Structure and Catalytic Behavior of Alumina Supported Bimetallic Au-Rh Nanoparticles in the Reduction of NO by CO

: Alumina-supported bimetallic AuRh catalysts, as well as monometallic reference catalysts, were examined with regard to their structural and catalytic properties in the reduction of NO by CO. Depending on the molar ratio of Au:Rh, the nanoparticles prepared by borohydride co-reduction of corresponding metal salt solutions had a size of 3.5–6.7 nm. The particles consisted of well-dispersed noble metal atoms with some enrichment of Rh in their surface region. NO conversion of AuRh / Al 2 O 3 shifted to lower temperatures with increasing Rh content, reaching highest activity and highest N 2 selectivity for the monometallic Rh / Al 2 O 3 catalyst. This behavior is attributed to the enhanced adsorption of CO on the bimetallic catalyst resulting in unfavorable cationic Rh clusters Rh + -(CO) 2 . Doping with ceria of AuRh / Al 2 O 3 and Rh / Al 2 O 3 catalysts increased the surface population of metallic Rh sites, which are considered most active for the reduction of NO by CO and enhancement of the formation of intermediate isocyanate (-NCO) surface species and their reaction with NO to form N 2 and CO 2 .


Introduction
Anthropogenic emissions of nitrogen oxides from fuel combustion processes pose a serious threat to human health. Since the reaction of CO and NO can eliminate two harmful pollution gases simultaneously, it has attracted a great deal of attention especially in automobile exhaust applications in a wide range of temperature and pressure [1,2]. Typically, noble metals such as Rh, Pt, and Pd are employed for the NO reduction and CO oxidation [3][4][5][6][7][8][9]. Among the studies reported so far, the NO reduction by CO and co-adsorption on Pd surfaces, as well as co-adsorption of NO and CO on platinum catalysts, have been extensively studied [6][7][8][9]. The formation and migration of isocyanate species are considered to be crucial for efficient catalytic cycles [8].
Due to its low reactivity, gold has long been disregarded as a candidate for heterogeneous catalysis until its nanoparticles supported on oxidic materials were discovered to catalyze the oxidation of CO even below room temperature [10]. The choice of metal oxide supports strongly affects the structural and chemical properties of Au-metal oxide interfaces and is therefore a crucial factor in the design of such catalysts [11][12][13][14][15][16]. The unique properties of gold, such as chemical stability, non-toxicity, and biocompatibility [11], make gold catalysts attractive for various reactions, including oxidations and hydrogenations [17][18][19][20]. Gold and its alloys with other metals show interesting performance in the NO reduction by CO, and thus, structurally optimized Au catalysts provide an attractive research direction [16]. Different Au-based catalysts synthesized by coprecipitation or deposition-precipitation methods have been tested for this reaction [15,[21][22][23][24], among which Au supported on Fe-loaded ceria shows high NO conversion and stability [21]. Au-Pd alloy catalysts exhibit high selectivity, compared to pure Pd catalysts [25], showing further potential of bimetallic gold-based catalysts.
Generally, the reaction steps of NO reduction by CO can be expressed as follows [7,26]: Reaction (1) directly forms N 2 while an alternative path requires reactions (2) and (3), where N 2 O is formed as an intermediate. Depending on the metal oxide support, Au nanoparticles offer high N 2 selectivity, whereas the main product is N 2 O over unsupported Au catalysts [22,25]. Therefore, the presence of pertinent Au-metal oxide interfaces promotes reactions (1) and (3). In situ IR spectroscopic studies revealed that CO and NO adsorb in various configurations, i.e., on-top Au δ+ -CO, CO on Au 0 particles (Au s 0 -CO), bridged NO (Au-NO), and on-top NO − species (Au-NO − ) [23,[27][28][29].
Their reaction on the Au surface forms isocyanate (-NCO) species, which are considered to be a key reaction intermediate that migrates to the oxide support [30,31].
In recent research on NO x reduction, ceria as a support material or as a promoter of the active metals has been the focus for activity enhancement [32,33]. Different dispersions and sizes of CeO 2 nanoparticles influence the catalytic performance [34]. The use of ceria was reported to induce strong metal-support interaction (SMSI) with active metal components. The dispersion of the metal was thus improved by the SMSI effect [32,35,36]. Besides, the oxygen vacancies, present at the interface between the ceria and the metal, impart high oxygen storage capacity to the catalyst and facilitate the transport of oxygen from the ceria to the metal surface, resulting in high activity for NO x reduction [37][38][39].
