2.1 Catalyst preparation characterization

Two types of commercial active carbons were used as catalyst's supports: activated pyrocarbon Sibunit (Novosibirsk, Russia) [37] and furnace black Vulcan XC-72 (Cabot Corporation) [38]. Before preparation of catalysts, the active carbon Sibunit was washed with a boiling mixture of concentrated HCl and HF, rinsed with large amounts of redistilled water and dried at 253 K in an air oven for 12 hours. Carbon Vulcan XC-72 was used without any purification. The properties of the supports are given in Table 1. It is evident that the Vulcan carbon is characterized by a larger average size of pores.

The catalysts consisting of the same palladium loading equal to 2 wt% Pd and the growing Au content were prepared by means of the reverse “water-in-oil” microemulsion method according to the procedure described previously [39,40]. In the PdAu/carbon catalysts the molar ratio of Pd:Au ranges were from 95:5 up to 70:30. The monometallic 2 wt%Pd/C (Pd₁₀₀) and 2 wt%Au/C (Au₁₀₀) catalysts (for catalyst notation, Table 2) were also prepared using the same preparation procedure. The reverse micellar solutions were prepared using surfactant dioctyl sulfosuccinate sodium salt (AOT, pure 98%, Fluka) and heptane (pure, Aldrich) as the oil phase. The aqueous solutions of PdCl₂ (99.9% Grade, Johnson Matthey) and HAuCl₄ (pure, Avantor) were used as the precursors. Hydrazine hydrate (reagent grade 98%, Aldrich) was applied as a reducing agent. Taking into account our previous studies [39,40] the value of molar ratio of aqueous solution (PdCl₂ and/or HAuCl₄) to AOT surfactant W = 5 was applied.

In the first stage of the preparation of the catalysts, AOT was dissolved in heptane, then an aqueous solution of precursor(s) was introduced to heptane-surfactant mixture and stirred to obtain a transparent yellow/orange suspension. To this microemulsion, hydrazine was added directly, the hydrazine to metal(s) molar ratio was equal to 20. The reduction of metal ions proceeded quickly and the colour of the liquid changed into black. Then, carbon support (Vulcan XC72 or Sibunit) was introduced into microemulsion and the system was kept under stirring for 1 h. Then, destabilization of microemulsion was carried out by introducing the THF solvent at a very slow rate, 0.25 cm³/min using an automatic syringe pump with the solution vigorously stirred. In the preparation of 1 g of catalyst with 2 wt % metal loading on carbon support it takes ca. 1–1.5 h. Upon addition of THF, the colour of liquid gradually changed from black to grey and finally the liquid phase was colourless, showing complete deposition of metal particles on the support. The obtained catalyst was separated by filtration, dried in air and washed with acetone, then with methanol to remove surfactant and heptane. Finally, the catalyst was washed with water to remove chloride ions and then dried at 393 K in air.

The list of the catalysts used in this work, their metal loadings, Pd–Au alloy composition, metal particle/crystallite size from SEM and XRD, respectively, are reported in Table 2. Characteristics of the Vulcan-supported Pd–Au catalysts were taken from [36].

Reduced (in a flow of 20%H₂/Ar mixture, 50 cm³/min, at 433 K) and post-reaction Pd–Au/Sibunit samples were investigated by X-ray diffractometry (Rigaku Denki, Ni-filtered Cu Kα radiation). The obtained XRD profiles were fitted to an analytical function of the PEARSON-VII type using commercial PEAKFIT software. For phase identification and calculation of the phase composition, centroids of the fitted profiles were used, and for the calculation of the crystallite size, the half-widths of the fitted profiles were taken.

Electron Microscopy (SEM) studies were performed by means of Field Emission Scanning Electron Microscope JEOL JSM-7500 F. Two detectors were used and the images were recorded in two modes. The secondary electron detector provided SEI images, and the backscattered electron detector provided BSE (COMPO) micrographs. To prepare the particle size distribution diagram and estimation of the average particle size the particles (100 or more) were manually counted using micrographs registered in BSE (COMPO) mode. The counting was carried out using several electron micrographs for each catalyst taken at 200,000 × magnification, where metal particles very well contrasted with the support and they were clearly observable. On these micrographs registered in BSE (COMPO) mode the accuracy of the particle size scale was 0.5 nm (the particle size 3 nm corresponds to 1.5 mm on the micrograph).

Temperature-programmed hydride decomposition (TPHD) experiments were performed in a flow system, with data acquisition by a microcomputer. After reduction in 10% H₂/Ar, the catalyst sample (∼0.25 g) was cooled to ∼273 K and subjected to subsequent temperature-programmed study in 10% H₂/Ar flow, ramping the temperature from ∼273 K to 393 K, at 8 K/min. Since the samples had already been reduced, the aim of such experiments was to monitor hydrogen evolution in the process of β-hydride decomposition [41].

