1 Introduction

Since 1980s, three-way catalysts (TWC) are efficiently used as exhaust emission abatement systems to purify the gasoline engine exhaust gases providing simultaneous oxidation of CO and unburnt hydrocarbons, and reduction of nitrogen oxides [1,2,3,4,5,6]. Modern TWC are mostly represented by a complex composition of various oxide supports (Al2O3, ZrO2, La2O3, CeO2, etc.) with deposited precious metals (Pd, Pt, Rh) [2, 4, 6,7,8,9,10,11,12,13,14,15,16]. Platinum and palladium are known to exhibit superior activity in oxidation reactions [17,18,19,20,21,22,23,24], while rhodium responsibility is NOx reduction [25]. Being supported on the oxide carriers in the form of disperse nanoparticles, small clusters or even single atoms, all these metals interact with the surface of the support, and the strength of such interaction depends on the surface and bulk properties of the exact oxide [22, 26,27,28,29,30,31,32,33,34]. For instance, palladium deposited on pure γ-Al2O3 and ZrO2 supports was reported to be stabilized on electron-donor sites on the surface in the form of small clusters, when metal loading does not exceed 0.5–0.8 wt% [35,36,37]. At larger Pd loadings, surface migration of palladium species occurs, thus resulting in agglomeration process and appearance of quite large Pd particles, particularly after aging at elevated temperatures. Another type of strong metal–support interaction was observed for Rh-containing catalysts. Thus, in the case of γ-Al2O3, rhodium ions Rh3+ being formed under the reaction conditions diffuse into the bulk of the support, where initiate local phase transformation and formation of corundum phase [38]. An encapsulation of Rh3+ ions within this phase makes such process of deactivation irreversible. In the case of CeO2-based supports, rhodium ions diffused onto the bulk, on the one hand, cause appearance of areas with increased compaction of the support, but, on the other hand, facilitate an improvement of the oxygen storage capacity [33]. Moreover, catalysts deactivated due to strong rhodium-ceria interaction can be reactivated by high-temperature oxidative treatment [39].

In order to prevent the undesirable metal–support interactions, like in the case of Rh/γ-Al2O3, and to stabilize the catalyst towards high-temperature deactivation processes, a concept of bimetallic Pd–Rh systems with enhanced metal–metal interaction was proposed [34, 40,41,42,43,44,45,46,47]. Both strong metal–support interaction between palladium and donor sites of alumina, and strong metal–metal interaction between palladium and rhodium were shown to have synergetic effect maintaining the catalytic activity on the appropriate level and significantly improving the thermal stability. Ratio of metals within the bimetallic Pd–Rh nanoparticles affects predictably the mechanism of the catalyzed reactions and overall catalytic performance. In general, the main reactions taking place during the overall process include oxidation of CO by oxygen, partial and complete oxidation of hydrocarbons by oxygen and nitrogen oxide, steam reforming of hydrocarbons and water–gas shift reaction [48, 49]. As we have reported recently [50], contribution of the reactions involving NO increases along with an increase in Rh content in the alloy. Therefore, the thermal stability of the bimetallic species and the surface concentration of rhodium are the key factors determining the catalytic performance of the samples.

It is worth noting that in most cases, pure model supports were used to study the effect of metal–support interaction on the catalytic behavior. On the other hand, such materials are of rare application due to their non-satisfactory exploitation properties. For instance, pure γ-Al2O3 is known to undergo phase transformations within a temperature range of 900–1100 °C, while this range is quite important for usage of TWC, and the support is required to provide appropriate resistance towards aging at such temperatures. From this point of view, thermally stabilized modifications of γ-Al2O3 are more attractive for manufacturers. Among them, alumina doped with lanthanum oxide is the most usually applied one [51,52,53,54,55]. It should be mentioned that introduction of dopant into alumina lattice also changes the mechanism of metal–support interaction. In particular, rhodium ions still diffuse into the bulk, but the preferable places of their location are quite different [56, 57]. In this case, rhodium is localized at nearest coordination environment of the La cations. Presence of La3+ ions within the alumina lattice creates new diffusion pathways for rhodium ions, thus accelerating the rate of their diffusion. Thereby, the high-temperature behavior of the bimetallic Pd–Rh nanoparticles supported on the La-doped alumina is of special interest.