Bimetallic gold-based alloys can offer interesting synergistic effects in catalysis [40]. While the reduction of NO by CO on bimetallic gold-based alloy catalysts containing Pt and Pd has been the focus of various studies [41][42][43][44], there exist only a few reports in the literature where the behavior of bimetallic AuRh catalysts has been investigated [45][46][47][48][49]. This may be attributed to the large miscibility gap of the AuRh system [50]. However, it has been shown that borohydride co-reduction of metal salt solutions yields nanoparticles in the bulk-immiscible AuRh system [47,51]. To the best of our knowledge, there exists no report where supported bimetallic AuRh catalysts have been examined for the reduction of NO by CO. Its alloying with Au could result in a favorable catalytic synergy effect, as it has been observed in other reactions [49,51]. With this in mind, we have investigated the structural and chemical properties of pure and ceria-promoted alumina-supported bimetallic AuRh alloy nanoparticles and examined their potential for the catalytic reduction of NO by CO.  [52]. Further addition of Rh caused a considerable broadening of the Au(111) reflection and a slight shift toward the Rh(111) reflection (2θ = 41.0 • ) [53,54]. There were no distinct diffraction and a slight shift toward the Rh(111) reflection (2θ = 41.0°) [53,54]. There were no distinct diffraction peaks of Au and Rh observed for AuRh(1:1)/Al2O3 and AuRh(1:3)/Al2O3 catalysts, implying that these catalysts consisted predominantly of well-dispersed bimetallic AuRh nanoparticles on Al2O3 [54].  Figure 2 shows STEM images and metal particle size distributions of various catalysts. Monometallic Au particles in the Au/Al2O3 catalyst were relatively large, whereas their alloying with Rh resulted in well-dispersed small nanoparticles on the Al2O3 surface. EDX line scans of selected bimetallic particles, displayed in Figure 3, showed that Au and Rh atoms were almost uniformly present in the bimetallic nanoparticles. The apparent absence of bimetallic particles with a distinct core/shell structure, e.g., that often observed for bimetallic PdAu particles [55], might be attributed to the high inter-diffusion rate of Rh and Au atoms in the bimetallic particles [56,57] and the fact that the catalysts were pretreated in H2 at 300 °C before characterization [55].  Figure 2 shows STEM images and metal particle size distributions of various catalysts. Monometallic Au particles in the Au/Al 2 O 3 catalyst were relatively large, whereas their alloying with Rh resulted in well-dispersed small nanoparticles on the Al 2 O 3 surface. EDX line scans of selected bimetallic particles, displayed in Figure 3, showed that Au and Rh atoms were almost uniformly present in the bimetallic nanoparticles. The apparent absence of bimetallic particles with a distinct core/shell structure, e.g., that often observed for bimetallic PdAu particles [55], might be attributed to the high inter-diffusion rate of Rh and Au atoms in the bimetallic particles [56,57] and the fact that the catalysts were pretreated in H 2 at 300 • C before characterization [55]. Table 1 summarizes the results of XPS analyses of the catalysts pretreated with pure H 2 (100 mL/min) at 300 • C for 1 h. The raw XPS spectra without background subtraction are shown in Figure S1 (Supplementary Materials). The Au 4f 7/2 photoelectron peak at a binding energy (BE) between 83.7 and 83.9 eV is attributed to the metallic Au 0 species. The reason for the slight shifts of the Au 4f 7/2 (84 eV) and Rh 3d 5/2 (307.6) signals is probably due to cluster size effects [58] or charging of the samples that mainly consisted of nonconductive Al 2 O 3 . As shown in Table 1, the atomic ratios of Au/(Au+Rh) were lower than those expected from the nominal bulk composition of the catalysts, indicating the presence of a segregated Rh-enriched phase in the near-surface region probably due to pretreatment with hydrogen.     Table 1 summarizes the results of XPS analyses of the catalysts pretreated with pure H2 (100 mL/min) at 300 °C for 1 h. The raw XPS spectra without background subtraction are shown in Figure  S1 (Supplementary Materials). The Au 4f7/2 photoelectron peak at a binding energy (BE) between 83.7 and 83.9 eV is attributed to the metallic Au 0 species. The reason for the slight shifts of the Au 4f7/2 (84 eV) and Rh 3d5/2 (307.6) signals is probably due to cluster size effects [58] or charging of the samples that mainly consisted of nonconductive Al2O3. As shown in Table 1, the atomic ratios of Au/(Au+Rh) were lower than those expected from the nominal bulk composition of the catalysts, indicating the presence of a segregated Rh-enriched phase in the near-surface region probably due to pretreatment with hydrogen. The behavior (NO conversion and N2 selectivity) of the AuRh/Al2O3 catalysts with different compositions is shown in Figure 4. The Au/Al2O3 catalyst showed poor performance and no N2 selectivity even at 250 °C. This is partly attributed to the relatively large size of the deposited gold

Catalytic Performance
The behavior (NO conversion and N 2 selectivity) of the AuRh/Al 2 O 3 catalysts with different compositions is shown in Figure 4. The Au/Al 2 O 3 catalyst showed poor performance and no N 2 selectivity even at 250 • C. This is partly attributed to the relatively large size of the deposited gold particles and poor activity for the reduction of the formed N 2 O by CO (reaction (3)). The higher the Rh content was, the better the activity was, while the N 2 selectivity was zero for temperatures up to 200 • C with all bimetallic catalysts (see also Table S1 in Supplementary Materials). The increase in the N 2 selectivity by the Rh addition is attributed to an increase of the rate of reaction (3). The only by-product was N 2 O, while NO 2 was not detected. Note that Rh/Al 2 O 3 outperformed all bimetallic AuRh catalysts. To gain some insight into the role of both metals, Au and Rh, in the NO reduction by CO, we used the in situ DRIFT spectroscopy. particles and poor activity for the reduction of the formed N2O by CO (reaction (3)). The higher the Rh content was, the better the activity was, while the N2 selectivity was zero for temperatures up to 200 °C with all bimetallic catalysts (see also Table S1 in Supplementary Materials). The increase in the N2 selectivity by the Rh addition is attributed to an increase of the rate of reaction (3). The only by-product was N2O, while NO2 was not detected. Note that Rh/Al2O3 outperformed all bimetallic AuRh catalysts. To gain some insight into the role of both metals, Au and Rh, in the NO reduction by CO, we used the in situ DRIFT spectroscopy.

