Novel Fe-Pd/SiO₂ catalytic materials for degradation of chlorinated organic compounds in water

Preparation of the supported metal nanoparticles

The Fe-Pd nanoparticles were synthesized inside the porous silica support. The granulated commercial silica carriers (Russian Federation) with a specific surface area of 300 m² g^–1 (HS-SiO₂) and 30 m² g^–1 (LS-SiO₂) were used as supports. The ammonium trioxalatoferrate trihydrate (NH₄)₃ Fe(C₂O₄)₃·3H₂O (98 %, Acrus Organics), tetraaminepalladium (II) chloride monohydrate [Pd(NH₃)₄]Cl₂·H₂O (41.42 % Pd; Aurate, Russia) and/or palladium (II) acetate trimer (Alfa Aesar) were used as precursors. The supported nanoparticles were prepared by direct deposition of the Fe and Pd precursors on the surface followed by the reduction to zero-valent metals. Fe and Pd loadings varied from 4.0 to 7.8 wt. % and from 0.7 to 4.4 wt. %, respectively. The metal precursors were deposited by the incipient wetness impregnation method simultaneously (SI procedure) or consecutively (CI procedure), unless indicated otherwise. The multiple impregnations with intermediate vacuum drying (MI procedure) were performed when Pd acetate was used as a precursor due to its insolubility in water and poor solubility in acetone. After impregnation, a sample was either dried at 60 °C to remove excess moisture or calcined in air to decompose the supported precursors (procedures SIC or CIC). The reference Fe-Pd/SiO₂ samples were prepared according to the known procedures in the presence of PdCl₂ and FeCl₃ as the precursors [21, 22] (procedures CIR and CICR). The dried or calcined sample was reduced in hydrogen flow and cooled to a room temperature. The sample was transferred into a glass vial and stored under an organic solvent to prevent oxidation of the supported particles. The prepared materials were denoted as xPd-yFe-A, where x and y referred to the metal loading (wt. %) in a reduced sample and A was the abbreviation of the preparation procedure.

Characterization of the samples

Thermal analysis of dried samples was performed by the thermogravimetric differential thermal analysis (TG-DTA) method using a thermoanalytical Derivatograph-C instrument (MOM Company). The sample (15 mg of the bulk or 20 mg of the supported sample) was placed in an alund crucible and heated in air from 20 to 500 °C at a heating rate 10 °C min^–1. The temperature programmed reduction (TPR) measurements were performed in a laboratory-constructed flow system described in [23] details. The sample (150 mg) was pretreated in Ar flow at 150 °C for 30 min and was cooled in Ar to –50 °C prior the TPR analyses. Heating from –50 to 400 °C was carried out in a 4.6 % H₂ in Ar flow (30 mL/min) at a heating rate of 10 °C min^–1. After cooling down to room temperature, a sample was subjected to oxidizing environment (5 % O₂/He flow; 40 mL/min) for 1 h (the molar ratio O₂ /Fe of 24) and kept overnight at the molar ratio of O₂/Fe of 0.2. The re-oxidized sample was reduced at 400 °C and kept at this temperature until hydrogen consumption ceased. The X-ray diffraction patterns were recorded using a DRON-2 diffractometer with Ni-filtered Cu Kα radiation (λ=0.1542 nm) in a step scanning mode, with a step of 0.02 ° and a counting time of 0.6 s per step in the range 2θ=10 °–60 °. Prior the analysis, samples were homogenized under toluene. To ensure that the sample is not oxidized during XRD analysis, the pattern was first recorded in the range 2θ=40 °–50 ° and then in the range 2θ=10 °–60 °. The major reflection of Fe⁰ was compared on both patterns. The crystal size was calculated from X-ray line broadening analysis. X-ray photoelectron spectra were obtained using an XSAM-800 spectrometer [24]. X-ray absorption spectra (XANES and EXAFS) measurements were carried out with the Hasylab X1 beamline (Hamburg) and the detailed description of relevant procedures could be found elsewhere [24]. Pd K-edge measurements were carried out on the same beamline (at 24 350 eV) in the transmission mode using a double-crystal Si(311) monochromator that was detuned to 50 % of the maximum intensity to exclude higher harmonics in the X-ray beam.

Testing the materials in 2-ClBP and PCE degradation

Degradation of 2-ClBP in the presence of as-prepared catalysts was performed as follows. The sealed glass bottle containing distilled water (32 mL) and a nanocatalyst (0.60 g) was prepared with a minimal headspace. To avoid an inside diffusion control during the activity test, catalyst granules were crushed in an agate mortar to the size below 0.1 mm. The stock solution of 2-ClBP (36 μL; 4.15 g·L^–1) prepared in ethanol was injected by a syringe in the bottle. The 2-ClBP content in the bottle was 0.15 mg, which corresponds to the 2-ClBP concentration in water 4.67 mg·L^–1. The blank bottle without any material was prepared as well. Two blank bottles and two parallel reaction mixtures were prepared and tested. All bottles were immediately mixed end-over-end at 350 rpm and room temperature (19–23 °C). In order to compare the catalytic activity of the materials, the reaction time of 24, 72 and 168 h were selected, because of the strong difference in the catalytic activity of the samples.

