Tailoring the crystal structure of CaTiO 3 by multielement doping for photo-assisted activation of NO

Although photocatalysis exhibits great prospects in selective catalytic reduction of NO x (photo-SCR), the principles toward informed design of photo-SCR catalysts are lacking. Herein, based on inert CaTiO 3 , we tailor a crystal structure for high-eciency deNO x by partially substituting Ca with Ce, and Ti with Fe and Mn. The pristine CaTiO 3 is hard to remove NO in 100–300°C, while Ce-Fe-Mn doped CaTiO 3 could achieve nearly 100% NO conversion at 135°C under light irradiation (GHSV = 72000 h − 1 ). The dopants enable CaTiO 3 to activate NO and harvest visible light. By selectively activating gaseous molecules, Ti, Fe, and Mn present signi�cant synergistic effects on accelerating the catalytic cycle of standard SCR. Moreover, it is �rmly shown that the photocarriers intensify NO oxidation rather than NH 3 -to-NH 2 transformation. This work provides new insights into the structure-activity relationship of perovskite-based catalysts, and deepens the understanding of the light-driven mechanism.

Mn present signi cant synergistic effects on accelerating the catalytic cycle of standard SCR.Moreover, it is rmly shown that the photocarriers intensify NO oxidation rather than NH 3 -to-NH 2 transformation.This work provides new insights into the structure-activity relationship of perovskite-based catalysts, and deepens the understanding of the light-driven mechanism.

Full Text
Cost-effective construction of NH 3 -SCR catalytic systems to realize NO x removal in the low-temperature range has become urgent given the increasingly stringent environmental requirements.Since Tanaka group 1,2 sparked the interest in photo-assisted selective catalytic reduction of NO (photo-SCR), a wealth of seminal studies [3][4][5][6] have proved that photoirradiation is pretty e cacious in ameliorating NO conversion and N 2 selectivity in the low-temperature range.Although great achievements have been made in this eld, to date, we still lack enough recognition of principles for constructing photo-SCR catalysts.
There are three fundamental issues that plague us.The rst is that the reaction mechanism of NH 3 -SCR is disputed.NH 2 NO and NH 4 NO 2 are the two principal intermediates of standard SCR, which form following the Eley-Rideal (E-R) and Langmuir-Hinshelwood (L-H) mechanisms, respectively.][9][10][11] Some researchers 10,12 proposed that whether NH 4 NO 2 prevails depends on the catalyst nature and reaction temperature.Inexplicably, even for catalysts of the same ingredients, different reaction pathways have been reported. 10,13The inherent complexity of NH 3 -SCR reaction mechanism makes the informed construction of catalyst di cult.
The second issue is that the functions of photocarriers on the catalytic redox are unclear. 5,14 he mainstream view 15,16 believed that light irradiation facilitates the transformation of chemisorbed NH 3 to NH 2 radical, thus accelerating the formation of the unstable intermediate NH 2 NO proceeding via Eley-Ride kinetics.Recently, by illuminated in situ DRIFTS, Gang et al. 17 found that photoirradiation could drive the oxdiation of NO into nitrite and nitrate over Fe-doped TiO 2 .However, two questions remain to be answered : (i) is this promotional effect applicable to other types of photocatalysts, such as perovskites?
(ii) could the photocarriers transform the chemisorbed NH 3 to NH 2 species?Therefore, this fundamental issue about the light-driven effect has not been systematically conclusive.
The last is the elusive correlation between surface activity and optical property, both of which are intrinsically determined by the electronic structure of the catalyst.For perovskite oxides of the ABO 3 formula, the electron occupancy in e g orbitals of the B-site element has been demonstrated as a useful descriptor to approximate the strength of adsorbate binding to the surface. 18Modifying the crystal structure also impacts the optical property, along with the quantum e ciency. 19,20The major challenge of photocatalyst design lies in how to tailor the crystal structure to harmonize the molecule adsorption and light response.
The perovskite structure is exibly tunable since both A and B sites are easily doped by other elements. 21,22Previously, our group 16,23 reported that perovskite-based catalysts have excellent photoactivity for NO removal, and doping perovskite with transition metal ions could evidently improve its performance.In this work, we choose the inert CaTiO 3 as a basal material and construct an e cient structure for photo-SCR deNO x by doping it with Ce, Fe, and Mn elements.From Mn doping to Ce-Fe-Mn doping, each step modi es the crystal structure of CaTiO 3 , making it increasingly elaborate and e cient.
