Simultaneous NO x and Particulate Matter Removal from Diesel Exhaust by Hierarchical Fe-Doped Ce-Zr Oxide.

Particulate matter and NO x emissions from diesel exhaust remains one of the most pressing environmental problems. We explore the use of hierarchically ordered mixed Fe-Ce-Zr oxides for the simultaneous capture and oxidation of soot and reduction of NO x by ammonia in a single step. The optimized material can effectively trap the model soot particles in its open macroporous structure and oxidize the soot below 400 °C while completely removing NO in the 285-420 °C range. Surface characterization and DFT calculations emphasize the defective nature of Fe-doped ceria. The isolated Fe ions and associated oxygen vacancies catalyze facile NO reduction to N2. A mechanism for the reduction of NO with NH3 on Fe-doped ceria is proposed involving adsorbed O2. Such adsorbed O2 species will also contribute to the oxidation of soot.

Introduction

Air pollution caused by exhaust gas emissions from various modes of transportation carries significant risk for human health and the environment.[1−4] Introduced in the 1970s, the three-way catalytic convertor has become a widespread technology for removing noxious gases from gasoline-fueled cars.[5] Precious group metals (PGMs) dispersed as nanoparticles on suitable oxide support materials can simultaneously oxidize CO and hydrocarbons and reduce NO to less harmful gases. This technology cannot be used to remove NO from the exhaust of diesel engines because it is too rich in oxygen. Aside from NO, diesel exhaust remains a major contributor to undesirable emissions of particulate matter (PM). Soot particles pose the most serious threat to human health. The major challenge in diesel exhaust cleanup is the removal of NO under lean (oxygen-rich) conditions.[6,7] Yoshida et al. were the first to propose the simultaneous removal of PM and NO by a single catalytic material.[8] Significant efforts have been made to develop suitable catalysts for this purpose.[9] Current commercial solutions combine a diesel oxidation catalyst (DOC) for the removal of CO and hydrocarbons, a catalyzed diesel particulate filter (CDPF) for soot filtration, and a selective catalytic reduction (SCR) step to remove NO using a reducing gas such as ammonia. These operations are carried out in different compartments, thereby increasing the size and cost of this technology. Another drawback is that in some steps expensive PGMs such as Pt are important catalyst ingredients.[10,11] Consequently, there is significant incentive to develop novel approaches that rely on more abundant elements and combine one or more pollutant conversion steps.[12]

A potential alternative is to combine the CDPF and SCR functions in a selective catalytic reduction and particulate filter (SCRPF). The particular challenge here is to achieve a high rate of soot oxidation in combination with substantial NO reduction at sufficiently low temperature. Therefore, it is necessary to identify materials with suitable redox abilities. Candidate materials are (mixed) metal oxides,[13,14] hydrotalcites,[15] and alkali oxides.[16] Besides high activity, increasing the contact area between the catalysts and solid reactant is a particular challenge in this field.[17,18] It is also important that the texture of these materials be suitable for capturing soot particles, which are typically larger than 25 nm. In such case, hierarchically structured oxides may be considered. Three-dimensionally ordered macroporous (3DOM) materials offer an ordered, interconnected macroporous structure with openings suitable for the capture of soot particles.[19]

Ceria is well-known for its excellent oxygen storage capacity.[20,21] The problem of low high-temperature stability of ceria structures can be overcome by introducing foreign elements into the ceria lattice, which also improves its redox properties.[22−25] For instance, Ce–Zr mixed oxides have been explored in the context of NO reduction and soot oxidation.[26−32] Other reports have already shown that doping Fe into ceria improves its reducibility, leading to more facile generation of oxygen vacancies at the surface important for soot oxidation and NO reduction.[33,34]

Herein, we report for the first time an Fe-doped 3DOM mixed Ce–Zr oxide material that can simultaneously remove PM and NO from diesel exhaust. Ammonia is used as a reductant for NO. We prepared the 3DOM mixed oxides by a carbon-templating method and varied the Fe content in the mixed oxide. Optimized materials show good performance in simultaneous removal of soot and NO at intermediate temperatures. The 3DOM mixed oxides are thermally stable and can be repeatedly regenerated without loss of activity. Density functional theory (DFT) calculations have been performed to understand the surface reducibility of the mixed oxides and gain insight into the role of Fe and surface oxygen vacancies in the reaction mechanism of NO reduction and soot oxidation.