A solvent extraction method with hexane (Merck, for HPLC ≥ 99.9 %) was applied to analyze the concentrations of organics in a reaction mixture and in a blank solution. The efficiency of extraction was found to vary from 88.0 to 99.9 %. The extracts from a 4 mL aliquot of a reaction solution (Probe 1) and from the bottle with a reaction mixture (Probe 2) were analyzed with GC (chromatograph KrystaLuxe-4000-M equipped with a 30 m capillary column OV-1 and a flame ionization detector) by the internal standard method with n-C₁₆ H₃₄ as a standard. The injector and detector temperature was 210 °C, and the thermostat temperature was set at 160 °C. The ratio of the peak areas 2-ClBP/C₁₆ H₃₄ or biphenyl (BP)/C₁₆ H₃₄ in the blank probe was considered as 100 %. The amount of 2-ClBP and BP in the reaction solution (Probe 1) or in the reaction mixture (Probe 2) was calculated as its relative portion. The 2-ClBP degradation efficiency was calculated based on analysis of Probe 1 and Probe 2, respectively. The conversion to BP was calculated as a molar ratio of the BP amount in the bottle after testing (calculated from analytical data for Probes 1–2) and the 2-ClBP amount in the initial blank solution.

The potential loss in catalysts reactivity was assessed in terms of 2-ClBP degradation during 1 and 3 days of experiments. After the test was over, the sample was separated from the reaction mixture by filtering a solution with a syringe, and a new portion of the reaction mixture was added to the sample.

The procedure used for testing the materials in PCE degradation was described elsewhere [24].

Phase composition of silica-supported nanoparticles

To evaluate the formation of Fe and Pd species, samples were characterized by TG-DTA, TPR-H₂, XRD (XANES + EXAFS) and X-ray photoelectron spectroscopy (XPS). According to TG-DTA, decomposition of the silica-supported (NH₄)₃ Fe(C₂O₄)₃·3H₂O proceeded in several steps, as can be seen from the derivative thermogravimetry (DTG) and especially DTA curve, which contained several distinct maxima (Fig. 1). The position of the most intensive DTG maximum differed from that of the initial compound. There were two exothermic peaks on the DTA curve observed in the temperature range of the DTG maximum, instead of the expected one peak based on the decomposition of the initial compound and the precursor supported on nonporous titania [23]. The intensity of these two peaks was lower than the intensity of the expected peak, which suggested the lack of oxygen in the pore volume inside the supported sample, which resulted in superposition of two decomposition pathways: exothermic one in an oxidative atmosphere [23, 25, 26] and endothermic one in an inert atmosphere [25]. Despite this variation, the mass loss was almost completed below 260 °C, which was similar to the bulk phase.

Fig. 1

DTG and DTA curves of the supported sample Pd(NH₃)₄ Cl₂/(NH₄)₃ Fe(C₂O₄)₃/SiO₂ (2) as compared with (NH₄)₃ Fe(C₂O₄)₃·3H₂O/SiO₂ (3) and [Pd(NH₃)₄]Cl₂·H₂O/SiO₂ (1).

The DTG-DTA curves for both, the free-standing and simultaneously supported Fe-Pd precursors exhibited no major features of Pd(NH₃)₄ Cl₂·H₂O decomposition indicating the absence of the phase with such composition in the samples. The DTG and DTA peaks were shifted to higher temperatures for the Fe-Pd sample as compared with the initial (NH₄)₃ Fe(C₂O₄)₃·3H₂O phase (Fig. 1).

This suggested the incorporation of [Pd(NH₃)₄]²⁺ ions into the trioxalatoferrate lattice during crystallization of the solids from the impregnating solution, which favored the formation of alloys during the reduction of the samples in hydrogen. On the contrary, there was an additional peak observed in the DTG curve (Fig. 1) of the supported Fe-Pd sample prepared by the two-step impregnation of HS-silica (in the presence of saturated ammonium trioxalatoferrate and saturated [Pd(NH₃)₄]Cl₂ solution). The position of this peak was close to the position of the major peak representing the decomposition of [Pd(NH₃)₄]Cl₂·H₂O supported on the same HS-silica, which could be attributed to the endothermic loss of ammonia and chlorine. Decomposition of silica-supported ammonium trioxalatoferrate in air occurred below 260 °C and introduction of Pd salt did not affect the decomposition. These results prove independent precipitation of the Fe and Pd precursors during drying at the two-step impregnation procedure, which could result in two separate metal phases in reduced samples. According to the aforementioned findings, temperature range from 240 to 260 °C resulted in the decomposition of a Fe precursor to oxide species before the actual reduction of the samples. Thus, to avoid the formation of iron oxide on silica support, samples were calcined at 240–260 °C for 4 h.

The reducibility of the calcined samples was studied by the TPR method, which was a useful tool to determine the temperature of reduction of the supported phases, as well as to differentiate various species. The TPR profiles of the samples are shown in Fig. 2.