In Situ DRIFTS Study
In situ DRIFTS spectra taken during the adsorption of 2000 ppm CO in He balance at 30 °C for 5 min (from the red line to the blue line at an interval of 1 min) are shown in Figure 5. The adsorption temperature was set to 30 °C to suppress disturbing desorption and disproportionation of CO and NO. At the initial stage of CO adsorption, a sharp band at 2069 cm −1 assignable to on-top CO on metallic Rh (Rh 0 -CO) emerged [59], followed by a gradual increase in bridged CO (1910 cm −1 ) [60,61] together with twin bands at 2091 cm −1 and 2025 cm −1 , which were assigned to asymmetric and symmetric stretching vibrations of gem-dicarbonyls on Rh cations, Rh + -(CO)2 [60]. As evident from the XPS data, the Rh particles existed as metallic Rh after H2 pretreatment. However, CO adsorption caused the decomposition of the Rh particles by dispersing cationic Rh atoms as Rh + -(CO)2 [62,63]. Interestingly, by adding Au, the intensity of these bands increased and reached a maximum for the AuRh(1:1)/Al2O3 catalyst. A similar tendency was also observed for on-top CO, indicating that the

In Situ DRIFTS Study
In situ DRIFTS spectra taken during the adsorption of 2000 ppm CO in He balance at 30 • C for 5 min (from the red line to the blue line at an interval of 1 min) are shown in Figure 5. The adsorption temperature was set to 30 • C to suppress disturbing desorption and disproportionation of CO and NO. At the initial stage of CO adsorption, a sharp band at 2069 cm −1 assignable to on-top CO on metallic Rh (Rh 0 -CO) emerged [59], followed by a gradual increase in bridged CO (1910 cm −1 ) [60,61] together with twin bands at 2091 cm −1 and 2025 cm −1 , which were assigned to asymmetric and symmetric stretching vibrations of gem-dicarbonyls on Rh cations, Rh + -(CO) 2 [60]. As evident from the XPS data, the Rh particles existed as metallic Rh after H 2 pretreatment. However, CO adsorption caused the decomposition of the Rh particles by dispersing cationic Rh atoms as Rh + -(CO) 2 [62,63]. Interestingly, by adding Au, the intensity of these bands increased and reached a maximum for the AuRh(1:1)/Al 2 O 3 catalyst. A similar tendency was also observed for on-top CO, indicating that the presence of Au promoted the CO adsorption on the Rh sites. Further addition of Au, as realized in the AuRh(3:1)/Al 2 O 3 catalyst, decreased the intensity by a factor of ca. 10, and the Au/Al 2 O 3 catalyst showed only weak bands at 2102 cm −1 and 2170 cm −1 , which were assigned to CO adsorbed on Au and Al 2 O 3 , respectively [64]. Remarkably, the addition of Au induced the formation of Rh + -(CO) 2 species, which are considered to be dispersed in the form of small clusters [60]. This phenomenon is in good agreement with the XRD data ( Figure 1), showing that the diffraction peaks were broadened by alloying Rh with Au. DRIFTS spectra measured during NO adsorption (gas mixture: 2000 ppm NO in He balance, total flow rate: 100 mL/min, temperature: 30 °C, and time duration: 5 min) are shown in Figure 6. A sharp band at 1684 cm −1 and a weak band at 1801 cm −1 , which can be assigned to on-top NO δ-and on-top NO, respectively, were observed for the Rh/Al2O3 catalyst [3,65]. Bridged NO was not observed in our study. The presence of gold promoted the adsorption of NO, especially for Rh-NO at ca. 1800 cm −1 but decreased the ratio of Rh-NO δ− . This was induced by the low electron density of cationic Rh (Rh + ) for AuRh alloy nanoparticles as evident from the CO adsorption in Figure 5. Therefore, the formation of cationic Rh suppressed the electron back-donation to NO molecules, which did not allow for the negatively-charged on-top NO δ− configuration. For the Au/Al2O3 study. The presence of gold promoted the adsorption of NO, especially for Rh-NO at ca. 1800 cm −1 but decreased the ratio of Rh-NO δ− . This was induced by the low electron density of cationic Rh (Rh + ) for AuRh alloy nanoparticles as evident from the CO adsorption in Figure 5. Therefore, the formation of cationic Rh suppressed the electron back-donation to NO molecules, which did not allow for the negatively-charged on-top NO δ− configuration. For the Au/Al 2 O 3 catalyst, a broad band at 1652 cm −1 assigned to asym (NO 2 ) of surface NO 2 species was observed [23]. The band at 1540 cm −1 is attributed to bridging nitrito (NO 2 − ) in the chelated state [66]. An additional band at 1894 cm −1 was assigned to NO adsorbed on Au [67]. Co-adsorption of NO and CO indicated that NO was adsorbed on Rh more strongly than CO (DRIFTS spectra are provided in Figure S2 in Supplementary Materials). However, the AuRh alloy catalysts tended to enrich CO on the surface. This might be the reason why the conversion decreased for the AuRh alloy catalysts due to high CO coverage suppressing the NO adsorption and dissociation.  