By combining experiments with DFT calculations, a systematic investigation into the relationships between intrinsic activity and crystal structure is implemented, as well as the light-driven mechanism.

Results
Synthesis and catalyst characterization.All the perovskite samples are extracted from Ti-bearing blast furnace slag (Ti-slag) using a molten salt method followed by acid leaching.The chemical compositions of Ti-slag are placed in supplementary Table 1.By roasting Ti-slag with NaOH at 1400°C, a molten system consisting of Na 2 O-CaO-MgO-TiO 2 -Al 2 O 3 -SiO 2 was acquired.With the decrease in temperature, CaTiO 3 crystallized from the molten salt, and the rest formed a zeolite phase (supplementary Fig. 1).Three metal oxides, including CeO 2 , Fe 2 O 3 , and MnO 2 , are also added into the mixture to modify CaTiO 3 .
EDS mapping analysis on the roasted slag in Fig. 1a illustrates that most of Ca, Ti, Ce, and Mn are preferentially concentrated into the perovskite, while Na, Mg, Al, and Si are enriched in the zeolite.The distribution of Fe shows little difference between perovskite and zeolite, perhaps due to its considerable solubility in both two phases.The zeolite is easily removed by acid leaching, and thus high-purity perovskite samples are obtained.
The perovskite obtained by introducing 10 wt% CeO 2 is designated as 10Ce, and so on.The chemical compositions of the resultant samples are listed in supplementary Table 2.All the adscititious metal oxides are well incorporated into the samples, such as 19.7 wt% CeO 2 of 10Ce, 12.0 wt% Fe 2 O 3 of 10Fe, and 10.4 wt% MnO of 10Mn.XRD analysis in Fig. 1b demonstrates that each sample is comprised of a perovskite phase.The (112) peaks from both 10Ce and 10Fe shift to the left relative to pure CaTiO 3 , showing that either Ce or Fe doping enlarges the unit cell.Conversely, Mn doping (10Mn) causes a right shift of (112) peak, lessening the lattice.
The morphology of the 5Ce5Fe10Mn specimen is investigated by HAADF-STEM and displayed in supplementary Fig. 2. In general, the particles from 5Ce5Fe10Mn are irregular and vary from dozens to several hundred nanometers.Elemental mapping on a typical particle (Fig. 1c) revealed that Ca, Ti, Fe, Ce, and Mn uniformly disperse in the particle at the nanometer scale.Hence, the dopants of Ce, Fe, and Mn are incorporated into the CaTiO 3 , forming a homogeneous solid solution rather than a mixture of metal oxides.
EXAFS spectra of 5Ce5Fe10Mn and 10Ce samples are recorded to reveal the positions of dopants in perovskite (Fig. 2).The Fourier-transformed k-space EXAFS spectra are displayed in supplementary Fig. 3, and the tting parameters are placed in supplementary Table 3-5.The tted coordination number of the Fe atom in 5Ce5Fe10Mn is 6.0.The coordination number of the Mn atom is constrained as 6.0 in the tting for Mn K-edge EXAFS spectra, and the R factor is returned as 1.5%.These results con rmed that both Fe and Mn substitute Ti in CaTiO 3 and occupy the B sites.Constraining the coordination number of Ce atom as 12.0 returns 0.5% of R factor, substantiating that Ce substitutes Ca and occupies the A site.A diagram is presented in Fig. 2d to illustrate the crystal structure of Ce-Fe-Mn doped CaTiO 3 .
The speci c surface areas of the samples are in a range of 14.9-16.5 m 2 /g (supplementary Table 6), and the pore volumes are in 0.022-0.033cm 3 /g.The two parameters are almost one magnitude less than the nanocomposites fabricated by homogeneous precipitation 24 , probably due to the fact that the samples are prepared by a high-temperature process.NH 3 -SCR activity of single-element doped CaTiO 3 .The pure CaTiO 3 is hard to remove NO x in 100-300°C (Fig. 3a).After introducing Ce, Fe, or Mn, the modi ed CaTiO 3 becomes active, showing that both A and B sites are responsible for the catalytic activity of perovskite.In terms of identical doping amount in mass ratio, 5Mn is superior to 5Ce and 5Fe.In particular, 10Mn could achieve 97.7% NO x conversion at 250°C, exhibiting outstanding advantage of Mn doping to gift CaTiO 3 activity for NH 3 -SCR deNO x .
