The Hydrothermal Stability and the Properties of Non- and Strongly-Interacting Rh Species over Rh/ γ , θ -Al 2 O 3 Catalysts

: The present work reports the effects of γ -, θ -phase of alumina on the hydrothermal stability and the properties of non- and strongly-interacting Rh species of the Rh/Al 2 O 3 catalysts. Comparing to γ -Al 2 O 3 , θ -Al 2 O 3 can not only reduce the amount of occluded Rh but also better stabilize Rh during hydrothermal aging treatment. When the aging time was prolonged to 70 h, all the non-interacting Rh was transformed into strongly-interacting Rh and occluded Rh. The XPS results indicated that non- and strongly-interacting Rh might exist in the form of Rh/Rh 3+ and Rh 4+ , respectively. CO-NO reaction was chosen as a probe reaction to research more information about non- and strongly-interacting Rh. The two Rh species had similar apparent activation energy ( E app ) of 170 kJ/mol, which indicated that non- and strongly-interacting Rh follow the same reaction path. The non-interacting Rh was removed from aged samples by the acid-treated method, and obtained results showed that only 2.5% and 4.0% non-interacting Rh was maintained in aged Rh/ γ -Al 2 O 3 and Rh/ θ -Al 2 O 3 .


Introduction
Three-way catalysts (TWCs) are widely used for the exhaust treatment in gasolineengine powered vehicles since 1979 [1,2]. The modern commercial TWCs contain Pt, Pd, and Rh for removing carbon monoxide (CO), hydrocarbons (HCs), and nitrogen oxides (NO x ). Rh is a crucial ingredient, aimed at eliminating NO x pollution [1,3]. The global consumption of Rh has increased rapidly in past decades and about 81% of Rh usages are in auto catalyst manufacturing in 2012 [4]. Based on the current knowledge, no material has been found to substitute for Rh component [1,5,6]; thus, the utilization efficiency of Rh should be improved greatly.
γ-Al 2 O 3 , with high surface area, is usually used as Rh-based catalyst support. However, Rh/γ-Al 2 O 3 becomes deactivated after hydrothermal treatment at high temperature in an oxidizing atmosphere [6,7], while α-Al 2 O 3 still shows a better ability to stabilize Rh until oxidizing treatment temperature over 900 • C [8][9][10]. The low surface area of α-Al 2 O 3 limits the usage as catalyst supports, but some researchers [11,12] synthesized α-Al 2 O 3 with a high surface area (>140 m 2 /g), which will allow α-Al 2 O 3 to play a more important role in catalyst supports in the future. Ceria-zirconia solid solutions are a kind of material with excellent oxygen storage capacity (OSC), used in the exhaust gas treatment of gasoline vehicles, and also used as the support of Rh-based TWCs according to published information [13]. However, recent researches showed that increasing cerium content may induce a 2. Results and Discussions 2.1. The Effect of γ, θ-Phase Alumina on the Hydrothermal Stability of Rh/γ, θ-Al 2

O 3 Catalysts
The catalytic performance of the Rh/γ, θ-Al 2 O 3 samples were tested, as shown in Figure 1. For fresh catalysts, it can be observed that Rh/γ-Al 2 O 3 shows a better activity for CO, C 3 H 6 , and NO conversions over 250 • C. It is worth noting that the fresh samples show superior deNO x performance [21].
Apparent catalyst deactivation was observed over Rh/γ, θ-Al 2 O 3 catalysts upon hydrothermal aging at 1050 • C for 10 h. In Figure 1, the light-off curves of CO, C 3 H 6 , and NO shift to higher temperatures. Noteworthy, compared to aged Rh/γ-Al 2 O 3 , aged Rh/θ-Al 2 O 3 still shows superior CO and C 3 H 6 oxidation activity. In addition, aged Rh/θ-Al 2 O 3 also shows a better NO conversion at low temperature, which may be due to more non-interacting Rh maintained in aged Rh/θ-Al 2 O 3 catalyst. The NO conversion of aged samples shows different trends with that of fresh samples. After hydrothermal aging at 1050 • C for 10 h, the Rh dispersion of Rh/γ, θ-Al 2 O 3 dramatically decreased. Only few active Rh sites were maintained to adsorb and dissociate NO. Many reductants were oxidized before reacting with NO; therefore, the NO conversion of aged samples decreased at high temperatures.
