*Article* **Hydrogen Production by Formic Acid Decomposition over Ca Promoted Ni**/**SiO2 Catalysts: E**ff**ect of the Calcium Content**

**B. Faroldi 1,2,\*, M. A. Paviotti 2, M. Camino-Manjarrés 3, S. González-Carrazán 3, C. López-Olmos <sup>1</sup> and I. Rodríguez-Ramos 1,\***


Received: 3 October 2019; Accepted: 23 October 2019; Published: 25 October 2019

**Abstract:** Formic acid, a major product of biomass processing, is regarded as a potential liquid carrier for hydrogen storage and delivery. The catalytic dehydrogenation of FA to generate hydrogen using heterogeneous catalysts is of great interest. Ni based catalysts supported on silica were synthesized by incipient wet impregnation. The effect of doping with an alkaline earth metal (calcium) was studied, and the solids were tested in the formic acid decomposition reaction to produce hydrogen. The catalysts were characterized by X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), temperature-programmed reduction (TPR), Fourier transform infrared spectroscopy (FTIR), transmission electron microscopy (TEM), and programmed temperature surface reaction (TPSR). The catalyst doped with 19.3 wt.% of Ca showed 100% conversion of formic acid at 160 ◦C, with a 92% of selectivity to hydrogen. In addition, all the tested materials were promising for their application, since they showed catalytic behaviors (conversion and selectivity to hydrogen) comparable to those of noble metals reported in the literature.

**Keywords:** hydrogen production; formic acid decomposition; nickel catalyst; calcium oxide promoter; silica support

#### **1. Introduction**

Hydrogen (H2) has significant advantages as an energy vector compared to petroleum or other conventional fossil fuels, although currently there are problems associated with its production, storage, and transportation that must be solved [1]. One possible solution for mobile applications, such as fuel cells, is to produce H2 in situ by a reaction such as the reforming of methanol, although these reactions have been associated with the generation of CO2, a greenhouse gas. There would be a significant advantage if H2 was produced from a chemical product derived from biomass since, in it, the CO2 formed in parallel must be considered a product of a carbon-neutral balance process. Formic acid is a chemical substance of relatively low specific volume and has limited uses, including its application as an antibacterial and antifungal agent. It can be produced by chemical methods such as the hydrolysis of methyl formate, but it is also obtained in equimolar proportions, together with levulinic acid, by hydrolysis of cellulose raw materials derived from biomass. Currently, with the increased interest in the production of levulinic acid and other valuable chemicals from biomass, it is important to

develop processes to use the derived formic acid, since, otherwise, it constitutes a waste material [1]. In this direction, the interest in the use of the decomposition reaction of formic acid to produce H2 has increased remarkably. Therefore, the challenge is to produce pure H2, with minimal CO content, at the lowest possible temperature. This demand can be achieved through the careful choice of the catalyst and the reaction conditions, so a lot of research is currently being done in this direction.

The production of H2 from formic acid using heterogeneous catalysts has been studied in liquid [2] and vapor phases [3], but, in most cases, formulations based on noble metals, such as Rh, Pt, Ru, Au, Ag, and Pd that are supported on C, Al2O3, and SiO2 have been investigated [1–3]. For the vapor-phase reaction, Solymosi et al. [3] found the following order of activity on a set of carbon-supported noble metals: Ir > Pt > Rh > Pd > Ru. They reported that the decomposition of formic acid started at and above 77 ◦C on all catalysts, and that decomposition was complete at 200–250 ◦C. On the other hand, in the case of non-noble metals, the activities of supported Ni catalysts have been measured, proving to be active at relatively higher temperatures (>220 ◦C) than noble metal catalysts [4].

Ni catalysts with different support matrices show adequate activity and selectivity for H2 production in various processes [5–8]. SiO2 stands out among the supports used due to its high surface area and low acidity [5,6,8]. Alkaline-earth metal oxides act as structural promoters by increasing the dispersion of the active phase and stabilizing the dispersed metallic phase against sintering [9–12]. Also, these additives act as chemical promoters by influencing the acid–base properties of support [13–15] or the electron density of dispersed metal crystallites [16,17]. In particular, the use of CaO as a promoter has exerted a positive effect on the increase in the interaction of the Ni with the support and the resistance of the Ni catalysts to the sintering [18–20].

In this work, Ni catalysts supported on a SiO2 matrix are used in the decomposition reaction of formic acid in the vapor phase. The effect of doping the support with different loadings of calcium is studied. The catalysts were characterized by X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), temperature-programmed reduction (TPR), Fourier transform infrared spectroscopy (FTIR), transmission electron microscopy (TEM), and programmed temperature surface reaction (TPSR).

#### **2. Materials and Methods**

#### *2.1. Catalysts Preparation*

Ni catalysts were supported on commercial SiO2 AEROSIL 200 (SBET = 200 m2/g, Degussa, previously calcined at 900 ◦C) or on Ca-SiO2 solids. The binary supports were prepared by incipient wetness impregnation of SiO2 with Ca(NO3)2·6H2O (Panreac Química SLU, Castellar del Vallès, Spain). Different Ca loadings were used (3.4, 6.8, and 19.3 wt.%). The Ca–SiO2 supports were maintained at room temperature for 12 h and then dried in an oven at 90 ◦C overnight. The solids thus obtained were finally calcined in flowing air, at 550 ◦C for 6 h. Samples are denoted as Ca(X)–SiO2, where X stands for the nominal Ca content in wt.%.

The Ni metal was incorporated by incipient wetness impregnation with a concentration of 5 wt.%, using Ni(NO3)2·6H2O (Alfa Aesar, Thermo Fisher Scientific, UK) as the precursor. These samples were subjected to a drying process in equal conditions to those of the binary support.

#### *2.2. Sample Characterization*

The surface area of the material was measured by BET analysis of the N2 adsorption isotherms collected at −196 ◦C (ASAP 2020 Micromeritics Instrument Corp., Norcross, GA, USA), with pretreatment at 200 ◦C for 2 h. The crystalline phases of the samples were examined by X-ray diffraction (XRD), using an X'Pert Pro PANalytical B.V., Almelo, The Netherlands. The TPR experiments were carried out in a conventional fixed-bed flow reactor, and the effluent gases were continuously monitored by online mass spectrometry (Pfeiffer/Balzers Quadstar GmbH, Asslar, Germany, QMI422 QME125); the samples were heated up in a 5% H2/Ar stream with a rate of 10 ◦C/min up to 800 ◦C. The XPS measurements were carried out using a multi-technique system (SPECS GmbH, Berlin, Germany) equipped with a dual Mg/Al X-ray source and a hemispherical PHOIBOS 150 analyzer. The catalysts were analyzed after two reduction treatments under H2 atmosphere, first at 400 ◦C for 1 h in a tubular quartz reactor and then at 400 ◦C for 15 min in the load-lock XPS chamber. Transmission electron microscopy (TEM) images of the reduced catalysts were acquired using a JEOL, Ltd., Tokyo, Japan, JEM 2100F field emission gun electron microscope equipped with an energy dispersive X-ray (EDX) detector. The fresh samples were reduced at 400 ◦C for 1 h in a pure H2 stream, while used samples were measured without any treatment. The particle size was determined by counting 300 particles. The temperature-programmed surface reaction (TPSR) measurements were carried out in conventional dynamic vacuum equipment coupled to a quadrupole mass spectrometer (RGA-200, SRS Inc., Sunnyvale, CA, USA). The catalysts were reduced before experiments in hydrogen flow at 400 ◦C and were degassed in a high vacuum at the same temperature. The adsorption was then carried out using a 40 Torr pulse of HCOOH at 40 ◦C. Once the gas phase was evacuated, the desorption step was carried out at a programmed temperature, and the gases released were analyzed with a mass spectrometer.

