In Situ TEM Study of Rh Particle Sintering for Three-Way Catalysts in High Temperatures

: One of the main factors in the deterioration of automobile three-way catalysts is the sintering of platinum group metals (PGMs). In this study, we used in situ tunneling electron microscopy (TEM) to examine the sintering of Rh particles as the temperature increases. Two types of environmental conditions were tested, namely, vacuum atmosphere with heating up to 1050 ◦ C, and N 2 with/without 1% O 2 at 1 atm and up to 1000 ◦ C. Under vacuum, Rh particles appeared to be immersed in ZrO 2 . In contrast, at 1 atm N 2 with or without 1% O 2 , the sintered Rh particles appeared spherical and not immersed in ZrO 2. The latter trend of Rh sintering resembles the actual engine-aged catalyst observed ex situ in this study. In the N 2 atmosphere, the sintering of support material (ZrO 2 or Y-ZrO 2 ) was ﬁrst observed by in situ TEM, followed by Rh particle sintering. The Rh particle size was slightly smaller on Y-ZrO 2 compared to that on ZrO 2 . To better understand these experimental results, density functional theory was used to calculate the systems’ junction energies, assuming three layers of Rh(111) 4 × 4 structures joined to the support material (ZrO 2 and Y-ZrO 2 ). The calculated energies were consistent with the in situ TEM observations in the N 2 atmosphere. Author Contributions: Conceptualization, M.N., and Y.S.; methodology, M.N., and Y.S.; software, Y.S.; validation, H.N., M.N., and Y.S.; formal analysis, H.N., M.N. and Y.S.; investigation, H.N., H.A., M.N. and Y.S.; resources, H.N., M.N. and Y.S.; data curation, H.N., M.N. and Y.S.; writing— original draft preparation, H.N., M.N. and Y.S.; writing—review and editing, H.A., M.N. and Y.S.; visualization, H.N., M.N. and Y.S.; supervision, Y.S.; project administration, M.N and Y.S.; All authors agreed version of the manuscript.


Introduction
Platinum group metals (PGMs) such as Pt, Pd, and Rh are commonly used as active sites in catalytic after-treatment applications in automobiles, especially the three-way catalyst (TWC) for gasoline engines. Such PGMs are dispersed on Al 2 O 3 and/or CZ (ceria-zirconia, a solid solution or composite of CeO 2 and ZrO 2 ). The PGM species and support material are optimized to maximize the degradation of hydrocarbons (HCs), CO, and NO x . To meet the regulations in each region [1], it is crucial to stabilize the dispersion of PGMs, which determines the active site density in the catalyst. To maintain highly dispersed PGMs after aging or the required life usage of the catalyst, many studies have focused on understanding the mechanism of PGM sintering and suppressing such behavior. One approach toward hindering PGM particle sintering is strengthening the interaction between the particles and the support material. For example, Pt sintering could be suppressed by loading Pt particles onto CeO 2 [2] or MgO [3]. Pd sintering was suppressed by loading Pd onto CeO 2 [4] or YSZ (Y-doped ZrO 2 ) [5]. Rh sintering was also suppressed by loading Rh onto Nd-doped La-ZrO 2 [6]. Another approach is the creation of a physical matrix around the PGM particles, such as the use of Al 2 O 3 material against Pt sintering [7] and Al 2 O 3 nanorod material [8] against Rh sintering. Regarding the PGM sintering mechanism, both macro-and micro-scale theories have been reported. Suzuki et al. developed multiscale theoretical methods for predicting the sintering behavior of Pt on various catalyst supports [9]. On the microscale, density functional theory (DFT)

In Situ TEM Observations
The TEM observation results and EDS mapping of Rh under vacuum conditions are shown in  for Rh/ZrO2, Rh/Y-ZrO2, Rh/CeO2, and Rh/Al2O3, respectively. In Figure 2 (Rh/ZrO2), changes in the morphology of ZrO2 and the sintering of oxide were observed at 900 °C. At 950 °C, remarkable Rh particle sintering was observed; however, the STEM image was not sufficiently clear to separate the Rh and oxide parts. Rather, EDS mapping information was required to identify the Rh. In addition, the Rh appeared to be immersed in ZrO2 rather than as spherical particles on the oxide. This phenomenon is discussed later in this paper. At 950 °C and above, the sintering of oxide became more severe. As shown in Figure 3 (Rh/Y-ZrO2), the morphology of Y-ZrO2 began to change, and oxide sintering was observed at 1000 °C. At 1050 °C, remarkable Rh particle sintering was also observed. Compared to Rh/ZrO2, Y-ZrO2 exhibited significantly more optimized resistance against oxide sintering, possibly because of the stabilizing effect of Y-doping on ZrO2. Figure 4 (Rh/CeO2) shows that the morphology of CeO2 began to change and oxide sintering was observed at 900 °C; remarkable Rh particle sintering was observed at 1050 °C. As shown in Figure 5 (Rh/Al2O3), the morphology of Al2O3 began to change and oxide and Rh particle sintering were observed at 1000 °C. Table 1 provides a summary of the sintering behavior of Rh and oxide for different material combinations.
TEM observation results and EDS mapping of Rh in Rh/ZrO2 under flowing gas at 1 atm (N2 or 1% O2 with balance N2) at various temperatures are shown in Figure 6. The morphology of ZrO2 began to change and oxide sintering was observed at 900 °C (N2 gas flow). At 950 °C, remarkable Rh particle sintering was observed, and the morphology of Rh particles was very different from that under the vacuum conditions ( Figure 2). In this study, the Rh particle appeared spherical and was deposited on ZrO2. At 1000 °C in 1% O2 (balance N2) gas flow, the ZrO2 material was severely sintered, and even the grain boundary of the crystal could not be seen.