The phase structures of all samples were characterized by XRD. In Figure 2, the XRD patterns of fresh catalysts show a good γand θ-phase, respectively. After hydrothermal aging, γ-Al 2 O 3 was transformed to θ-Al 2 O 3 (majority) and α-Al 2 O 3 (minority). θ-Al 2 O 3 was still mainly kept θ-phase with several weak peaks of α-Al 2 O 3 arising at 2θ = 25.7 • , 43.4 • , and 57.5 • [22]. No diffraction peaks of Rh species were observed in all catalysts, maybe due to the low Rh loading or high Rh dispersion on the surface of supports [15]. The collapse of pores appeared in the aging process as shown in Figure S1. The most probable pore size increases from 16 nm to 28 nm for Rh/γ-Al 2 O 3 , and from 28 nm to 37 nm for Rh/θ-Al 2 O 3 , respectively, but the smaller pore size is still preserved. The specific surface area of Rh/γ-Al 2 O 3 shows a large decrease, but the BET surface area of due to the low Rh loading or high Rh dispersion on the surface of supports [15]. The collapse of pores appeared in the aging process as shown in Figure S1. The most probable pore size increases from 16 nm to 28 nm for Rh/γ-Al2O3, and from 28 nm to 37 nm for Rh/θ-Al2O3, respectively, but the smaller pore size is still preserved. The specific surface area of Rh/γ-Al2O3 shows a large decrease, but the BET surface area of Rh/θ-Al2O3 only decreases from 83 to 72 m 2 /g (Table S1). Overall, θ-Al2O3 shows better hydrothermal stability.  The collapse of alumina pores also caused a large amount of Rh to be occluded. The amount of occluded Rh was determined by H2 temperature-programmed reduction  due to the low Rh loading or high Rh dispersion on the surface of supports [15]. The collapse of pores appeared in the aging process as shown in Figure S1. The most probable pore size increases from 16 nm to 28 nm for Rh/γ-Al2O3, and from 28 nm to 37 nm for Rh/θ-Al2O3, respectively, but the smaller pore size is still preserved. The specific surface area of Rh/γ-Al2O3 shows a large decrease, but the BET surface area of Rh/θ-Al2O3 only decreases from 83 to 72 m 2 /g (Table S1). Overall, θ-Al2O3 shows better hydrothermal stability.    The collapse of alumina pores also caused a large amount of Rh to be occluded. The amount of occluded Rh was determined by H2 temperature-programmed reduction The collapse of alumina pores also caused a large amount of Rh to be occluded. The amount of occluded Rh was determined by H 2 temperature-programmed reduction method according to the research of McCabe and Hwang [8,23]. Samples were treated with H 2 at 850 • C to reverse the strong interaction between Rh and support, followed by another oxidizing pretreatment and TPR experiment. Other samples hydrothermally aged at different temperatures (750 • C, 850 • C, and 950 • C) were prepared for comparing the changes of the amount of occluded Rh during the aging process in Rh/γ, θ-Al 2 O 3 catalysts. The occluded Rh covered with alumina layer cannot contact with H 2 [24], and it will not consume H 2 . Compared to fresh sample, the reduced H 2 consumption results from the occluded Rh species.
