The Attractiveness of the Ternary Rh-Pd-Pt Alloys for CO Oxidation Process

Ternary alloys of platinum group metals attract a growing interest due to their unique catalytic properties. The present research is aimed to synthesize a series of Rh-Pd-Pt alloys with varied ratios of metals using a single-source precursor approach. Rhodium and palladium are partly miscible metals, while each of these metals is unlimitedly miscible with platinum. Thermolysis of complex salts used as a precursor results in the formation of metastable systems. The 3D nanostructure alloys are being formed after the complete decomposition of the single-source precursor. High-resolution transmission electron microscopic studies have shown that the nanoalloys are composed of interconnected polycrystalline ligaments with a mean diameter of 50 nm. The single-phase composition is confirmed by an X-ray diffraction analysis. The ratio of metals plays an important role in determining the catalytic activity of alumina-supported alloys and their thermal stability. According to UV-vis spectroscopy data, the higher palladium portion corresponds to worse dispersion of initially prepared, fresh catalysts. Treatment of the catalysts under prompt thermal aging conditions (up to 800 ◦C) causes redispersion of palladium-rich alloyed nanoparticles, thus leading to improved catalytic activity and thermal stability.


Introduction
Currently, a growing interest is attracted by metallic nanoparticles consisting of two and more metals with a specified interaction between them. Such nano-objects exhibit unique properties in various fields of application [1][2][3][4][5][6][7]. Thereby, one of the actual challenges of modern materials science is the development of the preparation routes to obtain metallic nanoparticles with a predefined size, composition, and, finally, properties. Among the variety of metals efficiently used in catalytic processes, precious metals of platinum group (Pt, Pd, and Rh) are of special importance. The palladium-based catalysts are known as highly active and efficient in a wide range of heterogeneous reactions starting from fine organic synthesis to partial and complete oxidation, selective hydrogenation or dehydrogenation [8][9][10][11][12][13][14][15][16][17][18][19][20][21]. For instance, being deposited on oxide supports (ceria, zirconia, alumina, etc.), Pd provides complete oxidation of hydrocarbons and CO that makes it a significant component of the modern catalysts of the engine exhaust gases purification [22][23][24][25][26][27][28][29]. Platinum is widely applied in numerous industrially important catalytic processes including hydrogenation (hydrogenation of C=C, C=O, N=O, and C≡N bonds), dehydrogenation (cleavage of N-H and C-H bonds), reforming of oxygenates (alcohols, polyols, and other), CO oxidation, water gas shift reaction, conversion of CH 4 ,

Synthesis of Precursors
In the present work, single-source precursors [

Characterization Techniques
The powder X-ray diffraction (XRD) analysis of the prepared precursors and alloys was performed at room temperature and atmospheric pressure on a Shimadzu XRD-7000 diffractometer (Shimadzu Corp., Kyoto, Japan) operating with CuK α radiation, nickel filter in reflected beam, and scintillation detector with amplitude discrimination. The data were collected step-by-step in the range of angles 2θ = 5-135 • with a step of 0.05 • . The resulting patterns were analyzed and fitted using a PowerCell 2.4 and WINFIT 1.2.1 software packages, correspondingly [91,92]. The obtained coordination compounds were single-phase, which was identified by indexing the powder XRD patterns similarly to theoretical patterns for complexes studied by an X-ray structure analysis.
Elemental (C, H, and N) analysis was carried out on a Euro EA 3000 analyzer (Eurovector, S.p.a., Milan, Italy). The specimen was thermally destroyed in the presence of solid oxidizers with the addition of gaseous oxygen until the quantitative formation of final products: CO 2 , water, and nitrogen. The obtained gas mixture was separated chromatographically. The quantities of the components were used to calculate the portions (wt %) of carbon, hydrogen, and nitrogen in the analyzed sample.
The metal content in the samples was determined by the atomic absorption method using a Hitachi Z-800 spectrometer (Hitachi, Tokyo, Japan). The sample was dissolved in aqua regia, evaporated until the minimal volume, loaded into a graduated flask and diluted by distilled water until marker.
All the theoretical values of the element's content were calculated from the molecular weights of the coordination compounds. The metal loading in the supported catalysts was measured by an energy dispersive X-ray fluorescence (ED XRF) analysis. An M1 MISTRAL spectrometer (Bruker Nano Nano GmbH, Berlin, Germany) with a microfocus X-ray tube (W-anode) and a high-resolution silicon drift detector (SDD, energy resolution <150 eV at Mn-K α ) was used.