TPH-MS (temperature-programmed hydrogenation followed by mass spectrometry) runs with post-reaction catalyst samples were carried out in a flowing 20%H₂/He mixture (25 cm³/min) at a 10 K/min ramp to 1100 K and followed by mass spectrometry (MA200, Dycor-Ametek, Pittsburgh, USA). Principal attention was paid to m/z 36 and 38, which are suggestive of HCl liberation from catalysts used.

2.2 Catalytic hydrodechlorination of tetrachloromethane

The reaction of CCl₄ hydrodechlorination was carried out in the manner previously described [33]. Briefly, prior to the reaction, the catalyst charge (always ∼0.24 g) was dried at 373 K in an argon flow and reduced in flowing 20% H₂/Ar (50 cm³/min), with a temperature ramp from RT to 433 K, and keeping the catalyst at 433 K for 2 h.

The reaction of HdCl of tetrachloromethane, provided from a saturator maintained at 273 K and bubbled in a flow of H₂/Ar mixture (29 cm³/min), was carried out at 423 K and the H₂:CCl₄ ratio ∼14:1. The partial pressures of the reaction mixture were: CCl₄ 4.3 kPa, H₂ 60.5 kPa, Ar 36.5 kPa. Contact time (V/F) ∼0.5 s was applied. The HdCl reaction was followed by gas chromatography (HP 5890 series II with FID and a 5% Fluorcol/Carbopack B 10 ft from Supelco). A typical run lasted ∼20 h. The agreement in final conversion was fair, ±5%. Much better repeatability was found for product distribution, where final selectivities toward major products were in good agreement ±2%. Since all catalyst charges (∼0.24 g) contained nearly identical amounts of palladium, the conversion level of CCl₄ was the first direct measure of the catalytic activity per g_cat or per g_Pd. Very small catalyst grains (Table 1) allow us to neglect the possibility of internal diffusion limitations, which were found to play a role in the HdCl of 1,2-dichloroethane on carbon-supported Pd–Ag catalysts [42]. External diffusion limitations were also disregarded on the basis of additional kinetic experiments, as reported in [33]. Blank kinetic experiments with carbon supports showed very low conversions, ∼0.2%.

3.1 Characterization of carbon-supported Pd–Au catalysts

The description of Pd–Au/Vulcan catalysts in Table 2 comes from [36]. It appeared that the catalysts contained quite small metal particles (crystallites), between 4 and 9 nm. As expected, the preparation method used in that work led to preparation of well-mixed uniform Pd–Au alloy crystallites (EDS and STEM results), characterized by a relatively narrow particle size distribution [36].

X-Ray diffraction, electron microscopy and temperature-programmed hydride decomposition (TPHD) methods were employed to characterize the state of Sibunit carbon-supported Pd–Au alloys. Again, the reverse “water-in-oil” microemulsion method led to preparation of homogenized Pd–Au alloy crystallites characterized by a relatively narrow particle size distribution (Table 1). Fig. 1 shows XRD profiles of reduced (and post-reaction, discussed further in text) samples of Pd–Au/Sibunit catalysts. Maxima of the profiles very well reflect the nominal compositions of Pd–Au alloys (based on the Vegard's law). Symmetrical reflections also indicate good alloy homogeneity and their broadness with application of the Scherrer formula led to calculation of crystallite size which appeared reasonably comparable (with one exception for Pd₉₀Au₁₀/Sibunit) with respective data obtained from electron microscopy (Fig. 2 a–d, and Table 2).

Very good degree of Pd–Au alloying was also confirmed by the results of TPHD experiments, where decomposition of hydride phases was investigated. The well-known propensity of palladium to form β-hydride phases, characterized by respectable H/Pd bulk ratios (typically, up to ∼0.6 [43]) is expected to be gradually reduced by alloying with other elements, e.g. with gold. Luo et al. (Fig. 11 in [44]) measured the isotherms of hydrogen absorption and desorption for Pd–Au alloys at 303 K. Considering the plateau regions of different alloy compositions it is seen that the hydride phase no longer forms for Au contents higher than 15 at%. This is in agreement with the previously constructed phase diagram at 298 K, determined from the lattice parameters of the coexisting dilute (α) and hydride (β) phases [45]. Disappearance of the coexisting phases at 17 at% Au indicated the critical composition for the β-hydride formation at this temperature. Interestingly, the expansion of palladium lattice by alloying with gold results in the decrease of H₂ equilibrium pressure for desorption. This tendency corresponds to a higher stability of PdAuH phases compared to PdH, so under conditions of TPHD runs, when the experiments were carried out at a constant flow of 10% H₂/Ar mixture, PdAuH should be decomposed at higher temperature than PdH. In agreement with this expectation, Fig. 3 shows that indeed the decomposition of hydride phases is gradually reduced with alloying with gold. H/Pd ratio decreases from 0.29 for Pd100 to 0.18 and 0.12, for Pd₉₅Au₅ and Pd₉₀Au₁₀, respectively. Very low amounts of desorbed hydrogen uptake observed for Pd₈₅Au₁₅/Sibunit and Pd₇₀Au₃₀/Sibunit catalysts (∼0.03), are in line with the critical composition for PdAuH formation observed by Flanagan and co. [44,45]. In addition, the decomposition peaks are quite narrow, suggesting that the Pd–Au crystallites are uniform with respect to their bulk composition and particle size [41].

