Investigation of the interactions in CeO2–Fe2O3 binary metal oxides supported on ZSM-5 for NO removal by CO in the presence of O2, SO2 and steam

ZSM-5 series catalysts with different metal contents were prepared for NO removal by CO in the presence of O2, SO2 and steam via an impregnationmethod. The flash catalysts and used catalysts were characterized via XRD, BET, XPS, NH3-TPD, CO-TPR and in situ DRIFTS and their activities were tested in a fixed-bed reactor. The 10Ce–10Fe catalyst exhibited stable NO conversion of over 90% within the temperature range from 320 C to 650 C, and it has superior resistance to O2, SO2 and steam. The ZSM-5 structure was retained after loading Fe and Ce, and diffraction peaks of Fe2O3 and CeO2 were observed in the XRD spectra with an increase in the metal content. The specific surface area and pore volume of Ce–Fe/ ZSM-5 decreased, and its average pore diameter increased. The Ce–Fe/ZSM-5 catalysts possess chemisorbed oxygen and lattice oxygen, and their various cations (Fe/Fe and Ce/Ce) promote the production of oxygen vacancies, which are beneficial for the activation of N–O. The strong acid sites and some medium-strong acid sites are involved in the catalytic reaction as the active sites for Ce–Fe/ ZSM-5 in NH3-TPD. The interaction between Fe2O3 and CeO2 (Ce 3+ + Fe 4 Ce + Fe) in the Ce– Fe/ZSM-5 catalyst improves its catalytic performance, reducing property, resistance to O2, SO2 and steam, and service life.


Introduction
Nitrogen oxides (NO x ) are one of the predominant severe pollutants in the atmosphere, which cause acid rain, ozone depletion, photochemical smog, greenhouse effects, etc. 1 Currently, selective catalytic reduction (SCR) of NO x with NH 3 in 2-6 vol% O 2 is an efficient technology for the abatement of NO x from stationary sources, such as power plants. 2,3 The narrow temperature window (300-400 C), high toxicity of vanadium species (V 2 O 5 ), SO 2 oxidation to SO 3 , and formation of N 2 O at high temperatures are disadvantages for the widely used commercial catalyst, V 2 O 5 -WO 3 (MoO 3 )/TiO 2 . 4,5 Therefore, lowtemperature SCR with non-polluting catalysts has attracted increasing attention. Its reactor is generally located downstream of the electrostatic precipitator and desulfurization device to avoid catalyst deactivation caused by the high concentration of ash. 6 However, through extensive research on low-temperature SCR catalysts in recent years, it was found that the operating temperature window (150-300 C) for lowtemperature SCR catalysts 7,8 is slightly higher than the ue gas temperature (50-100 C) from the desulfurizer, and ammonium sulfate, which is formed in the presence of a small amount of SO 2 (about 50 ppm), results in catalyst deactivation. 4 Due to the ability of reducing SO 2 and its wide existence in ue gas, CO is a promising reductant for stationary sources, particularly coal-red power plants with a high CO content in their exhaust gas. CO was used in this study as a reducing agent to remove NO similar to automobile exhaust purication. 9,10 Compared with mobile source systems, CO oxidation by O 2 and the formation of hypertoxic COS by-products are the difficulties associated with this investigation and application. The main related reactions are represented by eqn (1.1)-(1.5): 2NO + 2CO / N 2 + 2CO 2 (1. Various catalysts for NO reduction by CO have been investigated, such as CuO-CoO/g-Al 2 O 3 , 11 CuO/CeO 2 -TiO 2 , 12 MnO x -CuO/Ce 0.67 Zr 0.23 O 2 (ref. 13) and CuO-CeO 2 , 14 but their resistance to O 2 and SO 2 is rarely considered. Pereira et al. 15 researched the stability of CuO/TiO 2 and FeO/TiO 2 catalysts during NO reduction by CO in the presence of O 2 and SO 2 . The results showed that the presence of O 2 leads to the direct oxidation of CO to CO 2 according to eqn (1.3), and the conversion of NO decreases drastically when SO 2 is added into the feed stream due to the formation of sulfates on the titania surface, which block the active sites. Li et al. 16 researched an Fe-based catalyst for the reduction of NO by CO under FCC regeneration conditions, and suggested that Fe-based catalysts enhance the reduction of NO by CO at a lower temperature and the reaction rate decreases with an increase in O 2 concentration.