In situ DRIFTS spectra of adsorbing NO + O 2 over CaTiO 3 and 10Mn were given in Fig. 3b.Three strong bands peaking at 1565 cm − 1 , 1423 cm − 1 , and 1343 cm − 1 from 10Mn are assigned to bidentate nitrate, nitro compounds, and bridged nitrites, respectively.However, all the three bands from CaTiO 3 are quite weak, indicating that pure CaTiO 3 is di cult to activate gaseous NO molecule.
DFT calculations were performed to further investigate the effect of Mn doping on CaTiO 3 activity.The (001)-plane termination exposing TiO 2 atoms is chosen as a basal surface of CaTiO 3 , which is relatively stable under the anoxic condition. 25,26As shown in Fig. 3c, pure CaTiO 3 is capable of adsorbing NH 3 on the ve-coordinated (Ti 5c ) site with an adsorption energy of -0.86 eV, but hard to capture NO as the adsorption energy is -0.19 eV (Fig. 3d).With the introduction of Mn into CaTiO 3 (supplementary Fig. 4), the adsorption energy of NH 3 slightly increases to -1.09 eV (Fig. 3e).Intriguingly, NO could bond strongly to the Mn 5c site of Ca(Ti, Mn)O 3 (E ads = -1.37 eV) as a bidentate anion carrying 0.20 electrons according to Bader charge analysis (Fig. 3f).
The adsorption con guration of NO on MnO 5 pyramids well satis es the requirement of spatially symmetric overlap between NO π * orbitals and MnO 5 e g orbitals. 27The redundant electrons in NO anion prefer occupying the antibonding orbital 2π of NO, weakening the N-O bond so that the chemisorbed NO is vulnerable to the lattice or chemisorbed oxygens to generate NO 2 (fast SCR) [28][29][30][31][32] , or the chemisorbed NH 3 on adjacent Ti 5c to yield intermediate NH 2 NO (L-H pathway).As depicted in supplementary Fig. 5, the formation of NH 2 NO on the surface of Mn-doped CaTiO 3 is energetically favorable.Therefore, the SCR activity of Ca(Ti, Mn)O 3 derives from the activation of NO by the Mn 5c site.NH 3 -SCR activity of multielement doped CaTiO 3 .Compared with 10Mn, the activity of 5Fe10Mn is signi cantly improved in 100-300°C (Fig. 4a).The NO x conversion over 5Fe10Mn reached 99% at 175°C, far more than the sum of these over sole 5Fe and 10Mn (Fig. 3a).Fe-Mn doping of CaTiO 3 presents synergistic effects on removing NO x .Nevertheless, if we continued doping 5Fe10Mn with 5 wt% CeO 2 , the performance showed little difference between 5Ce5Fe10Mn and 5Fe10Mn in 100-300°C (Fig. 4a).
XPS analysis in Fig. 4b showed that with the incorporation of Fe into 10Mn, the peak assigned to O 1s shifts from 529.0 eV to the higher energy of 529.3 eV, indicating that the surface oxygens from 10Mn carried more electrons than those from 5Fe10Mn.It is easy to understand that the substitution of B-site Ti 4+ by aliovalent Fe 3+ leads to the decrease in the valence electrons transferring to the oxygen anions (Charge distribution of lattice oxygen is placed in supplementary Fig. 6).
The redox behaviors of 10Mn and 5Fe10Mn were also evaluated by H 2 -TPR experiments (supplementary Fig. 7).The strongest peak from 10Mn is centered at 604.4°C, and that from 5Fe10Mn at 566.3°C.Fe addition shifts the reduction peak to the lower temperature, enhancing the diffusion of O 2− anion within the lattice. 33This might be partially ascribed to the oxygen vacancies introduced by the p-type doping of perovskite, which facilitates the movements of electrons and oxygen ions. 34The oxygen vacancies also provide energetically favorable connection channels for gaseous O 2 molecules, triggering the dissociation of O 2 into O atoms, further expediting the oxidation of NO. 35 However, considering that the 5Fe sample as a p-type doped perovskite possesses similar features, the synergetic effects between Fe and Mn dopants could not be ascribed to the oxygen vacancies.