In the present work, the most attention was paid to the investigation of the bimetallic Pd–Rh/(γ­Al2O3 + 4 wt% La2O3) system in comparison with monometallic Pd-only and Rh-only reference samples, as well as with bimetallic catalyst of the same composition obtained by mechanical mixing of monometallic samples with doubled metal loading (no metal–metal interaction). In order to elucidate the effect of metal–support interactions, the support was calcined at 800 °C prior the deposition of precious metals. Commercial bimetallic TWC containing the same amount of metals was used as a general reference sample. The support and the catalysts were characterized by a number of physicochemical methods.

2 Experimental

2.1 Synthesis of the Samples

Commercial La2O3-doped γ-Al2O3 (TM 100/150 L4, Sasol) containing 4 wt% of lanthanum oxide was taken as a support. The initial support was labeled as L4. Commercial solutions of rhodium and palladium nitrates (JSC «Krastsvetmet», Krasnoyarsk, Russia) were used as salts-precursors.

The samples of alumina-supported bimetallic catalysts with an atomic Pd:Rh ratio of 3:2 were prepared by an incipient wetness impregnation of the L4 support with a joint aqueous solution of rhodium and palladium nitrates. A portion of initially dry powder support was placed into beaker and then preliminary heated solution of the precursor was added dropwise at a constant stirring. The resulting suspension was transferred to a Petri dish and dried in air at room temperature for 12 h. Then, the sample was dried in a furnace at 105 °C for 6 h, heated in air at 550 °C with a temperature ramping rate of 10 °C/min, and maintained at this temperature for 1 h. After cooling, the calcined catalysts were homogenized via grinding in an agate mortar. The total metal loading was 0.35 ± 0.03 wt%. The reference monometallic samples were prepared similarly using single solutions of the corresponding nitrates. The loading of Pd and Rh in the monometallic samples was 0.21 and 0.14 wt%, accordingly. These values correspond to their content in the bimetallic samples. The bimetallic reference sample was obtained by mechanical mixing of specially prepared monometallic samples containing doubled amount of each metal (0.42 and 0.28 wt% of Pd and Rh, respectively). These monometallic samples were taken in equal amounts and mixed in an agate mortar with intensive communication. Additional bimetallic Pd–Rh/L4* sample was prepared by an impregnation of the L4 support preliminary calcined at 800 °C for 6 h (labeled as L4*) with a joint aqueous solution of rhodium and palladium nitrates followed by all the similar drying and calcination procedures.

Commercial reference sample (CRS) was provided by Ecoalliance LLC (Russia). The powder sample was obtained by stripping the catalytic coating from the honeycomb catalyst and grinding in an agate mortar. Total loading of Pd and Rh in the sample was also 0.35 wt%.

In order to prepare the monolith samples, each powder composition was mixed with water, and subjected to grinding in a beaded mill. The average particle size in the final suspension was about 6–7 microns (D90). The suspension contained 30–35 wt% of solids. Then the suspension was supported on a honeycomb cordierite monolithic substrate RD 4 × 4'' (101.6 mm in diameter and in length, 63 cells/cm2) by an aspiration procedure. The loading of the active composition in the monolith sample was 60 g/dm3. The monolith was dried at 120 °C and calcined at 550 °C for 1 h.

2.2 Characterization of the Samples

The textural parameters of the L4 support treated at different conditions (specific surface area, SSA; pore volume, Vp; average pore size, Dav) were determined by a low-temperature nitrogen adsorption/desorption method. Nitrogen adsorption isotherms were obtained at 77 K using an ASAP-2400 automated instrument (Micromeritics, USA). Before the measurements, the samples were kept at the degassing station in a nitrogen flow at 200 °C for 1 h and then cooled down during 0.5 h under the same conditions.

Powder XRD analysis of the L4 support treated at different temperatures was performed at room temperature on a Shimadzu XRD-7000 diffractometer (CuKα radiation, nickel filter in reflected beam, scintillation detector with amplitude discrimination). Data were collected step-by-step in the range of angles 2θ = 15–70° with a step of 0.1°.