Fig. 2

The TPR curves of the sample 3.0Pd-7.6Fe-SI after storage for 10 months (1) and the calcined samples Fe-O/SiO₂ (2); 3.0Pd-7.6Fe-SIC (3); 3.2Pd-7.8Fe-CIC (4). The dashed lines correspond to the samples re-oxidized in the TPR system.

The reduction curve of Fe-O/SiO₂ exhibited a broadened double peak below 400 °C with a maximum at 395 °C, and the hydrogen consumption at 400 °C proceeded for about 40 min. The ratio of the total hydrogen uptake to the Fe amount in the reduced probe (Fig. 3) corresponded to the reduction of Fe³⁺ to Fe²⁺ and most likely Fe⁰. Introduction of Pd as a tetrammine complex into the supported ammonium trioxalatoferrate resulted in increased hydrogen consumption (Fig. 3) and resulted in a low-temperature reduction below the reduction temperature of supported FeO_x species (Fig. 2).

Fig. 3

The H₂ consumption of the calcined samples and the samples after storage for 10 months (3.0Pd-7.6Fe-SI-10m).

The TPR profiles of the samples significantly varied from that of FeO_x species and highly depended on the support nature. The reduction curve at 70–300 °C with several maxima corresponded to other phases containing Feⁿ⁺. The ratio of the heights of these maxima was different for the samples prepared on HS- and LS-silica supports (Fig. 2, samples 3.0Pd-7.6Fe-SIC and 3.2Pd-7.8Fe-CIC). The TPR curve of the sample on the HS-silica support exhibited the additional hydrogen consumption peak at 50–100 °C as compared with the sample on the LS-silica support. Similar low-temperature peaks have been reported in literature [18–20, 22] for the silica-supported PdO-Fe₂O₃ samples with different Pd:Fe ratios, and these peaks were attributed to the different species, which reduction resulted in the Pd- or Fe-rich alloys and Fe⁰ even at 250 °C. Importantly, the higher the Pd content, the lower temperature of the maximum was obtained. The peak with a maximum at 100 °C was observed earlier for the Fe-Pd/MgO/SiO₂ sample calcined at 300 °C at x(Pd)=0.66 [18]. This peak has been attributed to the reduction of Fe oxide facilitated by Pd and resulted in amorphous alloys of the composition Pd_x Fe_1–x with 0.67 < x < 0.80 after reduction in hydrogen at 400 °C for 1 h. The dominating peak with a maximum at 135 °C was observed for Fe-Pd/SiO₂ at x(Pd) = 0.51–0.68 [22] and was ascribed to the Pd-Fe-O phase reduction, which after reduction in hydrogen at 250–450 °C for 1 h, resulted in the XRD detectable (average crystallite size 5–7 nm) alloy particles with Pd_0.87 Fe_0.13 composition [21, 22]. The average particle size decreased with the Fe loading in the sample, and phase did not appear on XRD patterns at x(Pd) ≤ 0.51.

Considering the fact that after calcination at 240–260 °C, the samples may still contain Pd(NH₃)₄ Cl₂, hydrogen consumption below 50 °C could be attributed to the reduction of Pd²⁺ to Pd⁰ and adsorption of hydrogen on the formed metallic Pd. The peak of hydrogen release at 50–70 °C due to the decomposition of the β-PdH phase supported this hypothesis. Indeed, there was no similar negative peak on the TPR curve of the calcined samples, which confirmed the absence of Pd⁰ particles and the formation of the Pd-Fe alloy particles suppressing the β-PdH phase formation as observed by other researchers [18, 20–22]. All these findings suggested the occurrence of a strong interaction between the Pd and Fe precursors and the products of their decomposition during preparation of the samples using (NH₄)₃[Fe(C₂O₄)₃] and [Pd(NH₃)₄]Cl₂.

Hydrogen consumption (Fig. 3) suggested that reduction of Feⁿ⁺ to Fe⁰ might be incomplete at the temperatures used, although the hydrogen consumption ceased. It is generally accepted that the duration of isothermal hydrogen consumption depends on the nature of the support and it is less or equal to 3 h. However, the reduction lasted 10 times longer than it was observed for pure magnetite [27]. The decreased reducibility of the silica-supported iron oxides indicated the interaction of Fe²⁺ ions with the silica surface during the reduction process.

Temperature programmed reduction (TPR) with hydrogen has been widely used to determine the oxygen content in iron nanoparticles. In the present study, TPR runs with intermediate exposition to oxygen were used, enabling the estimation of reduced species syability. After oxidation of the reduced Fe/SiO₂ sample in oxygen, the reduction of the formed Feⁿ⁺-O species started at the same temperatures as the reduction of the initial calcined sample (Fig. 2). The total hydrogen consumption was the same as for the initial sample (Fig. 3), indicating the complete re-oxidation of the reduction product, which was in agreement with the results presented in [17]. Such a redox behavior could be expected for Fe⁰ rather than for Fe-O species and could be explained by the reduction-oxidation only of the surface layers of iron oxide particles. Quite a different redox behavior was observed for the Fe-Pd samples. The intensity of the TPR profiles of the re-oxidized Fe-Pd samples decreased in comparison with the calcined samples (Fig. 2) and the total hydrogen uptake was considerably lower (Fig. 3). Thus, the addition of Pd(NH₃)₄ Cl₂ to (NH₄)₃[Fe(C₂O₄)₃] resulted in the reduced species with noticeably enhanced stability to oxidation with O₂. Whereas the reduced product of thermal decomposition of the silica-supported (NH₄)₃[Fe(C₂O₄)₃] phase was unstable to oxidation with oxygen at room temperature.