Zeolite ZSM-5, which has a porous structure and rich acid sites, is considered to be a perfect carrier for extensive use in many types of catalytic reactions, such as catalytic cracking and catalytic dehydration. 17,18 Cu/ZSM-5 and Fe/ZSM-5 are usually used for NH 3 -SCR, and it has been shown that the characteristics of Fe/ZSM-5 are conducive to NO reduction. The NO x conversion over Fe-CuO x /ZSM-5 catalyst reached 98% within the temperature range of 180 to 360 C. 19 Li et al. 20 indicated that Fe 2 O 3 is a good adsorbent for both NO and NO 2 , while H-ZSM-5 is not a good adsorbent for either NO or NO 2 . Fe-ZSM-5 only adsorbs NO 2 , which strongly demonstrates the existence of interactions between the Fe x O y clusters and the zeolite support. Nevertheless, the use of zeolite ZSM-5 for NO x removal by CO is unusual. Cheng et al. 21 discovered that NO x removal over Fe/ZSM-5 catalysts is very stable at 250-400 C in a simulated rotary reactor with 5% O 2 , and a higher CO concentration enhances the reduction efficiencies.
Ce-doped catalysts, including CuO/CeO 2 -TiO 2 , 22 CeWO x , 23 Ce-Cu/ZSM-5 (ref. 24) and Ce-Mn/ZSM-5, 25 have been reported recently because Ce can store and release oxygen via the Ce 3+ / Ce 4+ redox cycle. Deng et al. 22 showed that the strong synergistic effect between Ti 3+ , Ce 3+ and Cu + in CuO/CeO 2 -TiO 2 is benecial to improve activity due to more oxygen vacancies. Lai et al. 24 elucidated that the addition of Ce increases the Lewis acid sites on the surface of Cu/ZSM-5 and broadens the operation temperature window. However, studies about Ce and Fe modi-ed ZSM-5 are relatively scarce.
Herein, a series of Fe/ZSM-5 and Ce-Fe/ZSM-5 catalysts with different Fe and Ce contents were prepared via impregnation, and their catalytic activity for NO reduction by CO in the presence of O 2 and SO 2 was further evaluated. Then, the samples were characterized via XRD, BET, XPS, NH 3 -TPD, CO-TPR and in situ DRIFTS. The interactions in the CeO 2 -Fe 2 O 3 binary metal oxides supported on ZSM-5 were thoroughly investigated.

Catalyst preparation
ZSM-5 zeolites were synthesized via the seeding method in a hydrothermal system using silica sol, sodium aluminate and sodium hydroxide as the main raw materials. The ZSM-5 seed crystals were obtained from Nankai University, silica sol was procured by Guangzhou Suixin Chemical Industry Co., Ltd, and the other reagents were purchased from Sinopharm Chemical Reagent Co., Ltd, which were of analytical grade (AR). The HZSM-5 support was obtained aer ion exchange with ammonium nitrate.
The Fe/ZSM-5 samples were prepared by the conventional wet impregnation of HZSM-5 with an aqueous solution containing the required amount of Fe(NO 3 ) 3 $9H 2 O. The samples were dried at 110 C for 10 h and then calcined at 550 C in air for 4 h. The products were simply denoted as xFe/ZSM-5, where x represents the mass percent of Fe in the catalyst. 0Fe/ZSM-5 was pure HZSM-5. xFe/ZSM-5-A represents the used xFe/ZSM-5 catalyst aer catalytic reaction for 24 h. Similarly, the Ce-Fe/ ZSM-5 catalyst was prepared by the wet impregnation of HZSM-5 with Fe(NO 3 ) 3 $9H 2 O and Ce(NO 3 ) 3 $6H 2 O, which was denoted as yCe-xFe/ZSM-5, where x and y represent the mass percent of Fe and Ce in the catalyst, respectively.