Electron paramagnetic resonance (EPR) spectroscopy was performed using an ERS-221 EPR spectrometer operating in the X-band (ν = 9.3 GHz). EPR spectra were acquired at 20 dB attenuation with typical microwave power 3 mW. Radical anions of aromatic nitro compounds arising from their adsorption on the surface of oxide catalysts were used as spin probes for the study of surface electron-donor sites as described elsewhere [34, 35, 44]. A catalyst sample (~ 20 mg) was loaded in a quartz ampoule, heated at 170 °C for 12 h, and then cooled down to room temperature. 2 × 10−2 M 1,3,5-trinitrobenzene (TNB) solution in toluene was added to launch the formation of the anion radicals. The temperature was then maintained at 80 °C for 12 h for further acceleration of the radical anion formation to their maximum, which corresponds to the concentration of the surface electron-donor sites. The EPR spectra of generated radical anions were recorded at room temperature. The concentrations of the paramagnetic species were determined by a numerical double integration with baseline compensation.

Transmission electron microscopy (TEM) studies were performed using a JEOL JEM-2010 electron microscope (lattice plane resolution 0.14 nm at accelerating voltage of 200 kV). The samples for the TEM study were prepared on perforated carbon film mounted on a copper grid.

Ethane hydrogenolysis reaction was used as a testing reaction to characterize the surface concentration of supported precious metals in reduced state. As it was reported earlier [35, 38], the rate of this reaction linearly depends on the specific surface area of the supported metals. Each sample (100 mg, 0.25–0.5 mm fraction) was placed inside the quartz flow-through reactor. At the beginning of the experiment, a H2/He mixture was passed through the reactor for some time until the system reached a steady state. Then, ethane was added to the feed for a short period of time (3 min) and a probe was taken for chromatographic analysis. During next 10 min the sample was purged by the H2/He mixture flow in order to regenerate the initial steady state. This procedure was repeated for 5 times at each temperature within the studied temperature range with a step of 25 °C.

Diffuse reflectance UV–Vis spectra were recorded between 200 and 800 nm using a UV–Vis spectrometer UV–VIS 2501 PC (Shimadzu) with IRS-250A diffusion reflection attachment. The UV–Vis spectra were transformed into the Kubelka–Munk function F(R) calculated as F(R) = (1 −­ R)2/2R, where R is the experimentally measured reflectivity coefficient of the samples [58].

X-ray photoelectron spectra (XPS) were recorded on a SPECS spectrometer (Germany) using MgKα radiation (hν = 1253.6 eV). Initially, survey spectra of the samples were recorded in a range of binding energies (BE) from 0 to 1000 eV. Then, each spectrum was analyzed and the regions corresponding to the elements presented on the surface were chosen for more precise research. The binding energies were determined with respect to the Al2p line in alumina, which has the energy value equal to 74.5 eV in accordance with the literature [59, 60]. Atomic concentrations of elements were defined relatively to the aluminum concentration with regard to atomic sensitivity factors reported elsewhere [60].

2.3 Testing the Catalytic Activity and Thermal Stability

The catalytic performance and thermal stability of the powder catalysts were studied using a prompt thermal aging (PTA) procedure as described elsewhere [34, 44, 47, 58, 61]. CO conversion was used as a chemical response to follow the catalytic behavior and state of active components during the procedure. The sample (300 mg) was loaded into quartz flow reactor, and then the reaction mixture contained 0.15 vol% CO, 14.0 vol% O2 and N2 as balance was passed through the reactor. The flow rate was 334 ml/min. A temperature ramping rate was 10 °C/min. The samples underwent two light-off cycles at 50–320 °C, two cycles at 50–600 °C, two cycles at 50–800 °C, and finally one cycle at 50–400 °C. Changes in the CO concentration in the outlet flow were registered by a gas analyzer ULTRAMAT 6 (Siemens). The temperature of 50% CO conversion was used as a criterion to compare the stability of the catalysts.

