1 Introduction

Currently, more than 80 % of the Rh demand is for the three-way catalyst (TWC) because of its high efficiency to convert NO to N2 in the autoexhaust [1–5]. Rh is among the scarcest of precious metals so the development of catalysts with decreased Rh content is needed. Toward this aim, many attempts have involved the use of metal–support interactions to suppress sintering [6–18]. In particular, the interactions at the interface between Rh oxide and oxide support materials have attracted considerable attention. Tanabe et al. reported recently that Nd2O3-doped ZrO2 efficiently stabilizes Rh nanoparticles through Rh–O–Nd interfacial bonding [15]. Similar anchoring mechanisms have been reported for Pt and Pd supported on CeO2 [10–13, 17, 19], which are based on electrostatic interactions at the metal/oxide support interface under an oxidizing atmosphere. However, it should be pointed out that Rh oxide is difficult to reduce to active metallic Rh, as the metal–support bonding is strong. This trend leads to a trade-off between thermal stability and catalytic activity, which should be taken into consideration when designing active and thermostable Rh catalysts with a minimum Rh loading.

In a series of our preceding papers [20–26], another type of the Rh anchoring mechanism on metal phosphates, which stabilize Rh nanoparticles via Rh–O–P interfacial bonding, was reported. The Rh–O–P bonding is preserved under both an oxidizing atmosphere and a reducing atmosphere, whereas conventional metal oxide supports form the interfacial bonding only under oxidizing conditions [22]. In addition, the Rh–O–P linkage enables smooth reduction of Rh oxide to metallic Rh and improves resistance to reoxidation [23, 24]. Because the metallic Rh is more active than Rh oxide for most elementary reactions in the TWC [27–29], this resistance is beneficial for practical applications. The catalytic activity is also affected by the acidic nature of metal phosphates; hydrocarbons such as C3H6 are efficiently converted to aldehyde intermediates, which are highly reactive to NO and O2 adsorbed on Rh [23]. This feature is most obvious for the Rh/ZrP2O7 catalyst, which demonstrated pronounced NO x reduction efficiency under slightly lean conditions [24]. Conversely, the Rh–O–P linkage provides electron-deficient Rh due to electron-withdrawing effects from the acidic phosphate group [23]. This favors the decreasing back-donation from the Rh d-orbitals to the π* molecular orbitals of chemisorbed CO and/or NO, which has a negative impact on the catalytic activity for CO–O2, CO–H2O, and CO–NO reactions. To overcome this negative effect and to promote the desirable characteristics, the acidity of phosphate materials was controlled using metals with a higher basicity [25]. Consequently, LaPO4 was found to be the best material, which achieved both the high thermal stability and catalytic activity for TWC elementary reactions.

Toward further enhancement of the catalytic performance, a novel preparative approach for higher specific surface areas of LaPO4 is necessary. In this study, LaPO4/metal oxide nanocomposite materials were prepared as supports for Rh catalysts expecting a beneficial effect on the thermal stability compared with LaPO4 alone. Because Rh/LaPO4/SiO2 nanocomposites exhibited the highest catalytic activity for simulated NO–CO–C3H6–O2 reactions, the detailed catalytic performance was investigated as functions of Rh loading and thermal aging temperature. The local structure and metal–support interaction of supported Rh catalysts were characterized by transmission electron microscope (TEM), extended X-ray absorption fine structure (EXAFS), and chemisorption technique.