In Situ TEM Observations
The TEM observation results and EDS mapping of Rh under vacuum conditions are shown in  for Rh/ZrO 2 , Rh/Y-ZrO 2 , Rh/CeO 2 , and Rh/Al 2 O 3 , respectively. In Figure 2 (Rh/ZrO 2 ), changes in the morphology of ZrO 2 and the sintering of oxide were observed at 900 • C. At 950 • C, remarkable Rh particle sintering was observed; however, the STEM image was not sufficiently clear to separate the Rh and oxide parts. Rather, EDS mapping information was required to identify the Rh. In addition, the Rh appeared to be immersed in ZrO 2 rather than as spherical particles on the oxide. This phenomenon is discussed later in this paper. At 950 • C and above, the sintering of oxide became more severe. As shown in Figure 3 (Rh/Y-ZrO 2 ), the morphology of Y-ZrO 2 began to change, and oxide sintering was observed at 1000 • C. At 1050 • C, remarkable Rh particle sintering was also observed. Compared to Rh/ZrO 2 , Y-ZrO 2 exhibited significantly more optimized resistance against oxide sintering, possibly because of the stabilizing effect of Y-doping on ZrO 2 . Figure 4 (Rh/CeO 2 ) shows that the morphology of CeO 2 began to change and oxide sintering was observed at 900 • C; remarkable Rh particle sintering was observed at 1050 • C. As shown in Figure 5 (Rh/Al 2 O 3 ), the morphology of Al 2 O 3 began to change and oxide and Rh particle sintering were observed at 1000 • C. Table 1 provides a summary of the sintering behavior of Rh and oxide for different material combinations.
Al2O3/Rh   Figure 6. The morphology of ZrO 2 began to change and oxide sintering was observed at 900 • C (N 2 gas flow). At 950 • C, remarkable Rh particle sintering was observed, and the morphology of Rh particles was very different from that under the vacuum conditions ( Figure 2). In this study, the Rh particle appeared spherical and was deposited on ZrO 2 . At 1000 • C in 1% O 2 (balance N 2 ) gas flow, the ZrO 2 material was severely sintered, and even the grain boundary of the crystal could not be seen.

Effects of Rh-ZrO2 Interaction on the In Situ Sintering Behavior
As can be seen in Figures 2 and 6, the in situ status of the Rh particles on Rh/ZrO2 clearly depends on the atmosphere (vacuum or gas flow, respectively). After heating in vacuum, the Rh particle was not clearly visible and appeared to be immersed in ZrO2, whereas the Rh particle heated in gas flow appeared spherical and deposited on ZrO2. One possible explanation for this is the oxygen partial pressure. In Figure 3, because of the very low oxygen partial pressure at high temperatures, the electron beam irradiated during the STEM and EDS measurements may have reduced the ZrO2. While a quantitative discussion may be difficult, the electron irradiation of EDS is 90 times higher than that of STEM. Because frequent EDS measurements were performed under the vacuum conditions in addition to the STEM observations in Figure 2, a certain portion of ZrO2 may have been reduced, and the affinity between Rh and ZrO2 increased.
Next, the reduction of Rh2O3 to Rh in various gas conditions was examined by Differential thermal analysis-thermogravimetry (DTA-TG). As seen in Figure 7, Rh2O3 reduction was observed at 200 °C in H2, 800 °C in N2, and 1050 °C in 1% O2 (N2 balance). The amounts of weight loss in the high-temperature region in pure N2 and 1% O2 (N2 balance) were consistent with the amount of Rh reduction from Rh2O3. Thus, the reduction of Rh2O3 to Rh was considered to occur on ZrO2. This result also suggested that Rh on ZrO2 existed