In Figure 3a, the H/Rh ratios of fresh samples are nearly three, which shows that no Rh species was occluded, assuming the reduction of Rh 3+ to Rh. Raising hydrothermal aging temperature, the amount of occluded Rh increased for both Rh/γ, θ-Al 2 O 3 samples. In addition, the amount of occluded Rh in aged Rh/θ-Al 2 O 3 was always lower than that of aged Rh/γ-Al 2 O 3 . Specifically, 62.5% Rh was occluded by alumina layer in Rh/γ-Al 2 O 3 , but it was just 45.9% for Rh/θ-Al 2 O 3 after hydrothermal aging at 1050 • C. The pre-sintering treatment of support would reduce the amount of occluded Rh during hydrothermal aging, as reported by McCabe [8]. The lower amount of occluded Rh is beneficial to maintain a higher catalytic activity for aged Rh/θ-Al 2 O 3 . method according to the research of McCabe and Hwang [8,23]. Samples were treated with H2 at 850 °C to reverse the strong interaction between Rh and support, followed by another oxidizing pretreatment and TPR experiment. Other samples hydrothermally aged at different temperatures (750 °C, 850 °C, and 950 °C) were prepared for comparing the changes of the amount of occluded Rh during the aging process in Rh/γ, θ-Al2O3 catalysts. The occluded Rh covered with alumina layer cannot contact with H2 [24], and it will not consume H2. Compared to fresh sample, the reduced H2 consumption results from the occluded Rh species.
In Figure 3a, the H/Rh ratios of fresh samples are nearly three, which shows that no Rh species was occluded, assuming the reduction of Rh 3+ to Rh. Raising hydrothermal aging temperature, the amount of occluded Rh increased for both Rh/γ, θ-Al2O3 samples. In addition, the amount of occluded Rh in aged Rh/θ-Al2O3 was always lower than that of aged Rh/γ-Al2O3. Specifically, 62.5% Rh was occluded by alumina layer in Rh/γ-Al2O3, but it was just 45.9% for Rh/θ-Al2O3 after hydrothermal aging at 1050 °C. The pre-sintering treatment of support would reduce the amount of occluded Rh during hydrothermal aging, as reported by McCabe [8]. The lower amount of occluded Rh is beneficial to maintain a higher catalytic activity for aged Rh/θ-Al2O3.  Figure 3. (a) The H2 consumption of Rh/γ, θ-Al2O3 catalysts after H2 treated at 850 °C for 60 min, next oxidizing at 500 °C, and then H2 reduced by temperature-program reduction, and (b) the Rh dispersion of different Rh/γ, θ-Al2O3 catalysts. Aged-x: x represents the hydrothermal aging temperature. For example, aged-750 represents the fresh sample that was hydrothermally aged at 750 °C for 10 h.
The Rh dispersions of all samples were obtained by CO-chemisorption, as shown in Figure 3b. The fresh samples show a similar Rh dispersion. After hydrothermal aging at 750 °C, the Rh dispersions experienced a rapid decline. The Rh dispersion of Rh/γ-Al2O3 which decreased from 46.3% to 6.7% may be caused by the strong interaction between Rh and γ-Al2O3. Yao et al. [10] reported the dispersed Rh oxide can enter the bulk of γ-Al2O3 support when temperature is over 600 °C, which leads to small surface area of non-interacting Rh maintained in support surface. At that time, only ~20% Rh was occluded. Most of Rh had already become strongly-interacting before occluded by the collapsed alumina layers. Thus, the occluded Rh has little contribution on the poor activity of aged catalysts. Interestingly, θ-Al2O3 showed a better ability to stabilize Rh during hydrothermal aging process. The Rh dispersion of Rh/θ-Al2O3 is nearly two times that of Rh/γ-Al2O3 after hydrothermal aging at 1050 °C, which is consistent with the activity test results ( Figure 1).
As we know, Rh and γ-Al2O3 can form strong interaction upon heat treatment temperature above 600 °C in air but the temperature increased to ~900 °C for α-Al2O3 [9,10]. Thus, the interaction between Rh and Al2O3 is getting weaker as alumina phase changes from γ-to θ-phase, and then to α-phase. In a word, the superior TWCs activity of Rh/θ-Al2O3 is mainly decided by the weaker interaction between Rh and θ-Al2O3. Aged-x: x represents the hydrothermal aging temperature. For example, aged-750 represents the fresh sample that was hydrothermally aged at 750 • C for 10 h.