2 Experimental

2.1 Catalyst Preparation

Lanthanum phosphate (LaPO4) was prepared from the corresponding metal nitrates (Wako Pure Chemicals) and H3PO4 (Wako Pure Chemicals Ind. Ltd., 85 %) [25]. A H3PO4 solution (0.05 mol) in deionized water (50 mL) was added dropwise to the nitrate solutions (0.05 mol) of deionized water (50 mL) with vigorous stirring. After dropwise addition of aqueous ammonia (25 %) to adjust the pH to 9, the resulting slurry was hydrothermally treated at 200 °C for 18 h. As-formed white gel was recovered by centrifugation, washed several times with deionized water, and dried in air at 100 °C. The solid product was calcined in air at 900 °C for 5 h. Nanocomposite supports consisting of LaPO4 and metal oxides (γ-Al2O3, SiO2, or ZrO2) were also prepared by the hydrothermal synthesis and post-calcination in the same way. The metal oxides for this purpose were supplied by the Catalysis Society of Japan (JRC-ALO-8, JRC-SIO-9A, and JRC-ZRO-3) and were added to the slurry before the hydrothermal synthesis. The LaPO4 content in each composite solid was in a range of 5–40 mol%. Supported Rh catalysts (0.01–0.4 wt% as Rh metal) were prepared by impregnation of an aqueous solution of Rh(NO3)3 followed by air-drying at 100 °C and air calcination at 600 °C for 3 h. As-prepared catalysts were heated to 900–1100 °C for 25 h in flowing 10 % H2O/air in order to evaluate their thermal stability under an oxidizing atmosphere (hydrothermal aging). The catalysts thus aged at 900 °C were subsequently treated at 200–800 °C for 5 h in 20 % H2/He in order to evaluate their thermal stability under a reducing atmosphere.

2.2 Characterization

Powder X-ray diffraction analysis was performed using monochromatic Cu Kα radiation (30 kV, 20 mA, Multiflex, Rigaku), and the Rh content was determined by energy-dispersive X-ray fluorescence (EDXL300, Rigaku). The FE-TEM micrographs were acquired using an FEI TECNAI F20 transmission electron microscope operating at 200 kV. Brunauer–Emmett–Teller (BET) surface area (S BET) was calculated using N2 adsorption isotherms obtained at −196 °C (Belsorp-mini, Bel Japan). The Rh metal dispersion was determined by pulsed CO chemisorption at 50 °C (Belcat, Bel Japan) after a reduction treatment at 400 °C in 20 % H2/He. The metal dispersion is expressed in terms of the molar ratio of chemisorbed CO per loaded Rh (CO/Rh).

The local structure around Rh was investigated by EXAFS. The Rh loading for this purpose was increased to 2 wt% to ensure the spectral quality. The Rh K-edge EXAFS was recorded on an NW10A station at the Photon Factory Advanced Ring, High Energy Accelerator Research Organization (KEK) using a Si(311) double-crystal monochromator. The spectra were recorded at room temperature in a transmission mode by monitoring incident and transmitted X-rays with 17- and 31-cm-long ionization chambers filled with Ar and Kr, respectively. XAFS data were processed using the REX 2000 program (Rigaku). EXAFS oscillations were extracted by the fitting of a cubic spline function through the post edge region. The k 3-weighted EXAFS oscillation in the 30- to 138-nm−1 region was Fourier transformed. For the curve-fitting analysis, phase shifts and backscattering amplitude functions of the Rh–Rh and Rh–O–Rh shells were extracted from the EXAFS data obtained for Rh foil and Rh2O3. The curve-fitting analysis of the Rh–O–P and Rh–O–Si shells was performed using theoretical parameters.

2.3 Catalyst Performance Testing

Catalytic reactions were performed in a flow reactor at an atmospheric pressure. The as-prepared or thermally aged catalyst (50 mg, 20 mesh) was fixed in a quartz tube (4-mm inside diameter) using quartz wool at both ends of the catalyst bed. The temperature dependence of the catalytic activity was evaluated by heating the catalyst bed (W = 0.05 g) from room temperature to 600 °C at a constant rate of 10 °C min−1 while supplying a simulated exhaust gas mixture containing NO (0.050 %), CO (0.510 %), C3H6 (0.039 %), O2 (0.400 %), and He (balance) at F = 100 cm3 min−1 (W/F = 5.0 × 10−4 g min cm−3). This gas composition corresponds to the stoichiometric air-to-fuel ratio (A/F = 14.6). The gas feed and the catalyst bed were heated using an infrared image furnace equipped with two ellipsoidal reflectors (RHL-E25P, Ulvac Riko). The effluent gas was analyzed using a Pfeiffer GSD30101 mass spectrometer and Horiba VA3000 Non-dispersive infrared (NDIR) gas analyzers. Catalytic activities for eight elementary reactions were also evaluated in a similar manner [see Table S1 in Electronic supplementary material (ESM)].