3.2 Performance of carbon-supported Pd–Au catalysts in CCl₄ hydrodechlorination

All catalysts were screened in the reaction of CCl₄ hydrodechlorination at 423 K, i.e. at a temperature higher than that used in our previous studies [33] performed on Sibunit carbon-supported Pd–Au catalysts (363 K). The decision of increasing the reaction temperature arose from previous observation of Russian researchers that the time of stable catalyst performance generally increased as the reaction temperature was increased [7,8,10]. Our preliminary experiments confirm that observation. Kinetic runs performed at 363 K showed much less stable catalytic performance than those carried out at 423 K, so only the latter results shall be presented and discussed.

Fig. 4 shows the time on stream behaviour of all tested carbon-supported Pd–Au catalysts in CCl₄ hydrodechlorination. The overall conversion of CCl₄ on Au₁₀₀/Sibunit was found quite significant, higher than that of Pd₁₀₀/Sibunit. In the light of our previous experience [33] and the behaviour of Au/Vulcan (Fig. 3, inset), this unexpected (although repeatable) result acquires an unambiguous interpretation, supported by additional experimentation. Suspending this problem to future studies, we focus here on the effect of introducing gold to palladium.

Fig. 4 shows that irrespective of the kind of carbon support the performance of Pd₁₀₀ suffers from rapid deactivation, in line with other reports [7–15,33]. Palladium deactivation is paralleled by variations in product distribution. One observes a systematic growth of methane (Fig. 5) and the role of dimeric C₂H_xCl_y products which start to play a dominant role when the catalyst is gradually more deactivated (not presented). A similar situation is observed for the Pd₉₅Au₅/Sibunit (Figs. 4 and 5). Apparently, insufficient amount of added gold leads only to minor changes in the catalytic behaviour of palladium.

On the other hand, the samples of Pd₉₀Au₁₀/Sibunit, Pd₈₅Au₁₅/Sibunit and Pd₉₁Au₉/Vulcan exhibited very high total conversion of CCl₄, i.e. the result, which for nearly identical charges of catalyst (∼0.24 g) and comparable metal dispersions, directly indicates the superiority of these alloy compositions. Stabilized turnover frequencies given in Table 3 confirm a high value of these catalysts. Simultaneously, these catalysts showed the best selectivities towards C₂₊ hydrocarbons, which we regard as the most desired products [33]. Further addition of gold to palladium (to ∼30 at% Au) starts to depreciate the catalyst's performance, as it is seen for Pd₇₀Au₃₀/Sibunit and Pd₇₁Au₂₉/Vulcan. Both the overall conversion level (Fig. 4) as well as the product selectivities are inferior to the ones exhibited by Pd₉₀Au₁₀/Sibunit, Pd₈₅Au₁₅/Sibunit and Pd₉₁Au₉/Vulcan (Figs. 5 and 6).

Thus, we consider that these studies showed that only the introduction of small definite gold amounts (10–15%) to palladium results in a considerable improvement of the catalytic behaviour. Our previous work only marked that the overall activity, catalyst stability and product selectivity of palladium are positively influenced by gold addition [33]. Reflections about the origin of synergistic effects for Pd–Au alloys in CCl₄ HdCl are in the next subsection.

XRD studies of post-reaction catalysts showed a significant phase transformation for nearly all catalysts (excluding Au₁₀₀). A considerable shift of XRD profiles toward lower diffraction angles (Fig. 1) suggests carbon incorporation into Pd/PdAu bulk. This effect was observed (and widely discussed) for Pd₁₀₀/Sibunit catalysts subjected to CCl₄ hydrodechlorination at 363 K [33], but not noticed for Pd–Au/Sibunit catalysts. Evidently, bare C₁ species (from the CCl₄ molecule after stripping off all chlorine atoms) enter Pd bulk before being hydrogenated to methane. The increase of methane selectivity (Fig. 5) during the reaction suggests that palladium sink becomes gradually more “saturated”.