Light-off performance studies (temperature of 50% CO, CHx, and NOx conversion) as well as perturbation tests (maximum conversion in a pulse regime at 400 °C) for the monolith samples were carried out using a Horiba CTSJ-2003.12 analytical setup. The following reaction mixture was passed through the monolith catalyst: 1.15 vol% O2; 1.60 vol% CO; 1000 ppm NOx; 250 ppm C3H6; 250 ppm C3H8; 0.4 vol% H2; 9.35 vol% CO2; 10 vol% H2O; N2 as a balance. The gas hourly space velocity (GHSV) was 70000 h1. The sample was heated up in a range of 80–400 °C with ramping rate of 20 °C/min, and was then kept at 400 °C for 3 min.

The operation window of the monolith samples was measured using the same gas mixture as in the previous test (with the exception of O2-oscillation) by scanning Lambda (λ) from 0.97 to 1.01 at temperature of 500 °C. The borders of the operation window were limited by 80% conversion of at least one of three types of contaminants (CO, total hydrocarbons, NOx).

3 Results and Discussions

Since the thermal stability of the support is very important in general terms of the catalytic performance of the TWC and in particular to study the metal–support interactions, the initial commercial L4 support was thermally treated at different conditions. Calcination procedure was performed in a muffle furnace in an air atmosphere, while for aging procedure the sample was placed into flow-through reactor and heated up to the corresponding temperature in a flow of air containing 10 vol% of water vapors. Finally, the textural characteristics of the treated samples of the support were examined. The obtained values of specific surface area (SSA), pore volume (Vp), and average pore diameter (Dav) are summarized in Table 1. As seen from the table, each thermal procedure affects the parameters decreasing the SSA and Vp values, and increasing Dav. More noticeable changes are observed after treating at temperature of 1000 °C and above. Duration of the procedure also has an effect. In average, SSA drops down in 1.5 times, and Dav increases in 1.4 times.

The next important factor is a phase composition. Initially, the L4 support is represented by the γ-Al2O3 phase (space group Fd3̅ m). The corresponding XRD pattern (Fig. 1) demonstrates highly widened diffraction reflexes, and, along with it, the widening of the reflexes assigned to different families of crystalline planes differs strongly. Such behavior is quite typical for pure γ-Al2O3 support [46]. The XRD patterns for the L4 support calcined at 600 and 800 °C for 1 h are practically coincide with that for the initial sample. The only changes were observed for the sample calcined at 1000 °C. As seen, gamma-phase is particularly transformed into delta-phase (space group P4̅m2, PDF card #46-1131) labeled by diamond signs.

Table 1 Textural characteristics of the L4 support treated at different conditions
Full size table
Fig. 1
figure 1

XRD patterns of the initial L4 support and after calcination at 600, 800 and 1000 °C for 1 h. Reference pattern for γ-Al2O3 (JCPDS card No. 29-0063) is presented for comparison

Full size image

The spin probe method based on EPR spectroscopy was recently shown to be informative for characterization of electron-donor sites on the surface of alumina support [35, 46]. Corresponding EPR spectra for the initial and calcined L4 support are presented in Fig. 2. The concentration of radical anions calculated form these spectra are summarized in Table 2. Calcination of the support at 600 °C does not change the situation noticeably. The concentration of donor sites slightly increases, thus indicating the removal of hydroxyl groups initially adsorbed on the surface of the support. Calcination at 1000 °C affects the concentration of donor sites more significantly by decreasing this parameter in 2.6 times. Taking into account the correlation reported earlier [35], such decrease in the concentration of donor sites should diminish relatively the amount of palladium species stabilized on the surface.