3 Results and Discussion

3.1 Rh Supported on Nanocomposite LaPO4 Support Materials

To demonstrate a negligible influence of mass transfer and heat transfer contributions in our catalytic microreactor, the change in the conversion was measured as a function of gas flow rate (and thus the W/F value). When the conversion was below 50 %, the reaction rate was proportional to the W/F value, suggesting that the reaction was controlled by chemical kinetics. A Madon–Boudart criterion was applied to evaluate the effect of internal mass transport using one of our reference catalysts (see Fig. 3) with 0.4 and 0.2 wt% Rh loadings, which showed similar metal dispersions. The similarity of the TOF at each of two different temperatures indicates that the measured rates were not influenced by mass transport limitations. When a simulated stoichiometric NO–CO–C3H6–O2 reaction was performed in a light-off mode, the conversion increased very steeply to 100 % within a narrow temperature range (see Fig. S1 in ESM). In such a situation, it is difficult to compare the reaction rate at the same temperature under kinetically limited conditions. Therefore, the activity is simply expressed in terms of the light-off temperature (T 50) at which the conversion of NO x , CO, and C3H6 reached 50 %. The T 50 value is a reasonable parameter to compare the relative activity of the different catalysts. Table 1 summarizes the S BET, CO/Rh, and T 50 values for each catalyst before and after the hydrothermal aging at 900 °C for 25 h in 10 % H2O/air. The hydrothermal aging of Rh/LaPO4 decreased the metal dispersion (CO/Rh) from 22 to 3.7 % and thus increased the T 50 values. Because S BET of the catalyst decreased from 44.5 to 33.3 m2 g−1 during the hydrothermal aging, sintering of supports is a plausible deactivation mechanism. To improve the thermal stability of Rh/LaPO4, microstructural modification of the support material was investigated. Three composite materials containing 20 mol% LaPO4 and 80 mol% γ-Al2O3 (S BET = 150 m2 g−1), SiO2 (310 m2 g−1), or ZrO2 (38 m2 g−1), were prepared hydrothermally. The S BET values of all these composites were lower than expected because of oxide particles growth during the hydrothermal synthesis, while improved CO/Rh values greater than 5 % were obtained after the hydrothermal aging. The as-prepared Rh catalysts using these composite supports exhibited T 50 values comparable to those for Rh/LaPO4. After the hydrothermal aging, however, only the SiO2 composite could improve the activity, while the other composites with Al2O3 and ZrO2 exhibited more significant deactivation than LaPO4 alone.

Table 1 Surface area (S BET), metal dispersion (CO/Rh), and catalytic activity (T 50) of 0.4 wt% Rh supported on various supports
Full size table

The LaPO4/SiO2 composites with different mixing ratios (5, 10, 20, 30, and 40 mol% LaPO4) were studied. Figure 1 plots S BET, CO/Rh, and T 50 as a function of the content of LaPO4 for the catalysts thermally aged at 900 °C. As can be seen in Table 1 and the catalytic light-off curves (see Fig. S1 in ESM), the light-off of NO, CO, and C3H6 occurred at similar temperatures. The catalytic activity can therefore be expressed in terms of the T 50 for NO. The S BET was nearly constant at approximately 40 m2 g−1, while the CO/Rh value gave rise to the maximum at 20 mol% LaPO4, at which point the lowest T 50 was achieved. To elucidate this behavior, the line-broadening analysis of XRD data (see Fig. S2 in ESM for details) was performed to determine the LaPO4 crystallite size of these composites. Because the calculated size for 20 mol% LaPO4 (20 nm) was smaller than that of LaPO4 alone (24 nm), the SiO2 composite was effective in suppressing the grain growth of LaPO4 and thus Rh nanoparticles supported on LaPO4.