#### *2.3. Catalytic Test*

The catalytic activity measurements for the formic acid decomposition in the vapor phase were carried out in a conventional fixed-bed flow reactor. The catalysts were pretreated in H2 flux at 400 ◦C for 1 h and then cooled in N2 flux at the reaction temperature. A mixture of formic acid diluted with N2 was fed to the reactor using a saturator–condenser at 15 ◦C (HCOOH concentration equal to 6%, with a flow of 25 mL·min<sup>−</sup>1). The reactants and products were analyzed by gas chromatography with a Carboxen 1000 column and a TCD detector.

#### **3. Results**

#### *3.1. Catalysts Characterization*

Figure 1 shows the diffractograms obtained for the Ni/SiO2 and Ni over the binary supports after reduction in hydrogen at 400 ◦C for 1 h. From the XRD patterns, the characteristic diffraction broad peak centered on 2θ = 23◦ confirmed the amorphous nature of silica in Ni/SiO2 sample. No reflections from CaO species were observed in the diffraction patterns obtained for the Ca(X)-SiO2 supported catalysts.

**Figure 1.** X-ray diffractograms of reduced Ni/SiO2 and Ni/Ca(X)-SiO2 catalysts.

The XRD patterns of Ni/SiO2 and Ni/Ca(19.3)-SiO2 catalysts exhibit broad peaks which could be assigned to the Ni species. The XRD peaks at 2θ = 44.5◦, 51.9◦, and 76.4◦ indicate the presence of metallic nickel (JCPDS 04-0850), although small broad peaks at 2θ = 37.2◦, 43.3◦ and 62.9◦ reveal that there is also oxidized nickel phase (JCPDS 47–1049) [21].

The FTIR spectra of the reduced Ni/SiO2 and Ni/Ca(19.3)-SiO2 catalysts were analyzed (Figure 2). FTIR spectra show a broad band at 3528–3596 cm−1, which corresponds to the stretching vibration mode of the O-H bond from the silanol group (Si-OH). The band at 1050–1080 cm−<sup>1</sup> is assigned to the asymmetric stretching vibration of the siloxane bonds (Si-O-Si). The network Si-O-Si symmetric bond stretching vibrations are found at 620–900 cm<sup>−</sup>1, whereas the network O-Si-O bending vibration modes are observed at 469–481 cm−1. It is noted that the bands decrease in intensity as the content of Ca wt.% increases [20]. For the Ca promoted sample, with 19.3 wt.% of Ca, the broad band at 1458 cm<sup>−</sup>1, associated with the band at 876 cm−1, is assigned to asymmetric C-O stretch and out-of-plane deformation, respectively, of monodentate carbonate species on the CaO phase [22].

**Figure 2.** Fourier transform infrared spectroscopy (FTIR) spectra of reduced Ni/SiO2 and Ni/Ca(19.3)- SiO2 catalysts.

The XPS analysis of the reduced catalysts was carried out to study the presence of different surface Ni species (Figure 3). Both samples presented the Si 2s peak at 154.8 eV and the O 1s simple signal centered at 533 eV, corresponding mainly to the photoemission of the oxygen atoms presented on the siliceous support [23]. The Ca 2p3/<sup>2</sup> core level spectrum from the reduced sample shows binding energy of 348.2 eV [24].

Figure 3 shows a difference in the Ni 2p3/<sup>2</sup> spectra; the Ni catalysts exhibited a peak at 853.2 eV associated with reduced Ni species and another contribution related with octahedral Ni2<sup>+</sup> clusters at 856.3 and 857.1 eV for Ni/SiO2 and Ni/Ca(19.3)-SiO2, respectively [25]. This difference in the binding energy values could be related to a different interaction between the Ni and the support. In both catalysts, a reduced nickel fraction was observed under the treatment conditions carried out prior to the catalytic tests. This Ni0/Ni2<sup>+</sup> surface ratio was higher for the Ni/SiO2 catalyst. From the deconvolution of the spectra, the surface concentration of Ni<sup>0</sup> was estimated with respect to the total of surface Ni species, resulting in 70% and 13% for Ni/SiO2 and Ni/Ca(19.3)-SiO2, respectively.

**Figure 3.** X-ray photoelectron spectroscopy (XPS) spectra of Ni 2p3/<sup>2</sup> region of reduced Ni/SiO2 and Ni/Ca(19.3)-SiO2 catalysts.

The reducibility of the supported nickel catalysts was studied by temperature-programmed reduction (TPR). TPR is a powerful tool for the study of the reduction behavior of oxidized phase, as NiO, and obtainment of the strength of the oxide-support interaction. Figure 4 shows the TPR profiles of the catalysts, where two main reduction peaks can be observed. Peaks in the 200–300 ◦C range that are attributed to the reduction of superficial oxygen [11] are not detected in these solids. Peaks above 300 ◦C represent reductions in Ni(II) species with different interactions with the support. Peaks between 300 ◦C and 600 ◦C can be attributed to Ni(II) species with low support interaction, as NiO. Due to high mobility, this Ni(II) phase can be easily reduced and shows a low reduction temperature. Peaks above 600 ◦C refer to Ni(II) with moderate/strong support interaction [26]. Reduction peaks at higher temperatures appear as Ca is added, which suggests this addition increases the interaction of Ni(II) with the support. The reduction temperature increased with the Ca loading. These results are consistent with those observed by XPS experiments.

The TEM images of the undoped and doped Ca(19.3) materials are shown in Figure 5. It can be seen that the nickel particles are evenly distributed over the support. To estimate the average size, 300 particles were measured. Values of 5.1 and 4.8 nm for the undoped and doped catalyst, respectively, were obtained. Thus, the doping with Ca did not modify the average size of the particles, but did slightly modify the particle size distribution (see histograms in Figure 5).

**Figure 4.** Temperature-programmed reduction (TPR) profiles of Ni/SiO2 and Ni/Ca(X)-SiO2.

**Figure 5.** Transmission electron microscopy (TEM) images of reduced samples: (**a**) Ni/SiO2 and (**b**) Ni/Ca(19.3)-SiO2; the histograms were included.

Figure 6 shows the images obtained in the STEM mode and the EDX mapping of nickel (green), calcium (red), and silicon (white) revealed that Ni and CaO particles are evenly distributed on SiO2. In addition, the EDX images showed that Ni particles coincided in space with CaO phase supporting the metal-support interaction revealed by the XPS and TPR experiments.

**Figure 6.** Selected area for the EDX mapping in reduced Ni/Ca(19.3)-SiO2; mapping of nickel (green), calcium (red), and silicon (white).

#### *3.2. Catalytic Performance in Fixed-Bed Reactor*

Ni catalysts supported on SiO2 and on Ca-SiO2 were used in the decomposition reaction of formic acid to produce hydrogen. The catalytic activity of all materials was evaluated by operating the reactor in the experiments with a mass/flow ratio (W/F) equal to 5 <sup>×</sup> 10−<sup>5</sup> g·h·mL−1. The decomposition of formic acid can give the following as products:

$$\text{HCOOH(g)} \rightarrow \text{H}\_2 + \text{CO}\_2\tag{1}$$

$$\text{HCOOH(g)} \rightarrow \text{H}\_2\text{O} + \text{CO} \tag{2}$$

The values of reaction temperature for which the materials reach a 50% or 100% conversion of formic acid and its H2 selectivity are compared in Table 1.

**Table 1.** Catalytic activity of Ni solids: temperature reaction and H2 selectivity for 50% and 100% of HCOOH conversions (W/<sup>F</sup> <sup>=</sup> <sup>5</sup> <sup>×</sup> <sup>10</sup>−<sup>5</sup> <sup>g</sup>·h·mL<sup>−</sup>1; feed composition: 6% HCOOH/N2).