To explore the underlying mechanism, we built a CaTiO 3 model with Fe and Mn atoms situated in the B sites (supplementary Fig. 6c).In this model, the FeO 6 and MnO 6 octahedra are conjoined with each other by a bridged oxygen.DTF calculations reveal that for Mn-doped CaTiO 3 , the gaseous O 2 molecule prefers binding to the Mn 5c and Ti 5c sites with an adsorption energy of -1.26 eV (Fig. 4c).The chemisorbed O 2 could further combine with gaseous NO to produce monodentate nitrate binding to the Mn 5c site (ΔE = -1.65 eV) as shown in Fig. 4d (supplementary Video 1).By contrast, for Fe-Mn doped CaTiO 3 , the gaseous O 2 molecule selectively binds to the Fe 5c site rather than the Mn 5c or Ti 5c site (Fig. 4e, E ads = -1.12eV).
The subsequently generated nitrate is still attached to the Fe 5c site (Fig. 4f, E ads = -2.56eV, supplementary Video 2).The addition of Fe into Mn-doped CaTiO 3 converts the oxidation center of NO to nitrate from the Mn 5c site to the Fe 5c site.This conversion exerts profound in uence on the reaction pathway of the SCR process since nitrate is the precursor of NH 4 NO 2 .
Based on DFT calculations, we propose two reaction routes to describe the standard SCR over Mn and Fe-Mn doped CaTiO 3 .As shown in Fig. 5, the catalytic cycle over Ca(Ti, Mn)O 3 (the outer ring) comprises thirteen steps involving the NH 2 NO and NH 4 NO 2 intermediates. 9,11,13In this cycle, both two intermediates bind to the Mn 5c site thus their formations have to proceed by sequence.Fe-Mn doping makes O 2 selectively bind to the Fe 5c site, lightening the burden of the Mn 5c site, and also increasing the NO coverage on Mn 5c sites.More importantly, NH 4 NO 2 could form in tandem with the generation of NH 2 NO, markedly shortening the reaction pathway.The catalytic cycle over Ca(Ti, Mn, Fe)O 3 could be completed within eight steps (the inner ring in Fig. 5).
Photo-SCR activity of doped CaTiO 3 .Under light irradiation, the performance of perovskite for NO removal is signi cantly improved, especially in 100-200 °C (Fig. 6a).The 99%-conversion temperature over illuminated 5Ce5Fe10Mn is about 135°C, far lower than 175°C over ground-state 5Ce5Fe10Mn (Fig. 4a).As the light turns on at 125°C (Fig. 6b), NO x conversion over 5Ce5Fe10Mn rises from about 56-71% at 125°C without the decrease in N 2 selectivity, demonstrating that photocatalysis could serve as a green and effective technique to promote NO removal.
Pristine CaTiO 3 is hard to harvest the visible light due to its wide bandgap of 3.2 eV.By introducing dopants, the light wavelength absorbed by perovskite extends from the ultraviolet band to the visible band (Fig. 6c).The calculations on the projected density of states (Fig. 6d) disclose that Mn doping introduces an impurity band composed of Mn 3d and O 2p orbitals between the valance and conduction band.For Fe-Mn doped CaTiO 3 , there are four impurity bands, including the donor levels composed of Fe 3d, Mn 3d, and O 2p orbitals below Fermi level, and the acceptor levels composed of Mn 3d and O 2p orbitals above Fermi level.The impurity levels narrow the bandgap and enable the perovskite to harvest the visible light.
Although 5Ce5Fe10Mn shows little difference in NO removal with 5Fe10Mn in the dark (Fig. 4a), the former is notably superior to the latter under light irradiation (Fig. 6a).Li et al. 5 previously reported that PrFeO 3 doped with Ce displays a redshift of UV-vis absorption.Similar phenomena are observed in Fig. 6c.Relative to 5Fe10Mn, 5Ce5Fe10Mn exhibits enhanced light adsorption in the wavelength range from 500 nm to 800 nm.The optimization of optical property by Ce doping improves the perovskite photoactivity.