Fig. 1
figure 1

BET surface area (S BET), Rh metal dispersion (CO/Rh) and light-off temperature (T 50 for NO) of 0.4 wt% Rh catalysts supported on composites with different LaPO4 content after hydrothermal aging at 900 °C

Full size image

Figure 2 shows the TEM photographs of Rh/LaPO4/SiO2 before and after the hydrothermal aging. The as-prepared catalyst was composed of columnar LaPO4 crystallites of approximately 20 nm in thickness and 100 nm in length, which were dispersed in the macroporous SiO2 matrix with an amorphous morphology. After the hydrothermal aging at 900 °C, similar microstructure could be preserved and no significant growth of particles was observed. The sintering of support particles will be initiated by neck growth at around particle boundaries present in powder aggregates, and the rate of sintering should be determined by mass transport in the solid phase [30]. Therefore, dispersing LaPO4 crystallites within macroporous SiO2 matrix should be a promising approach to avoid direct contact between LaPO4 crystallites and thus suppress sintering. As shown in Table 1, SiO2 alone could also retain CO/Rh as high as 5 % after the hydrothermal aging at 900 °C, because of the highest S BET value (154 m2 g−1) among the catalysts studied in this study. Nevertheless, the T 50 values for Rh/SiO2 were much higher than those for Rh/LaPO4/SiO2, suggesting that improved dispersion of LaPO4 contributed significantly to the thermal stabilization of Rh.

Fig. 2
figure 2

TEM images of (a) as-prepared and (b) thermally aged (900 °C for 25 h in 10 % H2O/air) Rh/LaPO4/SiO2 (20 mol% LaPO4)

Full size image

3.2 Catalytic Properties of Rh/LaPO4/SiO2

Further studies were focused on the Rh/LaPO4/SiO2 catalyst to evaluate its catalytic performance, and compare it with that of Rh/LaPO4 and Rh/SiO2. Here, Rh/ZrO2 was used as another reference catalyst, because ZrO2 is widely used as a support for Rh in the commercial TWC [15]. The T 50 values (NO) for the catalysts thermally aged at 900 °C are plotted as function of Rh loading in Fig. 3. Two oxide-supported catalysts, Rh/SiO2 and Rh/ZrO2, exhibited steep increases of T 50 values as the loading decreased below 0.2 wt%, resulting in a significant activity drop, which is a common feature of Rh catalysts supported on oxides as was pointed out previously [20, 25], By contrast, LaPO4 and LaPO4/SiO2 composite could reduce the Rh loading to as low as 0.05 wt% without significant increase of T 50, which is the most prominent feature of metal phosphates as support materials for Rh catalysts in contrast to the oxide supports [20, 22, 25]. Among metal phosphates, LaPO4 particularly preserved the lowest T 50 value in a range of 0.05–0.4 wt% Rh loading [25]. A further improvement could be achieved by the LaPO4/SiO2 composite, which was entirely different from Rh/SiO2.

Fig. 3
figure 3

Catalytic activity (T 50 for NO) versus Rh loading for different supported catalysts after hydrothermal aging at 900 °C for 25 h in 10 % H2O/air

Full size image

Figure 4 plots the CO/Rh value as a function of Rh loading for LaPO4/SiO2 and ZrO2 after the hydrothermal aging at 900 °C. These supports showed contrasting behaviors; the value tends to increase with decreasing Rh loading for LaPO4/SiO2, which is in complete contrast to ZrO2. Decreasing dispersion at lower Rh loadings for Rh/ZrO2 may be associated with encapsulation of a part of the Rh oxide due to interactions at the RhOx–ZrO2 interface during the hydrothermal aging, as this phenomenon prevents the reduction of Rh oxide to metallic Rh or blocks the access of CO molecules. Contrary, Rh nanoparticles can be stabilized on the surface of LaPO4 via Rh–O–P interfacial bonding as in the case for Rh/AlPO4 and Rh/ZrP2O7 [22–25], which seems to be useful for higher dispersion of smaller amounts of Rh. This is in accordance with the retention of the higher catalytic activity at lower Rh loadings as shown in Fig. 3.