Fig. 2
figure 2

EPR spectra for the initial L4 support and after calcination at 600 and 1000 °C for 1 h

Full size image
Table 2 Concentration of radical anions for the initial L4 support and after calcination at 600 and 1000 °C for 1 h
Full size table

Samples of the catalysts with deposited precious metals were characterized by a testing reaction of ethane hydrogenolysis. As it was numerously reported by Sinfelt [62,63,64], kinetics of this reaction strongly depends on the amount of metal atoms in reduced state located on the surface and accessible for the reagents. The characterization technique based on this reaction allows studying the surface of the catalysts with supported metals, measuring of their surface concentration, and estimating the dispersity [35, 36, 38]. Temperature dependences of ethane hydrogenolysis over the studied samples are shown in Fig. 3. Rhodium is much more active in the testing reaction than palladium, and experimental curves corresponding to Rh-containing catalysts lie at significantly lower temperatures if compare with Pd-only sample. Curves for monometallic Rh/L4 and bimetallic Pd + Rh/L4 (obtained by mechanical mixing of separately deposited parts) samples are practically coincide, thus confirming absence of metal–metal interaction in the case of Pd + Rh/L4 sample. Curve for bimetallic Pd–Rh/L4 catalyst (obtained by joint impregnation) is shifted towards higher temperatures due to averaging of the catalytic properties in the bimetallic alloy [44]. Observed descent in conversion values at temperatures above 400 °C is caused by near-surface interactions of metal species with the support, which are reversible. It is important to note that the impregnation of L4* support preliminary calcined at 800 °C with joint solution of palladium and rhodium nitrates does not lead to the formation of Pd–Rh alloy. Catalytic behavior of this sample in the ethane hydrogenolysis reaction evidently illustrates that rhodium is not interacted with palladium. Moreover, the largest amount of metallic Rh on the surface was detected in this case.

Fig. 3
figure 3

Temperature dependences of ethane hydrogenolysis for bimetallic samples Pd–Rh/L4 and Pd–Rh/L4*, and reference samples Pd/L4, Rh/L4 and Pd + Rh/L4. Note that the L4* support was calcined at 800 °C for 6 h on air before impregnation with salts-precursors

Full size image

Catalytic performance of the studied samples in the target reactions is quite different. The results of PTA tests are presented in Fig. 4 and Table 3. Monometallic Pd/L4 catalyst shows high initial activity (Fig. 4a). After the first catalytic cycle, the T50 value increases by 33 °C due to changes of the catalyst’s surface state under the reaction conditions, which are mostly caused by dehydroxylation of the surface [37, 65]. Heating of the sample up to 600 °C leads to its deactivation reflected in a further shift of the light-off curves. Then, the sample demonstrates stable behavior, even after aging at 800 °C. Monometallic Rh/L4 sample is less active in comparison with Pd-only sample (Fig. 4b). Difference in the corresponding T50 values reaches 150 °C (Table 3). Contrary to Pd-only sample, catalytic activity of Rh/L4 sample in first five cycles is completely reproducible. Aging at 800 °C deactivates the sample, and the T50 values are increased by ~ 15 °C. Bimetallic reference sample Pd + Rh/L4 prepared by mechanical mixing exhibits all three deactivation steps (Fig. 4d): low-temperature dehydroxylation, and high-temperature deactivation after aging at 600 and 800 °C. Thermal stability of bimetallic Pd–Rh/L4 sample is similar to that for Pd-only sample, but catalytic activity is noticeably higher. The T50 value in the seventh cycle is 206 °C, which is better than for all other studied catalysts. It should be noted that, in terms of catalytic performance, Pd–Rh/L4 sample is very close to CRS (Table 3). Unusual catalytic behavior was shown by Pd­Rh/L4* sample (Fig. 4e): difference between all catalytic cycles is minimal, but activity (the T50 value) is the lowest among all the Pd-containing catalysts. As it follows from results of the ethane hydrogenolysis testing reaction, in this case, no alloy formation has occurred. It was recently reported that both palladium and rhodium prefer the same sites on the surface of the alumina support to interact with [35, 45]. These sites were found to be electron-donor ones. The strength of metal–support interaction and, therefore, an efficiency of stabilization of metals on the surface are determined by the amount of these sites. Calcination of the L4 support at elevated temperatures even for a short period of time, in its turn, leads to significant decrease in the concentration of donor sites (Table 2). Thereby, the conditions required for proper localization of both metals near the same sites, which is expected to be crucial for alloy formation, are not being realized. Finally, the tendency of appearance of more coarse particles predominantly consisting of single metals is prevailing. These particles are less susceptible to changes during the catalytic cycles, thus providing sustainable catalytic performance. Unfortunately, their activity is discontented.