Effects of Rh-ZrO 2 Interaction on the In Situ Sintering Behavior
As can be seen in Figures 2 and 6, the in situ status of the Rh particles on Rh/ZrO 2 clearly depends on the atmosphere (vacuum or gas flow, respectively). After heating in vacuum, the Rh particle was not clearly visible and appeared to be immersed in ZrO 2 , whereas the Rh particle heated in gas flow appeared spherical and deposited on ZrO 2 . One possible explanation for this is the oxygen partial pressure. In Figure 3, because of the very low oxygen partial pressure at high temperatures, the electron beam irradiated during the STEM and EDS measurements may have reduced the ZrO 2 . While a quantitative discussion may be difficult, the electron irradiation of EDS is 90 times higher than that of STEM. Because frequent EDS measurements were performed under the vacuum conditions in addition to the STEM observations in Figure 2, a certain portion of ZrO 2 may have been reduced, and the affinity between Rh and ZrO 2 increased.
Next, the reduction of Rh 2 O 3 to Rh in various gas conditions was examined by Differential thermal analysis-thermogravimetry (DTA-TG). As seen in Figure 7, Rh 2 O 3 reduction was observed at 200 • C in H 2 , 800 • C in N 2 , and 1050 • C in 1% O 2 (N 2 balance). The amounts of weight loss in the high-temperature region in pure N 2 and 1% O 2 (N 2 balance) were consistent with the amount of Rh reduction from Rh 2 O 3 . Thus, the reduction of Rh 2 O 3 to Rh was considered to occur on ZrO 2 . This result also suggested that Rh on ZrO 2 existed as Rh 2 O 3 after the pre-oxidation treatment. However, the weight loss in the H 2 atmosphere exceeded the reduction amount of Rh. It has also been reported that Rh 2 O 3 on ZrO 2 is reduced at low temperatures in an H 2 atmosphere [6]. Therefore, the weight loss above 300 • C was attributed to the reduction of a part of the ZrO 2 . In this study, the H 2 environment used for measurement was considered to have a very low oxygen partial pressure, similar to that under the vacuum conditions. From these results, the vacuum and high-temperature conditions (e.g., >900 • C) for the in situ TEM observation were considered sufficient for reducing both Rh oxide and ZrO 2 . Then, the interaction between Rh and ZrO 2 became sufficiently strong to generate a partial solid solution. Nonetheless, a certain amount of oxygen was present on the sample in the N 2 environment, even without the 1% O 2 . In these environments, at least the ZrO 2 is not severely reduced, resulting in less mutual affinity between Rh and ZrO 2 .
temperature conditions (e.g., >900 °C) for the in situ TEM observation were consid sufficient for reducing both Rh oxide and ZrO2. Then, the interaction between Rh and became sufficiently strong to generate a partial solid solution. Nonetheless, a ce amount of oxygen was present on the sample in the N2 environment, even without th O2. In these environments, at least the ZrO2 is not severely reduced, resulting in less tual affinity between Rh and ZrO2.
Compared to the ex situ STEM observations for the actual aged catalyst (Figur the in situ observations in vacuum ( Figure 2) varied slightly, while those in N2 or 1 (balance N2) ( Figure 6) exhibited greater similarity. This result shows that the gas c effective in the observation of real conditions, and the beam effects should be elimin in in situ TEM studies for catalytic analysis [20].

In Situ TEM Observation in N2 and O2 Atmospheres
Sets of in situ TEM images in N2 and O2 atmospheres are shown in Figure Rh/ZrO2 and Figure 9 for Rh/Y-ZrO2. In addition, the Rh particle size measured in ea situ TEM observation is shown in Figure 10. The Rh particle sizes were obtained usin TEM images from 940 to 950 °C in the N2 atmosphere, wherein the samples were assu to be in the Rh/ZrO2 state. For the Rh particles, the circle equivalent diameter was c lated by adjusting the threshold value of the mode method for the target Rh particle shown in Figure 10. In the case of Rh/ZrO2, aggregation of ZrO2 began at 918.1 °C, an particles began to precipitate. At higher temperatures, ZrO2 aggregation and Rh pa growth were observed. After reaching 950 °C, the switch to O2 atmosphere promote Compared to the ex situ STEM observations for the actual aged catalyst (Figure 1), the in situ observations in vacuum ( Figure 2) varied slightly, while those in N 2 or 1% O 2 (balance N 2 ) ( Figure 6) exhibited greater similarity. This result shows that the gas cell is effective in the observation of real conditions, and the beam effects should be eliminated in in situ TEM studies for catalytic analysis [20].