Fig. 4
figure 4

Rh metal dispersion (CO/Rh) versus Rh loading for Rh/LaPO4/SiO2 and Rh/ZrO2 catalysts after hydrothermal aging at 900 °C for 25 h in 10 % H2O/air

Full size image

Figure 5 plots the T 50 values for NO in the NO–CO–C3H6–O2 reaction after the hydrothermal aging in 10 % H2O/air at elevated temperatures. Although Rh/LaPO4 and Rh/LaPO4/SiO2 showed similar dependences on temperatures up to 900 °C, further increase of aging temperature led to a steeper rise of T 50 and thus a more pronounced deactivation for Rh/LaPO4 compared to Rh/LaPO4/SiO2. The enhanced thermal stability was also obvious when compared to Rh/SiO2, which exhibited the largest deactivation at 900 °C because SiO2 is well-known as one of the most weakly interacting support materials for Rh [31]. The catalytic activity was also evaluated after the hydrothermal aging at 900 °C in 10 % H2O/air and subsequent reduction treatment in 20 % H2/He at elevated temperatures to evaluate thermal stability under a reducing atmosphere (Fig. 6). The reduction treatment at 200 °C significantly lowered the T 50 value, because metallic Rh was formed which is more active than Rh oxide (Rh2O3) for NO x reduction. On the contrary, the T 50 increased with further elevation of reducing temperatures because of the sintering of metallic Rh. Again, the resulting deactivation was much more significant for Rh/SiO2 compared to that of LaPO4 and Rh/LaPO4/SiO2, because of weak interactions between metallic Rh and SiO2. Consistent with that, the metal dispersion (CO/Rh) for Rh/SiO2 decreased significantly from 4.0 % (reduction at 400 °C) to 1.3 % (800 °C), suggesting considerable sintering of Rh on SiO2. It is also implied that the formation of Rh silicide during the reduction pretreatment of Rh/SiO2 at high temperatures (>500 °C) may influence the catalytic activity [32, 33]. In contrast, the Rh/LaPO4/SiO2 catalyst achieved the highest activity after aging over a wide temperature range under both oxidizing and reducing atmospheres. The efficient thermal stabilization requires Rh particles that are mainly located on the surface of LaPO4 but not on the surface of SiO2.

Fig. 5
figure 5

Catalytic activity (T 50 for NO) of supported Rh catalysts (0.4 wt% Rh loading) after hydrothermal aging at each temperature for 25 h in 10 % H2O/air

Full size image
Fig. 6
figure 6

Catalytic activity (T 50 for NO) of supported Rh catalysts (0.4 wt% Rh loading) after hydrothermal aging at 900 °C for 25 h in 10 % H2O/air and subsequent reduction at each temperature for 5 h in 20 % H2/He