Fig. 4
figure 4

Light-off curves of CO conversion at PTA conditions for the studied samples Pd/L4 (a), Rh/L4 (b), Pd–Rh/L4 (c), Pd + Rh/L4 (d), Pd–Rh/L4* (e), and commercial reference sample CRS (f)

Full size image
Table 3 The T50 values in seven consecutive cycles at PTA conditions for all the tested samples
Full size table
Fig. 5
figure 5

Diffuse reflectance UV–Vis spectra for the samples Pd/L4, Pd–Rh/L4, and Pd + Rh/L4 calcined at 600 °C (a) and 1000 °C (b); diffuse reflectance UV–Vis spectra plotted in the coordinates of (F(R)·E)2 versus E for the same samples calcined at 1000 °C (c)

Full size image

In order to study the evolution of the palladium species state during the high-temperature treatment, the Pd-containing samples calcined at 600, 800 and 1000 °C were characterized by diffuse reflectance UV–Vis spectroscopy. Calcination of the samples was performed in a muffle furnace for 6 h, which corresponds to more drastic conditions required for realization of such processes as surface migration of palladium and bulk diffusion of rhodium. As it was recently reported [58], the spectra for the Pd/Al2O3 samples recorded under atmospheric conditions give overestimated values of Eg (band gap) for oxide clusters and small PdO particles, when calculated in accordance with the Tauc approach. To avoid it, the initial state of the catalysts was qualitatively analyzed comparing the shape of the d–d transition bands with normalization Gauss function with a characteristic value of FWHM = 0.71 eV that corresponds to the d–d transition band for isolated Pd2+ ions. The spectra shown in Fig. 5a were normalized in amplitude to 1.0 in the area of the d–d transition for Pd2+. As seen, a broadening of the d–d transition band in relation to Gauss function is observed for all the examined samples, thus testifying towards formation of small PdO particles. As it was already mentioned, Pd + Rh/L4 catalyst was prepared by mixing of separately synthesized Pd/L4 and Rh/L4 samples with doubled metal loading. In such case, the concentration of supporting metals can exceed the amount of electron-donor sites on the surface of the support. This leads to formation of PdO particles, which is seen in the UV–Vis spectra. No noticeable changes were observed for the samples calcined at 800 °C. After calcination of the samples at 1000 °C (Fig. 5b), the characteristic values of F(R) in the area of the d–d transitions for Pd2+ remain the same as for the samples calcined at 600 °C (~ 0.7). At the same time, the long-wavelength edge of the absorption band was restructured, which is typical for PdO particles. In this case, average size of the PdO particles can be characterized by the value of band gap (Eg) estimated in accordance with the Tauc approach. Corresponding UV–Vis spectra are presented in Fig. 5c. According to the data obtained, average size of PdO particles has increased in all the cases. However, the smallest value of band gap (Eg = 2.05 eV) was observed for Pd–Rh/L4 sample, while the largest value of 2.13 eV was found for Pd + Rh/L4 sample.

The samples aged at 1000 °C were investigated by TEM method. Figure 6 shows the TEM images obtained. Palladium was found to be distributed within Pd/L4 sample in a wide range of particle sizes: from small clusters of few nm (Fig. 6b) to relatively large agglomerates of tens nm (Fig. 6a). All this correlates well with the data of UV–Vis spectroscopy and results of the PTA tests showing agglomeration of the Pd species during the high-temperature aging. Rhodium in the Rh-only catalyst (Fig. 6c) is represented by particles of 2–3 nm in size, which are strongly interacted with the support. Sample prepared by the impregnation of the support with joint solution of Pd and Rh nitrates (Pd–Rh/L4) contains mostly bimetallic nanoparticles of 10–40 nm in size with Pd:Rh ratio varied from 5:4 to 4:1. An example of such particle consisting of 80% Pd and 20% Rh is shown in Fig. 6d. No large agglomerates as in the case of Pd/L4 sample were found for this catalyst. Thereby, TEM studies confirm a near surface stabilization of rhodium in Rh/L4 sample and a mutual anchoring of Pd and Rh in the bimetallic Pd–Rh/L4 sample.