Figure 7 shows DRIFTS spectra during the reduction of NO by CO at 300 °C for 10 min (from the red line (5 s) to the blue line (10 min) at an interval of 1 min; the green line represents a spectrum after 1 min of reaction). At this high temperature, desorption and reaction of NO and CO took place rapidly and the corresponding bands were greatly weakened. Instead, several new bands appeared, which are centered at 2254 cm −1 , 1634 cm −1 , and 1404 cm −1 , respectively. The catalyst surface was dominated by isocyanate species (-NCO) detected at 2254 cm −1 . The blue shift compared to previous reports results from the presence of Rh + sites [4]. The shoulder band at ca. 2230 cm -1 was assigned to Rh 0 -NCO, which completely disappeared for the AuRh(3:1)/Al2O3 catalyst. Because N2O was not detected in the gas phase, the contribution of IR bands of gaseous N2O in this region can be excluded. The intensity of isocyanate species first increased, reached a maximum after 4 min (green  isocyanate species (-NCO) detected at 2254 cm −1 . The blue shift compared to previous reports results from the presence of Rh + sites [4]. The shoulder band at ca. 2230 cm −1 was assigned to Rh 0 -NCO, which completely disappeared for the AuRh(3:1)/Al 2 O 3 catalyst. Because N 2 O was not detected in the gas phase, the contribution of IR bands of gaseous N 2 O in this region can be excluded. The intensity of isocyanate species first increased, reached a maximum after 4 min (green line), and then diminished (blue line), accompanied by a weak band at 2049 cm −1 assigned to carbonyl species adsorbed on Rh [3,4]. By adding Au, the intensity of isocyanate species increased and their decomposition also occurred rather rapidly. The decomposition process is better highlighted by switching the gaseous compositions to only 2000 ppm NO in He balance ( Figure S3 in Supplementary Materials). Rh-O + CO → Rh 0 + CO2, Rh-NCO + NO → Rh 0 + CO2 + N2.
A sharp band at 1404 cm −1 was assigned to nitrate species bonding to both Rh and oxidic support [4], which diminished by adding Au. The negatively charged Rh-(NO) δ− at 1680 cm −1 and 1634 cm −1 is considered as an active precursor for a key surface reaction step during NO-CO reaction [68,69]. The lowered wavenumber of Rh-NO δ− indicates a weakening of the N-O bond and strengthening of the Rh-N bond due to electron back-donation to the N-O antibonding orbital [68,69], resulting in higher propensity of NO molecules to be dissociated over reduced Rh surfaces to yield nitride and adsorbed oxygen atoms: Rh-N + Rh-NO → Rh-N2O.

Catalyst Characterization
From reactions (4)-(9) described above, it is obvious that redox reactions play a key role in the reduction of NO by CO, as indicated by the oxidation of Rh by NO dissociation and the reduction of Rh-O by CO. Cerium oxides are well known to facilitate redox processes by transformation between Ce2O3 and CeO2 [37][38][39]. With this in mind, we impregnated ceria onto the best performing binary Rh-N(CO) → Rh-NCO, Rh-NCO + NO → Rh 0 + CO 2 + N 2 .
A sharp band at 1404 cm −1 was assigned to nitrate species bonding to both Rh and oxidic support [4], which diminished by adding Au. The negatively charged Rh-(NO) δ− at 1680 cm −1 and 1634 cm −1 is considered as an active precursor for a key surface reaction step during NO-CO Catalysts 2019, 9, 937 9 of 20 reaction [68,69]. The lowered wavenumber of Rh-NO δ− indicates a weakening of the N-O bond and strengthening of the Rh-N bond due to electron back-donation to the N-O antibonding orbital [68,69], resulting in higher propensity of NO molecules to be dissociated over reduced Rh surfaces to yield nitride and adsorbed oxygen atoms:

Catalyst Characterization
From reactions (4)-(9) described above, it is obvious that redox reactions play a key role in the reduction of NO by CO, as indicated by the oxidation of Rh by NO dissociation and the reduction of Rh-O by CO. Cerium oxides are well known to facilitate redox processes by transformation between Ce 2 O 3 and CeO 2 [37][38][39]. With this in mind, we impregnated ceria onto the best performing binary catalyst, AuRh(1:1)/Al 2 O 3 catalyst. The resulting catalyst was designated henceforth as AuRh-CeO x /Al 2 O 3 and had a AuRh:Ce molar ratio of 1:1.