Full size image

3.3 Interaction between Rh and Support

Figure 7 compares the Fourier transforms of Rh K-edge EXAFS for Rh/LaPO4/SiO2 with those for Rh/LaPO4, Rh/SiO2, and Rh2O3. Because these data are shown without phase shift corrections, the observed peaks are shifted to shorter values with respect to the true atomic distances listed in Table 2. All of these samples showed an intense peak at approximately 2.0 Å, assignable to an Rh–O shell of Rh oxide. However, their second shell peaks were different from those of Rh2O3, suggesting interactions at the metal–support interface. As reported previously [25], Rh/LaPO4 showed the second shell peaks containing a contribution from the Rh–O–P interfacial bonding. The similar Rh–O–P shell was observed for Rh/LaPO4/SiO2, although the coordination numbers (CN) was smaller (0.64) than that of Rh/LaPO4 (1.1). Unlike these samples containing LaPO4, Rh/SiO2 exhibited another peak component assignable to Rh–O–Si, which formed at the interface between Rh oxide and SiO2. The outstanding thermal stability of Rh/LaPO4/SiO2 is closely associated with the Rh anchoring via Rh–O–P interfacial bonding as in Rh catalysts supported on LaPO4, AlPO4, and ZrP2O7 in our previous studies [22–25].

Fig. 7
figure 7

Fourier-transformed Rh K-edge EXAFS for as-prepared Rh/LaPO4, Rh/LaPO4/SiO2, Rh/SiO2, and Rh2O3

Full size image
Table 2 The fitting results obtained from Rh K-edge EXAFS analysis
Full size table

3.4 Activity Distribution for TWC Elementary Reactions

A variety of elementary reactions are involved in TWC, and the catalytic activity distribution for each reaction provides important information. Figure 8 plots the T 50 values for the eight elementary reactions over Rh catalysts supported on metal phosphates and reference oxides. As reported previously [23, 25], the activity distribution is strongly dependent on the basicity of the metal species on the phosphate supports. Compared to Rh/ZrO2, Rh/AlPO4 with a very low basicity was less active for the CO–O2, CO–H2O, and C3H6–H2O reactions, while its activities for the C3H6–O2 and NO–C3H6–O2 reactions were higher. This trend is closely associated with the Lewis acidity of AlPO4 and a resulting electron withdrawing effect, which leads to electron deficient Rh species and thus decreased back-donation from Rh d-orbitals to the antibonding π* orbitals of CO and NO [23]. Conversely, Rh/LaPO4 showed much higher activities for CO–H2O, CO–O2, NO–CO, and C3H6–H2O reactions in accordance with the higher basicity [25]. The similar activity distribution could be observed for the Rh supported on LaPO4/SiO2 nanocomposite. Another noteworthy feature of phosphate supports was the higher activities for C3H6 (NO–C3H6–O2 and C3H6–O2) than those of Rh/ZrO2. This trend is a common feature of phosphate supports and is correlated to the oxidative adsorption of C3H6 as aldehyde species on the phosphate surface, which are highly reactive toward O2 as well as NO [23–25]. The activities of the Rh catalysts supported on LaPO4 and LaPO4/SiO2 for these reactions were comparable. Although Rh/SiO2 also exhibited the similar activity distribution, the Rh loading dependence of activity (Fig. 3), the thermal behavior under reducing atmosphere (Fig. 6), and the EXAFS results (Fig. 7) suggest Rh anchored on LaPO4 plays a key role in the high catalytic performance for these elementary reactions.

Fig. 8
figure 8

Catalytic activity (T 50) for various TWC elementary reactions. T 50 values for underlined gases are plotted

Full size image

4 Conclusion

LaPO4/metal oxide nanocomposites were prepared as support materials of Rh catalysts for the TWC applications. Among SiO2, Al2O3, and ZrO2, SiO2 afforded a thermally stable nanocomposite, which achieved the highest metal dispersion and catalytic activity for a simulated NO–CO–C3H6–O2 reaction after the hydrothermal aging. In the LaPO4/SiO2 nanocomposite, columnar crystallites of LaPO4 provides a Rh anchoring effect via Rh-O–P bonding were highly dispersed in the macroporous SiO2 matrix. It is suggested that this microstructure decrease the direct contact between the LaPO4 crystallites and thus their sintering. Consequently, the Rh/LaPO4/SiO2 catalysts exhibited the higher thermal stability under both oxidizing and reducing atmosphere, compared to that of Rh catalysts supported on LaPO4 and SiO2 alone, and this could mostly reduce the threshold Rh loading without loss of catalytic activity.