In Situ TEM Observation in N 2 and O 2 Atmospheres
Sets of in situ TEM images in N 2 and O 2 atmospheres are shown in Figure 8 for Rh/ZrO 2 and Figure 9 for Rh/Y-ZrO 2 . In addition, the Rh particle size measured in each in situ TEM observation is shown in Figure 10. The Rh particle sizes were obtained using the TEM images from 940 to 950 • C in the N 2 atmosphere, wherein the samples were assumed to be in the Rh/ZrO 2 state. For the Rh particles, the circle equivalent diameter was calculated by adjusting the threshold value of the mode method for the target Rh particles, as shown in Figure 10. In the case of Rh/ZrO 2 , aggregation of ZrO 2 began at 918.1 • C, and Rh particles began to precipitate. At higher temperatures, ZrO 2 aggregation and Rh particle growth were observed. After reaching 950 • C, the switch to O 2 atmosphere promoted Rh particle growth, as many Rh particles were observed. In the case of Rh/Y-ZrO 2 , no remarkable aggregation of Y-ZrO 2 was observed between 900 and 950 • C; however, particles that were thought to be Rh were observed from 934.2 • C onwards. After reaching 950 • C, switching to the O 2 atmosphere promoted ZrO 2 aggregation and Rh particle growth, as many Rh particles were observed. The observed Rh particle shape was different from the in situ TEM measurement result in vacuum and resembled that in the engine-aged actual catalyst. As mentioned earlier, the observation of Rh particles by STEM was probably because the support material was not reduced and the Rh/oxide state was maintained during the in situ measurement. Overall, sintering of the support material (ZrO 2 and Y-ZrO 2 ) began first, followed by Rh particle sintering. A smaller Rh particle size was observed for Rh/Y-ZrO 2 compared to Rh/ZrO 2 , as detailed in Figure 10.
able aggregation of Y-ZrO2 was observed between 900 and 950 °C; however, particles that were thought to be Rh were observed from 934.2 °C onwards. After reaching 950 °C, switching to the O2 atmosphere promoted ZrO2 aggregation and Rh particle growth, as many Rh particles were observed. The observed Rh particle shape was different from the in situ TEM measurement result in vacuum and resembled that in the engine-aged actual catalyst. As mentioned earlier, the observation of Rh particles by STEM was probably because the support material was not reduced and the Rh/oxide state was maintained during the in situ measurement. Overall, sintering of the support material (ZrO2 and Y-ZrO2) began first, followed by Rh particle sintering. A smaller Rh particle size was observed for Rh/Y-ZrO2 compared to Rh/ZrO2, as detailed in Figure 10.   particle growth, as many Rh particles were observed. In the case of Rh/Y-ZrO2, no remarkable aggregation of Y-ZrO2 was observed between 900 and 950 °C; however, particles that were thought to be Rh were observed from 934.2 °C onwards. After reaching 950 °C, switching to the O2 atmosphere promoted ZrO2 aggregation and Rh particle growth, as many Rh particles were observed. The observed Rh particle shape was different from the in situ TEM measurement result in vacuum and resembled that in the engine-aged actual catalyst. As mentioned earlier, the observation of Rh particles by STEM was probably because the support material was not reduced and the Rh/oxide state was maintained during the in situ measurement. Overall, sintering of the support material (ZrO2 and Y-ZrO2) began first, followed by Rh particle sintering. A smaller Rh particle size was observed for Rh/Y-ZrO2 compared to Rh/ZrO2, as detailed in Figure 10.

DFT Calculation Results
DFT calculations were carried out with two purposes. The first was to examine the electronic state of a system of Rh particles embedded in the given supports, as shown in the STEM image in vacuum. The second purpose was to correlate the effect of different supports on the sintering behavior with the theoretical binding energies.
As mentioned earlier, the STEM images in vacuum showed Rh particles embedded in the ZrO 2 support (Figure 2d,e). This embedding may be attributed to the reduction of the support by the electron beam under vacuum and high-temperature conditions. We further examined the stability of this state of Rh particles embedded in a reduced ZrO 2 support using DFT calculations. We adopted the (111) surface of cubic ZrO 2 as the simplest model on which Rh particles were supported. First, a Rh 12 cluster was placed on a flat Zr-terminated surface, which is a simple model of a strongly reduced state. However, our energy and structural relaxation calculations failed to identify a stable adsorption state, suggesting that the Rh clusters do not adsorb stably on this surface. Next, we studied the adsorption of a Rh cluster on a depressed ZrO 2 surface, which was created by removing four Zr ions from the top surface, as shown in Figure 11. The slab of the support was constructed by stacking five layers of ZrO 2 with a 3 × 3 structure on the (111) surface of cubic ZrO 2 . Three layers of atoms from the surface of the slab and the Rh 12 cluster were structurally relaxed, while the atom positions of the remaining two layers of the slab were fixed at the position of the bulk crystal. The vacuum layer was set to exceed 15 Å. The energy and structural relaxation calculations showed that the Rh 12 cluster was strongly bound to the ZrO 2 surface (Figure 12). Some Rh atoms were located in surface depressions, and the Rh cluster embedded in the surface became a stable structure. This result is consistent with the STEM images obtained in vacuum. The binding energy between the Rh 12 cluster and the ZrO 2 slab was −15.3 eV. This binding energy was defined by subtracting the energies of the two isolated subsystems of the Rh 12 cluster and ZrO 2 slab (with the structures shown in Figure 12) from the total energy of the system. Because a total of six Rh ions are bound to the support, the binding energy per bond is −2.6 eV. This binding is stronger than that of the system with a Rh 12 cluster adsorbed on a flat oxygen-terminated surface. The latter structure is shown in Figure 13, wherein the settings of the vacuum layer and atoms being relaxed were the same as those for determining the structure in Figure 12. The Rh cluster is bound to the oxygen ion at the top surface through six bonds, and the binding energy per bond in this system is −1.5 eV.
Catalysts 2021, 11, x FOR PEER REVIEW 8 of 17 Figure 10. (a) Temperature dependence of the Rh particle size, which was obtained from tracking the corresponding Rh particle in the STEM images in Figures 8 and 9. (b) and (c) are STEM images of Rh/ZrO2 and Rh/Y-ZrO2 at 950 °C in N2 gas. The size of the Rh particles within the dashed circle was evaluated.