Thermogravimetric analyses (TGA) were carried out using a TG 209 F1 Iris thermobalance (NETZSCH-Gerätebau GmbH, Selb, Germany). Synchronic thermal analysis was performed using an STA 449F1 Jupiter analyzer equipped with a QMS 403D Aëolos quadrupole mass spectrometer (NETZSCH-Gerätebau GmbH, Selb, Germany). The measurements were made in a hydrogen-helium mixture (10% vol. H 2 in He). The parameters of the TGA experiments were the following: sample weight of 20 mg; Al-crucibles; the gas flow rate of 60 mL/min; and the temperature ramping rate of 10 • C/min.
The high-resolution transmission electron microscopy (HR TEM) measurements were performed using a JEM-2010 electron microscope (JEOL Ltd., Tokyo, Japan) with lattice plane resolution 0.14 nm at an accelerating voltage of 200 kV. Images of periodic structures were analyzed by the Fourier method. A local energy-dispersive X-ray analysis (EDXA) was carried out using an EDX spectrometer Xflash (Bruker Nano GmbH, Berlin, Germany) fitted with a Si detector with a resolution of 128 eV. The samples for the HR TEM study were prepared on a perforated carbon film mounted on a copper grid.
Diffuse reflectance UV-Vis spectra were recorded between 200 and 850 nm using a UV-vis spectrometer UV-VIS 2501 PC (Shimadzu Corp., Kyoto, Japan) 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 [93].

Testing the Catalytic Activity in the Oxidation of CO
The catalytic performance of the alumina-supported catalysts was examined in a model reaction of CO oxidation. Each sample was studied in seven consecutive cycles under prompt thermal aging (PTA) conditions as described elsewhere [90,[93][94][95][96]. The testing procedure was as follows. The sample (300 mg) was loaded into the quartz flow reactor, and then the reaction mixture contained 0.15 vol% CO, 14.0 vol% O 2 , 0.01 vol% NO, 0.01 vol% hydrocarbons (a mixture of methane, propylene, and toluene), and nitrogen (as balance) was passed through the reactor. The flow rate was 334 mL/min. A temperature ramping rate was 10 • C/min. The final temperature of the catalytic cycle was 320 • C for 1st and 2nd cycles, 600 • C for 3rd and 4th cycles, and 800 • C for the last three cycles. The CO concentration at the reactor outlet was registered by a gas analyzer ULTRAMAT 6 (Siemens). The temperature of 50% CO conversion (T 50 ) was used as a criterion to compare the thermal stability of the catalysts.

Characterization of the Precursors
At the first step of the study, all the prepared coordination compounds were characterized by elemental analysis, atomic absorption spectroscopy, and infrared spectroscopy.  Table 1. The infrared spectra of the prepared coordination compounds are shown in Figure 1a and the corresponding oscillation frequencies are summarized in Table 2. As seen, the measured values of elements content were very close to the calculated ones. The results of the infrared spectroscopy ( Table 2) allowed identifying of the functional groups in the compositions of the salts.       3 and their solid solutions with varied ratios of metals in the anionic part were easily isolated into the solid phase in the nonaqueous form. Both the purity and the single phase of all the synthesized products were confirmed by an XRD analysis method ( Figure 1b). As follows from the figure, the experimental patterns corresponded well with the simulated (calculated) one.
The thermal decomposition of [  3 composition in a reductive medium proceeded similarly, in few hardly separable stages accompanied by exothermic effects (Figure 2a). Along with this, the thermal stability of the solid solutions depends on the content of palladium in their composition (Figure 2b). The starting temperature point of decomposition regularly decreased with an increase of the palladium content due to the lower thermal stability of the Pd-containing coordination compounds. The final temperature point of decomposition was not changed noticeably and corresponded to 310-320 • C. The total weight loss and the weight of the products obtained by pyrolysis in a reductive atmosphere agreed well with the calculated content of metals in the composition of the initial single-source precursors (Table 3). an increase of the palladium content due to the lower thermal stability of the Pd-containing coordination compounds. The final temperature point of decomposition was not changed noticeably and corresponded to 310-320 °C. The total weight loss and the weight of the products obtained by pyrolysis in a reductive atmosphere agreed well with the calculated content of metals in the composition of the initial single-source precursors (Table 3).

Characterization of the Alloys
Some representative TEM images of the products of the precursors' thermolysis are shown in Figure 3. The decomposition of the precursors in a hydrogen atmosphere along with subsequent heating of the products up to 600 °C resulted in the formation of an alloy characterized by a porous sponge-like structure. In terms of morphology, the nanoporous alloys consisted of interconnecting ligaments with a mean diameter of 50 nm. As evidenced by HR TEM, the ligaments were polycrystalline with grain boundaries separated by a few nanometers.

Characterization of the Alloys
Some representative TEM images of the products of the precursors' thermolysis are shown in Figure 3. The decomposition of the precursors in a hydrogen atmosphere along with subsequent heating of the products up to 600 • C resulted in the formation of an alloy characterized by a porous sponge-like structure. In terms of morphology, the nanoporous alloys consisted of interconnecting ligaments with a mean diameter of 50 nm. As evidenced by HR TEM, the ligaments were polycrystalline with grain boundaries separated by a few nanometers.