It can be observed that the Ni/SiO2 catalyst reached the 50% conversion at a temperature of 148 ◦C and 100% at 180 ◦C, while the selectivity was 91% and 87%, respectively (Table 1). In the catalysts supported on binary systems, the selectivity was higher in all cases. It is important to note that the catalyst with the highest Ca content (19.3 wt.%) reached 100% conversion at 160 ◦C, this being 20 ◦C lower than the undoped one.

Liu et al. [27] reported the study of the effect of different temperature pretreatments and atmospheres on the catalytic behavior of Ni catalysts for the dry reforming of methane. They observed that materials treated in He compared with those treated in H2 achieved better yields. During the pretreatment with He, a small fraction of Ni particles was reduced. However, a short period of exposure to reactants was sufficient to achieve the formation of metallic Ni nanoparticles that are particularly active under reaction conditions [27]. This phenomenon could explain the high activity of the Ni/Ca(19.3)-SiO2 catalyst, even though a low proportion of surface metallic species was observed after the reduction treatment.

Figure 7 shows the conversion of formic acid as a function of the reaction temperature for the series of Ni catalysts. The light-off curves were made following the same procedure in all the samples. After the reduction of the catalytic material, it was cooled in N2 flux to 60 ◦C, and then the reaction mixture was fed with a concentration of HCOOH of 6% in N2. After the curve measured from 0% to 100% (1st evaluation—Figure 7), the temperature was lowered to leave it in isothermal conditions and to measure the stability of the samples (Figure 8). After 16 h of reaction, the temperature was lowered and the complete curve was again measured from 0% to 100% (2nd evaluation in Figure 7) of the light off curve. The behavior throughout the conversion range shows the same tendency observed in Table 1. The catalysts were relatively stable under the conditions tested, although it can be observed that the points corresponding to the 2nd evaluation are below those obtained in the 1st, probably due to a restructuring of the material at the temperature reached (160–180 ◦C) and with conversion values close to 100%. In the Ca(19.3)-SiO2 catalyst, this behavior is less marked.

**Figure 7.** Catalytic activity of Ni solids after reduction at 400 ◦C at different reaction temperatures. The HCOOH conversions are plotted as function of the reaction temperature (W/<sup>F</sup> <sup>=</sup> <sup>5</sup> <sup>×</sup> <sup>10</sup>−<sup>5</sup> <sup>g</sup>·h·mL<sup>−</sup>1; feed composition: 6% HCOOH/N2).

**Figure 8.** Stability test of the catalysts after reduction at 400 ◦C in the fixed-bed reactor. (Reaction temperature <sup>=</sup> <sup>145</sup> ◦C, W/<sup>F</sup> <sup>=</sup> <sup>5</sup> <sup>×</sup> <sup>10</sup>−<sup>5</sup> <sup>g</sup>·h·mL<sup>−</sup>1, feed composition: 6% HCOOH/N2.)

The doping of K in Pd catalysts supported over SiO2, Al2O3, and activated carbon was previously reported [28]. These authors observed a significant effect of improvement in the catalytic behavior of noble metal for the formic acid decomposition. As a reaction mechanism, they proposed, as a first step, the formation of a phase containing liquid formic acid condensed in the pores of the catalyst; this phase provides a reservoir for the formation of formate ions with the participation of K<sup>+</sup> ions that later decompose to form CO2 and H2. In our materials, since the support is a nonporous material, condensation of formic acid is not likely to occur in pores; however, formates could form in the alkaline earth oxyhydroxide phase in the doped catalysts, with these species being the reaction intermediates.

#### *3.3. Study of the Adsorbed Species Under Reaction Conditions: Temperature Programmed Surface Reaction and FTIR Experiments*

Temperature-programmed surface reaction (TPSR) experiments were carried out to try to understand the differences in the catalytic performance. The catalysts were reduced before experiments in hydrogen flow at 400 ◦C and were degassed in high vacuum at the same temperature. The adsorption was then performed using a pulse of 40 Torr of HCOOH at 40 ◦C. Once the gas phase was evacuated, the evolution of the masses desorbed as a function of temperature was followed by mass spectroscopy. The TPSR experiments for the Ni catalysts are shown in Figure 9. Among the detected gases are the evolution of H2, CO2, CO, H2O, and HCOOH (m/z = 2, 44, 28, 18, and 29, respectively). At lower temperature (<100 ◦C) the desorption of the unreacted HCOOH is observed, and, above 80 ◦C, the decomposition process begins to produce H2 and CO2 and minority CO and H2O. It can be observed that the undoped catalyst exhibits lower adsorption of HCOOH and, subsequently, lower production of H2 and CO2. In all the samples, the molar ratio CO2/H2 produced was equimolar, as corresponds to the decomposition of formic acid (Table 1). These values were calculated by integrating the H2 and CO2 signals and taking into account the relative calibration of these gases.

There are two regions marked on the profiles: the first part corresponds to the decomposition of formic acid (up to 180 ◦C), and the second part corresponds to the decomposition of surface or mass species (formates, bicarbonates, and carbonates). The second part is closely related to the basic component of the catalyst. These samples doped with Ca present the H2 and CO2 desorption at a higher temperature. This could indicate greater stability of the species, for example, formate or bicarbonate species, which store hydrogen.

**Figure 9.** Temperature-programmed surface-reaction profiles of Ni catalysts after 40 Torr pulse of HCOOH at 40 ◦C.

#### *3.4. Characterization of Used Catalysts*

The FTIR spectra of the used Ni catalysts were analyzed (Figure 10). Figure 10a shows FTIR spectrum of used Ni/SiO2 catalysts; the reduced one was included for comparison. It can be clearly observed that the spectra are identical before and after the catalytic test, and the signals described above are present. Figure 10b shows the spectrum of used Ni/Ca(19.3)-SiO2 catalysts; the reduced one was included for comparison.

**Figure 10.** Fourier transform infrared spectroscopy (FTIR) spectra of (**a**) reduced and used Ni/SiO2; (**b**) reduced and used Ni/Ca(19.3)-SiO2; and (**c**) used Ni/Ca(X)-SiO2 catalysts.

For all the samples, the fingerprints of SiO2 at 475, 805, and 1115 cm−<sup>1</sup> related to the Si-O-Si, Si-OH, and Si-O bonds, respectively, are observed. The characteristic bands of surface monodentate carbonate species at 1458 and 876 cm−<sup>1</sup> are not present in the used Ni/Ca(19.3)-SiO2 catalyst. In addition to SiO2 bands, characteristic bands of formate species at 2877 and 1377 cm−<sup>1</sup> are observed in the calcium doped solids used in reaction (Figure 10b,c). These bands are very weak in the spectrum of Ca(3.4)-SiO2, but they are more intense at high X values (Ca(19.3)-SiO2). The signal centered at 1645 cm<sup>−</sup><sup>1</sup> and the shoulder at 1240 cm−<sup>1</sup> revealed the presence of asymmetric C-O stretching a C-O-H bending modes, respectively, of bicarbonate species [22,29]. For all the Ca-SiO2 supported catalysts, the presence of formate and bicarbonate species in the catalysts after the reaction was confirmed by FTIR (Figure 10c) and were consistent with TPSR experiments.

Figure 11 shows the diffractograms obtained for the used Ni/SiO2 and Ni/Ca-SiO2 catalysts. The characteristic diffraction broad peak centered on 2θ = 23◦ confirmed the amorphous nature of the SiO2 support. After reaction experiments, the diffraction patterns obtained for the Ca(X)-SiO2 supported catalysts exhibit reflections from CaO species centered at 15.8◦, 26.5◦, and 30.7◦ [11]. The used Ni/Ca(19.3)-SiO2 catalyst, as well as the reduced one, present broad peaks assigned to Ni species, indicating the presence of both metallic and oxidizes Ni particles. This result is consistent with those observed through TPR and XPS experiments. For Ni/SiO2 catalyst, the peaks at 44.3◦, 51.7◦, and 76.2◦ are assigned to metallic Ni particles. Comparison with the fresh reduced samples (Figure 1) reveals that, in the case of the Ni/SiO2 sample, the reflections corresponding to Ni<sup>0</sup> became sharper, which suggest that, in the absence of the Ca promoter, nickel is affected by the reaction conditions.