Properties of the AuRh-CeO x /Al 2 O 3 catalyst and the corresponding reference catalysts with similar amounts of Ce, Au-CeO x /Al 2 O 3 and Rh-CeO x /Al 2 O 3 , are summarized in Table 2. The surface areas of the three ceria-promoted catalysts slightly decreased compared to those of the corresponding nonpromoted samples (Table 1). XRD analyses ( Figure 8) indicated that the addition of ceria induced the formation of crystalline gold domains in the bimetallic AuRh particles, as evident from the emergence of Au(111), Au(200), Au(220), and Au(311) reflections. In contrast, the intensity of diffraction patterns due to crystalline Rh stayed extremely weak and broad, indicating that the Rh nanoparticles were still well-dispersed on the surface. Reflections due to crystalline CeO 2 were not discernible for AuRh-CeO x /Al 2 O 3 , suggesting the formation of amorphous ceria dispersed on the surface. Figure 9 shows STEM images and particle size distributions of the AuRh-CeO x /Al 2 O 3 catalyst and the reference catalysts. The addition of CeO x had little influence on the size of the bimetallic AuRh particles (3.1 nm shown in Table 2), whereas it was significant for the Au-CeO x /Al 2 O 3 and Rh-CeO x /Al 2 O 3 catalysts. Table 2 also summarizes results of XPS analyses of Au 4f, Rh 3d, and Ce 3d. The corresponding raw XPS spectra are provided in Supplementary Materials ( Figure S4). A mixture of Ce 3+ and Ce 4+ oxidation states existed on the surface of the catalysts [70,71]. Hereafter, the cerium oxides are denoted as CeO x (x = 1.5-2). As shown in Table 2, the concentration of Ce 3+ in the AuRh-CeO x /Al 2 O 3 catalyst is higher than those of the monometallic reference catalysts and CeO x /Al 2 O 3 catalyst, indicating that the presence of bimetallic AuRh promotes the formation of Ce 3+ components as Ce 2 O 3 .
H 2 TPR profiles ( Figure S5 in Supplementary Materials) revealed that the presence of bimetallic AuRh particles lowered the reduction temperature of CeO x components from 400 to less than 200 • C. denoted as CeOx (x = 1.5-2). As shown in Table 2, the concentration of Ce 3+ in the AuRh-CeOx/Al2O3 catalyst is higher than those of the monometallic reference catalysts and CeOx/Al2O3 catalyst, indicating that the presence of bimetallic AuRh promotes the formation of Ce 3+ components as Ce2O3.
H2 TPR profiles ( Figure S5 in Supplementary Materials) revealed that the presence of bimetallic AuRh particles lowered the reduction temperature of CeOx components from 400 to less than 200 °C.   Figure 10 shows the NO conversion and N2 selectivity achieved in the reduction of NO by CO for the AuRh-CeOx/Al2O3 catalyst and the reference catalysts Au/CeOx/Al2O3 and Rh/CeOx/Al2O3. The comparison with the corresponding catalysts without ceria addition (Figure 4) showed that ceria  Figure 10 shows the NO conversion and N 2 selectivity achieved in the reduction of NO by CO for the AuRh-CeO x /Al 2 O 3 catalyst and the reference catalysts Au/CeO x /Al 2 O 3 and Rh/CeO x /Al 2 O 3 . The comparison with the corresponding catalysts without ceria addition (Figure 4) showed that ceria doping enhances the catalytic performance of AuRh/Al 2 O 3 and Rh/Al 2 O 3 catalysts, whereas that of the Au/Al 2 O 3 catalyst declines. Both NO conversion and N 2 selectivity curves of the AuRh/Al 2 O 3 and Rh/Al 2 O 3 catalysts are shifted towards lower temperature. In contrast, the catalytic performance declined with CeO x deposition on the Au/Al 2 O 3 catalyst. The N 2 selectivity of AuRh/Al 2 O 3 catalyst was zero at temperatures up to 200 • C but increased to 69% for the corresponding ceria-promoted AuRh-CeO x /Al 2 O 3 catalyst, as shown in Table S2 (Supplementary Materials). The only by-product was N 2 O. This implied that the addition of CeO x facilitated the reduction of N 2 O by CO, i.e., reaction (3) described in the Introduction section. Note that the N 2 selectivity of the AuRh)/Al 2 O 3 catalyst showed some fluctuation, depending on the reaction temperature. Since this behavior was perfectly reproducible, we supposed that it is due to a chemical or structural change of the catalysts, possibly leading to a mechanistic change of which clarification needs further investigations. Note that such behavior has earlier been observed in some cases for metal alloys [72] and metal-metal oxide nanocomposites [73,74]. leading to a mechanistic change of which clarification needs further investigations. Note that such behavior has earlier been observed in some cases for metal alloys [72] and metal-metal oxide nanocomposites [73,74].

In Situ DRIFTS Study
To unveil the role of CeOx in the reduction of NO by CO, in situ DRIFTS studies were carried out. Figure 11 shows DRIFT spectra measured during the adsorption of 2000 ppm CO in He balance at 30 °C for 5 min (from the red line to the blue line at an interval of 1 min). By the addition of CeOx, the intensity of the band of on-top CO at 2065 cm −1 became dominant while the bands of Rh + -(CO)2 at 2091 cm −1 and 2025 cm −1 were considerably weakened, indicating that there are less Rh + sites in the presence of CeOx. This alludes that Rh + sites are less active for the reaction and the presence of CeOx enhances the formation of Rh 0 sites, which facilitate the reduction of NO by CO. However, larger amounts of CeOx addition resulted in a dramatic decrease in the intensity of adsorbed CO due to higher coverage of CeOx on the bimetallic AuRh particles.