The sequence of thermal treatment steps during preparation of the Fe-Pd sample affected the redox properties of the supported components. The TPR profiles of the sample prepared by the direct reduction after drying and then slowly oxidized during 10 month storage were visually different from these samples that were preliminarily calcined before reduction and subsequently re-oxidized at room temperature (Fig. 2). The observed peak of hydrogen release below 0 °C indicated the formation of Pd-rich crystallites in the sample reduced after drying [21, 22]. There was no low-temperature peaks of hydrogen consumption observed, which were typical for Pd-Fe-O species. The obtained hydrogen consumption revealed that a long-term storage resulted in a partial (35 %) oxidation of the initial reduced sample. The easily re-oxidized portion estimated as the ratio of the hydrogen uptake after re-oxidation in the TPR system to that of the initial sample was about 17 %, whereas it reached 28 % for the sample prepared by the reduction of the calcined sample.

According to XRD results, samples with the same chemical composition however prepared by various methods varied in the phase composition and dispersion of the supported active components. In the samples prepared from [Pd(NH₃)₄]Cl₂ at the Pd loading below 2 wt. % or using Pd acetate at the Pd loading below 4 wt. %, the XRD analysis revealed no supported phases. This could be due to a small size of Pd⁰ or Fe⁰ crystallites distributed on the surface of silica, as well as due to the strong Pd interaction with Fe resulting in the formation of X-ray amorphous alloys as it was observed previously [21, 22]. The presence of highly dispersed Pd species in the samples was subsequently confirmed with X-ray absorption spectroscopy (XAS). The selected XRD patterns of calcined and reduced samples are shown in Fig. 4 and the calculated average particle size is listed in Table 1.

Fig. 4

XRD patterns of the samples after intermediate calcination at 240–260 °C (2 – 3.0Pd-7.6Fe-SIC; 3 – 3.2Pd-7.8Fe-CIC) and the reduced samples: 1 – 3.1Pd-I; 4 – 3.2Pd-7.8Fe-CIC; 5 – 3.0Pd-7.6Fe-SIC; 6 – 3.2Pd-7.8Fe-CI; 7 – 3.0Pd-7.6Fe-SI; 8 – 4.3Pd-6.8Fe-MI; 9 – 2.1Pd-4.4Fe-CIR; 10 – 2.1Pd-7Fe-CICR.

The position of the symmetric reflection of the XRD pattern (4.3Pd-6.8Fe-MI; Pd(111) indicated the formation of metallic Pd⁰ nanoparticles in the sample prepared by the sequential deposition of Pd acetate on the silica-supported (NH₄)₃[Fe(C₂O₄)₃] and poor incorporation of Fe atoms into the Pd crystal structure. The ammonium trioxalatoferrate is insoluble in acetone, therefore Pd acetate could only be adsorbed or deposited as separate particles on the surface employing this particular approach. The adsorbed Pd species can stabilize the fine metallic iron particles with the size below the XRD detection limit (6 nm [17]). The absence of any lines of the supported phases on the XRD patterns prepared with the Pd loading below 4 wt. % confirmed the formation of such particles.

The significant separation of metals was observed during calcination of the dried samples. The XRD analysis revealed the formation of the Pd⁰, α-Fe₂O₃, Fe₃O₄ phases in the calcined samples, as well as the large α-Fe, Pd and Pd_x Fe_1-x crystallites after the reduction (Fig. 4, samples 3.0Pd-7.6Fe-SIC, 3.2Pd-7.8Fe-CIC).

The XRD patterns of samples with the same chemical composition, but reduced after drying, exhibited only a weak reflection in the region of the Pd(111) (Fig. 4, samples 3.0Pd-7.6Fe-SI, 3.2Pd-7.8Fe-CI). On the XRD pattern of the reduced sample 3.0Pd-7.6Fe-SIC, the Pd(111) reflection was symmetric, yet the maximum was shifted towards the high angles by 0.2 ° (Fig. 4). This indicated the formation of a solid solution based on the metallic Pd phase. The XRD patterns of the reduced samples 3.0Pd-7.6Fe-SI, 3.2Pd-7.8Fe-CI, 3.2Pd-7.8Fe-CIC, 2.1Pd-4.4Fe-CIR (Fig. 4) showed high asymmetry of the Pd (111) reflections indicating the presence of at least two fcc phases: Pd⁰ and Pd_x Fe_1–x alloy, similar to previously observed for the samples prepared using PdCl₂ and FeCl₃ [21]. Furthermore, samples prepared using aforementioned precursors exhibited the largest shift (about 0.9 °) of the Pd(111) reflection among all the samples prepared in the current study (Fig. 4, samples 2.1Pd-4.4Fe-CIR, 2.1Pd-7.0Fe-CICR) or mentioned in literature. The significant shift of the maximum for the Pd(111) reflection confirmed that the solubility in water of both (NH₄)₃[Fe(C₂O₄)₃] and [Pd(NH₃)₄]Cl₂ favored coprecipitation of both metal ions during drying of the samples after simultaneous or two-step impregnation.