The TEM images of the undoped and doped Ca materials are shown in Figure 12. It can be observed that the nickel particles are evenly distributed over the support.

The estimated particle size using around 300 particles was 8.9 and 4.8 nm for the undoped and doped catalyst, respectively. The histogram of the Ni/SiO2 particles was modified during the catalytic test, and the distribution and average particle size are doubled with respect to those of the reduced sample. However, the doping with Ca modified the interaction of the metal with the support and the Ni particles remained stable during the catalytic test (see histograms in Figure 12). These findings are in agreement with the XRD results discussed above.

**Figure 11.** X-ray diffractograms of used Ni/SiO2 and Ni/Ca(X)-SiO2 catalysts.

**Figure 12.** TEM images of used samples: (**a**) Ni/SiO2 and (**b**) Ni/Ca(19.3)-SiO2; the histograms were included.

The stability in the average size of the nickel particles in the used Ni/Ca(19.3)-SiO2 catalyst could explain the difference in the behavior of this material with respect to the others presenting an equal performance in both catalytic tests (1st and 2nd evaluation in Figure 7). In addition, it can justify the highest stability of this catalyst during the long-term experiments of this reaction (Figure 8).

Figure 13 shows the images obtained in the STEM mode and the EDX mapping of nickel (green), calcium (red), and silicon (white) revealed that Ni and CaO particles are evenly distributed on SiO2 and coincide in occupying the same space on the support.

**Figure 13.** Selected area for the EDX mapping for used Ni/Ca(19.3)-SiO2 catalyst; mapping of nickel (green), calcium (red), and silicon (white).

#### **4. Conclusions**

Ni catalysts supported on SiO2 and on Ca-SiO2 were synthesized. These materials were employed in the formic acid decomposition reaction to produce hydrogen. On the catalytic system, calcium species act as structural promoters by stabilizing the dispersed metallic phase against sintering. In addition, this additive also acts as a chemical promoter by influencing the acid–base properties of support and the interaction metal-support.

The XRD patterns for Ni/SiO2 and Ni/Ca(19.3)-SiO2 catalysts indicated the presence of metallic and oxidized nickel particles after the reduction step at 400 ◦C. The reduction temperature increased with the Ca loading. These results are consistent with those of the XPS and XRD experiments. The average size of Ni particles was measured for the two samples, undoped and doped catalyst, being 5.1 and 4.8 nm, respectively. The incorporation of Ca did not modify the average size of the particles, but did slightly modify the particle size distribution.

The Ni/SiO2 catalyst reached 50% of formic acid conversion at a temperature of 148 ◦C and 100% at 180 ◦C, while the selectivity was 91% and 87%, respectively. For the catalysts supported on binary systems, the selectivity was higher in all cases. It is important to note that the catalyst with the highest Ca content (19.3 wt.%) reached 100% conversion at 160 ◦C, this being 20 ◦C lower than that of the undoped one. The doping with Ca modified the interaction of the metal with the support and the Ni particles remained stable during the catalytic test. However, for Ni/SiO2 catalyst the distribution and average particle size are doubled during reaction with respect to those of the reduced sample. The stability in the average size of the nickel particles in the used Ni/Ca(19.3)-SiO2 catalyst could explain the difference in the behavior of this material to the others. Moreover, this catalyst was relatively stable under the reaction conditions used, presenting an equal performance in two sequential catalytic tests.

The TPRS experiments reveal that, at lower temperatures (<100 ◦C), the desorption of the unreacted HCOOH is observed, and above 80 ◦C, the decomposition process begins to produce H2 and CO2 and minority CO and H2O. It can be observed that the undoped catalyst exhibits lower adsorption of HCOOH and subsequent lower production of H2 and CO2. In the Ca-SiO2 supported catalysts, the presence of formate and bicarbonate species in the catalysts after the reaction was confirmed by FTIR and were consistent with TPSR experiments.

**Author Contributions:** B.F. and I.R.-R. conceived of and designed the experiments; M.C.-M. and S.G.-C. carried out the synthetic experiments; M.A.P. conducted the FTIR measurements; C.L.-O. performed the TEM characterization; B.F. wrote the original draft manuscript; and all authors discussed the results and contributed to the manuscript.

**Funding:** This research was supported by the Spanish Agencia Estatal de Investigación (AEI) under project CTQ-2017-89443-C3-3-R. CONICET Postdoctoral External Scholarship awarded to B. Faroldi is acknowledged.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **(Ag)Pd-Fe3O4 Nanocomposites as Novel Catalysts for Methane Partial Oxidation at Low Temperature**

#### **Blanca Martínez-Navarro 1,2, Ruth Sanchis 3, Esther Asedegbega-Nieto 1, Benjamín Solsona <sup>3</sup> and Francisco Ivars-Barceló 1,2,\***


Received: 18 April 2020; Accepted: 18 May 2020; Published: 21 May 2020

**Abstract:** Nanostructured composite materials based on noble mono-(Pd) or bi-metallic (Ag/Pd) particles supported on mixed iron oxides (II/III) with bulk magnetite structure (Fe3O4) have been developed in order to assess their potential for heterogeneous catalysis applications in methane partial oxidation. Advancing the direct transformation of methane into value-added chemicals is consensually accepted as the key to ensuring sustainable development in the forthcoming future. On the one hand, nanosized Fe3O4 particles with spherical morphology were synthesized by an aqueous-based reflux method employing different Fe (II)/Fe (III) molar ratios (2 or 4) and reflux temperatures (80, 95 or 110 ◦C). The solids obtained from a Fe (II)/Fe (III) nominal molar ratio of 4 showed higher specific surface areas which were also found to increase on lowering the reflux temperature. The starting 80 m<sup>2</sup> g−<sup>1</sup> was enhanced up to 140 m2 g−<sup>1</sup> for the resulting optimized Fe3O4-based solid consisting of nanoparticles with a 15 nm average diameter. On the other hand, Pd or Pd-Ag were incorporated post-synthesis, by impregnation on the highest surface Fe3O4 nanostructured substrate, using 1–3 wt.% metal load range and maintaining a constant Pd:Ag ratio of 8:2 in the bimetallic sample. The prepared nanocomposite materials were investigated by different physicochemical techniques, such as X-ray diffraction, thermogravimetry (TG) in air or H2, as well as several compositions and structural aspects using field emission scanning and scanning transmission electron microscopy techniques coupled to energy-dispersive X-ray spectroscopy (EDS). Finally, the catalytic results from a preliminary reactivity study confirmed the potential of magnetite-supported (Ag)Pd catalysts for CH4 partial oxidation into formaldehyde, with low reaction rates, methane conversion starting at 200 ◦C, far below temperatures reported in the literature up to now; and very high selectivity to formaldehyde, above 95%, for Fe3O4 samples with 3 wt.% metal, either Pd or Pd-Ag.