The Rh dispersions of all samples were obtained by CO-chemisorption, as shown in Figure 3b. The fresh samples show a similar Rh dispersion. After hydrothermal aging at 750 • C, the Rh dispersions experienced a rapid decline. The Rh dispersion of Rh/γ-Al 2 O 3 which decreased from 46.3% to 6.7% may be caused by the strong interaction between Rh and γ-Al 2 O 3 . Yao et al. [10] reported the dispersed Rh oxide can enter the bulk of γ-Al 2 O 3 support when temperature is over 600 • C, which leads to small surface area of non-interacting Rh maintained in support surface. At that time, only~20% Rh was occluded. Most of Rh had already become strongly-interacting before occluded by the collapsed alumina layers. Thus, the occluded Rh has little contribution on the poor activity of aged catalysts. Interestingly, θ-Al 2 O 3 showed a better ability to stabilize Rh during hydrothermal aging process. The Rh dispersion of Rh/θ-Al 2 O 3 is nearly two times that of Rh/γ-Al 2 O 3 after hydrothermal aging at 1050 • C, which is consistent with the activity test results ( Figure 1).
As we know, Rh and γ-Al 2 O 3 can form strong interaction upon heat treatment temperature above 600 • C in air but the temperature increased to~900 • C for α-Al 2 O 3 [9,10]. Thus, the interaction between Rh and Al 2 O 3 is getting weaker as alumina phase changes from γto θ-phase, and then to α-phase. In a word, the superior TWCs activity of Rh/θ-Al 2 O 3 is mainly decided by the weaker interaction between Rh and θ-Al 2 O 3 .

The Researches of Non-and Strongly-Interacting Rh
To obtain more information about strongly-interacting Rh, fresh samples were hydrothermally aged at 1050 • C for 70 h to eliminate all non-interacting Rh, called deactivated samples. The activity test results of deactivated samples are shown in Figure 4. Both deactivated samples show a similar activity with each other. Except for strongly-interacting Rh, no non-interacting Rh can be maintained on the surface of supports after a long period of severe hydrothermal aging. Strongly-interacting Rh still has the capacity to catalyze CO and C 3 H 6 oxidation and NO reduction, but the activity of strongly-interacting Rh is much lower than that of non-interacting Rh.    Figure 5. XPS spectra (Rh 3d) of fresh and deactivated Rh/γ, θ-Al2O3.
CO-NO reaction was selected as a probe reaction to research more information about non-and strongly-interacting Rh species. Kinetics experiments were conducted to research the true active Rh sites for CO-NO reaction, as shown in Figure 6. Interestingly, similar apparent activation energy (Eapp) of 170 kJ/mol was observed in all the investigated catalysts, which is consistent with a previous report [31]. The similar Eapp indicated that the non-and strongly-interacting Rh have the same reaction path. The fresh and deactivated samples contain a large amount of non-and stronglyinteracting Rh on the surface of supports, respectively. The XPS analysis was performed to determine the chemical state of Rh in fresh and deactivated samples. As shown in Figure 5, the XPS spectra were fitted according to the work of Zimowska [25]. Three Rh species are detected on the surface of fresh samples.