In Situ DRIFTS Study
To unveil the role of CeO x in the reduction of NO by CO, in situ DRIFTS studies were carried out. Figure 11 shows DRIFT spectra measured during the adsorption of 2000 ppm CO in He balance at 30 • C for 5 min (from the red line to the blue line at an interval of 1 min). By the addition of CeO x , the intensity of the band of on-top CO at 2065 cm −1 became dominant while the bands of Rh + -(CO) 2 at 2091 cm −1 and 2025 cm −1 were considerably weakened, indicating that there are less Rh + sites in the presence of CeO x . This alludes that Rh + sites are less active for the reaction and the presence of CeO x enhances the formation of Rh 0 sites, which facilitate the reduction of NO by CO. However, larger amounts of CeO x addition resulted in a dramatic decrease in the intensity of adsorbed CO due to higher coverage of CeO x on the bimetallic AuRh particles. DRIFT spectra of NO adsorptions are displayed in Figure 12. In line with the CO adsorption, the band at a higher wavenumber (ca. 1800 cm −1 ) was weakened by the CeOx addition. The relative increase in the band at 1683 cm −1 indicated that the CeOx promoted the transformation of the adsorption configuration from Rh-NO to Rh-NO δ− . As described in Section 2.1.3, the enriched on-top NO δ− is indicative of the formation of metallic Rh and the strong electron back-donation from Rh to anti-bonding orbitals of adsorbed NO molecules. The rate constants for NO dissociation determines the activity of the catalysts [68]. Therefore, the dissociation of the negatively charged NO on metallic Rh is a key step for the reduction of NO with CO. A new band at 1890 cm −1 for catalysts promoted with ceria is assigned to adsorbed NO as Au n+ -NO (n ≈ 0) complex [68]. DRIFT spectra during co-adsorption of NO and CO demonstrated that CO molecules were adsorbed mainly on cationic Rh while NO molecules occupied metallic Rh ( Figure S6 in Supplementary Materials). DRIFT spectra of NO adsorptions are displayed in Figure 12. In line with the CO adsorption, the band at a higher wavenumber (ca. 1800 cm −1 ) was weakened by the CeO x addition. The relative increase in the band at 1683 cm −1 indicated that the CeO x promoted the transformation of the adsorption configuration from Rh-NO to Rh-NO δ− . As described in Section 2.1.3, the enriched on-top NO δ− is indicative of the formation of metallic Rh and the strong electron back-donation from Rh to anti-bonding orbitals of adsorbed NO molecules. The rate constants for NO dissociation determines the activity of the catalysts [68]. Therefore, the dissociation of the negatively charged NO on metallic Rh is a key step for the reduction of NO with CO. A new band at 1890 cm −1 for catalysts promoted with ceria is assigned to adsorbed NO as Au n+ -NO (n ≈ 0) complex [68]. DRIFT spectra during co-adsorption of NO and CO demonstrated that CO molecules were adsorbed mainly on cationic Rh while NO molecules occupied metallic Rh ( Figure S6 in Supplementary Materials).
In situ DRIFT spectra during the reduction of NO by CO at 200 • C for 10 min are shown in Figure 13. A broad band at 1835-1842 cm −1 and a band at 1910 cm −1 were assigned to a dinitrosyl complex and positively charged Rh-NO + species, respectively [3]. The intensity of isocyanate species (-NCO) reached a maximum after 5 min for the AuRh(1:1)/Al 2 O 3 catalyst, but only took 1 min for the AuRh-CeO x /Al 2 O 3 catalyst, showing that CeO x promoted the formation of isocyanate species. For the AuRh-CeO x (1:2)/Al 2 O 3 and AuRh-Ce(1:5)/Al 2 O 3 catalysts with AuRh:Ce molar ratios of 1:2 and 1:5, respectively, a new band at 2173-2196 cm −1 assignable to isocyanate species on CeO x emerged, suggesting the spill-over of isocyanate species from bimetallic AuRh particles to the CeO x surface. Note that, for the catalysts with a large amount of CeO x , the intensity of isocyanate species decreased greatly after reaching its maximum (highlighted with green and blue lines). This suggested that CeO x does not only promote the formation of isocyanate species, but also its decomposition to form N 2 and CO 2 . The rapid decomposition of isocyanate species was also confirmed by switching the gas stream from NO and CO to only NO ( Figure S7 in Supplementary Materials). For all the catalysts, only bands of Rh + -(CO) 2 at 2017-2090 cm −1 were observed under reaction conditions. The Rh + -(CO) 2 species seem stable and are most likely not involved in the reaction. This fact further substantiates the vital role of metallic Rh in the reaction. For the catalysts with high CeO x loadings, i.e., AuRh-CeO x (1:2)/Al 2 O 3 and AuRh-CeO x (1:5)/Al 2 O 3 catalysts, bands located in the region between 1500 and 1700 cm −1 , assigned to nitrate and carbonate species [3,4], grew stronger and more stable. However, the NO conversion and N 2 selectivity stayed almost the same for the AuRh-CeO x (1:2)/Al 2 O 3 and AuRh-CeO x (1:5)/Al 2 O 3 catalysts (Table S2 in Supplementary Materials). Therefore, these species are considered to be merely spectator species. In situ DRIFT spectra during the reduction of NO by CO at 200 °C for 10 min are shown in Figure 13. A broad band at 1835-1842 cm −1 and a band at 1910 cm −1 were assigned to a dinitrosyl complex and positively charged Rh-NO + species, respectively [3]. The intensity of isocyanate species (-NCO) reached a maximum after 5 min for the AuRh(1:1)/Al2O3 catalyst, but only took 1 min for the AuRh-CeOx/Al2O3 catalyst, showing that CeOx promoted the formation of isocyanate species. For the AuRh-CeOx(1:2)/Al2O3 and AuRh-Ce(1:5)/Al2O3 catalysts with AuRh:Ce molar ratios of 1:2 and 1:5, respectively, a new band at 2173-2196 cm −1 assignable to isocyanate species on CeOx emerged, suggesting the spill-over of isocyanate species from bimetallic AuRh particles to the CeOx surface. AuRh-CeOx(1:5)/Al2O3 catalysts (Table S2 in Supplementary Materials). Therefore, these species are considered to be merely spectator species.