The major line of Fe⁰ was observed only in the XRD patterns of the samples 3.0Pd-7.6Fe-SIC, 3.2Pd-7.8Fe-CIC, 2.1Pd-7.0Fe-CICR, which were calcined before reduction, and the sample 2.1Pd-4.4Fe-CIR prepared using FeCl₃ and PdCl₂. In other samples, Fe⁰ was present as a part of a Pd alloy, as well as the monometallic and bimetallic X-ray amorphous species. The existence of the later was confirmed by XRD analysis of the samples annealed in a purified Ar flow at 550 °C after reduction in hydrogen. The appearance of an intense Fe⁰ line and the shift of the Pd⁰ line on the XRD pattern of the sample after annealing in purified Ar (Fig. 5) indicated the reduction to Fe⁰ and formation of a disordered Fe-Pd phase in the sample reduced at 400 °C. The spacing distance d/n of the Fe⁰ line corresponded to the major reflection of α-Fe. Incorporation of Pd into metallic iron crystallites was very weak, and Fe crystallites were easily oxidized when exposed to air. This was confirmed by disappearance of the Fe⁰ line on XRD patterns for the annealed samples after exposing to air for 2 weeks (Fig. 5).

Fig. 5

XRD patterns of the sample on LS-SiO₂ (3.2Pd-7.8Fe-CI): 1 – reduced in hydrogen at 400 °C; 2 – annealed in purified argon at 550 °C after reduction in hydrogen; 3, 4 – after storage in air for 2 weeks (3 – annealed in argon, 4 – reduced in hydrogen).

X-ray absorption spectroscopy (XAS) was employed to study the bimetallic samples on the HS-silica support, since XRD analysis of these samples revealed no Fe and only some Pd phases. Two samples (1.7Pd-6.7Fe-SI and 3.0Pd-7.6Fe-SI) were studied after reduction at 400 °C followed by storage for 10 months under an organic solvent. The sample 1.7Pd-6.7Fe-SI was also studied after regeneration in hydrogen of the sample probe deactivated during the dechlorination reaction.

Pd K-edge

The Pd K-edge XANES spectra of the Fe-Pd samples and the standard reference compounds Pd-foil (Pd⁰) and PdO (Pd²⁺) are shown in Fig. 6.

Fig. 6

Pd K-edge XANES spectra of Fe-Pd containing samples and reference compounds: 1 – Pd-foil; 2 – PdO; 3 – 1.7Pd-6.7Fe-SI; 4 – 1.7Pd-6.7Fe-SI regenerated; 5 – 3.0Pd-7.6Fe-SI.

As can be seen from Fig. 6, the energetic position of the Pd K-edge in the spectra of all samples under study was lower than that observed for PdO and close to that observed in Pd-foil. However, the intensity of the oscillations was lower than that of the Pd-foil spectrum. Data suggest that the electronic state of palladium species in the samples was the same as in Pd-foil, and Pd metal particles were small. The radial distribution functions (without the phase and amplitude corrections) obtained by the Fourier transform (FT) of the oscillating portion of the spectra (EXAFS) are presented in Fig. 7.

Fig. 7

Fourier transforms (FT) of Pd K-edge EXAFS spectra of Fe-Pd containing samples and reference compounds: 1 – Pd-foil; 2 – PdO; 3 – 1.7Pd-6.7Fe-SI; 4 – 1.7Pd-6.7Fe-SI regenerated; 5 – 3.0Pd-7.6Fe-SI (Å).

In the spectrum of Pd foil, the first peak corresponded to the first coordination shell containing 12 Pd atoms at a 2.75 Å real distance from the central Pd atom. In the spectrum of PdO, the first peak corresponded to the first coordination shell containing 4 O atoms at 2.02 Å real distances from the central Pd atom. The second and third peaks corresponded to the second and third shells containing Pd atoms. The first peak of the Pd K-edge EXAFS spectra of all the samples was fitted in both r- and k-spaces with a one-shell model – a palladium shell around the central absorbing palladium atom. The results of the model fit of the Pd K-edge EXAFS spectra of the samples are given in Table 2.

Table 2

Parameters of the best model fit of Pd K EXAFS and Fe K EXAFS spectra.