The Rh 3d 5/2 peaks at 307.4 eV, 309.3-309.4, and 310.4-310.5 eV are assigned to nonionic Rh 0 , Rh 3+ , and Rh 4+ , respectively [18,25,26]. Additional Rh 3d XPS peak information is summarized in Table S2. Rh 3+ is the main Rh species in fresh samples, which accounts for 56.3% and 50.0% for fresh Rh/γ-Al 2 O 3 and Rh/θ-Al 2 O 3 , respectively. As for deactivated samples, Rh 4+ is the only detected Rh species at 310.3-310.4 eV. Weng-Sieh et al. [27] suggested that some Rh 4+ may exist in the form of RhO 2 . However, RhO 2 cannot be present in the deactivated samples because of the decomposition of RhO 2 at 680 • C under ordinary atmospheric condition [28]. Burch et al. [29] observed that Rh can incorporate into alumina structure at~500 • C. In addition, Zimowska et al. [25] thought that the Rh 4+ species was formed by the thermal evolution of Rh-O-Al xerogel; thus, we believed that Rh 4+ was formed when Rh 3+ inserted into the structure of Al 2 O 3 .    Figure 5. XPS spectra (Rh 3d) of fresh and deactivated Rh/γ, θ-Al2O3.
CO-NO reaction was selected as a probe reaction to research more information about non-and strongly-interacting Rh species. Kinetics experiments were conducted to research the true active Rh sites for CO-NO reaction, as shown in Figure 6. Interestingly, similar apparent activation energy (Eapp) of 170 kJ/mol was observed in all the investigated catalysts, which is consistent with a previous report [31]. The similar Eapp indicated that the non-and strongly-interacting Rh have the same reaction path. According to the XPS spectra, Rh 4+ is the only detected Rh species in deactivated samples. Thus, the strongly-interacting Rh may exist in the form of Rh 4+ . The surface of fresh samples contains Rh 0 (minority), Rh 3+ (majority), and Rh 4+ species, and fresh samples contained a large amount of non-interacting Rh. Further, Rh 0 and Rh 3+ are easier redox reagents compared with the Rh 4+ species [30]. Therefore, the non-interacting Rh may exist in the form of Rh/Rh 3+ .
CO-NO reaction was selected as a probe reaction to research more information about non-and strongly-interacting Rh species. Kinetics experiments were conducted to research the true active Rh sites for CO-NO reaction, as shown in Figure 6. Interestingly, similar apparent activation energy (E app ) of 170 kJ/mol was observed in all the investigated catalysts, which is consistent with a previous report [31]. The similar E app indicated that the non-and strongly-interacting Rh have the same reaction path. To further research the properties of active Rh sites, reaction order tests were conducted, as shown in Figure 7. The rate of NO reduction was determined under various partial concentrations of CO ( Figure 7a) and NO (Figure 7b), keeping those of NO and CO concentration constant, respectively. The reaction temperature was set to 240 °C for fresh samples and 290 °C for deactivated samples. The fresh and deactivated catalysts show the same reaction orders: 0.35 for CO and −0.60 for NO, respectively. The reaction could proceed through adsorbed CO and NO with the Langmuir-Hinshelwood mechanism. Moreover, the orders for NO are negative on all catalysts; an increase in NO concentration inhibits the adsorption of CO onto Rh active site. Therefore, we can conclude that the nonand strongly-interacting Rh in Rh/γ, θ-Al2O3 catalysts have the same reaction path. Temperature (ºC) Figure 6. Arrhenius plots of CO-NO reaction over fresh and deactivated Rh/γ, θ-Al2O3 catalysts. To further research the properties of active Rh sites, reaction order tests were conducted, as shown in Figure 7. The rate of NO reduction was determined under various partial concentrations of CO ( Figure 7a) and NO (Figure 7b), keeping those of NO and CO concentration constant, respectively. The reaction temperature was set to 240 • C for fresh samples and 290 • C for deactivated samples. The fresh and deactivated catalysts show the same reaction orders: 0.35 for CO and −0.60 for NO, respectively. The reaction could proceed through adsorbed CO and NO with the Langmuir-Hinshelwood mechanism. Moreover, the orders for NO are negative on all catalysts; an increase in NO concentration inhibits the adsorption of CO onto Rh active site. Therefore, we can conclude that the nonand strongly-interacting Rh in Rh/γ, θ-Al 2 O 3 catalysts have the same reaction path.