Au, Rh, and bimetallic AuRh nanoparticles supported on Al 2 O 3 were synthesized by aqueous-phase co-reduction of corresponding metal salts with NaBH 4 at room temperature [47]. In a typical synthetic method, 3.7 mL of deionized water, 200 mL of PEG-2000 (80 g/L), and 1 mL of 50 mM solutions of HAuCl 4 and RhCl 3 ·3H 2 O were stirred together in a 50 mL glass vial for 20 min. The Au:Rh molar ratios of the metal salts were adjusted 1:0, 3:1, 1:1, 1:3, and 0:1, keeping the total metal loading at a nominal value of 3 wt.%. Al 2 O 3 powder was then added to the solution under constant stirring. Afterwards, 100 mL of 20 g/L NaBH 4 aqueous solution were added slowly to the solution and the suspension was stirred for 1 h. The resulting particles in the suspension were then separated by centrifugation, washed several times with ethanol and dried at 60 • C overnight.
AuRh-CeO x /Al 2 O 3 catalysts with different molar ratios of AuRh:Ce (1:1, 2:1, and 1:5) were prepared by post-impregnation with a ceria precursor of the AuRh(1:1)/Al 2 O 3 catalyst. Ammonium cerium(IV) nitrate was dissolved in 100 mL of deionized water and stirred for 30 min. The AuRh(1:1)/Al 2 O 3 powder was then added to the solution and the suspension was stirred for 1 h. Afterwards, the suspension was transferred to a rotary evaporator and aged for 1 h at 60 • C, followed by the evaporation of the water at 35 • C under vacuum and drying at 60 • C overnight.
As the reference catalysts, Rh/Al 2 O 3 and CeO x /Al 2 O 3 with a molar amount of CeO x corresponding to that in AuRh-CeO x (1:1)/Al 2 O 3 were synthesized in the same manner. These samples were calcined at 500 • C for 3 h before reductive pretreatment.
All catalyst samples were pretreated with pure H 2 at a flow rate of 100 mL/min for 1 h at 300 • C (ramping rate: 5 • C /min) prior to characterization by nitrogen adsorption, X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), scanning transmission electron microscopy (STEM), energy-dispersive X-ray spectroscopy (EDX), and diffuse reflectance IR spectroscopy (DRIFTS) as well as catalytic tests.

Catalyst Characterization
Brunauer-Emmett-Teller (BET) surface areas were determined from N 2 adsorption isotherms measured at a liquid nitrogen temperature using a Quadrasorb SI instrument (Quantachrome, Boynton Beach, FL, USA). Prior to measurements, the as-prepared samples were degassed at 150 • C for 1 h to remove residual water adsorbed on the surface and in the pores. Temperature-programmed reduction (TPR) was carried out with a Quantachrome, ChemBET Pulsar TPR/TPD (temperature-programmed desorption) instrument. One-hundred milligrams of the as-prepared (unreduced) samples were loaded in a U-shaped quartz micro-reactor and then heated at a ramping rate of 10 • C/min in 10 vol% H 2 in Ar at a total flow rate of 15 mL/min. TPR signals were calibrated by injecting different amounts of 10 vol% H 2 in Ar.
XRD measurements were carried out ex situ using a PANalytical X'Pert diffractometer (Great Malvern, UK) with Pd-filtered Cu Kα radiation at 60 kV and 55 mA. The surface compositions and oxidation states of the catalysts were analyzed by XPS on a Thermo ESCALAB 250Xi instrument using Al Kα radiation (1486.6 eV, 15 kV, and 10.8 mA) to excite photoelectrons at a 50 eV pass energy at an energy scale calibrated versus Ag 3d 5/2 at 1 eV. The instrument was equipped with an in situ pretreatment chamber connected to the XPS chamber. The sample holder was transferred between the two chambers by a manipulator. The pretreatment conditions were the same as those described in Section 3.3. STEM (JEOL, JEM-2100, 200 kV, Tokyo, Japan) was used to examine the size distribution of the mono-and bimetallic nanoparticles deposited on the alumina support and EDX was employed to gain some information about the elemental distributions of Au and Rh in the nanoparticles. The samples previously pretreated with pure hydrogen for 1 h at 300 • C were used for both STEM and EDX measurements. Particle size distributions were based on analyzing more than 100 particles per sample.