An energy-dispersive spectroscopic study of the as-prepared samples confirmed that the Rh:Pd:Pt atomic ratio in alloys was consistent with the atomic ratio of metals in the respective precursors. The alloy nature of the porous products was also confirmed by estimation of the interplanar spacing (d 111 ) from the high-resolution TEM images ( Figure 3). The estimated mean values of the interplanar spacing correlated satisfactorily with the corresponding d 111 values calculated from the lattice parameters obtained by the XRD analysis method (Table 4).
It should be noted that the binary systems Pt-Pd, Rh-Pt, and Rh-Pd were characterized by a wide area of existence of the random solid solutions with face-centered cubic cell (fcc, space group Fm3m) [69,[97][98][99][100][101]. In all these binary systems the components were miscible at the temperatures above 900 • C in a whole range of compositions. The Rh-Pd system exhibited a miscibility gap with the critical point of solubility at 58 at.% Rh and 820 • C [69].  An energy-dispersive spectroscopic study of the as-prepared samples confirmed that the Rh:Pd:Pt atomic ratio in alloys was consistent with the atomic ratio of metals in the respective precursors. The alloy nature of the porous products was also confirmed by estimation of the interplanar spacing (d111) from the high-resolution TEM images ( Figure 3). The estimated mean values of the interplanar spacing correlated satisfactorily with the corresponding d111 values calculated from the lattice parameters obtained by the XRD analysis method (Table 4).
It should be noted that the binary systems Pt-Pd, Rh-Pt, and Rh-Pd were characterized by a wide area of existence of the random solid solutions with face-centered cubic cell (fcc, space group Fm3m) [69,[97][98][99][100][101]. In all these binary systems the components were miscible at the temperatures above 900 °C in a whole range of compositions. The Rh-Pd system exhibited a miscibility gap with the critical point of solubility at 58 at.% Rh and 820 °C [69].
Up to date, there is no information concerning an experimental study on the subsolidus area of the ternary Rh-Pd-Pt system. The only exception is the reported isothermal section at 1200 °C illustrating the unlimited miscibility of the components at this temperature in a whole range of compositions [102,103]. Therefore, the hypothetical scheme of the phase diagram for this ternary system can be based on the extrapolations from the binary Pt-Pd, Rh-Pt, and Rh-Pd systems only. Additionally, it should be taken into account that since the binary Rh-Pd alloy separated onto two phases at temperatures below 820 °C, the phase diagram of the ternary system was characterized by Up to date, there is no information concerning an experimental study on the subsolidus area of the ternary Rh-Pd-Pt system. The only exception is the reported isothermal section at 1200 • C illustrating the unlimited miscibility of the components at this temperature in a whole range of compositions [102,103]. Therefore, the hypothetical scheme of the phase diagram for this ternary system can be based on the extrapolations from the binary Pt-Pd, Rh-Pt, and Rh-Pd systems only. Additionally, it should be taken into account that since the binary Rh-Pd alloy separated onto two phases at temperatures below 820 • C, the phase diagram of the ternary system was characterized by a presence of the miscibility gap, which should be based on the Rh-Pd axis. The first attempt to draw the phase diagram for the Rh-Pd-Pt system was undertaken using the calculations of the heats of mixing by the Monte Carlo (MC) method combined with the semi-empirical modified embedded atom method (MEAM). The resulting miscibility gap for the temperature of 327 • C is presented in Figure 4. Taking into account the experimental temperature dependency of the equilibrium solubility of the solid solutions in the Rh-Pd system and the isothermal cross-section of the ternary phase diagram for the Rh-Pd-Pt system at 327 • C, the hypothetical miscibility gaps for the temperatures 600, 700, and 800 • C are drawn in Figure 4. As seen from the presumed ternary phase diagrams, the alloys Rh 2 (Pd 0.25 Pt 0.75 ) 3 , Rh 2 (Pd 0.5 Pt 0.5 ) 3 , and Rh 2 Pt 3 at 600 • C and above (filled triangles) were located in a homogeneous area and, therefore, should be in equilibrium. Contrary, the Rh 2 Pd 3 and Rh 2 (Pd 0.75 Pt 0.25 ) 3 alloys at the same temperatures (empty triangles) should be metastable. of the solid solutions in the Rh-Pd system and the isothermal cross-section of the ternary phase diagram for the Rh-Pd-Pt system at 327 °C, the hypothetical miscibility gaps for the temperatures 600, 700, and 800 °C are drawn in Figure 4. As seen from the presumed ternary phase diagrams, the alloys Rh2(Pt0.75Pd0.25)3, Rh2(Pt0.50Pd0.50)3, and Rh2Pt3 at 600 °C and above (filled triangles) were located in a homogeneous area and, therefore, should be in equilibrium. Contrary, the Rh2Pd3 and Rh2(Pt0.25Pd0.75)3 alloys at the same temperatures (empty triangles) should be metastable.  [104]. The colored squares correspond to experimental points for the binary Pd-Rh systems at temperatures of 600, 700, and 800 °C, while dash lines show the estimated miscibility gaps for the ternary Rh-Pd-Pt systems at these temperatures. Red triangles are experimental points for the alloys studied in the present work, where filled and empty triangles lie in areas of homogeneity and metastability, accordingly.    3 at 600 • C in a hydrogen atmosphere are presented in Figure 5. As seen, they contained only the fcc phase (Figure 5a). The shift in the position of the reflexes regarding pure Pd, Pt, and Rh was observed in a far angle region (Figure 5b). This indicates the formation of homogeneous solid solutions. The analysis of the reflexes position allowed the conclusion that an increase of the Pd content in the ternary alloy composition leads to the shifting of the reflexes towards a far angle region, thus testifying in a decrease in the parameter of the elementary cell of the solid solution. The lattice parameters of the alloys calculated from the positions of 331 and 420 reflexes are summarized in Table 4. Figure 6 shows the dependencies of the lattice parameters of the alloys on the atomic fraction in the Rh-Pt and Rh-Pd systems based on the literature data [69,101,106]. In these binary systems, an appropriate satisfiability of the Vegard's law was observed. The maximum deviation of the lattice parameters from the law did not exceed 0.002 Å. Taking all these into account, interpolating isotherms 'lattice parameter-composition' for the ternary Rh-Pd-Pt alloys are plotted in Figure 6 (dashed lines). position allowed the conclusion that an increase of the Pd content in the ternary alloy composition leads to the shifting of the reflexes towards a far angle region, thus testifying in a decrease in the parameter of the elementary cell of the solid solution. The lattice parameters of the alloys calculated from the positions of 331 and 420 reflexes are summarized in Table 4.  Figure 6 shows the dependencies of the lattice parameters of the alloys on the atomic fraction in the Rh-Pt and Rh-Pd systems based on the literature data [69,101,106]. In these binary systems, an appropriate satisfiability of the Vegard's law was observed. The maximum deviation of the lattice parameters from the law did not exceed 0.002 Å. Taking all these into account, interpolating isotherms 'lattice parameter-composition' for the ternary Rh-Pd-Pt alloys are plotted in Figure 6 (dashed lines). The measured values of lattice parameters of the ternary alloys ( Figure 6, triangles) matched the corresponding values on the plotted dependencies. Thereby, this confirms a coincidence of the composition of the synthesized homogeneous ternary solid solutions with the composition specified at the synthesis of the single-source precursors. The measured values of lattice parameters of the ternary alloys ( Figure 6, triangles) matched the corresponding values on the plotted dependencies. Thereby, this confirms a coincidence of the composition of the synthesized homogeneous ternary solid solutions with the composition specified at the synthesis of the single-source precursors.
from the positions of 331 and 420 reflexes are summarized in Table 4.  Figure 6 shows the dependencies of the lattice parameters of the alloys on the atomic fraction in the Rh-Pt and Rh-Pd systems based on the literature data [69,101,106]. In these binary systems, an appropriate satisfiability of the Vegard's law was observed. The maximum deviation of the lattice parameters from the law did not exceed 0.002 Å. Taking all these into account, interpolating isotherms 'lattice parameter-composition' for the ternary Rh-Pd-Pt alloys are plotted in Figure 6 (dashed lines). The measured values of lattice parameters of the ternary alloys ( Figure 6, triangles) matched the corresponding values on the plotted dependencies. Thereby, this confirms a coincidence of the composition of the synthesized homogeneous ternary solid solutions with the composition specified at the synthesis of the single-source precursors.