Fig. 6
figure 6

TEM images for the samples Pd/L4 (a, b), Rh/L4 (c), and Pd–Rh/L4 (d) after aging at 1000 °C

Full size image

The effect of the preliminary calcination of the support on the state of supported metals was more precisely studied by means of XPS technique. The XPS spectra recorded in a Pd3d region are shown in Fig. 7a. Despite the small loading of palladium in the samples, the Pd3d band is well resolved. The values of binding energy (BE) for the Pd3d5/2 band are 335.8 and 336.0 eV for Pd­Rh/L4 and Pd­Rh/L4* samples, respectively (Table 4). The samples aged at 1000 °C were used for this study. It should be noted that BE for metallic palladium corresponds to 335.1–335.3 eV, while for PdO oxide the values are 336.8–337.0 eV [29, 59, 66,67,68,69]. Thereby, the observed values of BE can be assigned to the relatively small particle of metallic palladium [70]. Despite both the samples contain the same amount of palladium, atomic ratios [Pd]/[Al] determined from the XPS spectra are strongly different (Table 5). Preliminary calcination of the support at 800 °C (sample Pd–Rh/L4*) leads to significantly much lower atomic ratio [Pd]/[Al] in the catalyst after aging at 1000 °C. Decrease in intensity of the XPS spectra at constant metal content is known to be typical for the case of the particles enlargement [71, 72]. Thereby, this testifies towards a presence of larger palladium particles, thus confirming less thermal stability of Pd particles on the support’s surface and their rapid agglomeration at elevated temperatures observed by previous characterization methods.

Table 4 Binding energies of the photoemission bands (eV)
Full size table
Table 5 Quantitative composition of the surface
Full size table

The XPS spectra for Rh3d region are shown in Fig. 7b. These spectra are represented by two duplet bands with BE Rh3d5/2 of ~ 307.4 and ~ 309.8 eV (Table 4). The first band can be assigned to nanosized particles of metallic rhodium [59, 60]. BE of the second band belongs to a range of 309.3–310.1 eV, which is typical for RhO2 oxide species [73, 74]. Since BE of Rh2O3 should correspond to an interval of 307.9–308.6 eV [73,74,75], absence of such bands allows one to exclude this oxide from the composition of the samples. Atomic ratio [Rh]/[Al] for total rhodium content in the samples and fraction of oxidized Rh species are summarized in Table 5. It is evident that after the high-temperature aging the most part of rhodium exists in the oxide form. Similarly to the case of palladium, atomic ratios [Rh]/[Al] for the samples are different. At the same time, this difference is not so high, and the effect of preliminary calcination of the support on the atomic ratio is insignificant.

Fig. 7
figure 7

XPS spectra for the Pd–Rh/L4 and Pd–Rh/L4* samples aged at 1000 °C: Pd3d region (a); Rh3d region (b); La3d region (c)

Full size image

Figure 7c demonstrates the XPS spectra for La3d region. The shape of the spectra is represented by two doublets with spin–orbit splitting in each of 16.8 eV that is typical for compounds with La3+, in particular, La2O3 [76]. The BE values for La3d5/2 bands in each doublet are almost the same (Table 4). Atomic ratios [La]/[Al], which characterize the relative surface concentration of lanthanum, are also similar and correspond to ~ 0.056–0.066 (Table 5). Note that according to the technical data, this ratio should be 0.012. Thereby, an enrichment of the surface with La takes place in both the cases. Since rhodium prefers to be localized near La3+ ions [56, 57], such redistribution of La within alumina matrix can be considered as an explanation of the near-surface localization of rhodium and its improved thermal stability.

