Nitrogen oxides (NO_x)—which are generated by many combustion processes, both stationary and mobile—are considered primary atmospheric pollutants [1, 2]. The emission of NO_x into the atmosphere can cause series environmental issues, including photochemical smog, acid rain, and ozone depletion [3, 4]. Currently, the selective catalytic reduction (SCR) of NO_x is considered one of the most efficient techniques for the removal of stationary-source NO_x [5-7]. The SCR of NO using CO as a reducing agent has been studied extensively in recent decades because NO and CO coexist in many industrial exhaust gases [8-10]. However, this approach presents several challenges such as ensuring low-temperature activation, resistance to O₂ and SO₂, and durable operation.

Noble metals have been used extensively as efficient catalysts for the CO-SCR reaction. However, owing to the high cost and poor resistance to sulfur/phosphate of noble metals [11], much effort has been devoted to developing transition metal oxide catalysts, and cerium-based catalysts have attracted the greatest interest. CeO₂ is an essential rare earth oxide with excellent oxygen storage capacity and redox behavior [12, 13], which may be attributed to the flexible transformation between valence states (Ce⁴⁺↔Ce³⁺) [14−16]. Many transition metals have been introduced into the CeO₂ lattice to modify its physical and chemical properties for various catalytic applications [17]. The synergistic effect arising from such modifications has become a primary focus in the catalysis research community [18-22]. Yao et al. [23] reported that a doped CeO₂ (Ce_xSn_1-xO₂) exhibited superior catalytic performance compared with undoped CeO₂ owing to the greater number of catalytic domains; Ma et al. [24]obtained a highly active Ce_xCu_1-xO₂ catalyst characterized by the presence of an intergrowth system comprising imperfect monoclinic CuO and cubic crystal CeO₂.

In the present study, we prepared a series of MO_x-CeO₂ catalysts (M = Co, Fe, or Cu) using a supercritical water (sc-H₂O) hydrothermal route [25]. The catalysts were subsequently characterized by X-ray diffraction (XRD), Brunauer-Emmett-Teller (BET) surface area measurement, and hydrogen temperature-programmed reduction (H₂-TPR). The CO-SCR reaction activity of the catalysts was evaluated using in situ diffuse reflectance Fourier transform infrared spectroscopy (DRFTIR). We expect that this work will provide an in-depth understanding of the effectiveness of transition metal-doped CeO₂ catalysts in the CO-SCR reaction for NO removal.

The transition metal-doped catalysts were prepared in our laboratory using a three-pump hydrothermal flow system (CHFS). The reactor, tubing, and components were made from grade 316 stainless steel (Swagelok™). The apparatus (Fig. S1) consisted of a metal salt solution high-performance liquid chromatography (HPLC) pump (P2), a base solution HPLC pump (P3), and a water HPLC pump (P1). Deionized water was pre-heated to the appropriate temperature (500 ℃) by pumping it through an electric coil (2.5 kW). The deionized water was then brought into contact with a metal salt solution containing Ce(NO₃)₃·6H₂O (0.1 mol·L^–1) and M(NO₃)_x·yH₂O (0.02 mol·L^–1), and a flow of NaOH solution (1 mol·L^–1) at a mixing point (a 1/4" counter-current mixer that was assisted by a 500 ℃ band heater), whereupon there was rapid co-precipitation of the crystalline products. The aqueous suspension was cooled using a water jacket, passed through a 7-µm in-line filter, and collected from the exit of a back-pressure regulator (BPR). Flow rates of 7.5, 7.5, and 30 mL·min^–1 were used for the metal salt solution, the base solution, and the water stream, respectively. The system pressure was maintained at approximately 23.0 MPa. Solids were recovered by centrifuging the suspension and freeze-drying to yield the final products.

The SCR of NO by CO was conducted in a fixed-bed quartz tubular reactor with a thermocouple at its center. The catalysts were pre-treated in a N₂ flow at 500 ℃ for 1 h before each test. The catalytic activity of each 40–60 mesh catalyst (2.2 mL) was examined and the experiments were performed at 60 to 300 ℃. The inlet gas comprised NO (600 ppm), CO (1200 ppm), and N₂ (the balance), and had a gas hourly space velocity (GHSV) of 40, 000 h^–1. NO, NO₂, and O₂ concentrations were monitored using a Testo 350 flue gas analyzer. NO conversion and selectivity were evaluated as follows:

where the "in" and "out" subscripts indicate the inlet and outlet concentrations of NO/N₂O in the steady state, respectively.