Experimental Section

Materials Synthesis

All starting chemicals were purchased from Sigma-Aldrich and used without further purification. Carboxy-modified poly(methyl methacrylate) (c-PMMA) spheres were prepared by a modified emulsifier-free biphasic emulsion polymerization technique using initiators for the water and oil phase.[35−37] Methyl methacrylate (MMA, 99%) was the monomer used for obtaining PMMA spheres. Addition of acrylic acid (AA, >99%) monomer to the mixture allowed for introducing carboxyl groups in the PMMA. Briefly, a four-necked, 1000 mL round-bottomed flask was filled a mixed solution of acetone (80 mL, >98%), distilled water (240 mL), and the monomers (120 mL). The resulting mixture was heated to 80 °C by a hot water bath. After approximately 30 min, 0.6 g of potassium persulfate (KPS, water-phase initiator, >99%) and 0.15 g of azodiisobutyronitrile (AIBN, oil-phase initiator, 98%) mixed with 40 mL of distilled water (preheated to 80 °C) were added. The whole solution was stirred at a constant speed of 350 min–1 for approximately 2 h with N2 bubbling. The obtained latex was cooled to room temperature and then centrifuged. The solid material was dried at room temperature (c-PMMA).

Three-dimensionally ordered macroporous (3DOM) Ce0.9–FeZr0.1O2 catalysts were prepared by carboxy-modified colloidal crystal templating (CMCCT). Ce(NO3)3·6H2O (99.5%), Fe(NO3)3·9H2O (99.99%), and ZrOCl2·8H2O (98%) were used as precursors for obtaining mixed metal oxides. Suitable amounts of Ce(NO3)3·6H2O, Fe(NO3)3·9H2O and ZrOCl2·8H2O were first dissolved in a mixture of ethylene glycol and methanol followed by vigorous stirring for 40 min. Then, this solution was contacted with the c-PMMA hard template for 12 h. After impregnation, the final material was subjected to vacuum filtration to remove excess precursor solution. The precipitate was dried at 50 °C in a vacuum oven, calcined in inert (Ar) atmosphere at 130 °C for 1 h, followed by increasing the temperature to 600 °C at a rate of 1 °C/min. After a dwell time of 5 h, the atmosphere was changed to air, and the sample was kept at 600 °C for another 3 h. The first step in Ar pyrolyzes the carbon: the sp2-hybridized carbon atoms are converted to a sturdy amorphous carbon material, which acts as the hard template for the in situ formation of the 3DOM mixed oxide. The carbon template was finally removed by calcination in air.

Catalyst Characterization

The crystal structure of the samples was investigated by powder X-ray diffraction (XRD) spectrometer (Shimadzu XRD 6000) with Cu Kα radiation (0.02° intervals in the range 5–90° at a rate of 4°/min). Nitrogen adsorption isotherms were measured using a Micromeritics TriStar-II 3020 instrument. SEM (FEI Quanta200F) was conducted to analyze the surface morphology of the samples. The microstructure and lattice parameters were analyzed by TEM (JEOL JEM 2100 electron microscope). Raman spectra were collected in the anti-Stokes range of 100–2000 cm–1 using an inVia Reflex-Renishaw spectrometer. The sample was excited using a He–Gd laser (532 nm excitation wavelength). X-ray photoelectron spectra were measured on an XPSPHI-1600 ESCA spectrometer using an Al Kα anode (hν = 1253.6 eV) as the X-ray source and using C 1s at 284.6 eV as an internal binding energy standard. Temperature-programmed desorption of ammonia (NH3-TPD) was carried out in a conventional flow apparatus using a thermal conductivity detector. Temperature-programmed reduction with H2 (H2-TPR) measurements was performed in an Autosorb IQ Quantachrome apparatus.