Characterization and Testing of the Alumina-Supported Catalysts
In the next step of the study, the synthesized precursors were supported on the commercial alumina powder via an incipient wetness impregnation. In order to control the content of the supported metals, the catalysts samples were characterized by an ED XRF method. The measured values along with the calculated ones are summarized in Table 5. As seen, the values coincided with a good enough accuracy. All the samples of alumina-supported catalysts were tested in a model reaction of CO oxidation. The testing procedure was performed under prompt thermal aging conditions, which allowed monitoring of the degradation of the catalysts during the high-temperature exposure. The corresponding light-off curves are shown in Figure 7a-e. To compare the samples evidently, the dependencies of the T 50 parameter on the catalytic run number are plotted in Figure 7f. The platinum-only catalyst was used as a reference sample. As follows from the obtained results, the metals in the alloy interacted with each other and influence the catalytic properties of the catalyst as a whole. For instance, Pt/γ-Al 2 O 3 sample demonstrated poor initial activity (T 50 = 286 • C) but undergoes reactivation at elevated temperatures. Most significantly this effect became apparent after heating up to 800 • C. In the case of the bimetallic Rh 2 Pt 3 /γ-Al 2 O 3 catalyst, the tendencies in catalytic activity and stability remained the same. In general, the presence of rhodium in the catalyst's composition made the catalyst a little bit more active. For the ternary alloy systems, a partial replacement of platinum with palladium much evidently enhanced the initial activity of the catalyst: more Pd content in the alloy corresponded to the lower value of the T 50 parameter. Since palladium is known to be less stable than platinum, the aging of the catalysts at 600 • C led to their noticeable deactivation. Then, being heated up to 800 • C, Rh 2 (Pd 0.25 Pt 0.75 ) 3 /γ-Al 2 O 3 and Rh 2 (Pd 0.5 Pt 0.5 ) 3 /γ-Al 2 O 3 catalysts show reactivation similarly to the samples without palladium. Oppositely, the sample Rh 2 (Pd 0.75 Pt 0.25 ) 3 /γ-Al 2 O 3 containing a higher amount of Pd continued to deactivate even after treatment at 800 • C. Taking into account the mentioned oppositely directed trends, it could be concluded the most promising system is a ternary alloy Rh 2 (Pd 0.5 Pt 0.5 ) 3 with an equivalent amount of palladium and platinum. At a higher content of Pd, the "metal-metal" interactions in the alloy were assumed to become weaker due to partial immiscibility of palladium and rhodium, and the further catalytic behavior of the catalyst was determined by Pd. If platinum dominates over palladium, then the catalyst performs worse activity.
As we have reported recently, optical spectroscopy methods are useful and informative tools allowing characterizing of the state of the supported precious metals and their alloys as well as the changes taking place during high-temperature aging [93]. Therefore, all the studied binary and ternary catalysts and the monometallic reference samples were examined by a diffuse reflectance UV-vis spectroscopy. To exclude the effect of "metal-metal" interactions, some additional reference samples obtained by mechanical mixing of corresponding monometallic catalysts were prepared and studied. As we have reported recently, optical spectroscopy methods are useful and informative tools allowing characterizing of the state of the supported precious metals and their alloys as well as the changes taking place during high-temperature aging [93]. Therefore, all the studied binary and ternary catalysts and the monometallic reference samples were examined by a diffuse reflectance UVvis spectroscopy. To exclude the effect of "metal-metal" interactions, some additional reference samples obtained by mechanical mixing of corresponding monometallic catalysts were prepared and studied. Figure 8 presents the UV-vis spectra for the monometallic Pt/γ-Al2O3 and Rh/γ-Al2O3, samples and the bimetallic Rh2Pt3/γ-Al2O3 (alloy) and Rh+Pt/γ-Al2O3 (mechanical mixture) catalysts. To study    To study the effect of the surface hydroxylation under natural conditions, the Pt/γ-Al 2 O 3 sample was additionally calcined in argon at 400 • C for 4 h immediately before the measurements. Such a low temperature of pretreatment was chosen to minimize any possible sintering effects. As known, the spectra of the Rh 3+ (4d6) and Pt 4+ (5d6) ions in the low-spin octahedral complexes are characterized by a charge transfer band and d-d transitions 1 A 1g → 1 T 1g , 1 T 2g . The representative value of the energy-gap width (E g ) of Rh 3+ complexes for the Rh/γ-Al 2 O 3 sample was equal to 2.49 eV (Figure 8, spectrum 2). Koffyberg has reported that the E g value for Rh 2 O 3 particles decreases from an asymptotic value of 2.59 eV for the isolated Rh 3+ complexes to 1.2 eV for the bulk oxide Rh 2 O 3 [107]. additionally calcined in argon at 400 °C for 4 h immediately before the measurements. Such a low temperature of pretreatment was chosen to minimize any possible sintering effects. As known, the spectra of the Rh 3+ (4d6) and Pt 4+ (5d6) ions in the low-spin octahedral complexes are characterized by a charge transfer band and d-d transitions 1 A1g → 1 T1g, 1 T2g. The representative value of the energygap width (Eg) of Rh 3+ complexes for the Rh/γ-Al2O3 sample was equal to 2.49 eV (Figure 8, spectrum 2). Koffyberg has reported that the Eg value for Rh2O3 particles decreases from an asymptotic value of 2.59 eV for the isolated Rh 3+ complexes to 1.2 eV for the bulk oxide Rh2O3 [107].
In the case of the Pt/γ-Al2O3 sample, the energy-gap width for Pt 4+ complexes was about 2.62 eV. This value did not change noticeably after the dehydroxylation procedure in an argon flow at 400 °C, despite a slight increase in the intensity of the 1 A1g → 1 T1g band. Unfortunately, there is no information on the Eg value for the Pt 4+ complexes in the literature. The experimental Eg values for the α-PtO2 particles were varied in a range from 1.3 [108] , to 1.8 eV [109].