Sample	Path	r (Å)	CN	σ2 × 10–3 (Å2)	ΔE (eV)	DPd (Å)
1.7Pd-6.7Fe-SI	Pd-Pd	2.74 ± 0.06	5.6 ± 0.8	3 ± 1	5 ± 1	7.4 ± 1.1
	Fe-O	1.89 ± 0.01	0.8 ± 0.1	5 ± 1	–1 ± 1
	Fe-O	2.43 ± 0.01	3.2 ± 0.2	11 ± 1	–2 ± 1
	Fe-Fe	2.89 ± 0.01	0.8 ± 0.1	8 ± 1	–7 ± 1
1.7Pd-6.7Fe-SI regenerated	Pd-Pd	2.76 ± 0.01	6.7 ± 0.7	6 ± 1	5 ± 1	9.0 ± 1.5
	Fe-O	1.79 ± 0.03	0.4 ± 0.1	1 ± 1	–8 ± 1
	Fe-O	2.41 ± 0.01	3.3 ± 0.3	4 ± 1	–6 ± 1
	Fe-Fe	2.77 ± 0.01	1.1 ± 0.2	3 ± 1	–6 ± 1
3.0Pd-7.6Fe-SI	Pd-Pd	2.74 ± 0.01	7.8 ± 0.9	4 ± 1	5 ± 1	11.6 ± 3.2
	Fe-O	1.89 ± 0.01	0.5 ± 0.1	12 ± 1	–1 ± 1
	Fe-O	2.47 ± 0.01	5.0 ± 0.1	11 ± 1	–1 ± 1
	Fe-Fe	2.90 ± 0.01	0.9 ± 0.1	9 ± 1	–8 ± 1

Based on these results, the Pd particle size in the sample 1.7Pd-6.7Fe-SI was estimated below 1 nm (Table 2). This explained the absence of a Pd reflection on the XRD pattern. The larger average palladium metal particle size was calculated for the sample 3.0Pd-7.6Fe-SI. Moreover, the presence of the long distance peaks in the radial distribution spectra of this sample indicated that even the larger particles were available in the sample, which was in agreement with XRD results. It is evident from Fig. 7 and data presented in Table 2 that the local structure of Pd species in the regenerated sample 1.7Pd-6.7Fe-SI differed from that of the initial sample. In the regenerated sample, the distance between Pd atoms and the average CN in the first shell were higher.

Fe K-edge

Figure 8 shows the Fe K-edge XANES spectra of the samples and standard reference compounds.

Fig. 8

Fe K-edge XANES spectra of Fe-containing samples and the reference compounds: 1 – Fe-foil; 2 – Fe₃O₄; 3 – Fe₂O₃; 4 – 1.7Pd-6.7Fe-SI; 5 – 1.7Pd-6.7Fe-SI regenerated; 6 – 3.0Pd-7.6Fe-SI.

As it could be seen from Fig. 8a, the energetic position of Fe K-edge X-ray absorption in the spectra of all samples was higher than that observed for Fe-foil, lower than that observed for Fe₂O₃ and close to that observed in Fe₃O₄. At the same time, the shape of the absorption spectra and the intensity of the white line in the spectra of all the samples (Fig. 8b) was lower than that observed in the spectra of Fe₂O₃ and Fe₃O₄, and higher than that observed in the Fe-foil spectrum. Data indicated that the electronic state of iron species in Fe-containing samples was different from that observed in Fe₂O₃, Fe₃O₄, and Fe-foil; though the iron species existed in the samples in the reduced and oxidized state, most likely as a mixture of Fe⁰, Fe²⁺ and Fe³⁺. The FT patterns of EXAFS oscillations of the iron-containing samples and Fe reference compounds are presented in Fig. 9. The first peak in the Fe-foil spectrum represented the Fe-Fe atomic pair with a coordination number of 8 at 2.499 Å real distances. The first peak in the Fe₃O₄ spectrum represented the Fe-O atomic pair with a coordination number of 6 at 2.060 Å real distances. The wide first peak in the Fe₂O₃ spectrum represented two O shells containing three atoms at a 1.946 Å real distance and other 3 atoms – at 2.116 Å real distances. As could be seen in Fig. 9, the local structure of iron species in the Fe-containing samples varied in comparison to Fe₂O₃, Fe₃O₄, and Fe-foil. The first peaks of the Fe K-edge EXAFS spectra of all samples were fitted in both r- and k-spaces with a three-shell model – the first one and the second one: an oxygen shell around the central absorbing iron atom, and the third one: an iron shell around the central absorbing iron atom. The results of the model fit of the Fe K-edge EXAFS spectra of all Fe-containing samples under study are given in Table 2.

Fig. 9

Fourier transformed Fe K EXAFS spectra of Fe-containing samples and reference compounds: 1 – Fe-foil; 2 – Fe₃O₄; 3 – Fe₂O₃; 4 – 1.7Pd-6.7Fe-SI; 5 – 1.7Pd-6.7Fe-SI regenerated; 6 – 3.0Pd-7.6Fe-SI.

The analysis of the EXAFS oscillations showed that the local structure of iron species could be presented as follows: the nearest neighbors of the central Fe atom were O atoms with an average coordination number (CN) 0.4–0.8 at 1.79–1.91 Å real distances for all the samples in the current study. Next neighbors were O atoms with average coordination numbers varying from 3.2 to 5.0 at 2.41–2.47 Å real distances. The third neighbors were Fe atoms with average coordination numbers 0.8–1.6 at 2.77–2.90 Å real distances. In the spent and regenerated samples, the distance between Fe and first O atoms was shorter (1.79 vs. 1.89 Å) and the average CN in the first shell was two-fold lower (0.4 vs. 0.8). Also, the real distance between two Fe atoms was shorter as well (2.77 vs. 2.89), whereas the average Fe-Fe CN remained roughly the same. Based on the data, regeneration resulted in the partial reduction of oxidized iron species.