Catalytic Activity Measurements

Catalytic activity measurements were taken in a fixed-bed reactor. Printex U carbon black (Orion Engineered Carbons) was used as a model for particulate matter. This carbon black has an average particle size of 25 nm and surface area of 100 m2/g. Prior to each catalytic activity test, 100 mg of catalyst and 10 mg of Printex U were mixed gently with a spatula (loose contact mode). Thereafter, the mixture was placed between quartz wool plugs in a quartz tubular reactor with an inner diameter of 10 mm. The reactor feed was comprised of 1000 ppm of NO, 1000 ppm of NH3, and 3% O2 with N2 as the balance gas. In some cases, 5% H2O was added to the reactor feed to evaluate the influence of moisture. The gas hourly space velocity (GHSV) was 25,000 h–1 with a total flow of 100 mL/min at standard pressure and temperature. The performance of the optimum catalyst was also evaluated at higher GHSV by decreasing the catalyst amount. The concentrations of NH3, NO, NO2, N2O, CO2, and CO were monitored at the outlet by online infrared spectroscopy (Thermo Is50 FTIR equipped with a 2.4 m gas cell). For quantification, a robust method for multicomponent gas analysis was used implementing TQ Analyst software and making use of calibration curves based on mixtures of the relevant gases in different concentration ranges.[38] Before each catalytic activity measurement, the catalyst sample was first swept by a flow of 100 mL/min N2 for approximately 45 min prior to collecting a background IR spectrum of the reactor effluent. Afterward, effluent IR spectra were recorded of the reactor feed consisting of 1000 ppm of NH3, 1000 ppm of NO, and 3% O2 in N2 for 30 min. Catalytic activity tests were carried out by heating the reactor bed from 30 to 600 °C at a rate of 3 °C/min. The stability of the catalyst was evaluated by repeatedly evaluating its performance in this manner. For this purpose, 10 mg of Printex U was mixed with the catalyst bed. The absence of mass transfer limitations for the NO reduction reaction was verified by applying the Koros–Nowak criterion, and the absence of heat transfer due to soot oxidation was evaluated by Mears’ criteria (see the Supporting Information).

Computational Methods

DFT calculations were performed using the VASP code employing the GGA-PBE exchange-correlation potential.[39] The valence electrons (5s, 4f, 3d for Ce; 2s, 2p for O; and 4s, 3d for Fe) were expanded in a plane-wave basis set with a cutoff energy of 400 eV. The projector augmented wave method (PAW) was used to describe the effect of core electrons.[40,41] The bulk equilibrium lattice constant of ceria (5.49 Å) previously calculated by PBE + U (Ueff = 4.5 eV) was uSed.[42] Then, a 3 × 3 surface unit cell was used for the CeO2 (111) surface. Fe atoms and the six top atomic layers of the ceria slab were allowed to relax, whereas the three bottom layers were kept fixed to their bulk positions. The vacuum gap thickness was set to 15 Å. Because of the large size of the slab model (11.64 Å × 11.64 Å), a Monkhorst pack 1 × 1 × 1 mesh was used for Brillouin zone integration. All structures were relaxed until the forces acting on each atom were smaller than 0.05 eV/Å. To improve the description of the on-site Coulomb interactions in the Ce-f states and Fe-d states, a Hubbard correction was added. For Ce, a value of Ueff = 4.5 eV was used for its 4f orbital.[43−45] For Fe, a value of Ueff = 3.8 eV was used for its 3d orbital.[46] The location and energy of transition states were calculated with the climbing-image nudged elastic band method (CINEB).[47] Adsorption energies are expressed with reference to the adsorbing molecule in vacuum. The energies of all gas species were determined in a 15 Å cubic box with a cutoff energy of 400 eV at the Γ-point.

Results and Discussion

Preparation and Characterization

The carboxy-modified variation of colloidal crystal templating using poly(methyl methacrylate) (c-PMMA) spheres is schematically depicted in Figure .[35] The sturdy amorphous carbon material derived from PMMA pyrolysis can be used as a hard template for the fabrication of structured metal oxides.[48] The carboxy modification of PMMA using acrylic acid as a comonomer was necessary to obtain a well-mixed Ce–Zr oxide structure. We prepared c-PMMA spheres by copolymerization of MMA and AA using suitable initiators. Centrifuging and drying of the latex resulted in a highly ordered c-PMMA material. The structured oxides were obtained by impregnation of the solid organic template with a mixture of suitable precursor salts dissolved in a mixture of ethylene glycol and methanol followed by pyrolysis at 600 °C in inert and calcination in air to remove the organic part. TEM images show the ordered texture of the optimum Ce0.8Fe0.1Zr0.1O2 mixed oxide with macropores and uniformly sized walls (Figure b–d) interconnected by smaller windows.[49] All materials have the fluorite structure of ceria independent of the Fe and Zr content, and no separate iron or zirconium oxide phases were detected by XRD (Figure S2). Small shifts in the main diffraction peaks for the mixed oxides compared to CeO2 evidence inclusion of Fe3+ and Zr4+ into the fluorite structure of ceria. High-resolution TEM images show (111) surface terminations with the d-spacing being consistent with that of ceria (Figure e).