Characterizations
X-ray diffraction (XRD) patterns were recorded on a Dutch X'pert HighScore Plus X-ray diffractometer using Cu Ka radiation in the 2q range of 5-90 with a step of 2 min À1 . The X-ray tube was operated at 40 kV and 40 mA. The quantitative analysis of TFe and FeO was performed via a chemical titration method.
Nitrogen adsorption-desorption isotherms were measured at 77 K on a Micrometrics ASAP-2020 adsorption apparatus. Samples were placed under vacuum for 4 h at 400 C as a pretreatment. Specic surface area was determined using the Brunauer-Emmett-Teller (BET) method. The pore volume and aperture size were calculated using the Barrett-Joyner-Halenda (BJH) formula.
X-ray photoelectron spectroscopy (XPS) was performed on a Thermo Scientic Escalab 250Xi using Al Ka as the radiation source at 150 W. The binding energies (BEs) were referenced to the adventitious C 1s at 284.6 eV. This reference gave BE values with an accuracy of AE0.1 eV. NH 3 temperature-programmed desorption (NH 3 -TPD) was carried out on a PX200 apparatus by Tianjin Pengxiang Technology Ltd. About 100 mg sample was placed in a quartz reactor and pretreated at 500 C under a ow of N 2 (70 mL min À1 ) for 1 h. The amount of NH 3 desorbed was detected by a thermal conductivity detector (TCD).
Carbon monoxide temperature-programmed reduction (CO-TPR) was performed on a Micromeritics Autochem 2920 equipped with TCD. Initially, 50 mg sample was pretreated under N 2 ow at 300 C for 1 h. Aer cooling to room temperature, the sample was heated to 800 C (10 C min À1 ) under a CO-N 2 mixture (10% CO by volume).
In situ diffuse reectance infrared Fourier transform spectra (in situ DRIFTS) were collected from 400 to 4000 cm À1 at a spectral resolution of 4 cm À1 (number of scans: 32) on a Nicolet IS50 FT-IR spectrometer equipped with a high-sensitive MCT detector. The sample was placed on a high temperature cell and pretreated with N 2 at 300 C for 1 h. The background of each target temperature was collected during the cooling process; then, the sample was exposed to a controlled stream of CO-Ar (10% vol. CO) or/and NO-Ar (5% vol. NO) at a rate of 10 mL min À1 for 30 min. Desorption/ reaction studies were performed by heating the adsorbed species and the spectra were recorded at target temperatures from room temperature to 350 C at the rate of 10 C min À1 by subtracting the corresponding background reference.

Evaluation of catalytic performance
Catalyst activity tests were carried out in a xed-bed reactor with 5 mL sample (40-60 mesh) between 200 C and 650 C. The gas mixture contained 600 ppm NO, 600 ppm CO, and N 2 as diluents with a space velocity of 36 000 h À1 . The product was analyzed via gas chromatography (GC9310) with TCD and two columns (Porapak Q and Molecular 13X). The 10Fe/ZSM-5 and 10Ce-10Fe/ZSM-5 catalysts were also evaluated in the reduction of NO with CO at 400 C in the presence of 1000 ppm O 2 60 ppm or 120 ppm SO 2 , and 3% water steam. The conversion (h) of NO was calculated using eqn (2.1).