In situ IR spectroscopic studies were carried out using a Fourier-transform infrared spectrometer (Bruker Optics, Vertex 70, Billerica, MA, USA) equipped with a liquid N 2 -cooled mercury cadmium telluride (MCT) detector at a spectral resolution of 4 cm −1 and a scanning velocity of 60 kHz. A ceramic cup with 20 mg of the samples was placed in an in situ DRIFTS reactor with three ZnSe windows, connected to an automated 4-way Valco valve (VICI AG International) and mass flow controllers (Azbil Control Solution Co., Ltd., Tokyo, Japan). The DRIFTS compartment and all the IR beam lines were purged with dry air supplied by a compressor/dryer-combined system (EKOM) to remove the H 2 O vapor, of which IR bands overlapped with the IR fingerprint regions. Prior to in situ DRIFTS measurements, the catalysts were exposed to pure H 2 at a flow rate of 100 mL/min for 1 h at 300 • C (ramping rate: 5 • C /min) in the DRIFTS reactor. For all the experiments, 2000 ppm NO and/or 2000 ppm CO in He balance (He purity: 99.9999%) were utilized at a total flow rate of 100 mL/min. Before all IR measurements, background spectra were taken under helium flow.

Catalytic Reaction
The catalytic performance was examined employing a fixed-bed micro-flow reactor system. Three-hundred milligrams of catalysts were packed into the stainless-steel tube reactor with an inner diameter of 8 mM. Prior to catalytic tests, the catalysts were pretreated with pure H 2 (flow rate: 100 mL/min) at 300 • C for 1 h at a ramping rate of 5 • C/min. Residual H 2 in the gas lines and adsorbed hydrogen on the catalyst surface were removed by keeping the catalyst in a He stream at 300 • C for 10 min. Afterwards, the reactor was cooled to 50 • C and a gas mixture (2000 ppm CO and 2000 ppm NO in He balance) was added at a total flow rate of 150 mL/min, affording a gas hourly space velocity (GHSV) of 18,000 h −1 (length of catalyst bed: 1 cm, catalyst bed volume: 0.5 cm 3 ). The gaseous compositions were analyzed using a gas chromatograph (Fili, 9750 GC, Shanghai, China) equipped with a TCD detector and a molecular sieve 5A column and a NO-NO 2 -NO x analyzer (Thermo Scientific, 42i-HL, Waltham, MA, USA). Catalytic performance data were recorded after the gaseous compositions at the reactor outlet reaching a steady-state (30-70 min) at each measurement temperature. The same experiments were repeated three times to ensure the reproducibility of the data (less than ±1.0%). Blind tests using the empty tube reactor showed no significant activity.

Conclusions
Bimetallic AuRh nanoparticles with different Au:Rh ratios (3:1, 1:1, and 1:3) and mean sizes (3-7 nm) were prepared by aqueous-hase co-reduction of corresponding noble metal salts with NaBH 4 and deposited on alumina as a support. The nanoparticles were made up of well-distributed noble metal atoms forming an alloy with Rh enriched in the surface region. The AuRh/alumina catalysts were examined, concerning their potential for the reduction of NO by CO. The activity for NO reduction decreased with the increasing Au:Rh ratio, most likely due to unfavorable formation of cationic Rh clusters Rh + -(CO) 2 in the presence of the reactant CO, as uncovered by in situ DRIFTS. To explore the feasibility of enhancing the performance of the bimetallic AuRh/Al 2 O 3 catalysts, cerium oxide was added to the AuRh(1:1)/Al 2 O 3 catalyst by post-impregnation. DRIFTS, XPS, and XRD analyses showed that Ce was present as Ce 3+ and Ce 4+ in the form of well-dispersed CeO x . The catalytic performances of the AuRh-CeO x /Al 2 O 3 catalyst as well as that of the Rh/Al 2 O 3 catalyst were strongly enhanced compared to that of the parent AuRh(1:1)/Al 2 O 3 catalyst, contrasting the behavior of the Au-CeO x /Al 2 O 3 catalyst where ceria doping had no beneficial effect on the catalytic performance. The reason for the activity enhancement is ascribed to the redox properties of CeO x facilitating the creation of metallic Rh sites, which are considered to be crucial for the reduction of NO by CO. Furthermore, in situ DRIFTS investigations during the reduction of NO by CO at 200 • C proved that the presence of CeO x enhances the formation of intermediate isocyanate species and their reaction with NO to produce N 2 and CO 2 . Finally, it should be stressed that it might be possible to further enhance the catalytic performance the AuRh-CeO x /Al 2 O 3 catalyst for the reduction of NO by CO by optimizing their composition. Overall, the study shows that alumina-supported bimetallic AuRh catalysts hardly provide a better catalytic performance in the reduction of NO by CO, compared to the well-known monometallic Rh/alumina catalyst. Doping with ceria enhances the catalytic performance of both catalysts.