Figure 1

(a) Schematic representation of the synthesis of the 3DOM mixed Ce–Fe–Zr oxide and its catalytic function in diesel exhaust cleanup; (b) SEM and (c–e) TEM images at different magnifications showing the macroporous structure (c, d) and d-spacing of CeO2(111).

Introduction of Fe and Zr into the CeO2 lattice did not alter the 3DOM structure as long as the Fe substitution level was kept below 0.2 (Figures S3 and S4). Introduction of more Fe led to segregated iron oxides observable in high-resolution TEM images (Figure ).[50] Raman spectra of the 3DOM Ce0.9–FeZr0.1O2 samples contain an absorption band at ∼460 cm–1 due to the F2g mode of CeO2 (Figure ).[24,51] Only at higher Fe content (x ≥ 0.2) did Raman bands at 215 and 280 cm–1 typical of Fe–O stretching vibrations in Fe oxides appear. The nitrogen adsorption–desorption isotherms show a nearly linear correlation between the relative pressure and absorbed volume (Figure S4), which is the consequence of unrestricted monolayer-multilayer adsorption. The presence of an H3 hysteresis loop is further indication of the macroporous structure. Although the pure ceria material has a surface area of approximately 12 m2/g (Table S1), the mixed oxides have higher surface area, which is in part due to the presence of mesopores evident from the hysteresis in the p/p0 range between 0.4 and 0.8. These mesopores are likely occluded in the walls of the macroporous material.

Figure 2

Representative TEM images of 3DOM materials: (a) CeO2, (b) Ce0.85Fe0.05Zr0.1O2, (c) Ce0.8Fe0.1Zr0.1O2, (d) Ce0.7Fe0.2Zr0.1O2, (e) Ce0.6Fe0.3Zr0.1O2, and (f) Ce0.5Fe0.4Zr0.1O2.

Figure 3

Raman spectra of the of 3DOM materials: (a) CeO2, (b) Ce0.85Fe0.05Zr0.1O2, (c) Ce0.8Fe0.1Zr0.1O2, (d) Ce0.7Fe0.2Zr0.1O2, (e) Ce0.6Fe0.3Zr0.1O2, and (f) Ce0.5Fe0.4Zr0.1O2.

Catalytic Activity Measurements

Compared with ceria, mixed Ce–Zr oxides display better thermal stability and oxygen storage capacity, which is beneficial for PM combustion.[52] In general, it is a challenge to reduce NO under the oxygen-rich conditions required to oxidize PM into CO2.[53] As NO2 is more effective in soot oxidation than O2, soot is usually first oxidized in the NO/O2 exhaust gas, followed by ammonia-assisted NO reduction using, for instance, Cu/zeolites placed downstream of the PM combustion zone.[54−57] Ammonia can be conveniently supplied to the after-treatment system by hydrolyzing urea. It has been demonstrated before that Fe is an active ingredient for NO reduction.[58]

We optimized the Fe and Zr content of the 3DOM mixed Fe–Ce–Zr oxide toward low-temperature NO reduction and complete soot oxidation. For this purpose, model soot particles with an average size of 25 nm were loosely mixed with the 3DOM mixed oxide catalysts and exposed to a simulated diesel exhaust feed containing 1000 ppm of NO, 1000 ppm of NH3, and 3% O2 with balance N2 fed at a GHSV of 25,000 h–1. The loose contact mode provides a better approximation of soot trapping in a DPF than tight contact conditions involving grinding the components in a mortar.[7]Figure shows the transient behavior of the catalyst during temperature-programmed reaction. CO2 is produced by combustion of the model soot particles. The effluent CO2 concentration decreases at high temperature, as combustion of the model soot near completion. NO conversion at high temperature is limited because of the oxidation of NH3 (Figure S5). In low temperature NH3–SCR, NO oxidation to NO2 is crucial to improve the rate of NO removal via the fast SCR reaction.[59,60] Moreover, NO2 is also a more active soot oxidant than NO.[61] The 3DOM Ce0.8Fe0.1Zr0.1O2 catalyst shows excellent activity in the oxidation of NO to NO2 (Figure S6). The optimal catalyst Ce0.8Fe0.1Zr0.1O2 is effective for reducing NO by 90% in the range of 265–420 °C and for completely oxidizing soot to CO2 at approximately 375 °C. Among the 3DOM mixed Fe–Ce–Zr–O catalysts (Table ), the optimized material is able to oxidize coke below 400 °C. When the Fe content is too high, the performance was much lower because segregated Fe oxides block the surface of the solid Fe–Ce–Zr–O solution.[62] Consistent with this, Fe2O3 itself showed low activity in PM oxidation and NO SCR.