Considering an increased stability to oxidation and long-term removal efficiency, the samples 3.0Pd-7.6Fe-SIC and 3.2Pd-7.8Fe-CIC tested against dechlorination for 4 months, were studied with XRD and XPS. Thus, XPS survey spectra of the samples under study are shown in Fig. 10.

Fig. 10

XPS survey spectra of 3.0Pd-7.6Fe-SIC (1) and 3.2Pd-7.8Fe-CIC (2) samples.

In all the survey spectra, the C 1s, O 1s, O 2s, Si 2p, Si 2s, Pd 3d electron lines and the C KLL and O KVV Auger-electron lines were observed. The photoelectron line Fe 2p was observed in the survey spectrum of sample 3.2Pd-7.8Fe-CIC. Iron in a trace amounts was found in the surface and subsurface layers of sample 3.0Pd-7.6Fe-SIC. XPS spectra of Pd 3d and Fe 2p electron regions are presented in Fig. 11.

Fig. 11

XPS Pd 3d region (a) and Fe 2p region (b) of 3.0Pd-7.6Fe-SIC (1) and 3.2Pd-7.8Fe-CIC (2) samples.

The binding energy of XPS electronic lines and the surface atomic ratios are given in Table 3.

Table 3

XPS binding energy and surface atomic ratio.

Sample	Binding energy (eV)				Atomic ratio
	O 1s	Si 2p	Fe 2p3/2	Pd 3d5/2	Pd/Si	Fe/Si	Fe/Pd
3.0Pd-7.6Fe-SIC	530.0	103.7	–	335.7	0.0072	–	–
3.2Pd-7.8Fe-CICa	530.1	103.7	711.9	335.8	0.01	0.02	2.0

^aThe sample prepared on the LS-SiO₂ support.

The XPS data obtained for the sample 3.2Pd-7.8Fe-CIC indicated the formation of iron oxide species on the surface. The binding energies Fe 2p_3/2 = 711.9 eV and Fe 2p_1/2 724.9 eV were attributed to Fe³⁺ ions in Fe₃O₄, α-Fe₂O₃ or α-FeO(OH) [28], and no photoelectron peak of Fe⁰ at 707 eV [29] was observed. The binding energies of the Pd 3d_5/2 line found in both 3.0Pd-7.6Fe-SIC and 3.2Pd-7.8Fe-CIC samples (335.7 and 335.8 eV, respectively) were characteristic of small Pd⁰ particles (Fig. 11a, Table 3). The Pd loading was 1.07 times higher in the sample 3.2Pd-7.8Fe-CIC than in 3.0Pd-7.6Fe-SIC, yet the Pd/Si atomic ratio in the surface layers of this sample appeared to be higher by a factor of ca. 1.3. The Pd/Fe atomic ratio in the surface layers of the sample 3.2Pd-7.8Fe-CIC was 0.50 in comparison to 0.21 in the bulk sample. Thus, there was nearly two-fold higher enrichment of the surface layers with palladium, especially for the sample 3.0Pd-7.6Fe-SIC. The obtained results were quite opposite to the data presented in the literature for deactivated Pd/Fe nanoparticles [30]. According to Yan and colleagues [30], the Pd 3d_5/2 signals in XPS spectra were severely attenuated after a 1-day exposure to an aqueous environment due to shielding the Pd particles underneath the formed iron oxide layer. This different behavior can be explained by the weak interaction of the Pd nanoparticles deposited by the redox method (2–5 nm) with the Fe⁰ surface in the samples prepared in that particular study. The other reason could be the presence of significantly more dispersed Pd-rich phases that were stable to oxidation and covered Fe-rich phases in the samples of the current study.

PCE and 2-ClBP degradation

PCE degradation did not occur on the supports without nanoparticles, whereas Fe-, Fe-Pd-, and Pd-containing samples exhibited significant catalytic activity (Table 1, Fig. 12).

Fig. 12

Variation in the PCE degradation depending on the Pd atomic content in the supported Fe-Pd nanocomposite for the samples prepared using [Pd(NH₃)₄]Cl₂.

The Fe-Pd-containing samples prepared by the reduction of the supported precursors were significantly more active than the samples with only Fe or Pd nanoparticles (Fig. 12). The effectiveness of Fe-Pd catalysts was more than an order of magnitude higher than that of Fe/SiO₂ and Pd/SiO₂ samples with the same Pd content. The results indicated the synergetic effect due to the strong interaction between Fe and Pd atoms in the supported samples prepared by the aforementioned methods. Also, the catalytic activities of the Fe-Pd samples prepared by simultaneous and two-step introduction were comparable (Fig. 12). Furthermore, PCE conversion was lower or even not observed in the presence of samples prepared with intermediate calcination, which may be attributed to lesser amount of the active sites formed due to the lower temperature throughout the process (Table 1, samples 3.2Pd-7.8Fe-CIC, 2.1Pd-7.0Fe-CICR).