**Keywords:** methane; oxidation catalysis; formaldehyde; magnetite iron oxide; Fe3O4; heterogeneous catalysis; palladium; Pd; silver; Ag; low-temperature activity; nanocomposite; Raman; TG in air; TG in hydrogen; XRD; electron microscopy; EDS

#### **1. Introduction**

Transformation of methane into valuable compounds is one of the hardest challenges in petrochemistry. The large abundance of methane in natural gas and the recent discoveries of new shale gas reserves around the world make methane functionalization even more interesting. If partial oxidation of methane to compounds with high value, such as methanol and especially formaldehyde, could take place with high efficiency, it could be competitive against the indirect process in several steps from synthesis gas [1]. Unfortunately, no catalytic system has shown promising values

neither regarding reaction rates nor selectivity at medium and high methane conversions. The main drawback is related to the high stability of the methane molecule, which presents strong methyl C–H bonds, which are not easy to activate. Thus, it is usual that this reaction takes place at high reaction temperatures so that in those conditions the partial oxidation products are quickly oxidized into carbon oxides. As in mild reaction conditions, the reaction rates are low, and in harder conditions the carbon oxide formation is predominant; an intermediate trade-off working seems to be the best option [2]. In any case, the direct oxidation of methane into methanol, formaldehyde, and other oxygenated products is still very far from being competitive for commercial implementation [2,3]. Among the catalysts employed for direct oxidation of methane to formaldehyde using molecular oxygen, those based on vanadium, molybdenum [4–7], and especially iron [3,8–17] have shown the most promising performance. Nevertheless, the yield of formaldehyde reported with these catalysts does not exceed 5% [2,4,8,18–20], although higher yields can be found in the literature for other catalysts. A significant example of this is the 14–17% yield of formaldehyde patented for Mo-based heteropoly acid catalysts at 600–650 ◦C [8], never again mentioned, with CH4 conversions and formaldehyde selectivity within the 20–23% and 65–84% ranges, respectively.

In the case of iron-based catalysts, bulk FePO4 systems and mainly iron supported on a range of siliceous materials (standard silica, MCM-41, SBA-15, and others) have been studied in this reaction [10,14–17,21]. Although it is accepted that iron sites must not be highly aggregated, otherwise formaldehyde readily decomposes, there is not a wide consensus about the exact nature of the iron active and selective species. It has been proposed that isolated tetrahedral Fe3<sup>+</sup> sites are the most selective sites [22] whereas other authors have suggested that Fe with higher aggregation (2D FeOx oligonuclear sites) is more effective than isolated Fe3<sup>+</sup> sites [16]. In any case, the reaction temperatures employed in these articles exceed 400 ◦C and are often over 500 ◦C [3,8,12]. Interestingly, recent work by Zhao et al. showed that iron species supported on different zeolites (ZSM-5, beta, and ferrierite) activate methane at 350 ◦C using N2O as an oxidant, reaching methane conversions until 3%, with the formation of different partial oxidation products (methanol, formaldehyde, dimethyl ether) but also carbon oxides, obtaining up to 10% selectivity to formaldehyde [9]. It must be mentioned that the transformation of methane at room temperature has also been studied but with no competitive results or using reaction conditions not viable for commercial applications [23,24]. In this vein, the direct transformation of methane into formaldehyde has been interestingly demonstrated to be possible at room temperature using gaseous [Al2O3] <sup>+</sup> clusters, although with low rates [25].

On the other hand, palladium catalysts are, together with platinum ones, usually employed in the total oxidation of methane into carbon dioxide [26–29], a very much applied reaction from an environmental viewpoint. Thus, palladium sites can activate methane at low temperatures and high conversions can be easily reached, although the methane transformation is directed towards a total combustion product of limited interest. These works mainly pay attention to the light-off curves in order to see the temperatures required for all methane to be converted and to follow the stability with the time on stream. However, due to the massive CO2 formation, no special interest has been paid to see what takes place at low methane conversions.

Within this context, the aim of the current work is to rationally design and develop catalytic systems with potential for the partial oxidation of methane at ambient pressure and relatively mild reaction temperatures. For such a purpose, we have drawn inspiration from materials, chemical species, and results, all known from the scientific literature as well as from our own experience. Moreover, we kept the premises of using simple preparation methods and raw materials not especially expensive, in order to develop a multifunctional catalyst which was not specifically reported for partial oxidation of methane. Thus, Fe3O4-based nanostructures supporting, in turn, smaller Pd nanoparticles, were the catalytic system selected and developed. To complement this, a systematic study on some synthesis parameters in order to improve the surface area of the iron oxide substrate without using sophisticated methods was successfully performed. An additional purpose was also to prove the use of Ag as

a dopant to tune the high reactivity of Pd toward moderate catalytic behaviors rather than its usual propensity for methane combustion [30,31].

The structural and elemental compositions, textural properties, morphology, and reactivity against reduction and oxidation for the developed (Ag)Pd-Fe3O4 nanocomposites were investigated by several physicochemical techniques, as well as their catalytic properties for methane partial oxidation at temperatures not higher than 250 ◦C. Promising results were achieved from the preliminary catalytic screening which will be discussed and contrasted with the characterization results throughout the manuscript.

#### **2. Materials and Methods**

Nanostructured magnetite iron oxides were synthesized by the coprecipitation of the corresponding iron hydroxide species formed in aqueous NH4OH medium, starting from the mixture of FeCl2·4H2O and FeCl3·6H2O reagents dissolved in deionized water, according to the following reaction:

$$\text{FeCl}\_3 + \text{FeCl}\_2 + 8\text{NH}\_4\text{OH} \rightarrow \text{Fe}\_3\text{O}\_4 + 8\text{NH}\_4\text{Cl} + 4\text{H}\_2\text{O} \tag{1}$$

Magnetite-based solids differing in their specific surface areas were obtained using a nominal Fe3+/Fe2<sup>+</sup> molar ratio of 2 or 4, and different coprecipitation temperatures (within the 80–110 ◦C range) in an inert atmosphere (N2) under reflux conditions. NH4OH was added at the corresponding synthesis temperature to get the basic medium for the iron hydroxides coprecipitation. In all cases the reflux temperature was maintained for 30 min. The precipitates were filtered, washed with deionized water, and dried at 100 ◦C/12 h.

(Ag)Pd-Fe3O4 nanocomposite materials were prepared by an impregnation procedure of Pd or Ag and Pd cations over the magnetite solid obtained by reflux at 80 ◦C with a nominal Fe<sup>3</sup>+/Fe2<sup>+</sup> molar ratio of 4. The impregnation was carried out using the proper amount of the corresponding metal nitrate salt/s to get a metal load of 1, 2, or 3 wt.% Pd or 3 wt.% Ag-Pd. In the latter case, no sequential method was followed but both metals were simultaneously incorporated, employing an Ag:Pd molar ratio of 2:8. The impregnation procedure consisted of dissolving the Pd(NO3)2 and AgNO3 amounts needed in acetone and adding the magnetite solid into the acetone metal solution which was kept under stirring for 10 min, followed by an ultrasonication treatment for 30 min. This two-step cycle was repeated 4 times for every sample in order to favor a homogeneous dispersion of the noble metal/s over the magnetite. After the last cycle, the mixture was stirred at room temperature until the acetone was completely evaporated. The magnetite solid containing Pd or Ag/Pd was dried at 100 ◦C/12 h, and afterwards treated at 200 ◦C for 1 h, under a reducing gas stream consisting of an 80 mL min−<sup>1</sup> total flow of H2/N2 (50/50), in order to selectively reduce the noble metal cations to their metallic state, avoiding the reduction of iron from the magnetite which starts over 350 ◦C.

XRD patterns were collected using an X'Pert PRO diffractometer with θ–2θ configuration, from PANalytical (Almelo, The Netherlands), equipped with an X'Celerator fast detector, operating at 45 kV and 40 mA and using a nickel-filtered Cu Kα radiation source which produces a nearly monochromatic X-ray beam (λ = 1.5406 nm). Diffractograms were obtained within the 4–90◦ 2θ range employing 20 s cumulative time.