Figure 4

(left) CO2 concentration and (right) NO conversion as a function of temperature upon exposure of 3DOM Ce0.9-xFeZr0.1O2 catalysts loosely mixed with model soot particles in a gas feed containing 1000 ppm of NH3, 1000 ppm of NO, 3% O2 and balance N2 at a gas hourly space velocity of 25,000 h–1.

Table 1

Performance of Structured Oxides in Simultaneous NO Reduction and PM Combustiona

catalyst	Tmax,CO2b (°C)	Tmax,NOc (°C)
Fe2O3	514	526
CeZrO2	477	418–523
Ce0.85Fe0.05Zr0.1O2	398	387–438
Ce0.8Fe0.1Zr0.1O2	375	285–410
Ce0.7Fe0.2Zr0.1O2	409	372–448
Ce0.6Fe0.3Zr0.1O2	433	404–510
Ce0.5Fe0.4Zr0.1O2	442	408–523

Catalyst (100 mg) loosely mixed with 10 mg of Printex U model soot particles, 1000 ppm of NO, 1000 ppm of NH3, and 3% O2 and balance N2 at a gas hourly space velocity of 25,000 h–1.

Temperature of maximum CO2 concentration.

Temperature range where NO conversion is complete.

We also evaluated the performance of the catalyst in the presence of water. Adding 5% H2O to the reactor feed, the catalytic performance for PM oxidation was decreased, whereas that for NO reduction was only slightly lower in comparison to the experiments without water (Figure S7). Complete reduction of NO was achieved in the 343–426 °C range, and soot was completely combusted at 421 °C. Clearly, water had a negative effect on low-temperature NO conversion but improved NO reduction rate at high temperature. The strong influence of water on NO reduction is due to competitive adsorption of NH3 and H2O. This limits NH3 adsorption on acid sites at low temperature, thus decreasing low temperature NO reduction. On the other hand, at high temperature, the inhibiting effect of H2O slows NH3 oxidation, resulting in a higher NO reduction rate.

As the used space velocity was relatively low with respect to diesel exhaust gas treatment, we also evaluated the performance of the optimum 3DOM Ce0.8Fe0.1Zr0.1O2 at higher space velocities (GHSV of 50,000 and 100,000 h–1). Figure S8 shows that under these more stringent conditions catalytic performance was decreased. Complete NO conversion was still obtained in the 338–420 °C temperature range at a GHSV of 50,000 h–1, whereas at the highest GHSV, the maximum NO conversion was limited to 80%. The PM combustion rate displayed maxima at 407 and 435 °C for GHSV values of 50,000 and 100,000 h–1.

Figure shows that the optimized 3DOM Ce0.8Fe0.1Zr0.1O2 catalyst can be reused without loss of activity for five consecutive cycles with fresh model soot being added after each cycle. Because the ceramic materials may be exposed to high temperatures in real applications, we also aged the optimum 3DOM mixed oxide at 900 °C in air for 5 h. This had only a minor effect on the catalytic performance (Figure S9) with the maximum rate of soot combustion being observed at 398 °C and full NO conversion in the 327–420 °C range. SEM shows that the texture of the 3DOM mixed oxide is largely retained, emphasizing its good thermal stability (Figure S10). Comparison of the catalytic performance of the optimum 3DOM catalyst to literature data emphasizes the outstanding performance in combined soot oxidation and NO reduction (Table S2). SEM images of the original and the catalyst used in five consecutive cycles demonstrates that the structured mixed oxide is thermally stable in the experiments (Figure ).

Figure 5

Reuse of the optimal 3DOM Ce0.8Fe0.1Zr0.1O2 catalyst during five consecutive cycles (the spent catalyst was mixed with new Printex U model soot particles and re-evaluated under similar conditions; GHSV = 25,000 h–1, 1000 ppm of NH3, 1000 ppm of NO, 3% O2 in N2, 0.1 model soot/catalyst mass ratio).

Figure 6

SEM of the (a) fresh and (b) spent optimal 3DOM Ce0.8Fe0.1Zr0.1O2 catalyst after five cycles; TEM images of (c) Printex U and (d) 3DOM Ce0.8Fe0.1Zr0.1O2 catalyst mixed with Printex U after temperature-programmed oxidation until 250 °C (reaction conditions: GHSV = 25,000 h–1, 1000 ppm of NH3, 1000 ppm of NO, 3% O2 in N2, 0.1 model soot/catalyst mass ratio).