The catalytic activity towards PCE degradation of the most effective samples was higher than that of free standing Pd/NZVI (sample 5Pd-95Fe in Table 1) and significantly higher than the catalytic activity reported elsewhere [7]. The complete conversion of PCE was obtained only at the low PCE:Fe ratios (6.3×10^–3 mol/mol) [7], whereas for the samples prepared in the present study the complete PCE conversion was observed at the significantly higher PCE:Fe ratios (0.07–0.3 mol/mol).

The adsorption and desorption of BP and 2-ClBP on the HS-SiO₂ support and in the presence of the selected catalysts was studied prior testing the catalytic activity of the Fe-Pd/SiO₂ nanocomposites to dechlorinate 2-ClBP. Adsorption of both BP and 2-ClBP on the SiO₂ support was very low (below 4 %) for the first 13 h and only reached 12–15 % in two days. No conversion of 2-ClBP to BP was observed for this period. Adsorption of BP on the catalysts was significantly faster and larger and reached its steady values (40–60 %) in 1–2 h. Furthermore, weak reversible adsorption of 2-ClBP and its dechlorination products on the used silica supports enabled the detection of two previously un-indentified organic products in the presence of highly active Fe-Pd/SiO₂ samples. As their peaks appeared on the chromatograms before BP, it was assumed that they were the products of BP hydrogenation on Pd.

The 2-ClBP conversion on the supported Fe samples was below the detection limit. The Fe-Pd-containing samples were significantly more active, and the conversion of 2-ClBP to BP was detected at the 2-ClBP:Fe molar ratio of 1×10^–3–3.5×10^–3 in 1 to 3 days (Table 1, Fig. 13). The 2-ClBP:Fe molar ratio was significantly higher in comparison to bulk Pd/Fe particles [15] and rather similar to Pd/Fe/AC samples [31].

Fig. 13

Variation in the 2-ClBP degradation depending on the Pd loading in the sample for the samples prepared using [Pd(NH₃)₄]Cl₂ or Pd acetate.

Catalytic properties of Fe-Pd/SiO₂ samples highly depend on the Pd loading and the preparation procedure. The Pd/Fe/SiO₂ samples prepared using Pd acetate (MI procedure) exhibited relatively large adsorption and the high initial conversion to BP. The conversion to BP increased as the Pd loading reached 1.5 % and varied within the experimental error for the samples with the Pd loading up to 3.3 % (Fig. 13). At the higher Pd loading, the 2-ClBP conversion decreased. Adsorption of 2-ClBP on the samples prepared by reduction of the supported Fe oxalate and Pd amino complexes (SI and CI procedures) was large. Nevertheless, the significant conversion to BP was observed only for the samples with the Pd loading of 2.3–3.0 wt. % (Fig. 13), and their catalytic activity was lower than the activity of the samples prepared by the MI procedure. The catalytic activity of the Fe-Pd samples on HS-silica increased with the Pd loading rather than with the Pd molar content in the Fe-Pd supported nanocomposites. The sample on LS-silica (3.2Pd-7.8Fe-CI) exhibited the highest activity among the samples prepared by this procedure. The efficiency of the catalysts was similar to 5 %Pd/14 %Fe/AC reported elsewhere [31]. The subsequent results indicated the decrease in the catalytic activity after several days of experiments (Table 1). The catalytic activity of the samples increased more than an order of magnitude when the dried samples were calcined to decompose the precursors before the reduction with hydrogen (Table 1, samples 3.0Pd-7.6Fe-SIC, 3.2Pd-7.8Fe-CIC; 2.1Pd-7.0Fe-CICR). The results of the second catalytic run (Table 1) exhibited the higher stability of these samples as compared to others prepared in the current work. It should be mentioned that no Fe or Pd leaching into the solution was detected from the samples prepared using Fe oxalate and Pd amino complexes, whereas significant leaching was observed, when chlorides were used as the reaction precursors [21, 22].

Regeneration of the spent catalysts

Previously the attempts have been made to regenerate the 0.1 %Pd/Fe nanoparticles aged in water for 24 h [32]. Regeneration with HCl partially restored the reactivity of the aged Pd/Fe sample, while the NaBH₄ reduction treatment did not rejuvenate the activity of the spent catalyst, the zerovalent state of the surface iron was reinstated. A poor recovery of the reactivity was explained by covering up the active Pd islets with iron formed from an iron oxides film. The catalysts prepared in this work using (NH₄)₃[Fe(C₂O₄)₃] and [Pd(NH₃)₄]Cl₂ were regenerated by reduction in hydrogen at 400 °C. The results presented in Table 4 demonstrated that reduction in hydrogen restored dechlorination activity of the completely deactivated sample. Although the adsorption of 2-ClBP decreased, the conversion was comparable with that of the samples with the higher Fe and Pd loading, perhaps due to a slight increase in the Pd particle size after regeneration.