The specific surface areas were determined by the Brunauer–Emmett–Teller (BET) method from N2 adsorption isotherms at 77 K measured in a Micromeritics ASAP2020 instrument (Norcross, GA, USA). A desorption treatment at 150 ◦C (10 ◦C min−<sup>1</sup> heating rate) for 5 h, at high vacuum range conditions, was conducted for every sample just prior to the BET analysis.

The scanning electron microscopy analyses were performed employing a field emission scanning electron microscope (FESEM; model GeminiSEM 500, ZEISS, Oberkochen, Germany). Two different electron detection modes were used for FESEM imaging: (i) a secondary electron In-Lens detector, located inside the electron column, which works with low-energy secondary electrons and provides images with high resolution, and (ii) a energy selective backscattered (EsB) in-lens detector, independent

of the secondary in-lens detector, which provides a pure backscattered signal with no secondary electron contamination and very low acceleration potential, providing a higher Z-contrast image than any other backscattered detector. Elemental chemical microanalyses in selected areas from the scanning electron micrographs were carried out with an X-ray energy-dispersive detector (EDS) from Oxford Instruments (Abingdon, UK), coupled with the FESEM microscope (ZEISS, Oberkochen, Germany).

A 200 kV analytical electron microscope equipped with a field emission electron gun (FEG; model JEM-2100F, JEOL, Tokio, Japan) allowing ultrahigh resolution in scanning transmission mode (STEM) through the sub-nanometric highly stable and bright electron probe (0.2 nm) provided by the FEG, was employed for the high-angle annular dark-field (HAADF) imaging in STEM mode. A high-resolution CCD (charge-coupled device) camera (2048 × 2048 pixels; model SC200, GATAN, Pleasanton, CA, USA), was used for image acquisition controlled by Digital Micrograph software. Image processing was performed by ImageJ free-software (Windows 10, 64-bit, version) [32]. The JEOL JEM-2100F microscope was equipped with a detector for X-ray energy-dispersive spectroscopy (EDS; model X-Max 80, Oxford Instruments, Abingdon, UK). The EDS detector, with 127 eV resolution and 0.5–2.4 nm spot size range, was used for elemental chemical analysis of punctual sites and selected areas from the STEM imaging.

Selected samples were measured by confocal micro-Raman spectroscopy using a LabRam-IR HR-800 model instrument (HORIBA Jobin Yvon, Edison, NJ, USA) coupled to an optical microscope model Olympus BX41. Excitation was provided by a He-Ne laser operating at a wavelength of 632.8 nm through a 100× objective lens, and with the laser power set to 0.6 mW, below the limit reported to induce phase transformations in iron oxides [33–36]. The regular Raman spectra were collected within the 100–1670 cm−<sup>1</sup> region with 1 cm−<sup>1</sup> spectral resolution, using an integration time of 8–16 s and 8–16 accumulations. In these conditions the lateral (xy) spatial resolution on the sample was 1 μm. For a high-resolution spectral profile, the integration time and the number of accumulations were increased to 32 s and 64, respectively, within a measuring region reduced to 320–777 cm<sup>−</sup>1.

Thermogravimetric analyses (TGA) were carried out in an SDT Q 600 apparatus (TA Instruments, New Castle, DE, USA) under reducing (H2) or oxidizing (synthetic air) conditions. Prior to analysis, samples were pre-treated at 150 ◦C for 1 h under a He stream (30 mL min−<sup>1</sup> flow) which was maintained thereafter during the cooling down to room temperature. Following this pre-treatment, and according to the desired reducing or oxidizing experiment, the gas was switched from He to H2 (5 mL min−<sup>1</sup> flow) or synthetic air (24 mL min−<sup>1</sup> flow), respectively. After the adequate stabilization time (ca. 30–40 min.), the temperature was raised, at a 5 ◦C min−<sup>1</sup> rate, to 200 ◦C; then until 800 ◦C using a different heating rate of 10 ◦C min<sup>−</sup>1. The evolution of species (CO, CO2, H2O, O2, N2, and H2) in the gas phase produced during the thermogravimetric experiments was followed with a quadrupole mass spectrometer (Pfeiffer Vacuum Omnistar™ GSD 301, Asslar, Germany) coupled with the TGA equipment. For this purpose, the following mass fragments (m/z) were measured: 2 (H2), 28 and 16 (CO), 44 (CO2), 18 and 17 (H2O), 32 and 16 (O2), 28 and 14 (N2).

The heterogeneous catalysis experiments for partial oxidation of methane were carried out in a fixed-bed quartz tubular flow reactor at atmospheric pressure, within the 100–250 ◦C temperature range, and using 100 mg of catalyst diluted with SiC (mg catalyst/SiC = <sup>1</sup> <sup>2</sup> volume ratio). The feed consisted of a molar ratio CH4/O2/N2 = 32/4/64, for a total flow of 50 mL min<sup>−</sup>1. The separation and analysis of reactants and products were carried out online using a gas chromatograph equipped with two different chromatographic columns (3 m length molecular sieve 5 Å column and 30 m length RT-U-bond 0.53 mm i.d. column) and a thermal conductivity detector (TCD).

Blank reaction experiments up to 300 ◦C with no catalyst in the reactor tube showed no conversion at all. In all cases, carbon balances were 100 ± 4%, irrespective of the catalyst and mass used. Each data point for the assessment of catalytic performance was determined by averaging three different analyses of reactants and products. The minor differences observed among the repetitive analyses confirmed the steady-state conditions.

#### **3. Results**

#### *3.1. Development of Nanostructured Fe3O4 Materials*

From the XRD patterns of the solids obtained by coprecipitation of the corresponding iron hydroxides during the reflux syntheses at different temperatures and employing aqueous mixtures of Fe2<sup>+</sup> and Fe3<sup>+</sup> chloride salts with Fe<sup>3</sup>+/Fe2<sup>+</sup> molar ratio 2 or 4 (Figure 1), the presence of magnetite structure as the main phase formed in all cases can be confirmed. Thus, diffraction peaks at 2θ = 30.2, 35.7, 43.4, 53.7, 57.3, and 62.9◦ are present, with a similar intensity ratio among all the samples, according to the standard magnetite (Fe3O4) powder diffraction patterns previously reported [37–39].

**Figure 1.** XRD patterns of magnetite solids prepared by coprecipitation employing a Fe3+/Fe2<sup>+</sup> aqueous molar ratio of 2 (**a–c**) or 4 (**d–f**), at different synthesis temperatures: 110 ◦C (**a,d**), 95 ◦C (**b,e**), or 80 ◦C (**c,f**). Symbols: Fe3O4 (- - -), α-Fe2O3 (- · - · -).

It is important to point out that similar 2θ positions are found in powder XRD patterns of cubic maghemite (γ-Fe2O3) which moreover presents a medium intensity 2θ peak at 32.17◦ among others of lower intensity such as the ones at 15.0 and 23.8◦ [37–40]. The absence of those peaks confirms the assignment of magnetite as the major phase present in every iron oxide synthesized (Figure 1). Without limiting the foregoing, the XRD patterns of the solids prepared with a nominal Fe<sup>3</sup>+/Fe2<sup>+</sup> molar ratio of 4 show an additional low-intensity peak at ca. 33.1◦, which is not observed for those obtained from a nominal Fe3+/Fe2<sup>+</sup> molar ratio of 2. The 2θ peak at 33.1◦ can be assigned to the presence of hematite (α-Fe2O3) as a minority phase since it appears as the highest relative intensity diffraction for the main rhombohedral hematite ICSD (Inorganic Crystal Structure Database) references (01-079-0007, 00-024-0072, 01-072-0469, 01-085-0599). Indeed, the hematite content was found below 10% from a semiquantitative analysis of the XRD data by X'Pert Highscore Plus software.