To unravel photocarriers roles in NO removal, in situ IR spectra of 5Ce5Fe10Mn with gaseous molecules adsorbed were recorded in the dark and light at 125°C.The dark spectra of NH 3 -adsorbed 5Ce5Fe10Mn (black line in Fig. 6e) have four strong peaks.Two peaks around 931 cm − 1 and 966 cm − 1 are assigned to weakly absorbed NH 3 , 36 the one centered at 1250 cm − 1 corresponds to NH 3 coordinated on Lewis acid site, 37 and that around 1366 cm − 1 belongs to NH 4 + bonded to Bronsted acid site. 38The illuminated spectra (red line in Fig. 6e) vary little in comparison to the dark counterpart.Most peaks remain unchanged except for those at 1250 cm − 1 and 1366 cm − 1 .The intensity of the former increases while the latter becomes a bit weak, suggesting that under light irradiation, the adsorption of NH 3 on the Lewis acid site is enhanced while the adsorption on the Bronsted acid site is mildly inhibited.
The characteristic absorption peaks of NH 2 species are generally distributed in the range of 1505-1580 cm − 1 . 39Nevertheless, none of such peaks appears in the dark or illuminated spectra, clearly substantiating that the N-H bond from the chemisorbed NH 3 on 5Ce5Fe10Mn is hard to break at 125°C, even assisted by light irradiation.
There are four strong peaks in the dark spectra of 5Ce5Fe10Mn with NO + O 2 adsorbed (black line in Fig. 6f).These located at 1600 cm − 1 and 1630 cm − 1 are assigned to the bridged nitrate, 17 and those at 1850 cm − 1 and 1960 cm − 1 are ascribed to the weakly adsorbed NO species.The illuminated spectra (red line in Fig. 6f) show apparent differences from the dark counterpart.On the one hand, the intensities of the above four peaks become stronger.On the other, a series of strong peaks appear in the range from 1495 cm − 1 to 977 cm − 1 .The peak at 1100 cm − 1 is assigned to nitrito monodentate species, 40 those around 1200 cm − 1 , 1332 cm − 1 , and 1465 cm − 1 are attributed to nitrites, 33,41 and that at 1495 cm − 1 corresponds to monodentate nitrate. 36These results manifest that light irradiation dramatically promotes NO oxidation rather than NH 3 -to-NH 2 transformation.

Discussion
The conventional fabrication of Ti-based perovskite always uses expensive tetrabutyl titanate as a precursor. 42Nowadays, titanium extraction from minerals still needs to go through complicated procedures with high costs.In this work, CaTiO 3 is synthesized by a molten-salt method with sodium aluminosilicate as the ux.This method provides a possibility to directly extract high-purity Ti-based perovskite from the Ti-bearing minerals because most titanium in the earth exists in the form of oxides.According to our experiments, a series of Ti-based perovskite could be fabricated by this method at low cost, such as PbTiO 3 , BaTiO 3 , LaTiO 3 , and so on.
No matter by Fe-Mn doping or light irradiation, the deNO x activity of modi ed CaTiO 3 mainly originates from the activation of NO.Hence, constructing photo-SCR catalysts from Ti-based perovskite should focus the principle issue on how to activate NO.Jonathan et al. 27 proposed that the binding strength of NO on perovskite surface depends on the B-site e g lling, speci cally, increasing with the decrease in e g lling in the range from 1 to 0.5.According to our results, NO adsorption rises as the e g lling increases from 0 of CaTiO 3 to 0.06 of Ca(Ti, Mn)O 3 , indicating a volcano shape of NO adsorption as a function of e g lling in the range from 0 to 1.The results also suggest that doping Ti-based perovskite with metal ions of e g lling greater than 0, such as Co 3+ , and Ni 3+ , could augment NO adsorption.Furthermore, doping CaTiO 3 simpli es the regulation of electronic structure on the optical property so that more attention could be put into how to activate NO.This strategy also extends the material range for photo-SCR deNO x to semiconductors of the wide bandgap.
It is widely accepted that NH 2 NO is a requisite in the cycle of standard SCR, but the involvement of NH 4 NO 2 is in dispute.Powerful evidence by the time-resolved spectroscopy from Marberger et al. 10 showed that NH 3 coordinated to Bronsted site (B NH3 ) of V A lot of enlightening studies 12,13,43,44 reported that catalysts composed of different metal oxides are always superior to single metal oxide in terms of SCR performance.Most of the authors believed that the redox cycle between multivalent metal ions promotes NO-to-NO 2 oxidation, triggering the fast SCR reaction.Our results demonstrate that Fe-Mn doping of CaTiO 3 acts on the standard SCR by shortening the reaction pathway, providing new insights into the synergistic effect between multivalent metal ions.