Figure also displays the performance of a nontemplated mixed oxide of the same composition as the optimal one. Soot combustion is delayed too much at higher temperatures, presumably because of the much less efficient contact of the soot particles with the surface of the mixed oxide. PM oxidation can enhance NO reduction by involving C=O groups on soot, which are intermediates in the complete oxidation of soot.[63] Compared with the large pores of the 3DOM structure, the average pore size of the nontemplated mixed oxide is only 15.8 nm, too small for the model soot particles to enter. Thus, the soot particles can only interact with a much smaller portion of the mixed oxide surface. The strong influence of the texture together with the use of the loose mixing method suggests that the model soot particles will enter the pores of the 3DOM structure during the performance test. This supposition is confirmed by studying a 3DOM sample that was only heated to 250 °C. Figure shows TEM images of the model soot as well as a soot particle trapped in the large pores of the 3DOM structure after heating to 250 °C. On the other hand, the rate of reduction of NO was substantially lower for the nonstructured mixed oxide catalyst. As its surface area is higher than that of the 3DOM mixed oxide, the lower performance suggests that the surface of the nontemplated mixed oxide has a different composition, likely containing fewer Fe sites.

Temperature-programmed reduction (H2-TPR) profiles of 3DOM Ce0.9–FeZr0.1O2 catalysts demonstrate the better reducibility of the Fe-doped Ce–Zr mixed oxides compared with CeO2 and Ce–Zr oxide (Figure ). The 3DOM ceria sample shows two reduction maxima at 550 and 820 °C due to surface and bulk reduction. The mixed CeZrO2 sample shows one reduction feature at 591 °C. Inclusion of Zr in the ceria lattice is known to increase the reducibility of the bulk of ceria.[64,65] An additional low-temperature reduction feature in the 425–476 °C range appears in the Fe-doped mixed oxides. It occurs at the lowest temperature for the best-performing Ce0.8Fe0.1Zr0.1O2 sample. The high-temperature reduction features observed for samples at higher Fe content are due to reduction of Fe2O3.[66,67] In line with the low NO SCR performance of the nontemplated mixed oxide, TPR shows that the surface reduction occurs at relatively high temperature and with relatively low hydrogen consumption. This suggests that a relatively small part of Fe is built into the ceria, indicating that the CMCCT method is conducive to generating highly dispersed Fe species in the ceria surface.

Figure 7

H2-TPR traces of the 3DOM materials: (a) CeO2, (b) CeZrO2, (c) Ce0.85Fe0.05Zr0.1O2, (d) Ce0.8Fe0.1Zr0.1O2, (e) Ce0.7Fe0.2Zr0.1O2, (f) Ce0.6Fe0.3Zr0.1O2, and (g) Ce0.5Fe0.4Zr0.1O2, and (h) nontemplated Ce0.8Fe0.1Zr0.1O2.

DFT Calculations

To better understand how doping with Fe enhances the reducibility of ceria and catalytic performance, we performed DFT + U calculations using a CeO2 (111) surface model in which one Ce atom was substituted by an Fe atom. We choose the (111) surface, as it is the most stable termination of ceria. We studied the oxygen formation energy of Fe-doped ceria as well as a reaction mechanism for the oxidation of NO to N2 by NH3 and O2. Finally, we also discuss the role of adsorbed O2 in the oxidation of coke.

Compared with the high oxygen vacancy formation energy of the stoichiometric (111) surface of ceria (2.1 eV, 1 eV ≈ 96 kJ/mol), removing an oxygen atom from the Fe-doped CeO2(111) surface is exothermic by −0.10 eV. This result implies that the ceria surface will already contain oxygen vacancies. The energy to remove a second O atom close to the first one is 1.39 eV, which is still substantially lower than the oxygen vacancy formation energy of the stoichiometric ceria surface. Thus, the first reduction feature observed in the H2-TPR traces of the Fe-doped samples is because of the removal of a second O atom close to the Fe substitution. The DFT calculations predict that removing this O atom results in two Ce3+ surface atoms (Figure a). In keeping with this, XPS confirms that the Fe-containing samples contain more Ce3+ than the Fe-free reference sample (Table ). The highest Ce3+/Ce4+ ratio was observed for the most active sample. XPS also demonstrates that the surface contains the highest amount of surface adsorbed oxygen in the form of O22– and O–. We speculate that the oxygen species at higher binding energy are, because of molecular oxygen, strongly adsorbed on oxygen vacancies in close proximity to the Fe dopant in the ceria surface. DFT calculations show that molecular O2 strongly adsorbs to the defective Fe-substituted ceria surface, forming O22– species (Eads = −0.71 eV, Figure b).