Catalytic activity
NO conversion as a function of reaction temperature between 200 C and 650 C on the series of ZSM-5 catalysts with different metal contents is shown in Fig. 1(a), the reaction temperature (T 10% , T 90% , and T 100% ) and maximal conversion are listed in Table 1. The 0Fe/ZSM-5 catalyst presented almost no activity in the working temperature window. The 2Fe/ZSM-5 catalyst exhibited signicant catalytic activity at the T 10% of 379 C and the highest conversion of NO reached 89.24%. The T 10% of the 5Fe/ZSM-5 catalyst decreased to 318 C, and the NO conversion reached 100% at 476 C. However, with an increase in temperature to 580 C, the NO conversion declined. The NO conversion curve of 10Fe/ZSM-5 clearly shied to a low temperature, and T 90% decreased to about 80 C compared with that of 5Fe/ZSM-5. Besides, the NO conversion remained at 100% up until 650 C, which broadened the operating temperature window. Briey, with an increase in Fe content in the catalyst, the NO conversion signicantly increased and the catalytic reaction temperature gradually decreased. Aer adding Ce, the NO conversion curve of the 5Ce-10Fe/ZSM-5 catalyst also shied to low temperature, and T 90% dropped by nearly 20 C to 338 C. The T 90% of the 10Ce-10Fe/ZSM-5 catalyst further deceased to 317 C, which showed the best performance. However, the 10Ce/ZSM-5 catalyst exhibited poor activity, and 445 C was too high for T 10% , which shows that the catalytic activity of the Fe species is better than Ce. Above all, these results illustrates that Ce has a promoting effect on the catalytic reaction. The N 2 selectivity curves ( Fig. 1(b)) shied to a low temperature with an increase in the metal contents, which are similar to the NO conversion curves. The N 2 selectivities increased when the reaction temperature increased as a result of the rapid decomposition of absorbed NO on the catalyst surface at high temperature.
The performances of resistance to O 2 , SO 2 and steam on the 10Fe/ZSM-5 and 10Ce-10Fe/ZSM-5 catalysts are shown in Fig. 2. The main components of the initial feed stream were 1200 ppm CO and 600 ppm NO. The NO conversion over the 10Fe/ZSM-5 catalyst was 100% at 400 C. When 60 ppm SO 2 was added, the conversion of NO to N 2 constantly remained 100%. However, the conversion of CO to CO 2 slowly increased due to the redox reaction between CO and SO 2 according to eqn (1.4), and returned to the original value when the SO 2 feed was interrupted. The NO conversion rapidly reduced to 53.8% and CO conversion clearly increased to 87.5% aer 1000 ppm O 2 and 60 ppm SO 2 were introduced into the stream. There was a competitive relationship between CO + NO and CO + O 2 , and O 2 favored CO oxidation to CO 2 . When the O 2 and SO 2 streams were off, the NO conversion recovered to 85.4% but could not reach 100%, indicating that the catalyst was deactivated as a result of some irreversible changes. The NO conversion dropped drastically to 45.8% because the high SO 2 concentration of 120 ppm in the ue gas inhibited the catalyst activity. For   the 10Ce-10Fe/ZSM-5 catalyst, the resistance curve to O 2 and SO 2 was similar to that for 10Fe/ZSM-5. However, in the presence of 60 ppm SO 2 , the CO conversion increased and remained steady. The NO conversion decreased to 67.7% (higher than that for 10Fe/ZSM-5) and the CO conversion increased to 72.5% when 1000 ppm O 2 and 60 ppm SO 2 were added, and the NO conversion recovered to 100% aer the O 2 and SO 2 streams were turned off. The NO conversion declined to 51.8% in the presence of 120 ppm SO 2 , which exhibits preferable resistance to a high SO 2 concentration. Fig. 2(c) shows that the NO conversions on 10Fe/ZSM-5 and 10Ce-10Fe/ZSM-5 at 400 C remained almost unchanged (over 95%) in the presence of steam. In contrast, the 10Ce-10Fe/ZSM-5 catalyst exhibited superior resistance to O 2 , SO 2 and steam, and could not be deactivated under the research conditions.   Table 2. The low Fe 2+ content (about 0.09-0.10 wt%) implies that the Fe x O y species was primarily Fe 2 O 3 . As shown in Fig. 3(a), when the Fe content was lower, the diffraction peaks of Fe 2 O 3 were hardly observed in the 2Fe/ZSM-5 catalyst, which implies that Fe 2 O 3 was highly dispersed on the internal and external surfaces of the carrier. The evident characteristic diffraction peaks of Fe 2 O 3 were observed when the Fe content was 5%, and peaks of Fe 2 O 3 gradually strengthened with an increase in Fe content. The literature 24 shows that metal oxide is amorphous at a low loading content, and under a high loading content, enriched metal oxide on the surface could be crystalline, which has evident characteristic peaks. In the XRD spectra of the used 10Fe/ZSM-5-A catalyst, the typical Fe 2 O 3 peaks disappeared, but the characteristic diffraction peaks of Fe 3 O 4 appeared, which is in agreement with the increase in Fe 2+ content to 2.54 wt% in Table 2. This could be due to the reduction of Fe 2 O 3 to Fe 3 O 4 under the CO atmosphere. It can be seen from Fig. 3(b) that there were no typical diffraction peaks of CeO 2 over 5Ce-10Fe/ ZSM-5, and the CeO 2 peaks appeared when the Ce content reached 10%, which is similar to Fig. 3(a). In addition, the characteristic peaks of Ce 2 O 3 were not detected due to its instability. The Fe 2+ content only increased slightly to 0.12 wt% in Table 2, and the Fe 3 O 4 peaks in 10Ce-10Fe/ZSM-5-A catalyst were weaker than that in 10Fe/ZSM-5-A, suggesting that Ce inhibits the reduction of Fe 2 O 3 .