Emphasizing the synergistic effect between Fe and Mn does not mean that Ti is dispensible in this system.Instead, Ti of Ca(Ti, Mn, Fe)O 3 has two momentous effects in the SCR process.First, the Ti 5c site selectively adsorbs NH 3 , initiating the catalytic cycle by reacting with chemisorbed NO on the Mn 5c site.
The cooperation between Ti and Mn boosts the formation of NH 2 NO, and is more e cient than that between two Mn sites since the high adsorption strength of NO on the Mn 5c site, to some extent, inhibits NH 3 adsorption along with NH 2 NO formation.Second, Ti 3d orbitals are the main components of the valence band, dictating the optical property of perovskite.Owing to the unique function of Ti, CaTiO 3 is a suitable framework to construct e cient catalytic units for photo-SCR deNO x .
NH 3 -SCR deNO x involves three feeding gases, namely NO, NH 3 , and O 2 .A single type of active site may lead to excessive coverage of the catalyst surface by speci c gas molecules, thus lowering the reaction e ciency.Compared with modulating gas adsorption strength, constructing disparate sites to selectively activate gas molecules could be a more feasible alternative to heighten the SCR reaction rate.

Methods
Materials.The perovskite-based catalyst was directly extracted from Ti-bearing blast furnace slag (Tislag), which is a kind of metallurgical solid waste.More details about Ti-slag could be acquired in the supporting material.The Ti-slag was provided by Pangang Group Co., Ltd., China, and its chemical analysis is shown in Table 1.The slag was crushed, ground and sieved through a 200-mesh sieve before use.Unless otherwise speci ed, all of the rest reagents used in this study were of analytical grade and provided by Sinopharm Chemical Reagent Co., Ltd., China.
Catalyst preparation.To obtain the target perovskite from Ti-slag, a process involving mineral phase reconstruction followed by acid leaching was adopted.To be speci c, a powder mixture composed of 100.0 g of Ti-slag, 23.0 g of NaOH, 10.0 g of MnO 2 , 5.0 g of Fe 2 O 3 , and 5.0 g of CeO 2 was placed in the MgO crucible and roasted at 1400°C for one hour.As the sintered sample was cooled to room temperature, it was crushed and ground into ne particles for acid leaching and characterization.
15.0 g of the sintered sample power was placed in a glass beaker of 500 mL, and then 225 mL of 8% hydrochloric acid was added into the beaker.The mixture was stirred with a Te on agitator at 25°C for 60 min.In this way, almost all the aluminosilicates were dissolved and transferred into the solution.After the slurry was ltered, the solid residue was dried at 105 °C for four hours for further analysis.The asobtained sample was denoted as 5Ce5Fe10Mn, in which the number represents the mass ratio of the following oxide to Ti-slag.Likewise, the label '10Mn' meant that the mass ratio of MnO 2 to Ti-slag in the powder mixture was 10%, and neither CeO 2 nor Fe 2 O 3 was incorporated into the mixture.
Catalyst characterization.X-ray diffraction (XRD) patterns were collected by the Rigaku X-ray diffractometer (Ultima ) with Cu Kα radiation.Mineralogy analysis of the specimen was performed by eld emission scanning electron microscopy (SEM, JSM-6490LV) equipped with an Oxford energy dispersive X-ray spectroscope (EDS).Morphology analysis was carried out by scanning transmission electron microscopy in high-angle annular dark eld (HAADF-STEM, FEI TalosF200x).The high-resolution mapping of elements was implemented by Super-X EDS.