Despite this minor structural difference dependent on the nominal molar ratio used, a significant variation was found in the specific surface areas of the prepared iron oxides (Table 1) which additionally showed a dependence with the reflux temperature employed (varied from 80 to 110 ◦C). Thus, the BET

surface area obtained for the Fe3O4 precipitates from the synthesis employing a nominal Fe3+/Fe2<sup>+</sup> molar ratio of 4 was higher than those from a ratio of 2. For both ratios, the lower reflux temperature induced a higher specific surface area (Table 1).

**Table 1.** Specific surface areas of magnetite solids precipitated at different reflux temperature and nominal Fe3+/Fe2<sup>+</sup> molar ratio.


<sup>1</sup> The standard deviation of ±0.2 m2 g−<sup>1</sup> was estimated for the BET surface areas calculated from the N2 adsorption isotherms.

Therefore, the highest BET area (138 m<sup>2</sup> g−1) was achieved for the magnetite solid prepared at 80 ◦C employing a nominal Fe<sup>3</sup>+/Fe2<sup>+</sup> molar ratio of 4, which was selected as the support for the following incorporation of Pd or Ag and Pd in order to prepare the target nanocomposite materials that would be catalytically tested for methane partial oxidation.

From the analysis by field emission scanning electron microscopy (FESEM), a nanometric spherical morphology was found for the crystalline Fe3O4 particles prepared. An example is displayed in Figure 2 for the highest surface magnetite substrate selected to incorporate the noble metals in the next step. The elemental chemical microanalyses by X-ray energy dispersive spectroscopy (EDS), from a number of FESEM images showing similar homogeneous distribution of nanosized spherical particles, including those in Figure 2, confirmed an averaged Fe:O stoichiometry of 3.0 (±0.3):4.1 (±0.3) consistent with the magnetite Fe3O4 structure assigned from the powder XRD results (Figure 1). A representative statistic study of the crystallite size distribution from the FESEM micrographs of the highest surface magnetite, whose resulting histogram is displayed in Figure 2, illustrates an average Fe3O4 particle diameter of 16.2 ± 3.2 nm.

**Figure 2.** (**a**) Field emission scanning electron microscopy (FESEM) micrograph of magnetite solid precipitated from a reflux synthesis at 80 ◦C using a nominal Fe3+/Fe2<sup>+</sup> molar ratio of 4. Image conditions: in-lens mode, work distance 2.7 mm, energy selective backscattered (EsB) grid of 300 V and electron high tension (EHT) of 1 kV. (**b**) Statistic distribution of particle diameter from several FESEM micrographs of the magnetite solid shown in (a).

#### *3.2. Development of Nanocomposite (Ag)Pd-Fe3O4 Materials*

The incorporation of 1, 2, and 3 wt.% Pd, or 3 wt.% Ag-Pd (molar Ag:Pd ratio of 2:8) to the selected magnetite substrate by impregnation and subsequent reduction treatment (250 ◦C/1 h under 50% H2 in N2 stream), gave rise to the nanostructured (Ag)Pd-Fe3O4 composite materials that were tested as catalysts for methane partial oxidation. The XRD patterns of these nanocomposites are shown in Figure 3. All display the 2θ peaks common to those found for the (Ag)Pd-free iron oxide substrate (Figure 1f); i.e., the peaks' set of 30.2, 35.7, 43.4, 53.7, 57.3, and 62.9◦ assigned to magnetite [37–39], along with the one at 33.1◦ from traces of hematite (ICSD: 01-079-0007). However, low-intensity 2θ peaks at 71.2 and 74.3◦ clearly appear for all the (Ag)Pd-containing magnetite samples (Figure 3). These two peaks can also be glimpsed, to a greater or lesser extent, in some XRD patterns of the pristine iron oxides synthesized from Figure 1, especially in those obtained from Fe3+/Fe2<sup>+</sup> = 2 synthesis. A thorough analysis, comparing some representative references for indexed XRD patterns of Fe3O4 structures, allows us to identify that the set of the given main six peaks assigned above to Fe3O4 structure is common for both the cubic (ICSD: 01-076-0957, 01-075-0449) and the orthorhombic (ICSD: 01-075-1609) crystal systems of magnetite. Nevertheless, the low-intensity 2θ peaks at 71.2 and 74.3◦ appear specifically characteristic of orthorhombic magnetite.

**Figure 3.** XRD patterns of nanocomposite catalysts based on magnetite (from 85 ◦C synthesis using Fe3+/Fe2<sup>+</sup> molar ratio of 4) doped with Pd or Pd-Ag: 1% Pd-Fe3O4 (**a**), 2% Pd-Fe3O4 (**b**), 3% Pd-Fe3O4 (**c**), 3% AgPd-Fe3O4 (**d**). Symbols: (*---*) Fe3O4 (ICSD: 01-075-1609); the highlighted strip defines the region where the highest intensity diffractions for pure Pd, pure Ag, and AgPd alloys are found [41–44].

Furthermore, semiquantitative analysis of the XRD data by X'Pert Highscore Plus software confirmed the major presence of an orthorhombic Fe3O4 structure in all (Ag)Pd-Fe3O4 catalysts from Figure 3. Only for 1% Pd-Fe3O4 and 3% AgPd-Fe3O4 samples did the semiquantitative phase analysis provide a minor coexistence with the cubic Fe3O4 structure (ca. 10% and 20%, respectively).

On the other hand, an additional low-intensity peak at 2θ = 40.1◦ can be observed for the 3% Pd-Fe3O4 catalyst (Figure 3c), which corresponds to the (111) plane of metal Pd [41,42,44]. This low-intensity peak turns into a small broadband within the 2θ range from 38.7 to 41.6◦ for the bimetallic 3% AgPd-Fe3O4 sample. This band, along with the glimpsed wide bump centered at 45.5◦, is in full agreement with previously reported XRD patterns of bimetallic AgPd alloys presenting fcc crystalline structure [41–44], especially with those containing Ag:Pd molar ratio of 2:8 [42] or between 3:7 and 1:9 [41]. The main XRD diffraction of metal Pd is barely observed for the 2% Pd-Fe3O4 sample and the region becomes totally flat for the 1% Pd-Fe3O4, which is coherent to the quantitative detection limit (<2%) associated to the XRD technique for multiphase mixtures [45].

The monometallic Pd-Fe3O4 samples were investigated by field emission scanning electron microscopy (FESEM) using an energy selective backscattered (EsB) electron in-lens detector which can select the electrons according to their energy. Thus, the EsB detector is very sensitive to variations in the atomic number (Z) of atoms, allowing Z-contrast images (i.e., brighter image appearance for atoms with a higher Z), since the higher the atomic number the more electrons are scattered due to greater electrostatic interactions between the nucleus and the microscope electron beam at high angles. Indeed, Figure 4 shows Z-contrast micrographs for 1% Pd-, 2% Pd-, and 3% Pd-Fe3O4 samples, all with glaring bright spots consistent with metal Pd clusters, over a matt grey matrix assigned to Fe3O4. Congruently, a higher population density of these metal clusters is displayed as the nominal Pd load increases in the magnetite (Figure 4).

The average diameter of the metal Pd clusters was statistically determined for each Pd-Fe3O4 specimen. By direct measurement using ImageJ software [32], over more than 150 shiny clusters from several micrographs were acquired by EsB mode. Thus, the results in Table 2 show an average diameter of 3.3 ± 1.2 nm for the Pd clusters in 1% Pd-Fe3O4 catalyst, which remains practically unchanged for 2% Pd-Fe3O4, while it slightly increases to 4.4 ± 1.6 nm for the highest Pd load (i.e., 3% Pd-Fe3O4). It should be mentioned that the EsB detector employed gives a higher Z-contrast than any other backscattered detector, enabling differentiation between elements that are only distinguished by a few atoms.