Pore analysis
The BJH pore diameter distribution curves of the catalysts are shown in Fig. 4, and the textural properties of the catalysts are listed in Table 3. 0Fe/ZSM-5 possesses mesoporous molecular sieves with a large specic surface area of 341.49 m 2 g À1 . It can be observed from Fig. 5 that the pores in the ZSM-5 carriers are between 2-4 nm, and the apertures distributed in range of 10-100 nm could be secondary pore structures between carriers or active components loaded on the surface of the carriers. Aer Fe loading, the number of 2-4 nm mesopores slightly declined, and the secondary pores between 10-100 nm relatively increased. Therefore, the specic surface area and pore volume decreased, and the average pore diameter gradually increased.
With an increase in Fe content, there was almost no change in the amount of mesopores and secondary pores in 10Fe/ZSM-5, similar with 10Ce/ZSM-5. For 10Ce-10Fe/ZSM-5, the change in number of mesopores was not evident, but the number of secondary pores remarkably increased, leading to a further reduction in specic surface area and pore volume, and the increase in average pore diameter. Combined with the ash catalyst, the mesopores in the used catalyst slightly decreased, and the change in secondary pores was relatively signicant, so the average pore diameter of the used catalyst became smaller.

XPS results
The XPS spectra are shown in Fig. 5, and Table 4 lists the binding energy and calculated relative percentage of ions by    consistent with the conclusion from XRD. In addition, the lattice oxygen content in 10Fe/ZSM-5-A was 0.18, which decreased by 30% compared to 0.26 for the ash catalyst, which shows that irreversible changes occurred in the physical and chemical properties of the catalyst during the denitration process. However, the lattice oxygen content in 10Ce-10Fe/ZSM-5-A was reduced by only 8% as a result of the interaction between Fe 2 O 3 and CeO 2 (Ce 3+ + Fe 3+ 4 Ce 4+ + Fe 2+ ), which implies that the lattice oxygen was effectively restored and the resistance to O 2 and SO 2 improved. Fig. 6 shows the NH 3 -TPD proles of the catalysts, in which all the catalysts display three peaks. The peak centered at about 100 C belongs to NH 3 adsorbed on Lewis acid sites and physically adsorbed on the weak surface acid sites. The peak between 150-300 C is ascribed to NH 3 adsorbed on the Lewis acid sites, which are medium-strong acid sites. The peak between 300-500 C is attributed to NH 3 adsorbed on Bronsted acid sites, which are strong surface acid sites. The peak intensity in the proles represents the total quantity of acid, which slightly changed aer Fe loading. Aer Ce loading, the acid quantity signicantly decreased. The temperature and area percent of desorption peaks are shown in Table 5. The percentage of weak acid on the catalyst reduced aer the Fe and Ce loading, and the strong acid sites increased. In addition, the total quantity of acid for the used 10Fe/ZSM-5-A and 10Ce-10Fe/ ZSM-5-A catalyst evidently decreased, but the acid quantity of 10Fe/ZSM-5-A was relatively lower than that in the 10Ce-10Fe/ ZSM-5-A catalyst, which illustrates that the interaction between Ce and Fe was helpful to remain active. The percentage of strong acid on the 10Fe/ZSM-5-A catalyst sharply decreased compared with that in the ash catalyst. The percentage of strong acid on the 10Ce-10Fe/ZSM-5-A catalyst also declined, but the percentage of medium-strong acid also slightly decreased. It was conjectured that the strong acid sites and some of the medium-strong acid sites were both involved in the catalytic reaction as the active sites for the 10Ce-10Fe/ZSM-5 catalyst. It was also inferred that there was an interaction between Ce and Fe, which is consistent with the XPS analysis.