DFT Calculation Results
DFT calculations were carried out with two purposes. The first was to examine the electronic state of a system of Rh particles embedded in the given supports, as shown in the STEM image in vacuum. The second purpose was to correlate the effect of different supports on the sintering behavior with the theoretical binding energies.
As mentioned earlier, the STEM images in vacuum showed Rh particles embedded in the ZrO2 support (Figure 2d,e). This embedding may be attributed to the reduction of the support by the electron beam under vacuum and high-temperature conditions. We further examined the stability of this state of Rh particles embedded in a reduced ZrO2 support using DFT calculations. We adopted the (111) surface of cubic ZrO2 as the simplest model on which Rh particles were supported. First, a Rh12 cluster was placed on a flat Zr-terminated surface, which is a simple model of a strongly reduced state. However, our energy and structural relaxation calculations failed to identify a stable adsorption state, suggesting that the Rh clusters do not adsorb stably on this surface. Next, we studied the adsorption of a Rh cluster on a depressed ZrO2 surface, which was created by remov- Figure 10. (a) Temperature dependence of the Rh particle size, which was obtained from tracking the corresponding Rh particle in the STEM images in Figures 8 and 9. (b) and (c) are STEM images of Rh/ZrO 2 and Rh/Y-ZrO 2 at 950 • C in N 2 gas. The size of the Rh particles within the dashed circle was evaluated.
Rh ions are bound to the support, the binding energy per bond is −2.6 eV. This binding is stronger than that of the system with a Rh12 cluster adsorbed on a flat oxygen-terminated surface. The latter structure is shown in Figure 13, wherein the settings of the vacuum layer and atoms being relaxed were the same as those for determining the structure in Figure 12. The Rh cluster is bound to the oxygen ion at the top surface through six bonds, and the binding energy per bond in this system is −1.5 eV.   stronger than that of the system with a Rh12 cluster adsorbed on a flat oxygen-terminated surface. The latter structure is shown in Figure 13, wherein the settings of the vacuum layer and atoms being relaxed were the same as those for determining the structure in Figure 12. The Rh cluster is bound to the oxygen ion at the top surface through six bonds, and the binding energy per bond in this system is −1.5 eV.   A more strongly bound system (i.e., the one with Rh clusters embedded in a surface depression) is expected to display strong metal-support interaction (SMSI) effects. Figure 14 shows the differential electron density distributions for the two systems of Rh clusters embedded in the depressed Zr-terminated surface and adsorbed on the flat oxygenterminated surface. The differential electron density is given as ∆ρ = ρ(Rh 12 /support)ρ(Rh 12 )-ρ(support), where ρ(material) is the electron number density of the given isolated material. ∆ρ represents the change in the number of electrons due to binding of the Rh cluster to the support. Rh atoms in the ∆ρ > 0 region have increased ability to donate electrons due to this binding, while Rh ions in the ∆ρ < 0 region have a stronger ability to accept electrons. An increase in electron donating/reverse-donating capacity due to the bonding of Rh clusters to the support material could be one of the SMSI effects. In a system with Rh clusters bound to the depressed support, the region of non-zero ∆ρ extends not only to the position of the bound Rh cluster but also in the region between the Rh ions near the apex of the Rh cluster. This means that the region of higher catalytic activity is more extensive in the case where the Rh clusters are embedded in surface depressions. A more strongly bound system (i.e., the one with Rh clusters embedded in a surface depression) is expected to display strong metal-support interaction (SMSI) effects. Figure  14 shows the differential electron density distributions for the two systems of Rh clusters embedded in the depressed Zr-terminated surface and adsorbed on the flat oxygen-terminated surface. The differential electron density is given as Δρ = ρ(Rh12/support) -ρ(Rh12) -ρ(support), where ρ(material) is the electron number density of the given isolated material. Δρ represents the change in the number of electrons due to binding of the Rh cluster to the support. Rh atoms in the Δρ > 0 region have increased ability to donate electrons due to this binding, while Rh ions in the Δρ < 0 region have a stronger ability to accept electrons. An increase in electron donating/reverse-donating capacity due to the bonding of Rh clusters to the support material could be one of the SMSI effects. In a system with Rh clusters bound to the depressed support, the region of non-zero Δρ extends not only to the position of the bound Rh cluster but also in the region between the Rh ions near the apex of the Rh cluster. This means that the region of higher catalytic activity is more extensive in the case where the Rh clusters are embedded in surface depressions.  Figures 12b and 13b, respectively. The yellow-green isosurface represents positive Δρ = ρ(Rh12/support) − ρ(Rh12) − ρ(support), and the sky blue isosurface represents negative Δρ. Δρ ~ 0 in the region where no isosurface is displayed.
We calculated the binding energies between Rh and various types of supports using DFT and compared them with our STEM observations of Rh sintering. In particular, we focused on the supports for which we measured the temperature-dependent Rh particle size, namely, ZrO2 and Y-ZrO2, as shown in Figure 10. Cubic ZrO2, Y-ZrO2, oxygen-deficient Y-ZrO2, and CeO2 were chosen as supports, and the (111) surface of each support was adopted as the surface for Rh binding. The same crystal symmetry was assumed to compare the binding energy on different support materials. Structural relaxation with small Rh clusters on the support surface resulted in multiple local stable states with different cluster shapes that have different numbers of bonds between Rh and oxygen in the support material. To eliminate such multiple states, we calculated the junction energy at the Rh-support interface, instead of the binding energy between the Rh cluster and the support. Figure 15a is a top view of the five-layered 3 × 3 structure of the (111) plane of oxygenterminated ZrO2. The slab structure, which was structurally relaxed along the c-axis, was based on the lattice parameters of structurally relaxed cubic ZrO2. Figure 15b is a top view of the three-layered 4 × 4 structure of the (111) plane of fcc-Rh. The blue lines in Figure  15a indicate a good match between the ab-planes of the surface of the ZrO2 slab and the Rh 4 × 4 structure. After structural relaxation, some Rh atoms in the bottom layer are expected to bind to the oxygen of the support, and those not bound to oxygen should be stabilized by the metal bonds in Rh. The same may be true for the binding of support materials to the Rh particles, a few nanometers in size, measured in our STEM experiments. Structural relaxation calculations were performed for the three-layered structure of Rh and the interfacial three-layered structure of ZrO2, and the junction energy at the interface was −6.24 eV. The junction energy was defined here by subtracting the sum of the energy of the two isolated subsystems (Rh layers and ZrO2 slab, with the calculated structures in the entire system) from the total energy of the entire system. The junction energy per Rh number at the interface was −0.39 eV.  Figures 12b and 13b, respectively. The yellow-green isosurface represents positive ∆ρ = ρ(Rh 12 /support) − ρ(Rh 12 ) − ρ(support), and the sky blue isosurface represents negative ∆ρ. ∆ρ~0 in the region where no isosurface is displayed.