The Amount of Non-Interacting Rh in Aged Samples
Determining the amount of non-interacting Rh contained in aged samples is an interesting and challenging work. McCabe et al. [8] ascertained the amount of non-interacting Rh by H2-TPR method in Rh/δ, α-Al2O3 after different treatment, but the obtained figure has a large measurement error. As we know, the amount of non-interacting Rh is small for aged samples and then the H2 consumption peak is extremely weak. Leaching is a good method to remove nanoparticles from catalyst supports [32][33][34]. Luo et al. [35] removed the finely dispersed CuO species from the CuO-CeO2 catalysts by nitric acid and found the activity declined dramatically after being acid-treated. The non-interacting Rh

The Amount of Non-Interacting Rh in Aged Samples
Determining the amount of non-interacting Rh contained in aged samples is an interesting and challenging work. McCabe et al. [8] ascertained the amount of non-interacting Rh by H 2 -TPR method in Rh/δ, α-Al 2 O 3 after different treatment, but the obtained figure has a large measurement error. As we know, the amount of non-interacting Rh is small for aged samples and then the H 2 consumption peak is extremely weak. Leaching is a good method to remove nanoparticles from catalyst supports [32][33][34]. Luo et al. [35] removed the finely dispersed CuO species from the CuO-CeO 2 catalysts by nitric acid and found the activity declined dramatically after being acid-treated. The non-interacting Rh has little interaction with the alumina support, which may be removed easily from the catalyst support.
Here, we treated aged samples with HCl-KBr solution to remove non-interacting Rh from the supports; 2.5% and 4.0% Rh were removed by inductively coupled plasma (ICP) test from aged Rh/γ-Al 2 O 3 and aged Rh/θ-Al 2 O 3 , respectively. In addition, the catalytic activity for CO, C 3 H 6 , and NO conversions of aged samples after being acid-treated had a significant decline, especially at low temperature, as shown in Figure 8. Compared to strongly-interacting Rh, non-interacting Rh has higher catalytic activity of NO reduction. The activity of acid-treated catalysts is close to that of deactivated catalysts ( Figure S2). Thus, non-interacting Rh was indeed removed from catalyst support, and only 2.5% and 4.0% non-interacting Rh were maintained in aged Rh/γ-Al 2 O 3 and Rh/θ-Al 2 O 3 catalysts.
In Figures 1c and 8c, only fresh samples reached~20% NO conversion at 250 • C, while the NO conversion of other samples were neglected. The non-interacting Rh should be responsible for the activity of NO reduction at low temperatures. The hydrothermal aging treatment turns most of the non-interacting Rh into strongly-interacting Rh and occluded Rh. However, at high temperatures, an interesting phenomenon occurs in which the fresh and aged Rh/γ, θ-Al 2 O 3 samples show a close NO conversion at 325-400 • C as the latter samples contain trace non-interacting Rh. When 2.5% Rh was removed from the aged Rh/γ-Al 2 O 3 , the NO conversion decreased from 88.2% to 48.7%. According to the activity of these samples, the non-interacting Rh in aged samples must get a higher turnover frequency (TOF) of NO reduction. The reaction rates of fresh samples and aged samples hydrothermally aged at different temperatures were measured at 350 • C, as shown in Figure 9. The TOF is 3.48 s −1 at 350 • C for fresh Rh/γ-Al 2 O 3 catalyst. After hydrothermal aging treatment, the TOF indeed increased significantly. In total, the TOF of NO reduction of different aged samples was four times larger than that of fresh sample at least for Rh/γ-Al 2 O 3 . The phenomenon of TOF increasing after hydrothermal aging treatment also appeared in Rh/θ-Al 2 O 3 catalysts, as shown in Figure 9b. Thus, the state of non-interacting Rh must have a change during hydrothermal aging treatment.
has little interaction with the alumina support, which may be removed easily from the catalyst support.
