomes dominant with an increase in the amount of tungsten, leading to blocking of the surface redox-active V sites and thus resulting in a decrease in activity.

### Effects of water on the NH3-SCR cycle

The activity test clearly showed a positive effect of tungsten substitution in the presence of water. We also conducted operando IR measurements under a wet condition (2 vol% water) to check the effect of water on the acid sites for 0W–V and 3.5W–V. Although the IR peaks of adsorbed ammonia species on Lewis and Brønsted acid sites (NH3 and NH4+, respectively) were both observed under a dry condition, the absorption band attributed to the NH4+ species was exclusively seen in the presence of water (Fig. 8a, b). The areas of the IR absorption band for Lewis and Brønsted acid sites under dry and wet conditions are shown in Fig. 8e, f. The area of the IR absorption band for NH4+ (Brønsted acid sites) increased from 10.0 (dry) to 18.7 (wet) and that for NH3 (Lewis acid sites) decreased from 1.21 (dry) to 0.27 (wet) in the case of 3.5W–V when water vapor was introduced (3.5W–V; Fig. 8d–f). On the other hand, 0W–V showed only a slight change in the distribution of acid sites (0W–V; Fig. 8c, e, f). On a metal oxide surface, Lewis acid sites (coordinatively unsaturated metal cations) and Brønsted base sites (oxygen ions) can be changed into Brønsted acid sites by dissociative adsorption of water58. In a previous study, it was theoretically predicted that Brønsted acid sites (hydroxyl groups) are newly created on the V2O5–WO3 solid solution by dissociative adsorption of water on tungsten-exposed sites and bridging oxygen sites under a wet atmosphere (Fig. 8g)48. On the other hand, water could be adsorbed on V2O5 without tungsten, but a Brønsted acid site does not form47,48. The observed changes in the surface state of vanadium oxide with and without tungsten can reflect such theoretical predictions. The area of the IR spectra (NH4+) decreased with time of NO + O2 flowing both for 3.5W–V and 0W–V (Fig. 8c, d, red and blue lines), and ammonia species were consumed by the reaction under a wet condition. Considerable N2 production was confirmed for 3.5W–V under a wet condition, but the N2 production of 0W–V was smaller than that under a dry condition (Fig. 8c, d, mass spectra). These results suggest that NH3-SCR of bulk vanadium oxide without tungsten is strongly inhibited by water, while bulk tungsten-substituted vanadium oxide proceeds with the reaction by newly created protonic Brønsted acid sites that are less affected by water.

Operando UV–Vis measurements were also conducted for 3.5W–V with and without water vapor to check the redox cycle (Fig. 9 and Supplementary Fig. 13). NO (500 ppm) + NH3 (500 ppm) and O2 (8%) were repetitively introduced three times to confirm the redox cycle. During the measurements under wet and dry conditions, the pseudo-absorbance (KM unit) at λ = 700 nm similarly increased and decreased under NO + NH3 and O2 flowing, showing reduction (V5+ → V4+) and oxidation (V4+ → V5+) half-cycles (Fig. 9a–c for a wet condition and Supplementary Fig. 13 for a dry condition). The times of change in pseudo-absorbance (KM unit) at λ = 700 nm are shown by blue (wet) and red (dry) solid lines in Fig. 9d. Repetitive production of N2 with a change in ΔKM700 was confirmed under both dry and wet atmospheres (Fig. 9d). Notably, considerable N2 production was seen even in the presence of water (Fig. 9d, mass spectra). These results demonstrate that 3.5W–V proceeds with NH3-SCR by the redox cycle and the N2 production is not affected by water in a transient state.

### Difference between supported and bulk vanadium oxide

Operando IR spectra were observed for 3.5W–V and V–W/TiO2 to consider the difference between bulk and supported tungsten-substituted vanadium oxide catalysts. The measurement was conducted at 200 °C at which V–W/TiO2 is sufficiently active and NO was flowed into samples after adsorption of NH3 to obtain the information of acid sites next to redox sites (Supplementary Fig. 14). After the adsorption of NH3 under a wet condition, NH4+ species adsorbed on Brønsted acid sites were solely confirmed for 3.5W–V (Fig. 10a), which was a similar result obtained at 150 °C in the presence of water. NH4+ species were rapidly consumed and significant N2 production was confirmed when NO flowed (Fig. 10c). On the other hand, NH3 species adsorbed on Lewis acid sites were mainly observed in addition to NH4+ species for V–W/TiO2 (Fig. 10b). However, the consumption of NH4+ and NH3 species were dull and N2 production was less (Fig. 10c). The results demonstrate that Brønsted acid sites of 3.5W–V are more reactive than Brønsted and Lewis acid sites of V–W/TiO2. After the above operando IR measurement under NO flowing for 3.5W–V, the sample was oxidized and NO flowed again to check if the residual NH4+ reacts with NO. Although the N2 production and the consumption of NH4+ were attenuated under the first NO flowing (Fig. 10c), NH4+ species on Brønsted acid sites were consumed again and N2 production was confirmed under the second NO flowing after oxidation (Supplementary Fig. 15). If Operando UV–Vis measurement was conducted in the same procedure (NH3 → 1st NO → O2 → 2nd NO), the similar N2 production behavior was observed accompanied by redox cycle (V5+ ↔ V4+) under the first and second NO flowing (Supplementary Fig. 16). These results show that V sites reduced by the reaction of NH4+ (Brønsted acid site) with NO are re-oxidized, and remaining ammonia species move to Brønsted acid sites which are adjacent to redox V sites, then they react with NO by the re-oxidized V sites again. According to the previous reports on supported vanadium-based catalysts, Brønsted acid sites are not directly involved in the catalytic cycle and they play a role in an NH3 pool to supply NH3 to the Lewis acid sites20,52. In the case of bulk tungsten-substituted vanadium oxide catalyst, Brønsted acid site (B) would be located next to redox-active V5+ site and it would directly react with NO as following reduction half-cycle:

$${\mathrm{V}}\left( {\mathrm{V}} \right) = {\mathrm{O}} + {\mathrm{NH}}_3\left( {\mathrm{B}} \right) + {\mathrm{NO}} \to {\mathrm{V}}\left( {{\mathrm{IV}}} \right) - {\mathrm{OH}} + {\mathrm{N}}_2 + {\mathrm{H}}_2{\mathrm{O}}.$$
(2)

Furthermore, we conducted temperature-programmed reaction measurements under NO + NH3 flow for 3.5W–V and V–W/TiO2 (Supplementary Fig. 17) to observe the reducibility. The reduction of 3.5W–V started at ambient temperature, while a higher temperature was needed to reduce V–W/TiO2. Bulk tungsten-substituted vanadium oxide catalyst has water-tolerant Brønsted acid sites and greater reducibility originated from bulk characteristics, resulting in high NH3-SCR activity at a low temperature in the presence of water vapor.

## Methods

### Reagents

NH4VO3 and oxalic acid were purchased from FUJIFILM Wako Pure Chemical Corporation. Ammonium metatungstate ((NH4)6[H2W12O40nH2O, n 6) was purchased from Nippon Inorganic Colour & Chemical Co., Ltd. All reagents were used without further purification.

### Synthesis of tungsten-substituted vanadium oxide catalysts

First, NH4VO3 and oxalic acid (11.9 g, 131.7 mmol) were dissolved in 50 mL of water and the aqueous solution was stirred for 10 min for completion of the change in color of the solution. Next, ammonium metatungstate was added to the solution. Then, the solution was heated overnight to evaporate the water on a hotplate. Finally, the resulting solid was calcined twice at 300 °C for 4 h each time. The sample was denoted as xW–V, where x is the molar ratio (mol%) of W to V. The amounts of reagents are shown in Supplementary Table 1.

### Synthesis of TiO2-supported V2O5/WO3 (1 wt% V2O5–5 wt% WO3/TiO2)

First, vanadium and tungsten precursor aqueous solutions were prepared. For the vanadium precursor aqueous solution, 0.020 g of NH4VO3 (0.17 mmol) and 0.045 g of oxalic acid (0.51 mmol) were dissolved in 2 mL of water and the aqueous solution was stirred for 10 min. For the tungsten precursor aqueous solution, 0.084 g of ammonium metatungstate hydrate (0.028 mmol) and 0.089 g of oxalic acid (0.99 mmol) were dissolved in 2 mL of water and the aqueous solution was stirred for 10 min. Then. 1.4 g of TiO2 (P25) was added to 20 mL water and the vanadium and tungsten precursor aqueous solutions were added to the dispersion. The dispersion was then heated at 120 °C to evaporate the water. Finally, the resulting solid was calcined at 300 °C for 4 h.

### Catalyst characterization

STEM images were obtained by using FEI Titan Cubed G2 60–300. For TEM measurements, samples were deposited on a carbon film-coated mesh copper grid. XRD patterns were collected by SmartLab (Rigaku) with Cu Kα radiation. Rietveld analyses were carried out using Reflex program in Materials Studio 2017.

### Catalytic activity test

The NH3-SCR activity of vanadium oxide catalysts was measured using a fixed-bed flow reactor (Supplementary Fig. 1). The reaction gas mixture, 250 ppm NO, 250 ppm NH3, 4 vol% O2 and 5–20 vol% water (when used) in Ar (250 mL min−1), was fed to the catalyst (0.375 g). The outlet gases were analyzed by an IR spectrometer (JASCO FT/IR-4700) equipped with a gas cell (JASCO LPC12M-S). NO conversion and N2 selectivity were calculated by the following equations:

$${\mathrm{NO}}\,{\mathrm{conversion}}\left( {\mathrm{\% }} \right) = \frac{{{\mathrm{NO}}_{{\mathrm{in}}} - {\mathrm{NO}}_{{\mathrm{out}}}}}{{{\mathrm{NO}}_{{\mathrm{in}}}}} \times 100,$$
(3)
$$\begin{array}{l}{\mathrm{N}}_2\,{\mathrm{selectivity}}\left( {\mathrm{\% }} \right)\frac{{2 \times {\mathrm{N}}_{2{\mathrm{out}}}}}{{({\mathrm{NO}}_{{\mathrm{in}}} + {\mathrm{NH}}_{3{\mathrm{in}}}) - ({\mathrm{NO}}_{{\mathrm{out}}} + {\mathrm{NH}}_{3{\mathrm{out}}})}}\\ \left( {2 \times {\mathrm{N}}_{2{\mathrm{out}}} = ({\mathrm{NO}}_{{\mathrm{in}}} + {\mathrm{NH}}_{3{\mathrm{in}}}) - ({\mathrm{NO}}_{{\mathrm{out}}} + {\mathrm{NH}}_{3{\mathrm{out}}} + {\mathrm{NO}}_{2{{\mathrm{out}}}} + 2 \times {\mathrm{N}}_2{\mathrm{O}}_{{\mathrm{out}}})} \right).\end{array}$$
(4)

Reaction rate and reaction order of the catalysts were determined by adjusting the amount of catalysts and the flow rate for NO conversion to be below 20%.

### Operando FT-IR measurements

Operando FT-IR spectra were recorded at 150 °C using a JASCO FT/IR-4200 with a TGS (triglycine sulfate) detector. Samples (each 40 mg) were pressed to obtain self-supporting pellets (ϕ = 20 mm), which were placed in a quartz IR cell with CaF2 windows connected to a conventional gas flow system. Prior to measurements, the sample pellets were heated under a flow of 10% O2/He (100 mL min−1) at 300 °C for 10 min. NH3/He (0.1%) was then introduced for 0.5 h, and subsequently, the gas was switched to He (20 min) in order to let the residual NH3 gas out. After taking the first spectrum, 500 ppm NO, 8% O2 (when used)/He was introduced to the sample at a flow rate of 100 mL min−1. For the measurement under a wet condition, 2% H2O was introduced. Spectra were measured by accumulating 20 scans at a resolution of 4 cm−1. A reference spectrum taken at 150 °C under He flow was subtracted from each spectrum. An MS (BELMass, MicrotracBEL Corp.) was used for the analysis of N2 gas.

### Operando UV–Vis measurements

Diffuse reflectance UV–Vis measurements were conducted at 150 °C with a UV–Vis spectrometer (JASCO V-670) connected to an operando flow cell with a quartz window. The light source was led to an integrating sphere through an optical fiber. Samples (each 10 mg) were placed in the sample cell connected to a gas flow system. Reflectance was converted to KM units using the KM function. A background spectrum was corrected by measuring BaSO4. Prior to measurements, the sample pellets were heated under a flow of 10% O2/He (100 mL min−1) at 300 °C for 10 min. Then, the sample pellets were cooled down to 150 °C. The measurements were performed by the following steps: (i) The sample pellets were purged with He flow to remove residual oxygen and the initial UV–Vis spectra were measured (spectra 1). (ii) A gas mixture (500 ppm NH3 + 500 ppm NO/He) was fed to a sample at a flow rate of 100 mL min−1 for a given time. Then, UV–vis spectra were recorded (spectra 2). (iii) After turning off the gas flow, the sample pellets were purged with He flow followed by UV–Vis spectra acquisition (spectra 3). (iv) Then, a gas mixture (8% O2/He) was fed to a sample at a flow rate of 100 mL min−1 for a given time and UV–Vis spectra were obtained (spectra 4). Steps (i)–(iv) were regarded as one cycle. The change in the KM unit at λ = 700 nm (ΔKM700), which was assigned to the dd transition of V4+, was measured to obtain the relative amount of redox sites. ΔKM700 values for reduction (V5+ reduced by NO + NH3) and oxidation (V4+ oxidized by O2) half-cycles were calculated by the following equations:

$$({\Delta}{\mathrm{KM}}_{700}\,{\mathrm{for}}\,{\mathrm{reduction}}\,{\mathrm{half}}\,{\mathrm{cycle}}) = \left( {{\mathrm{KM}}_{700}\,{\mathrm{in}}\,{\mathrm{spectra}}\,2} \right) - \left( {{\mathrm{KM}}_{700}\,{\mathrm{in}}\,{\mathrm{spectra}}\,1} \right),$$
(5)
$$({\Delta}{\mathrm{KM}}_{700}\,{\mathrm{for}}\,{\mathrm{oxidation}}\,{\mathrm{half}}\,{\mathrm{cycle}}) = \left( {{\mathrm{KM}}_{700}\,{\mathrm{in}}\,{\mathrm{spectra}}\,4} \right) - \left( {{\mathrm{KM}}_{700}\,{\mathrm{in}}\,{\mathrm{spectra}}\,3} \right).$$
(6)

The production of N2 in the outlet gas mixture was monitored by an MS (BELMass, MicrotracBEL Corp.).

### Others

The crystal structure was drawn by VESTA59.