As evident from Figure 8, the sum of spectra 1 and 2 (Pt/γ-Al2O3 and Rh/γ-Al2O3 samples, accordingly) corresponded well to the spectra for the bimetallic samples Rh2Pt3/γ-Al2O3 and Rh+Pt/γ-Al2O3, despite the lower intensity. Such an approach was recently reported as efficient for the approximation of spectra of the bimetallic Rh2Pd3/γ-Al2O3 model catalysts [95]. Good agreement of the spectra for the binary systems with the spectra for the monometallic reference samples indicates that there were no significant distortions of the octahedral surrounding taking place during the formation of Rh2Pd3 oxide clusters. Long-wave edge of d-d transitions of the bimetallic samples at the present ratio of metals was mainly determined by Rh 3+ ions. The values of the energy-gap width for the samples Rh2Pt3/γ-Al2O3 and Rh+Pt/γ-Al2O3 were equal to 2.47 and 2.34 eV, respectively. The sample Rh+Pt/γ-Al2O3 was characterized by a larger particle size in comparison with the sample Rh2Pt3/γ-Al2O3, which possessed practically the same Eg value as the reference sample Rh/γ-Al2O3. Figure 9a shows the UV-vis spectra for the samples of the reference catalyst Pt/γ-Al2O3 treated at 400 °C (spectrum 2), aged under PTA conditions (spectrum 1), and treated in air at 400 °C for 6 h after aging under PTA conditions (spectrum 3). Left part of these spectra presented in Figure 9b demonstrates the charge transfer band for the same samples. Treatment in air was aimed to oxidize the reduced part of platinum and recover the initial hydroxyl coverage on the surface without any sintering effects. The observed appearance of an intensive band of the plasmon resonance in the UV In the case of the Pt/γ-Al 2 O 3 sample, the energy-gap width for Pt 4+ complexes was about 2.62 eV. This value did not change noticeably after the dehydroxylation procedure in an argon flow at 400 • C, despite a slight increase in the intensity of the 1 A 1g → 1 T 1g band. Unfortunately, there is no information on the E g value for the Pt 4+ complexes in the literature. The experimental E g values for the α-PtO 2 particles were varied in a range from 1.3 [108] , to 1.8 eV [109].
As evident from Figure 8, the sum of spectra 1 and 2 (Pt/γ-Al 2 O 3 and Rh/γ-Al 2 O 3 samples, accordingly) corresponded well to the spectra for the bimetallic samples Rh 2 Pt 3 /γ-Al 2 O 3 and Rh+Pt/γ-Al 2 O 3 , despite the lower intensity. Such an approach was recently reported as efficient for the approximation of spectra of the bimetallic Rh 2 Pd 3 /γ-Al 2 O 3 model catalysts [95]. Good agreement of the spectra for the binary systems with the spectra for the monometallic reference samples indicates that there were no significant distortions of the octahedral surrounding taking place during the formation of Rh 2 Pd 3 oxide clusters. Long-wave edge of d-d transitions of the bimetallic samples at the present ratio of metals was mainly determined by Rh 3+ ions. The values of the energy-gap width for the samples Rh 2 Pt 3 /γ-Al 2 O 3 and Rh+Pt/γ-Al 2 O 3 were equal to 2.47 and 2.34 eV, respectively. The sample Rh+Pt/γ-Al 2 O 3 was characterized by a larger particle size in comparison with the sample Rh 2 Pt 3 /γ-Al 2 O 3 , which possessed practically the same E g value as the reference sample Rh/γ-Al 2 O 3 . Figure 9a shows the UV-vis spectra for the samples of the reference catalyst Pt/γ-Al 2 O 3 treated at 400 • C (spectrum 2), aged under PTA conditions (spectrum 1), and treated in air at 400 • C for 6 h after aging under PTA conditions (spectrum 3). Left part of these spectra presented in Figure 9b demonstrates the charge transfer band for the same samples. Treatment in air was aimed to oxidize the reduced part of platinum and recover the initial hydroxyl coverage on the surface without any sintering effects. The observed appearance of an intensive band of the plasmon resonance in the UV region of the spectrum 1 testifies towards the formation of Pt 0 particles, while d-d transitions indicate the presence of oxidized Pt 4+ species. The shift of the 1 A 1g → 1 T 2g band to the short-wave region can be assigned to the part exchange of the ligands in the Pt 4+ complexes as a result of aging in the PTA regime, and to their interaction with metallic particles. The oxidative treatment of the catalyst in air leads to a decrease in intensity of the band in the UV region of the spectrum down to the initial level, thus indicating the oxidation of platinum. On the other hand, an increase in the intensity of d-d transitions of the Pt 4+ complexes along with a shift of the absorption band's edge towards the long-wave region allows the conclusion about the formation of relatively large α-PtO 2 particles characterized by the E g value of 1.65 eV.