XRD patterns were recorded using a Rigaku D/Max RA diffractometer with Cu-K_α radiation (λ = 0.15418 nm) at 40 kV and 150 mA. All XRD patterns were obtained with scattering angles (2θ) ranging from 10° to 80° at a step size of 0.026°. The crystal phases of the catalysts were identified according to the Joint Committee on Powder Diffraction Standards (JCPDS) database.

The BET surface areas (BETs) were determined using N₂ physisorption at −196 ℃ using Micrometrics ASSP 2020 equipment. Prior to N₂ adsorption, each catalyst was degassed for 2 h under vacuum at 200 ℃.

H₂-TPR was conducted in TP-5089 equipment (Tianjin Xianquan Industry and Trade Development Co., Ltd.); before each test, the catalysts (0.05 g) were maintained at 400 ℃ for 1 h and purged with 3% O₂/He gas at a flow rate of 50 mL·min^–1, then cooled to 100 ℃. Subsequently, they were preheated at 100 ℃ for 40 min, then further heated to 900 ℃ at a rate of 10 ℃·min^–1 while purging with 5% H₂/N₂ gas at a flow rate of 35 mL·min^–1. The variation in H₂ concentration was recorded using a thermal conduction detector (TCD).

Fourier-transform infrared spectroscopy (FTIR) spectra were acquired using an in situ DRFTIR cell equipped with a gas flow apparatus. The DRFTIR measurements were obtained using ZnSe windows coupled to Bruker tensor 27 FTIR spectrometers. Inside the DRFTIR cell, the catalysts were pretreated in a He atmosphere at 300 ℃ for 2 h, then cooled to 100 ℃. The background spectrum was obtained using flowing He, and was subtracted from the spectrum of each catalyst.

The phase identities and purities of the M-doped CeO₂ catalysts (M = Co, Fe, or Cu) were evaluated by XRD. As shown in Fig. 1, all the catalysts displayed distinct XRD reflections at approximately 28.6°, 33.2°, 47.5°, 56.4°, and 76.9° for cubic fluorite CeO₂ (PDF-ICDD81-0792). No extra peaks attributable to FeO_x, CoO_x, or CuO species were observed. This may have been because the ultrafine transition metal oxides were highly dispersed throughout the CeO₂, or because the CeO₂ lattice was fully doped with the metal ions [17, 26]. All the catalysts had a higher 2θ shift than pure CeO₂ (see the magnified patterns in Fig. 1(b)). This suggests that the transition metal ions were all incorporated into the CeO₂ lattice and formed solid solutions, consistent with previous reports [27, 28]. Such incorporations led to the shrinkage of the CeO₂ lattice because each metal ion had a lower ionic radius than Ce⁴⁺ (0.094 nm) [29, 30]. The BET surface areas of the FeO_x-CeO₂, CoO_x-CeO₂ and CuO-CeO₂ catalysts were 147.6, 86.5 and 106.8 m²·g^–1, respectively.

As shown in Fig. 2(a), the pure CeO₂ exhibited the lowest SCR activity measured in the absence of light; its T₅₀ (i.e., 50% NO conversion) was approximately 233 ℃ and its T₉₀ (i.e., 90% NO conversion) was approximately 300 ℃. The doping of CeO₂ with transition metals greatly improved its SCR activity, and the CuO-CeO₂ catalyst had the highest SCR activity of the catalysts investigated (i.e., a T₅₀ of only 83 ℃ and a T₉₀ of approximately 126 ℃). The synergic interaction between the transition metals and the cerium accounts for the improvement in performance. Compared with many state-of-the-art catalysts, which usually have a T₅₀ in the range 200–300 ℃ [31], the CuO-CeO₂ catalyst exhibited superior SCR activity with regard to the reduction of NO by CO. We prepared a CuO-CeO₂ catalyst with an identical composition using a conventional hydrothermal route for comparison. As shown in Fig. S2, the catalyst produced by the supercritical water hydrothermal (sc-H₂O) route exhibited much higher SCR activity than that produced by the conventional hydrothermal route (ht-CuO-CeO₂), further confirming the superiority of the sc-H₂O route. The sequence of SCR activity was: FeO_x-CeO₂ < CoO_x-CeO₂ < CuO-CeO₂. This was not consistent with the surface areas of the catalysts, which suggests that surface area is not the crucial determinant of CO-SCR reactivity. Fig. 2(b) shows N₂ selectivity in CuO-CeO₂ and CeO₂ catalysts. The CuO-CeO₂ catalyst also exhibited a superior N₂ selectivity; its T₅₀ was approximately 147 ℃ and its T₉₀ was approximately 179 ℃ with regard to N₂ selectivity.