Figure 8

(a) Structure of Fe-doped CeO2(111) as the stoichiometric surface and with one and two oxygen vacancies; (b) adsorption of NH3 and NO on Fe-doped CeO2(111) with one oxygen vacancy (Fe1Ce1–O2–(111)); Fe-doped CeO2(111) with the oxygen vacancy preadsorbed by O2 (O2*Fe1Ce1–O2–(111)). Color scheme: white, Ce4+; green, Ce3+; red, O; orange, O to be removed; purple, Fe; blue, N; bright white, H.

Table 2

Surface Composition and Oxidation State for the 3DOM Materials As Probed by XPS

				O 1s envelope
	Ce 4f envelope			surface O		lattice O
catalyst	Ce3+ (%)	Ce4+ (%)	Ce3+/Ce4+	O–(%)	O2–(%)	O2–(%)	ratioa
CeZrO2	21.1	78.9	0.267	7.4	19.4	73.2	0.366
Ce0.85Fe0.05Zr0.1O2	26	74.1	0.351	4.6	30.5	64.9	0.541
Ce0.8Fe0.1Zr0.1O2	26.3	73.7	0.357	4.7	30.7	64.6	0.548
Ce0.7Fe0.2Zr0.1O2	24.9	75.1	0.331	11.8	22.6	65.6	0.524
Ce0.6Fe0.3Zr0.1O2	23.5	76.5	0.307	8.9	23.8	67.3	0.486
Ce0.5Fe0.4Zr0.1O2	23	77	0.299	6.2	24.5	69.3	0.443

Ratio of surface and lattice oxygen.

The combined results of surface characterization and catalytic testing emphasize the unique properties of Fe atoms doped into ceria toward NO reduction with NH3 combined with soot oxidation. To gain better insight into the role of ceria doping with Fe, we investigated the mechanism of NO reduction by NH3 through DFT calculations (Figure ). We started the catalytic cycle from the stable surface under oxygen-rich conditions, i.e., the surface that contains O2 adsorbed on the oxygen vacancy close to the Fe site (state i). NO strongly adsorbs on the exposed Lewis acid Fe site (state ii, ΔEads = −2.15 eV). The adsorbed NO molecule will easily react with a ceria lattice O atom to form nitrite (state iii). The activation barrier determined by the climbing image nudged elastic band method is 0.39 eV. Although state iii is slightly less stable than state ii, the formation of the nitrite species allows NH3 to adsorb on the Lewis acid Fe site. This adsorption is strong with ΔEads = −1.14 eV (state iv). The formation of nitrite stores NO on the surface and alleviates the competition between NO and NH3 for adsorption on the catalytic surface. This Langmuir–Hinshelwood mechanism is entropically favored over the Eley–Rideal alternative involving direct reaction of NO from the gas phase with a lattice O atom. NH3-TPD confirms that ammonia is more strongly adsorbed on the Fe-doped samples than on Fe-free samples (Figure S12).

Figure 9

Potential energy diagram of NO reduction with key reaction intermediates. State v and v′ represent N–H dissociation in adsorbed NH3 by lattice O and adsorbed O2, respectively.

An aspect worth discussing is that the adsorption of O2 on the oxygen vacancy oxidizes Fe2+ to Fe3+. Consequently, NH3 adsorbs stronger on the surface in the presence of coadsorbed O2 (ΔEads,NH = −1.25 eV) than in its absence (ΔEads,NH = −0.66 eV) (Figure b). Furthermore, NO adsorption is stronger on Fe3+ (ΔEads,NO = −2.15 eV) than on Fe2+ (ΔEads,NO = −0.28 eV) (Figure b). Both effects are expected to increase the rate of the NO SCR reaction.