Fe and Mn K-edge EXAFS measurements were carried out at the BL01B1 beamline at Spring-8 of the Japan Synchrotron Radiation Research Institute (JASRI).Ce L -edge EXAFS measurement was collected at the Beamline of TPS44A1 in National Synchrotron Radiation Research Center (NSRRC).All the measurements were performed at room temperature, and XAFS data were recorded in the uorescent mode.The X-ray energy was calibrated using Fe and Mn foils, and the data analyses were performed utilizing Athena and Artemis in the Demeter software package.X-ray photoelectron spectra (XPS) were recorded on a Thermo Scienti c K-Alpha spectrometry using Al Kα irradiation.Raman spectra were obtained by the Renishaw Raman spectrophotometer.UV-vis diffuse re ectance spectra (DRS) of the specimens were recorded by a PerkinElmer Spectrometer (Lambda 950).The speci c surface area of the specimen was measured by Brunauer-Emmett-Teller (BET) method.Both temperature-programmed reduction of H 2 (H 2 -TPR) and desorption of NH 3 (NH 3 -TPD) were implemented on a Huasi instrument (DAS-7000).
In situ DRIFT experiments were performed on a Nicolet 6700 spectrometer in the wavenumber range of 2000 to 1000 cm − 1 with 4 cm − 1 resolution.In situ IR spectra were collected on a PerkinElmer spectrometer in the wavenumber range of 2000 to 800 cm − 1 with 4 cm − 1 resolution.The detailed procedures for in situ IR spectrum are as follows.First, the specimen was treated at 300 °C under N 2 ow for 0.5 h to remove the species adsorbed on the specimen surface.Second, the specimen was regulated to the target temperature to obtain a background spectrum which should be subtracted from the specimen spectrum.Then, the specimen was exposed to a ow of 1000 ppm NH 3 or 1000 ppm NO + 3 vol% O 2 for one hour, and the IR spectrum in the dark was recorded.Finally, the Xe lamp was turned on, and the IR spectrum under light irradiation was collected 15 min later.
Catalytic test.The SCR activity was tested in a xed-bed reactor equipped with a Xenon lamp.The simulated gas consisted of 1000 ppm NO, 1000 ppm NH 3 , 3 vol% O 2 , and a balance of N 2 , corresponding to a gas hourly space velocity (GHSV) of 72000 h − 1 .Both NO and NO 2 were determined by a BUV150 ue gas analyzer, NH 3 determined by a BLD200 analyzer, and N 2 O by a BGM250 analyzer.During the catalytic process, the reactor was cooled by circulating water to offset the temperature uctuation caused by light irradiation.The NO conversion percentage and N 2 selectivity were calculated by the following formula: Computational details.The structural relaxation and singlet energy point energy calculations were performed using the density functional theory (DFT) method as implemented in Vienna Ab initio Simulation Package (VASP) 45,46 .The generalized gradient approximation (GGA) in the form of Perdew-Burke-Ernzerhof (PBE) functional was used to evaluate the exchange-correlation potential 47 .The projectaugmented wave approach was employed to represent the core-electron interaction.The atomic orbitals treated as valence states were as follows: 3s3p4s of Ca, 3d4s of Ti, 3p3d4s of Mn, 3p3d4s of Fe, and 2s2p of O. Hubbard corrections with U eff = 3.9 eV and 5.3 eV were applied to describe the 3d-orbitals of Fe and Mn atoms, respectively.
The lattice constant of relaxed bulk CaTiO 3 was 5.36×5.46×7.62Å in reasonable agreement with the experimental value of 5.38×5.44×7.64Å (PDF 078-1013) 48 .A periodic CaTiO 3 (001) slab separated by a vacuum layer of 15 Å was built from this lattice.In total, there are 80 atoms in each slab, containing 16 Ca, 14 Ti, 1 Mn, 1 Fe, and 48 O.The self-consistent eld method with the tolerance of 1.0×10 − 6 eV/atom and the plane wave basis sets with an energy cutoff of 500 eV were adopted.The structural optimization was relaxed until the Hellman-Feynman force on each ion was less than 0.02 eV/Å.
Adsorption energies of gas molecules, E ads , in this paper were calculated as the following equation:

Figure 5 Proposed
Figure 5

Figure 6
Figure 6 2 O 5 -WO 3 -TiO 2 does not participate in the catalytic cycle.NH 3 binding to Lewis sites (L NH3 ) could react with gaseous NO to form NH 2 NO, generating a Bronsted acid site.However, if L NH3 could combine with this as-obtained Bronsted site, a new B NH3 would be produced.This B NH3 could be consumed by forming NH 4 NO 2 so that the total amount of B NH3 remains constant, also explaining the results of Fig. 1 in ref 10.Therefore, the reaction mechanisms involving both NH 2 NO and NH 4 NO 2 are also supported by Marberger's results.