regime, and to their interaction with metallic particles. The oxidative treatment of the catalyst in air leads to a decrease in intensity of the band in the UV region of the spectrum down to the initial level, thus indicating the oxidation of platinum. On the other hand, an increase in the intensity of d-d transitions of the Pt 4+ complexes along with a shift of the absorption band's edge towards the longwave region allows the conclusion about the formation of relatively large α-PtO2 particles characterized by the Eg value of 1.65 eV. The UV-vis spectra for the bimetallic catalysts Rh2Pt3/γ-Al2O3 and Rh+Pt/γ-Al2O3 and monometallic reference sample Pt/γ-Al2O3 after the prompt thermal aging experiments and subsequent treatment in air at 400 °C are shown in Figure 10. Corresponding values of the energygap width for the initial and aged samples are summarized in Table 6. It was evident that both the bimetallic samples were more thermally stable towards sintering if compared with the Pt/γ-Al2O3 reference, and possessed very close values of Eg 1.97-2.01 eV.  The UV-vis spectra for the bimetallic catalysts Rh 2 Pt 3 /γ-Al 2 O 3 and Rh+Pt/γ-Al 2 O 3 and monometallic reference sample Pt/γ-Al 2 O 3 after the prompt thermal aging experiments and subsequent treatment in air at 400 • C are shown in Figure 10. Corresponding values of the energy-gap width for the initial and aged samples are summarized in Table 6. It was evident that both the bimetallic samples were more thermally stable towards sintering if compared with the Pt/γ-Al 2 O 3 reference, and possessed very close values of E g 1.97-2.01 eV.
regime, and to their interaction with metallic particles. The oxidative treatment of the catalyst in air leads to a decrease in intensity of the band in the UV region of the spectrum down to the initial level, thus indicating the oxidation of platinum. On the other hand, an increase in the intensity of d-d transitions of the Pt 4+ complexes along with a shift of the absorption band's edge towards the longwave region allows the conclusion about the formation of relatively large α-PtO2 particles characterized by the Eg value of 1.65 eV. The UV-vis spectra for the bimetallic catalysts Rh2Pt3/γ-Al2O3 and Rh+Pt/γ-Al2O3 and monometallic reference sample Pt/γ-Al2O3 after the prompt thermal aging experiments and subsequent treatment in air at 400 °C are shown in Figure 10. Corresponding values of the energygap width for the initial and aged samples are summarized in Table 6. It was evident that both the bimetallic samples were more thermally stable towards sintering if compared with the Pt/γ-Al2O3 reference, and possessed very close values of Eg 1.97-2.01 eV. Figure 10. The diffuse reflectance UV-vis spectra of samples Pt/γ-Al2O3, Rh2Pt3/γ-Al2O3, and Rh+Pt/γ-Al2O3 aged at 800 °C and treated at 400 °C. All the spectra are corrected concerning the support.  The UV-vis spectra for the ternary systems Rh 2 (Pd 0.25 Pt 0.75 ) 3 /γ-Al 2 O 3 , Rh 2 (Pd 0.5 Pt 0.5 ) 3 /γ-Al 2 O 3 , and Rh 2 (Pd 0.75 Pt 0.25 ) 3 /γ-Al 2 O 3 before and after the aging under PTA conditions are shown in Figure 11. In the initial state, the values of the energy-gap width for these catalysts were very close ( Table 6). The sample Rh 2 (Pd 0.25 Pt 0.75 ) 3 /γ-Al 2 O 3 containing the highest amount of platinum was characterized by a better dispersion of the alloy nanoparticles (Figure 11a). The aging of the samples at 800 • C changed significantly the situation. In this case, the samples could be ranked by a decreased dispersion as Rh 2 (Pd 0. 75   The UV-vis spectra for the ternary systems Rh2(Pd0.25Pt0.75)3/γ-Al2O3, Rh2(Pd0.5Pt0.5)3/γ-Al2O3, and Rh2(Pd0.75Pt0.25)3/γ-Al2O3 before and after the aging under PTA conditions are shown in Figure 11. In the initial state, the values of the energy-gap width for these catalysts were very close ( Table 6). The sample Rh2(Pd0.25Pt0.75)3/γ-Al2O3 containing the highest amount of platinum was characterized by a better dispersion of the alloy nanoparticles (Figure 11a). The aging of the samples at 800 °C changed significantly the situation. In this case, the samples could be ranked by a decreased dispersion as Rh2(Pd0.75Pt0.25)3/γ-Al2O3 > Rh2(Pd0.5Pt0.5)3/γ-Al2O3 > Rh2(Pd0.25Pt0.75)3/γ-Al2O3. The Eg values for the samples Rh2(Pd0.75Pt0.25)3/γ-Al2O3 and Rh2(Pd0.5Pt0.5)3/γ-Al2O3 were 2.3 and 2.17 eV, accordingly. Thereby, in terms of dispersion of the active component, the ternary alloy systems surpassed the bimetallic and monometallic catalysts. Figure 11. The diffuse reflectance UV-vis spectra of samples Rh2(Pd0.25Pt0.75)3/γ-Al2O3, Rh2(Pd0.5Pt0.5)3/γ-Al2O3, and Rh2(Pd0.75Pt0.25)3/γ-Al2O3 treated at 400 °C: (a) initial and (b) aged at 800 °C under PTA conditions. All the spectra are corrected concerning the support.