Here, we treated aged samples with HCl-KBr solution to remove non-interacting Rh from the supports; 2.5% and 4.0% Rh were removed by inductively coupled plasma (ICP) test from aged Rh/γ-Al2O3 and aged Rh/θ-Al2O3, respectively. In addition, the catalytic activity for CO, C3H6, and NO conversions of aged samples after being acid-treated had a significant decline, especially at low temperature, as shown in Figure 8. Compared to strongly-interacting Rh, non-interacting Rh has higher catalytic activity of NO reduction. The activity of acid-treated catalysts is close to that of deactivated catalysts ( Figure S2). Thus, non-interacting Rh was indeed removed from catalyst support, and only 2.5% and 4.0% non-interacting Rh were maintained in aged Rh/γ-Al2O3 and Rh/θ-Al2O3 catalysts. In Figures 1c and 8c, only fresh samples reached ~20% NO conversion at 250 °C, while the NO conversion of other samples were neglected. The non-interacting Rh should be responsible for the activity of NO reduction at low temperatures. The hydrothermal aging treatment turns most of the non-interacting Rh into strongly-interacting Rh and occluded Rh. However, at high temperatures, an interesting phenomenon occurs in which the fresh and aged Rh/γ, θ-Al2O3 samples show a close NO conversion at 325-400 °C as the latter samples contain trace non-interacting Rh. When 2.5% Rh was removed from the aged Rh/γ-Al2O3, the NO conversion decreased from 88.2% to 48.7%. According to the activity of these samples, the non-interacting Rh in aged samples must get a higher turnover frequency (TOF) of NO reduction. The reaction rates of fresh samples and aged samples hydrothermally aged at different temperatures were measured at 350 °C, as shown in Figure   Figure 8. The profiles of (a) CO, (b) C 3 H 6 , (c) NO conversion, and (d) N 2 O formation over aged Rh/γ, θ-Al 2 O 3 acid-treated catalysts during TWC reactions. The activity of aged Rh/γ, θ-Al 2 O 3 is also drawn as a reference. Feed stream: 1% CO, 1000 ppm HCs (C 3 H 6 :C 3 H 8 = 2:1), 1000 ppm NO, 0.917% O 2 , 12% CO 2 , 3% H 2 O, and N 2 balance. Space velocity: 6,000,000 cm 3 ·g cat −1 ·h −1 for Rh/γ-Al 2 O 3 .
sts 2021, 11, x FOR PEER REVIEW 9 of 13 9. The TOF is 3.48 s -1 at 350 °C for fresh Rh/γ-Al2O3 catalyst. After hydrothermal aging treatment, the TOF indeed increased significantly. In total, the TOF of NO reduction of different aged samples was four times larger than that of fresh sample at least for Rh/γ-Al2O3. The phenomenon of TOF increasing after hydrothermal aging treatment also appeared in Rh/θ-Al2O3 catalysts, as shown in Figure 9b. Thus, the state of non-interacting Rh must have a change during hydrothermal aging treatment.  Rh dispersed well in fresh catalysts according to the results of CO-chemisorption. The TOF of NO reduction of fresh Rh/θ-Al2O3 is nearly two times larger than that of fresh Rh/γ-Al2O3 at 350 °C, and the two fresh samples have closed Rh dispersion. Therefore, the Rh dispersed well in fresh catalysts according to the results of CO-chemisorption. The TOF of NO reduction of fresh Rh/θ-Al 2 O 3 is nearly two times larger than that of fresh Rh/γ-Al 2 O 3 at 350 • C, and the two fresh samples have closed Rh dispersion. Therefore, the TOF of NO reduction at high temperature is related to catalyst supports which could lead to the different surrounding of non-interacting Rh. Fernández et al. [36] reported that subnanometric Pt clusters exhibit better catalytic performance than single Pt atoms for CO-NO reaction. The subnanometric Rh cluster may generate during the aggregation of Rh atoms at high temperature, and is possibly responsible for the higher TOF of NO reduction at high reaction temperature.