Essentially identical positions of (111) fcc reflections for post-reaction samples of all Pd and Pd–Au catalysts at 2∼38.9° suggest the presence of a PdC_0.13(0.15) phase [46,47]. This, in turn, suggests considerable phase changes. If palladium escapes from the alloy to form a carbide phase, then the rest of bimetallic material should become richer in gold. Since the PdC_0.13 phase does not seem to be active in this reaction [33], then the resulting, Au-enriched, Pd–Au alloy must constitute a catalytically active phase. This hypothesis needs substantiation by additional HRTEM/EDS studies.

Finally, Fig. 7 shows the results of temperature-programmed hydrogenation (TPH-MS) of all catalysts subjected to a prolonged CCl₄ reaction (ca. 20 h on stream). Large amounts of chlorine desorbed from Pd₁₀₀ and Pd₉₅Au₅ catalysts indicate that, apart from carbonaceous coke, retained chloride species is also a possible cause of the catalyst's deactivation. It is in line with other results [48–51].

3.3 Synergistic effects in CC₄ hydrodechlorination on Pd–Au catalysts

Fig. 8 shows the relation of the turnover frequency with Pd–Au alloy composition, calculated on the basis of stabilized conversion, i.e. at the end of catalytic runs (all runs with 0.24 g samples). The overall activity pattern obtained with two series of carbon-supported Pd–Au catalysts shows a maximum at ∼10 at% Au. The fact that the results of both series of Pd–Au/carbon catalysts, characterized by so different porous structures, form a kind of an universal relationship is in line with our conclusion (in §Experimental) about the absence of internal diffusion limitations. Present results also confirm our previous results obtained on Pd–Au/Sibunit catalysts prepared by reductive deposition [33], where some bimetallic catalysts were more active, stable and selective (towards hydrocarbon products) than the monometallic Pd/Sibunit catalyst. It should be admitted that a beneficial HdCl behaviour of gold introduction to palladium was earlier reported [32].

Synergistic effects observed for Pd–Au alloys are interpreted by the occurrence of two effects: ensemble (geometric) and ligand (electronic) ones [52]. The ensemble effect is a dilution of surface Pd by Au, leading to the disappearance of contiguous Pd site ensembles and the creation of isolated Pd sites. The maximum in activity at about 10 at% Au suggests that the geometric effect does not operate here, unless we are dealing with a significant gold segregation to the alloy surface. XPS studies were not undertaken in this work, however XPS studies of Pd–Au/Vulcan catalysts showed no special surface enrichment with gold [36]. The ligand effects occurring via direct charge transfer or by affecting bond lengths cause the Pd d band to be more filled, moving the d-band centre away from the Fermi level. This makes Pd more “atomic like” which binds reactants and products more weakly. It would relieve catalyst deactivation caused by self-poisoning and enhances the activity/selectivity. This effect would explain the better resistance to deactivation of Pd–Au alloys. Improvement in deactivation resistance to chloride species in HdCl (of mainly trichloroethene) reactions by for gold-supported palladium nanolayers was observed in numerous papers from the group of Wong [30,53–59], where both ensemble and ligand effects were invoked to describe the superiority of Pd–Au combinations.

Very small amounts of gold introduced to palladium make a pronounced and beneficial difference in the catalytic behaviour. Synergistic effects observed for 10–15 at% Au in the overall activity (Fig. 8), stability (Fig. 4) and selectivity to C₂₊ hydrocarbons (Fig. 6) suggest that a larger group of Pd surface atoms must be electronically modified by single Au atoms. However, such modification should not be too deep as a gold-rich surface would be inactive in providing active (dissociated) hydrogen needed for hydrodechlorination and removal of chloro-carbonic residues.

Our recent XPS data for Pd–Au/Vulcan catalysts showed the binding energy of Pd 3d_5/2 within the range of 335–337 eV [table 2 in [36]). The monometallic Pd/Vulcan showed the Pd state of binding energy 335.3 eV corresponding to the Pd-metal. In the spectra of all Au-containing catalysts (starting from Pd₉₁Au₉/Vulcan), the binding energy of Pd-metal peak component was slightly shifted but also a less intense peak component at slightly higher energy of ca. 336–337 eV appeared. Since the Pd binding energy values slightly above 336 eV are commonly ascribed to the partially oxidized Pd^σ+ species [60], therefore these XPS data could indicate a Pd–Au electronic interaction. A simplified picture of the Pd₆Au₁ ensemble is proposed in Fig. 9. The large size of the molecule of CCl₄ (kinetic diameter ∼0.6 nm) implies its interaction with such a large surface ensemble.