Reduction behavior of the catalysts (CO-TPR)
The CO-TPR proles of the catalysts showed CO consumption in the entire range of temperature used in the analysis. There are three reduction peaks shown in Fig. 7(a). The peak at about 100 C is related to the reduction of reactive oxygen species (O À or O 2À ). The second peak between 100-500 C is ascribed to the reduction of Fe 3+ to Fe 2+ , and the peak above 500 C is attributed to the reduction from Fe 2+ to Fe 0 . The CO-TPR prole of the 0Fe/ZSM-5 catalyst without any reduction peaks shows that the 0Fe/ZSM-5 catalyst had no reduction activity. When the Fe content increased from 5% to 10%, the reduction temperature range from 455.4-621.0 C broadened to 292.5-715.9 C, indicating that the increase in the active components enhanced the  reducing ability, and promoted the reduction from Fe 3+ to Fe 2+ and the formation of oxygen vacancies. This is the reason of decrease in T 10% (Fig. 1). For the 10Fe/ZSM-5-A catalyst used in the catalytic reaction at 650 C, the reduction peak of Fe 3+ to Fe 2+ almost completely disappeared. The broad peak for 10Ce/ ZSM-5 in Fig. 7(b) at 429.7 C is related to Ce 4+ to Ce 3+ . Aer the Ce loading, the reduction peaks of Fe 3+ to Fe 2+ disappeared, indicating that the interaction of Ce and Fe was helpful to restrain the Fe 3+ reduction, which is consistent with the results of XPS analysis. The peak for Ce 4+ to Ce 3+ conversion in 10Ce-10Fe/ZSM-5 shied to a low temperature as a result of this interaction (Ce 3+ + Fe 3+ 4 Ce 4+ + Fe 2+ ). 10Ce-10Fe/ZSM-5-A retained the peaks of the ash catalyst, which suggests that the service life of the catalyst could be improved.

CO or/and NO interaction with the 10Ce-10Fe/ZSM-5 catalyst
The in situ DRIFTS results of CO adsorption on the 10Ce-10Fe/ ZSM-5 catalyst are shown in Fig. 8(a). The peak at 1636 cm À1 is attributed to bidentate bicarbonate, two peaks at 1551 cm À1 and 1473 cm À1 belong to the surface carbonate species, and the peak at 1366 cm À1 is assigned to bidentate formate. 22 Their intensities all decreased with an increase in temperature and disappeared completely at 250 C due to decomposition. Similarly, the peaks 2175 cm À1 and 2117 cm À1 attributed to M-CO became weaker with the increase in temperature, which disappeared at 300 C due to desorption. The intensity of the peaks attributed to gaseous CO 2 centered at 2360 cm À1 and 2341 cm À1 increased as the temperature increased. This indicates that Ce 4+ or/and Fe 3+ in a high state were reduced to Ce 3+ or/and Fe 2+ under the CO atmosphere, and the released oxygen vacancies oxidized CO to CO 2 . There were a variety of different NO adsorption species at low temperature ( Fig. 8(b)). The peak at 1857 cm À1 is attributed to M-NO, which disappeared when the temperature reached 150 C. The peaks for the bridging bidentate nitrates (1602 cm À1 ), chelated nitrates (1577 cm À1 ), hyponitrites (1344 cm À1 ), linear nitrites (1315 cm À1 ) and chelated nitrites (1284 cm À1 ) all reduced greatly with the increase in temperature and disappeared completely at 200 C. The similar peaks of the bridging monodentate nitrates (1533 cm À1 and 1186 cm À1 ) disappeared completely at 300 C. 10 The peaks for the bridged nitro at 1544 cm À1 and nitrate at 1382 cm À1 could be produced in the decomposition process, 19 and their disappearance at high temperature also conrms this assumption. Furthermore, at 250 C, a new peak appeared at 1566 cm À1 , which is assigned to chelated nitrates, and its intensity strengthened with the increase in temperature. From the shi in this peak, it was surmised that the NO species adsorbed on the catalyst surface were not only attributed to simple adsorption and decomposition, but the species could have been further restructured.