Catalyst Preparation
The γ-, θ-Al 2 O 3 were obtained by calcining commercial Al 2 O 3 (Sasol) at 400 • C and 1100 • C for 4 h in air, respectively. X-ray diffraction examination of these samples showed a good γ-, θ-Al 2 O 3 phase ( Figure S3). Rh loading was 0.6 µmol Rh atoms/m 2 (BET). The surface areas of γ, θ-Al 2 O 3 supports were 142 m 2 /g and 83 m 2 /g, respectively. The nominal Rh loading of Rh/γ-Al 2 O 3 and Rh/θ-Al 2 O 3 was 0.87 wt. % and 0.51 wt. %, respectively. Supported Rh/γ, θ-Al 2 O 3 catalysts were prepared by the incipient wetness impregnation method, followed by air-drying at 100 • C and calcination at 550 • C for 2 h. The obtained samples were called fresh samples. The fresh catalysts were hydrothermally aged at 1050 • C in flowing air with 10% H 2 O for 10 h and 70 h, called aged and deactivated samples, respectively. During the hydrothermal aging process, the amount of Rh could be occluded by Al 2 O 3 layers. In order to compare changes in the amount of occluded Rh, the fresh samples were hydrothermally aged at 750 • C, 850 • C, and 950 • C for 10 h, and the obtained samples denoted as "aged-x." The x represents the hydrothermal aging temperature. For example, aged-750 represents the fresh sample that was hydrothermally aged at 750 • C for 10 h.
To remove the non-interacting Rh from aged samples, 0.5 g aged catalyst and 0.48 g KBr were immersed in 25 mL 12 M HCl solution at 100 • C for 2 h and then mixed with 100 mL deionized water before filtering. The first filtrate was collected for ICP test, and the residue was washed with 80 • C H 2 O to remove Cl − until there was no white precipitate formation when the filtrate was mixed with AgNO 3 solution. The residue was dried at 100 • C overnight and calcined at 250 • C for 2 h. Only 2.5% and 4.0% Rh were removed from aged Rh/γ-Al 2 O 3 and Rh/θ-Al 2 O 3 , respectively, and the dissolved amounts of Al 2 O 3 support were both less than 6.6 wt. %.

Catalytic Activity Tests
The steady-state catalytic performance tests and kinetics measurements were performed in a packed-bed flow microreactor. The steady-state reaction feed contained 1% CO, 1000 ppm HCs (C 3 H 6: C 3 H 8 = 2:1), 1000 ppm NO, 0.917% O 2 , 12% CO 2 , 3% H 2 O, and N 2 balance. The space velocity was 6,000,000 cm 3 ·g cat −1 ·h −1 for Rh/γ-Al 2 O 3 . Typically, each test sample contained 0.087 mg Rh, and the weight of the mixture of catalyst and diluted SiO 2 was 250 mg. Prior to the activity test, the catalyst was treated with 5% H 2 /N 2 at 350 • C for 30 min. The inlet and outlet gas compositions were analyzed with an MKS MultiGas 2030 FT-IR. CO-NO reaction kinetics measurements reaction feed contained 1000 ppm CO, 1000 ppm NO, 3% H 2 O, and N 2 balance. The space velocity was 7,800,000 cm 3 ·gcat −1 ·h −1 for Rh/γ-Al 2 O 3 . Prior to the kinetics measurements test, the catalyst was treated with reaction feed at 500 • C for 1 h. The reaction orders were measured at 240 • C for fresh samples and 290 • C for deactivated samples to keep the conversions of NO below 20%. The CO concentration was varied between 1000 and 4000 ppm, and the NO concentration was varied between 1000 and 4000, respectively. The NO reaction rate of Rh/γ, θ-Al 2 O 3 catalysts were determined at 350 • C in TWC reaction feed stream. The flow rate was 3500 mL/min, and the mass of catalysts varied from 3 mg to 18 mg. The NO conversion