We calculated the binding energies between Rh and various types of supports using DFT and compared them with our STEM observations of Rh sintering. In particular, we focused on the supports for which we measured the temperature-dependent Rh particle size, namely, ZrO 2 and Y-ZrO 2 , as shown in Figure 10. Cubic ZrO 2 , Y-ZrO 2 , oxygendeficient Y-ZrO 2 , and CeO 2 were chosen as supports, and the (111) surface of each support was adopted as the surface for Rh binding. The same crystal symmetry was assumed to compare the binding energy on different support materials. Structural relaxation with small Rh clusters on the support surface resulted in multiple local stable states with different cluster shapes that have different numbers of bonds between Rh and oxygen in the support material. To eliminate such multiple states, we calculated the junction energy at the Rh-support interface, instead of the binding energy between the Rh cluster and the support. Figure 15a is a top view of the five-layered 3 × 3 structure of the (111) plane of oxygen-terminated ZrO 2 . The slab structure, which was structurally relaxed along the c-axis, was based on the lattice parameters of structurally relaxed cubic ZrO 2 . Figure 15b is a top view of the three-layered 4 × 4 structure of the (111) plane of fcc-Rh. The blue lines in Figure 15a indicate a good match between the ab-planes of the surface of the ZrO 2 slab and the Rh 4 × 4 structure. After structural relaxation, some Rh atoms in the bottom layer are expected to bind to the oxygen of the support, and those not bound to oxygen should be stabilized by the metal bonds in Rh. The same may be true for the binding of support materials to the Rh particles, a few nanometers in size, measured in our STEM experiments. Structural relaxation calculations were performed for the three-layered structure of Rh and the interfacial three-layered structure of ZrO 2 , and the junction energy at the interface was −6.24 eV. The junction energy was defined here by subtracting the sum of the energy of the two isolated subsystems (Rh layers and ZrO 2 slab, with the calculated structures in the entire system) from the total energy of the entire system. The junction energy per Rh number at the interface was −0.39 eV.  Figure 16 shows the six-layered 3 × 3 structure for the (111) plane of oxygen-terminated Y-ZrO2. Each layer contains one Y atom and eight Zr atoms. Because the ionic radius of Y is larger than that of Zr, the a-and b-axis lengths of the unit cell of the structurally relaxed Y-ZrO2 are slightly longer than those of ZrO2. However, the matching between the ab-plane of the surface of the Y-ZrO2 slab and the Rh 4 × 4 structure remains adequate. Structural relaxation calculations were performed, and the junction energy per Rh atom at the interface was found to be −1.06 eV. The bonding of Rh to Y-ZrO2 at the interface was thus illustrated to be stronger than the bonding to ZrO2. This is consistent with the STEM observation showing that the Rh particles supported on Y-ZrO2 were less sintered than those supported on ZrO2 ( Figure 10). Along with the two results presented so far, Table 2 shows the junction energies of the system with three layers of Rh(111) 4 × 4 structures joined to oxygen-deficient Y-ZrO2 and CeO2 slabs. One oxygen between the Y-Zr layers connected to the two Y atoms was  Figure 16 shows the six-layered 3 × 3 structure for the (111) plane of oxygen-terminated Y-ZrO 2 . Each layer contains one Y atom and eight Zr atoms. Because the ionic radius of Y is larger than that of Zr, the a-and b-axis lengths of the unit cell of the structurally relaxed Y-ZrO 2 are slightly longer than those of ZrO 2 . However, the matching between the ab-plane of the surface of the Y-ZrO 2 slab and the Rh 4 × 4 structure remains adequate. Structural relaxation calculations were performed, and the junction energy per Rh atom at the interface was found to be −1.06 eV. The bonding of Rh to Y-ZrO 2 at the interface was thus illustrated to be stronger than the bonding to ZrO 2 . This is consistent with the STEM observation showing that the Rh particles supported on Y-ZrO 2 were less sintered than those supported on ZrO 2 ( Figure 10).  Figure 16 shows the six-layered 3 × 3 structure for the (111) plane of oxygen-terminated Y-ZrO2. Each layer contains one Y atom and eight Zr atoms. Because the ionic radius of Y is larger than that of Zr, the a-and b-axis lengths of the unit cell of the structurally relaxed Y-ZrO2 are slightly longer than those of ZrO2. However, the matching between the ab-plane of the surface of the Y-ZrO2 slab and the Rh 4 × 4 structure remains adequate. Structural relaxation calculations were performed, and the junction energy per Rh atom at the interface was found to be −1.06 eV. The bonding of Rh to Y-ZrO2 at the interface was thus illustrated to be stronger than the bonding to ZrO2. This is consistent with the STEM observation showing that the Rh particles supported on Y-ZrO2 were less sintered than those supported on ZrO2 ( Figure 10). Along with the two results presented so far, Table 2 shows the junction energies of the system with three layers of Rh(111) 4 × 4 structures joined to oxygen-deficient Y-ZrO2 and CeO2 slabs. One oxygen between the Y-Zr layers connected to the two Y atoms was Along with the two results presented so far, Table 2 shows the junction energies of the system with three layers of Rh(111) 4 × 4 structures joined to oxygen-deficient Y-ZrO 2 and CeO 2 slabs. One oxygen between the Y-Zr layers connected to the two Y atoms was chosen as the deficient site. The binding strength with the Rh layer could be ranked as ZrO 2 < oxygen-deficient Y-ZrO 2 < Y-ZrO 2 . This order can be attributed to the difference in valence between Y and Zr. The oxygen at the top surface coupled to Y receives more electrons from Rh than do those coupled to Zr. Hence, the bond between Rh and the terminating oxygen in Y-ZrO 2 is relatively stronger. When an oxygen deficiency occurs near Y, sufficient electrons are supplied from that Y to the top surface oxygen. As a result, the coupling between the top surface oxygen and the Rh layer is weakened. Figure 17 shows the structural relaxation results of the bond between the Rh layer and various support slabs. The binding of oxygen to Rh increases near the Y site in Y-ZrO 2 . However, when an oxygen deficiency is introduced near the Y site, the tight coupling between Rh and oxygen is relaxed. Table 2. Junction energies of the system with three-layered Rh(111) 4 × 4 structures joined to various support slabs. The energy is given per Rh atom at the interface.