Fig. 3 illustrates H₂-TPR profiles of the catalysts in the temperature range 100–800 ℃. Pure CeO₂ had two H₂ consumption peaks at approximately 470 and 700 ℃, which we ascribed to the reduction of surface and bulk CeO₂, respectively [32]. The CoO_x-CeO₂ catalyst exhibited a series of consumption peaks in the range 100–350 ℃. Apparently, various cobalt oxides, including CoO, Co₃O₄, and Co₂O₃, coexist in this catalyst. The FeO_x-CeO₂ catalyst exhibited three peaks centered at approximately 356, 535, and 789 ℃. We attributed the latter two peaks to CeO₂, whereas we believe the former peak was due to the reduction of Fe-O-Ce [33]. The CuO_x-CeO₂ catalyst had the highest redox potential of the catalysts investigated; we observed two H₂ consumption peaks at approximately 142 and 158 ℃, which we ascribed to the reduction of finely dispersed CuO and Cu-O-Ce, respectively [32, 34−37]. Apparently, the redox potentials obtained from the H₂-TPR analysis followed the sequence FeO_x-CeO₂ < CoO_x-CeO₂ < CuO-CeO₂, which was in accordance with the corresponding activity measurements.

Fig. 4 illustrates the DRFTIR spectra of NO adsorption onto the MO_x-CeO₂ catalysts (M = Co, Fe, or Cu) at 100 ℃. After NO was introduced, several bands appeared at the surface. The bands at 1012 and 1187 cm^–1 were assigned to bidentate NO₃^– and nitrosyl, respectively [38-40]. The bands at 1245 and 1613 cm^-1 were due to the bridging NO₃^–, and the band at 1558 cm^–1 derived from the bridging nitro [38, 41, 42]. With regard to CoO_x-CeO₂ (Fig. 4(a)), the intensity of the 1428 cm^–1 band (NO₂^–) increased over time, which suggests that the NO species were consistently adsorbed onto the catalyst. With regard to FeO_x-CeO₂ (Fig. 4(b)), the intensity of the band at 1187 cm^–1 (nitrosyl) initially increased and then decreased after 25 min. The intensities of the 1245, 1613 cm^–1 (bridging NO₃^–), and 1558 cm^–1 (bridging nitro) bands kept increasing after 10 min [38, 43, 44]. This suggests that the anionic nitrosyl gradually transformed into bridging NO₃^– and bridging nitro species over time. The CuO-CeO₂ catalyst (Fig. 4(c)) exhibited two additional bands, which were attributed to chelating nitro (1218 cm^–1) and monodentate NO₃^– (1503 cm^–1) [45]. The DRFTIR results allowed us to deduce that all the catalysts had a high capacity for NO adsorption because the intensities of the bands attributable to adsorbed NO all increased over time.

The in situ DRFTIR spectra for CO adsorption onto the MO_x/CeO₂ (M = Co, Fe, or Cu) catalysts are shown in Fig. 5. Bands appeared at 1032, 1052, 1215, 1295, 1390, 1466, 1535, and 1565 cm^–1 after introducing CO for 1 min. The band at 1052 cm^–1 was associated with CO_x, which originated from the reduction of cerium [46]. The bands at 1032 and 1466 cm^–1 were attributed to monodentate CO₃²⁻, whereas the band at 1295 cm^–1 was related to CO₃²⁻ vibrations (v_s(CO₃²⁻)) [23, 45, 47-49]. The bands at 1390 and 1535 cm^–1 were attributed to the symmetrical and asymmetrical stretching vibration modes of COO^–, respectively [43]. The broad band at approximately 1565 cm^–1 was due to bidentate CO₃²⁻, whereas the weak band at 1215 cm^–1 was attributed to the formation of HCO₃^– [23, 43]. With regard to the MO_x-CeO₂ catalysts (M = Co, Fe, or Cu), the CO adsorption bands (e.g., CO₃^–, HCO₃^–, COO^–) all appeared. However, we observed an additional band at 2100 cm^–1 in CuO-CeO₂ (Fig. 5(d)). This band has been attributed to the linear Cu⁺-CO species formed by CO molecules adsorbed at Cu⁺ sites [23, 50, 51].