In situ DRIFTS was performed under simulative reaction conditions, which provides information about the changes in the surface absorbed species in order to investigate the  interaction of the reactant gas with the catalyst as shown in Fig. 8(c). At low temperature, the peaks attributed to CO adsorption species were not signicant, and the major adsorption states were NO adsorption species. The differences between the spectra in Fig. 8(b) were that: (1) the peak for chelated nitrites was not observed; (2) the peak for the bridging bidentate nitrates (1691 cm À1 ) increased at 100 C, and did not disappear; (3) the peak intensities of the chelated nitrates (1573 cm À1 ) and linear nitrites (1310 cm À1 ) increased at 100 C then greatly decreased. When the temperature reached 150 C, peaks for CO absorption species appeared gradually, such as gaseous CO 2 (2360 cm À1 ), M-CO (2185 cm À1 and 2125 cm À1 ), bidentate bicarbonate (1627 cm À1 ) and carbonate species (1437 cm À1 and 1531 cm À1 ). These peaks rst strengthened and then weakened as the temperature increased.
The following conclusions can be deduced through comprehensive analysis: (1) NO was preferentially adsorbed on the catalyst surface due to unpaired electrons, NO (g) / NO (ads) ; (2) the NO absorption species gradually decomposed with an increase in temperature, and CO adsorbed on the released active sites, CO (g) / CO (ads) ; and (3) Ce 4+ or/and Fe 3+ in a high state were reduced to Ce 3+ or/and Fe 2+ by CO, and the released oxygen vacancies were advantageous for the activation of N-O and promoted the decomposition of NO, NO (ads) / N (ads) + O (ads) . Then, the adsorption species on the catalyst surface were restructured, CO (ads) + O (ads) / CO 2(ads) , NO (ads) + N (ads) / N 2 O (ads) , N (ads) + N (ads) / N 2(ads) and O (ads) + O (ads) / O 2(ads) . Finally, the products (CO 2 , N 2 O, N 2 and O 2 ) were obtained aer desorption, CO 2(ads) / CO 2(g) , N 2 O (ads) / N 2 O (g) , N 2(ads) / N 2(g) , O 2(ads) / O 2(g) .

Conclusions
The Ce-Fe/ZSM-5 catalysts prepared via an impregnation method showed better catalytic performances than Fe/ZSM-5 for NO removal by CO in the presence of O 2 , SO 2 and steam. The 10Ce-10Fe/ZSM-5 catalyst exhibited stable NO conversion of over 90% within the temperature range from 320 C to 650 C, and it had superior resistance to O 2 , SO 2 and steam. As the contents of Fe and Ce increased, the ZSM-5 structure was still retained, diffraction peaks of Fe 2 O 3 and CeO 2 were observed via XRD, the specic surface area and pore volume decreased, and the average pore diameter increased. From the XPS results, the Ce-Fe/ZSM-5 catalysts possessed chemisorbed oxygen and lattice oxygen, and various cations (Fe 3+ /Fe 2+ and Ce 4+ /Ce 3+ ) promoted the production of oxygen vacancies, which were advantageous for the activation of N-O and in agreement with the CO-TPR results. Not only strong acid sites, but also some medium-strong acid sites were involved in the catalytic reaction as the active sites for the 10Ce-10Fe/ZSM-5 catalyst in NH 3 -TPD. The interaction between Fe 2 O 3 and CeO 2 (Ce 3+ + Fe 3+ 4 Ce 4+ + Fe 2+ ) in the 10Ce-10Fe/ZSM-5 catalyst, which was conrmed by XPS, NH 3 -TPD and CO-TPR, improved the catalytic performance, reducing property, resistance to O 2 , SO 2 and steam, and the service life of the catalyst.

Conflicts of interest
There are no conicts to declare.