Support
Junction Energy (eV) Catalysts 2021, 11, x FOR PEER REVIEW 13 of 17 chosen as the deficient site. The binding strength with the Rh layer could be ranked as ZrO2 < oxygen-deficient Y-ZrO2 < Y-ZrO2. This order can be attributed to the difference in valence between Y and Zr. The oxygen at the top surface coupled to Y receives more electrons from Rh than do those coupled to Zr. Hence, the bond between Rh and the terminating oxygen in Y-ZrO2 is relatively stronger. When an oxygen deficiency occurs near Y, sufficient electrons are supplied from that Y to the top surface oxygen. As a result, the coupling between the top surface oxygen and the Rh layer is weakened. Figure 17 shows the structural relaxation results of the bond between the Rh layer and various support slabs. The binding of oxygen to Rh increases near the Y site in Y-ZrO2. However, when an oxygen deficiency is introduced near the Y site, the tight coupling between Rh and oxygen is relaxed.

Catalyst Preparation
Using Rh nitrate as the precursor, Rh was loaded at 3 wt% on different support materials (ZrO2, Y-ZrO2, CeO2, and Al2O3) by the conventional incipient wetness impregnation method, followed by calcination at 600 °C for 30 min in air. The details of each support material are summarized in Table 3.

Catalyst Preparation
Using Rh nitrate as the precursor, Rh was loaded at 3 wt% on different support materials (ZrO 2 , Y-ZrO 2 , CeO 2 , and Al 2 O 3 ) by the conventional incipient wetness impregnation method, followed by calcination at 600 • C for 30 min in air. The details of each support material are summarized in Table 3. For comparison with the Rh/oxide powder catalyst, an actual TWC was prepared and engine-aged for ex situ STEM observations. The TWC was a Pd/Rh commercial modified double-layer catalyst in which Pd was loaded on Al 2 O 3 in the top layer and Rh was loaded on both Al 2 O 3 and ZrO 2 /CeO 2 materials in the bottom layer. The Pd loading was 1.8 g/L, and the Rh loading was 0.2 g/L. The catalyst was coated on cordierite honeycomb and engine-aged with a lean/rich cycle at temperatures higher than 900 • C for 50 h. The lean/rich cycle was exothermic fuel cuts (F/C) every 20 s followed immediately by 5 s rich spikes (air/fuel ratio λ = 0.83).