As shown in Fig. 6, the MO_x-CeO₂ catalysts (M = Co, Fe, or Cu) produced several bands related to CO₃²⁻ (1032, 1295, 1466, and 1565 cm^–1), HCO₃^– (1215 cm^–1), COO^– (1390 and 1538 cm^–1), and ad-CO species (2100 cm^–1) following pretreatment with CO for 30 min. When NO was introduced, the CO adsorption bands gradually disappeared, indicating the reaction between NO and CO. In particular, the characteristic band at 2100 cm^–1 attributable to Cu⁺-CO species in CuO-CeO₂ rapidly vanished once NO was introduced, indicating high reactivity of the species in the CO-SCR reaction. Over time, only the adsorption bands attributable to NO were retained, i.e., NO₃^– (1012, 1245, 1280, and 1613 cm^–1), nitro species (1572 cm^–1), and nitrosyl (1187 cm^–1) [23, 37].

As shown in Fig. 7, bands attributable to bidentate NO₃⁻ (1012 cm^–1), bridging NO₃^– (1245, 1613 cm^–1), monodentate NO₃⁻ (1280, 1503 cm^–1), and bridging nitro (1572 cm^–1) appeared following adsorption of NO for 30 min. These bands were retained after the introduction of CO, and no other bands representative of oxidized CO intermediates (e.g., CO₃^– and COO^–) were observed. This suggests that there was competitive adsorption of CO and NO in the catalysts. NO seemed to be adsorbed in preference to CO by the catalysts. Of the MO_x-CeO₂ catalysts (M = Co, Fe, or Cu), only CuO-CeO₂ exhibited CO-Cu⁺ species at 2122 cm^–1 after 10 min. This band gradually increased over time, suggesting that the Cu⁺ retained CO adsorption in the catalyst during the CO-SCR reaction.

As shown in Fig. 8, we found that after exposure of the MO_x-CeO₂ catalysts (M = Co, Fe, or Cu) to CO and NO at 100 ℃, the NiO-CeO₂, CoO_x-CeO₂, and FeO_x-CeO₂ catalysts preferentially adsorbed NO on their surfaces, forming NO₃^– (1012, 1245, 1280, and 1613 cm^–1), nitrosyl (1187 cm^–1), and nitro species (1215, 1572 cm^–1). In comparison, only the CuO-CeO₂ catalyst retained a distinct ad-CO band (2100 cm^–1). This result further confirms that Cu⁺ serves as an active site for CO adsorption, ensuring the CO-SCR reaction in the CuO-CeO₂ catalyst.

The DRFTIR results suggested that the CO-SCR reaction of the CoO_x-CeO₂ and FeO_x-CeO₂ catalysts proceeded according to an Eley-Rideal reaction mechanism (i.e., a reaction between absorbed NO and gaseous CO) due to the lack of CO adsorption by these catalysts. With regard to CuO-CeO₂, the presence of Cu⁺ species preserved the adsorption of CO. Therefore, this catalyst adopted a Langmuir-Hinshelwood SCR reaction mechanism. This explains why the CuO-CeO₂ catalyst exhibited the highest SCR activity in the SCR of NO by CO.

In the present study, we used an sc-H₂O hydrothermal route to synthesize MO_x-CeO₂ catalysts (M = Co, Fe, or Cu). This resulted in the formation of M_xCe_1-xO₂ solid solutions. H₂-TPR analyses revealed the following redox potential sequence: FeO_x-CeO₂ < CoO_x-CeO₂ < CuO-CeO₂. This was in accordance with the activity measurements for the catalysts with regard to the SCR of NO by CO. The DRFTIR results suggested that the MO_x-CeO₂ catalysts (M = Co or Fe) mainly followed an Eley-Rideal reaction mechanism for the CO-SCR of NO, whereas the CuO-CeO₂ catalyst adopted a Langmuir-Hinshelwood SCR reaction mechanism owing to the presence of Cu⁺ species, which ensured the effective adsorption of CO. This explains why CuO-CeO₂ exhibited the best catalytic activity with regard to the SCR of NO by CO.