It was already mentioned that the Pt-only catalyst underwent noticeable reactivation after the aging at 800 °C. Its catalytic activity in the 6th and 7th runs was comparable with that for the binary and ternary catalysts (Figure 7). On the other hand, as it follows from the UV-vis spectroscopy data (Figure 9a), the active Pt species after the PTA aging were predominantly represented by the Pt 4+ complexes. For the ternary systems, the higher amount of these forms is observed in the case of Rh2(Pd0.5Pt0.5)3/γ-Al2O3 ( Figure 11, spectrum 2). Possibly, this explains the better catalytic performance (high initial activity and appropriate stability) of this sample.
Summarizing all the data presented above, it should be mentioned that the concept of a singlesource precursor is applicable and quite efficient for the preparation of the supported ternary Rh-Pd-Pt catalysts. The thermolysis of the solid solutions of the double complex salts used as such precursors allows for obtaining of the ternary alloys with a predefined metal ratio. Thus, the final catalytic properties of the catalysts can be easily tuned. It was already mentioned that the Pt-only catalyst underwent noticeable reactivation after the aging at 800 • C. Its catalytic activity in the 6th and 7th runs was comparable with that for the binary and ternary catalysts (Figure 7). On the other hand, as it follows from the UV-vis spectroscopy data (Figure 9a), the active Pt species after the PTA aging were predominantly represented by the Pt 4+ complexes. For the ternary systems, the higher amount of these forms is observed in the case of Rh 2 (Pd 0.5 Pt 0.5 ) 3 /γ-Al 2 O 3 ( Figure 11, spectrum 2). Possibly, this explains the better catalytic performance (high initial activity and appropriate stability) of this sample.
Summarizing all the data presented above, it should be mentioned that the concept of a single-source precursor is applicable and quite efficient for the preparation of the supported ternary Rh-Pd-Pt catalysts. The thermolysis of the solid solutions of the double complex salts used as such precursors allows for obtaining of the ternary alloys with a predefined metal ratio. Thus, the final catalytic properties of the catalysts can be easily tuned.

Conclusions
Precious metals supported on different supports (alumina, ceria, and zirconia) are widely attractive due to their excellent catalytic properties in various industrially and environmentally important processes. For instance, one of the areas of their application is three-way catalysis aimed to reduce the emissions from gasoline engines. The requirements imposed for such catalytic converters are an appropriate activity in the catalyzed reactions (oxidation of CO and hydrocarbons, and reduction of nitrogen oxides) and long-term stability including the resistance towards the elevated temperatures. The CO oxidation reaction is commonly used as a model reaction for screening the prospective catalytic compositions. The present work is purposed in studying the applicability of a single-source precursor concept to prepare the ternary nanoalloys with improved catalytic characteristics. A series of the catalysts were prepared by an incipient wetness impregnation of the alumina support with an aqueous solution of the corresponding complex salt containing one, two, or three metals (Rh, Pd, and Pt). The formation of alloys as a result of the precursor's thermolysis in a reductive medium was confirmed by the XRD and TEM methods. The phase diagram for the ternary Rh-Pd-Pt system at temperatures 600, 700, and 800 • C was proposed for the first time. The prepared samples were examined in the CO oxidation reaction performed under prompt thermal aging conditions. It was shown that the introduction of palladium into the composition of the active particles improves the initial activity along with appropriate stability at high temperatures. The best catalytic performance was exhibited by the Rh 2 (Pd 0.5 Pt 0.5 ) 3 /γ-Al 2 O 3 sample with equal amounts of platinum and palladium. The characterization of the catalysts by a diffuse reflectance UV-vis spectroscopy has revealed that palladium affects the dispersion of the nanoparticles: the higher the amount of Pd that is introduced, the better dispersion is. The extraordinary activity of the Rh 2 (Pd 0.5 Pt 0.5 ) 3 /γ-Al 2 O 3 sample is assigned to the presence of Pt in the form of the Pt 4+ complexes.