In Situ and Ex Situ STEM Observations
In situ and ex situ STEM observations and EDS elemental analysis and mapping were carried out using a JEOL JEM-ARM200F NEOARMex instrument (JEOL Ltd., Tokyo, Japan) with a maximum resolution of 82 pm, together with two different in situ chips. Details of the conditions for STEM and EDS observations are shown in Table 4. The following two in situ chips were supplied by Protochips (Protochips, Morrisville, NC, USA): Fusion TM for vacuum observation at a maximum temperature of 1200 • C, and Atmosphere TM at 1 atm gas flow with a maximum temperature of 1000 • C. During the in situ STEM observation in vacuum (ca. 2.0 × 10 −5 Pa) with Fusion TM , the temperature was rapidly ramped up to 300 • C. Then, EDS analysis was performed, and the results were used to select a site with an adequate amount of Rh. STEM imaging and EDS element mapping were carried out at the selected site. The temperature was ramped to 900 • C, followed by STEM imaging and EDS element mapping. The same procedure was repeated after ramping from 900 to 950 • C, from 950 to 1000 • C, and from 1000 to 1050 • C. To fix the observation point, STEM was observed in real time during the measurement.
During the in situ STEM observation in N 2 gas with Atmosphere TM , pure N 2 initially flowed to the chip at 30 mL/min. The temperature was rapidly ramped up to 300 • C in N 2 , and EDS analysis was performed. The subsequent steps were similar to those described above for the vacuum environment; however, EDS analysis was not performed at temperatures above 300 • C in order to avoid the electron beam effect. Further, upon reaching 950 • C, the gas was changed from pure N 2 to 1% O 2 with N 2 balance. STEM movies were created for each heating period. The purities of both the N 2 and O 2 gas cylinders used in the experiments were over 99.99995%.

DTA-TG Measurement
Differential thermal analysis-thermogravimetry (DTA-TG) was carried out on a Thermo plus EVO2 device (Rigaku corporation, Tokyo, Japan), into which various gases can be flowed. Measurements were performed on Rh/ZrO 2 in three different atmospheres: N 2 , 7%O 2 /N 2 , and 2%H 2 /N 2 . The sample (15 mg) was placed on a platinum pan (5 mm diameter and 5 mm height). The temperature was ramped up to 200 • C at a rate of 30 • C/min, and then to 1100 • C at 10 • C/min.

DFT Calculations
Electronic state calculations based on density functional theory (DFT) were performed using the Vienna Ab-Initio Simulation Package (VASP) (VASP Software GmbH, Vienna, Austria), version 5 [21]. The generalized gradient approximation method of Perdew-Burke-Ernzerhof (GGA-PBE) [22] and the projector augmented wave method [23] were employed. Following single-point structural relaxation calculation, Brillouin zone integration was performed with a 3 × 3 × 1 Monkhorst-Pack mesh [24] to obtain the total energy and electron density distribution. The cut-off energy of the plane wave was set to 400 eV. The convergence condition of energy relaxation in the self-consistent calculation of the wavefunction and the charge density was smaller than 1.0 × 10 −5 eV. The convergence condition of the structural relaxation calculations was assumed to be less than 0.03 eV/Å for all forces. The slab models and electron number densities of the calculated systems were visualized using VESTA(Visualization for Electronic Structural Analysis) [25].

Conclusions
In this study, in situ STEM and EDS observations of Rh on various oxide supports were obtained and then compared with each other, as well as with ex situ observations of an actual TWC catalyst aged by an engine. Both the in situ STEM and EDS techniques achieved Rh particle sintering on ZrO 2 , yttria-doped ZrO 2 , CeO 2 , and Al 2 O 3 support materials. Observation via the two in situ observation techniques showed that the sample was more strongly reduced under vacuum and high temperature than under N 2 or 1% O 2 (balance N 2 ) gas flow conditions. The Rh sintering phenomenon was further examined in real time using Rh/ZrO 2 and Rh/YZrO 2 as examples, by tracking Rh particles in a N 2 atmosphere using in situ TEM. General sintering of the support material (ZrO 2 and Y-ZrO 2 ) began, followed by Rh particle sintering. The Rh particle size was slightly smaller on Rh/Y-ZrO 2 compared to that on Rh/ZrO 2 . These experimental results were supported by DFT calculations. In the calculation results, some Rh atoms were located in surface depressions, and the Rh cluster embedded in the surface stabilized. These predictions are consistent with the images obtained by STEM experiments in vacuum. In addition, strong metal-surface interaction was suggested by the calculated electronic density difference data. The junction energies of the system with three layers of Rh(111) 4 × 4 structures joined to various support materials were also calculated using DFT techniques. The computational Rh-substrate interactions showed remarkable correlation with the in situ TEM observations in an N 2 atmosphere. Considering these results, the thermal durability of the support material and the junction energy obtained by DFT calculations are important for preventing Rh sintering of TWCs under high-temperature conditions such as engine aging. This study is expected to be very useful not only for the development of TWCs but also for the development of specific catalyst analysis methods. Funding: This research received no external funding Data Availability Statement: The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.