The catalytic cycle continues by reaction of the nitrite species with adsorbed ammonia. First, one of the N–H bonds of chemisorbed NH3 is activated by a basic O atom to form adsorbed OH and NH2 surface species. Because of its higher basicity, H abstraction by a ceria lattice O2– ion is preferred (state v, ΔEreaction = 0.99 eV) over abstraction by coadsorbed O22– (state v′ represented in Figure , ΔEreaction = 1.55 eV). The resulting NH2 radical will then react with NO to form ONNH2 as a reactive intermediate. The activation barrier for this process is very low (ΔEbarrier = 0.21 eV). For the decomposition of this complex, we follow the mechanism identified in gas phase cluster studies of VO3 and V2O5 with NO and NH3.[68−70] By abstraction of another H atom and proton transfer from the OH group (ΔEreaction = 0.92 eV), the HONNH surface intermediate is obtained, which weakly binds via its OH moiety to the Fe site (state vii, ΔEads = 0.20 eV). Such intermediates are very unstable[71] and decompose without activation barrier to gaseous N2 and, in this case, two OH groups, one bridging between two Ce ions and one coordinating to the Fe cation (state viii). These reaction events are very exothermic (ΔEreaction = −3.76 eV). The surface then contains three OH groups (the three H atoms originate from ammonia; the O atom is one of the OH groups from NO). One OH group and one proton are removed as water (state ix, ΔEdes = 1.4 eV). The proton left behind will remove an O atom from the surface as water together with another proton obtained in a subsequent similar reaction cycle. The energetics of subsequent cycles should be very similar to the above-described cycle.[72] Finally, the resulting O vacancies will be filled by dissociating O2. Taken together, these reactions amount to the overall 4 NO + 4 NH3 + O2 → 4 N2 + 6 H2O reaction stoichiometry. The potential energy diagram for the formation of the first part of the cycle is shown in Figure . Candidate rate-controlling steps are the two proton abstraction steps (iv → v and vi → vii) and water desorption (ix → x), as the latter step is facilitated by the entropy gain of water desorbing from the surface. Therefore, the present data suggest that the proton abstraction steps from ammonia to the ceria surface are the most likely reaction steps that control the overall reaction rate.

A second aspect of doping ceria with Fe relates to the oxidation of soot. Routine soot oxidation in CDPF is undergone before NO reduction because NO2 produced by NO oxidation in the first step is a stronger oxidant than O2. To evaluate the influence of NO removal during soot oxidation, we carried out a soot oxidation experiment without NO and NH3 in the feed (Figure S13). The activity of the catalyst was slightly lower in this way, as evidenced by the small shift in the CO2 production maximum to higher temperature (405 °C). Nevertheless, the performance of the Ce0.8Fe0.1Zr0.1O2 catalyst under these conditions was still outstanding compared to that of reference systems. This result implies that the substitution of Fe into the ceria surface leads to activated oxygen species that are involved in the oxidation of soot. Although a thorough computational analysis of these aspects is beyond the scope of this study, electronic analysis of adsorbed O2 on the defective Fe-substituted ceria model (state i) and stoichiometric ceria shows nearly similar energetics with a formal O2– state. However, comparison of the density of states (Figure S14) shows more O 2p states close to the Fermi level for O2 adsorbed on the defective Fe-substituted ceria model, which will enhance oxidation of aromatics. Another relevant aspect is the much higher density of O vacancies in Fe-doped ceria as compared with the stoichiometric ceria surface, which should also contribute significantly to the improved soot oxidation performance.

Conclusions

We demonstrate that 3DOM mixed Fe–Ce–Zr oxides are suitable for the simultaneous oxidation of soot and selective catalytic reduction of NO in SCRPF technology. NO is reduced by >90%, and soot is completely combusted in the 265–420 °C temperature range. The addition of Fe and Zr to ceria lowers the temperature of soot combustion to a level that is typically achieved by more expensive Pt catalysts. The 3DOM texture is suitable for trapping soot particles, and the presence of Fe in the ceria surface gives rise to high activity in NO reduction and soot oxidation at intermediate temperatures. The importance of the open macroporous 3DOM texture in soot capture and combustion was demonstrated by comparison to a mesoporous mixed oxide of the same composition. Surface characterization and DFT calculations show that substitution of Fe in the structured mixed Ce–Zr oxide increased the number of oxygen vacancies. A mechanism is explored for the reduction of NO with NH3 involving adsorbed O2 as a catalytic surface intermediate. Such adsorbed O2 species may also be important in soot oxidation. These structured mixed oxides may find application in diesel particulate filters, e.g., by inclusion in wall flow filters constituting ceramic honeycomb structures plugged to force the exhaust flow through the walls. One may, for instance, consider integrating mixed oxide developed here with the base corderite ceramic used in such filters.