Bimetallic catalysts Pd–Rh/L4 and Pd + Rh/L4, obtained by joint impregnation and mechanical mixing, correspondingly, were compared with CRS in the monolith tests. Light-off performance of the prepared monolith samples is shown in Fig. 8. As seen from Fig. 8a, both the studied samples surpass CRS in catalytic activity: temperatures of 50% conversion of CO, NOx and total hydrocarbons (THC) are noticeably lower for these samples. Comparing the samples with each other it can be noticed that mechanical mixture (Pd + Rh/L4 sample) is slightly more active than Pd–Rh/L4 catalyst in oxidation of hydrocarbons. In terms of CO oxidation and NOx reduction activity, the situation is opposite. The values of THC, CO, and NOx conversions at 400 °C are compared in Fig. 8b. In relation to this parameter, Pd + Rh/L4 and CRS catalysts coincide in THC and CO conversions, while Pd–Rh/L4 sample demonstrates better performance. In the case of NOx conversion, both studied samples surpass the reference one.

Fig. 8
figure 8

Light-off performance of the Pd–Rh/L4, Pd + Rh/L4, and CRS monolith catalysts: T50 values of THC, CO, and NOx conversion (a); conversion of THC, CO, and NOx at 400 °C (b)

Full size image

Another important parameter characterizing the efficiency of TWCs is operation window. As it follows from Fig. 9, the widths of the operation window for the Pd–Rh/L4 and Pd + Rh/L4 samples are comparable. Both presented windows are limited by NOx and CO conversions. The curve for CO conversion in the case of bimetallic Pd–Rh/L4 sample is mildly sloping, which causes shortening of the window width, thus making this catalyst slightly less effective. On the other hand, the difference is not crucial. For instance, the width of operation window for (0.5 wt%) Rh/CeZrYLaO2 catalyst aged at similar conditions was recently reported to be just 1.3 units [39]. Thus, it can be concluded that the application of bimetallic Pd–Rh/L4 catalysts of alloyed type looks very promising in terms of appropriate catalytic performance and excellent thermal stability.

Fig. 9
figure 9

Operation window of the monolith catalysts: Pd–Rh/L4 (a); Pd + Rh/L4 (b)

Full size image

4 Conclusions

Bimetallic Pd–Rh catalysts with strong metal–metal and metal–support interactions attract a growing interest due to enhanced catalytic performance and thermal stability. On the other hand, La-doped γ­Al2O3 system used as one of the components of modern three-way catalysts is not precisely studied yet. Thereby, the present work was devoted to detailed study of Pd–Rh/(γ­Al2O3 + 4 wt% La2O3) catalyst in comparison with monometallic Pd-only and Rh-only reference samples. In order to investigate the effect of metal–metal interaction, bimetallic sample of the same composition was prepared by mixing separately prepared samples with doubled loading of each metal. The effect of metal–support interaction was studied by comparing the samples with another reference sample synthesized by an impregnation of the preliminary calcined at 800 °C support with a joint solution of rhodium and palladium nitrates. Catalytic performance of the prepared samples was referred to that for the sample of commercial catalyst.

Doping of alumina with lanthanum oxide is known to be efficient for improvement of the thermal stability of the support. At the same time, properties of thermally stabilized alumina are being changed after calcination or aging at temperatures of 1000 °C and above. In particular, the value of specific surface area is dropped down significantly, reflexes of the δ-phase are appeared in the XRD patterns, and the concentration of the surface electron-donor sites is decreased. Changes in the support’s surface properties during the thermal treatment affect the localization of metals on the impregnation stage and the catalytic properties of the final catalysts. Thus, in the case of initial support, its impregnation with joint solution of Pd and Rh nitrates results in formation of the Pd–Rh alloyed nanoparticles strongly interacted with the support. Preliminary calcination of the support at already 800 °C changes the preferable localization of metals during the same procedure, and the alloy is not being formed, since metals do not interact with each other. Moreover, interaction of metals with the support became significantly weaker. Such catalyst exhibits the worse catalytic activity and thermal stability. Another bimetallic Pd + Rh/L4 sample (mechanical mixture) with no metal–metal interaction was found to be characterized by good enough initial activity, but quite poor thermal stability. Contrary, bimetallic Pd–Rh/L4 catalyst of alloyed type has shown excellent stability and catalytic activity comparable with the commercial reference sample. Thereby, it can be concluded that the concept of bimetallic Pd–Rh nanoalloys can be efficiently applied in the case of La-doped alumina used in the real three-way catalysts.