**Table 2.** Characteristics of the metal nanoparticles (NPs) on the magnetite-based catalysts.

<sup>1</sup> Pd or Ag-Pd metal load nominally introduced by wet impregnation. <sup>2</sup> Atomic ratios employed for wet impregnation. <sup>3</sup> An averaged atomic ratio determined by chemical analysis (standard deviation = 0.1) of selected areas employing X-ray energy-dispersive spectroscopy (SA-EDS) coupled to a 200 kV field emission analytical electron microscope employing high-angle annular dark-field (HAADF) imaging in scanning transmission mode (STEM) mode. <sup>4</sup> The average diameter of metal nanoparticles (± SD), measured by EsB detector in HRFESEM (1–3% Pd-Fe3O4) or HAADF imaging in STEM (3% AgPd-Fe3O4).

For the bimetallic 3% AgPd-Fe3O4 sample, the Z-contrast microscopy studies were carried out differently, employing high-angle annular dark-field (HAADF) imaging in scanning transmission electron mode (STEM), using an analytical electron microscope equipped with a field emission electron gun (FEG). In a similar way to the EsB detector from the high-resolution field emission scanning electron microscopy (HRFESEM) instrument, the HAADF in STEM senses a greater signal from atoms with a higher atomic number (i.e., causing the metal atoms to shine brighter in the resulting image). Indeed, HAADF-STEM micrographs of 3% AgPd-Fe3O4, as the example displayed in Figure 5, show dull grey spheres within the diameter range of 11–20 nm, accordingly with Fe3O4 nanospheres, underneath some brighter and smaller metal particles with an average diameter of 4.3 nm (Table 2). The chemical analysis pointed to these metal particles (Figure 5) using selected area X-ray energy-dispersive spectroscopy (SA-EDS) with 0.5–2.4 nm spot size range (Table 2) and confirmed the coexistence of Ag and Pd with an averaged Ag:Pd stoichiometry of 0.2:0.8 (±0.1), consistent with the Ag0.2Pd0.8 alloy structure observed by XRD (Figure 3).

**Figure 4.** Z-contrast micrographs obtained by high-resolution field emission scanning electron microscopy (HRFESEM) using an energy selective backscattered (EsB) electron in-lens detector for the Fe3O4-based catalysts doped with 1% Pd (**a**), 2% Pd (**b**), and 3% Pd (**c**).

**Figure 5.** HAADF-STEM micrograph of 3% AgPd-Fe3O4 sample, with SA-EDS analysis spots marked and numbered: chemical analyses in brighter particles (1–5 spots) showed AgPd wt.% >12, with an averaged Ag:Pd atomic ratio of 0.2:0.8 (0.1 SD), while in 6–10 spots AgPd wt.% <0.9.

#### *3.3. Catalytic Tests for Methane Partial Oxidation*

The catalytic properties of the developed nanocomposite (Ag)Pd-Fe3O4 materials were tested in gas phase heterogeneous catalysis for the partial oxidation of methane at moderate reaction conditions (i.e., ambient pressure and temperatures up to 250 ◦C). Interestingly, using these mild conditions, methane gets activated although with low reaction rates. The catalytic results, summarized in Figure 6, show methane conversion at 200 ◦C for all the catalysts tested, leading to the formation of formaldehyde as the main reaction product (>74%), with CO2 as the only secondary product detected. In general, the bimetallic 3% AgPd-Fe3O4 catalyst shows a catalytic activity significantly higher than the monometallic Pd-Fe3O4 catalysts; the latter ones presenting a maximum activity for the 2% Pd-Fe3O4 sample (Figure 6a). Therefore, no linear correlation is observed between the catalytic activity and the formal content of Pd which could be initially considered as the major source of methane activation at such a low temperature, according to the literature [30,31]. On the other hand, the selectivity to formaldehyde rises with the increase in the Pd load, showing a maximum above 97% for both the monometallic 3% Pd-Fe3O4 catalyst and the bimetallic 3% AgPd-Fe3O4 one at 200 ◦C (Figure 6b). At 250 ◦C the methane conversion slightly increases, while the selectivity to the partial oxidation product proportionally drops for all the monometallic Pd-Fe3O4 catalysts, with no absolute inverse correlation between conversion and selectivity observed. On the contrary, the high selectivity to formaldehyde hardly changes for the bimetallic 3% AgPd-Fe3O4 catalyst from 200 to 250 ◦C at which a formaldehyde productivity of 43 gCH2O kgcat−<sup>1</sup> h−<sup>1</sup> is reached with a turnover frequency (TOF) of 24.7 h−<sup>1</sup> (reaction conditions are shown in the footnote of Figure 6).

**Figure 6.** Catalytic activity (**a**) and selectivity to formaldehyde (**b**) for methane partial oxidation over the (Ag)Pd-Fe3O4 nanocomposite catalysts at 200 ◦C (light color bars) and 250 ◦C (dark color bars). Reaction conditions: 100 mg catalyst mass, CH4/O2/He molar ratio of 32/4.3/63.7, and contact time (W/F, at standard conditions) of 2.6 gcat. h molCH4<sup>−</sup>1.

The pristine metal-free iron oxide substrate was comparatively tested for the partial methane oxidation upon identical reaction conditions used in the (Ag)Pd-Fe3O4 nanocomposite catalysts, detecting no methane conversion, likewise the blank results obtained no catalyst. Moreover, the catalytic results were confirmed by testing twice, under equivalent reaction conditions, all the nanocomposite catalysts displayed in Figure 6. The catalysts used in the first test were employed in a second catalytic cycle showing no apparent change in their performance (i.e., similar catalytic activity and selectivity).

#### **4. Discussion**

According to the results presented, the experimental research of this work consisted of three basic stages: (i) the development of a high surface nanostructured Fe3O4 substrate; (ii) its coating with noble metal nanoparticles to form mono- or bi-metallic (Ag)Pd-Fe3O4 nanocomposites; (iii) to assess the potential of the noble-metal/iron-oxide nanocomposites developed as catalysts for gas-phase partial oxidation of methane at ambient pressure and mild temperatures.

For the first stage, a reflux-assisted coprecipitation was the chosen method to carry out the preparation of high surface optimized magnetite. In this sense, a systematic study is presented here on the influence of both the reflux temperature (80–110 ◦C range) and the aqueous Fe<sup>3</sup>+/Fe2<sup>+</sup> molar ratio (2 or 4) in the specific surface area of the resulting magnetite precipitate. The data analysis confirmed a higher Fe3O4 surface area is strongly favored for Fe3+/Fe2<sup>+</sup> molar ratio of 4, rather than 2, in the aqueous synthesis solution (Table 2). Regardless, the magnetite surface area was also observed to increase as the reflux temperature was lowered within the 110–80 ◦C range, for both nominal ratios assayed (Table 2). At this point, it should be mentioned that an aqueous Fe3+/Fe2<sup>+</sup> molar ratio of 2, matching the Fe3O4 stoichiometry, has been traditionally employed in the coprecipitation-based methods reported for magnetite preparation [37,46–49]. In fact, no peer-reviewed publication could be found, with which to compare our results, that claimed the use of Fe<sup>3</sup>+/Fe2<sup>+</sup> = 4 for Fe3O4 synthesis. Nevertheless, the gain in the specific surface area achieved by the Fe3O4-based solid prepared by the higher nominal Fe<sup>3</sup>+/Fe2<sup>+</sup> ratio in synthesis (Table 2) is in good agreement with a size decrease reported to be observed for Fe3O4 particles in suspension while increasing Fe<sup>3</sup>+/Fe<sup>2</sup> ratios [49]. However, along with this statement, in the first known work employing a coprecipitation method for magnetite preparation, no numeric values for Fe3+/Fe2<sup>+</sup> ratios larger than 2 are overtly mentioned.

Notwithstanding the excess of Fe3<sup>+</sup> present in the synthesis media when using an Fe<sup>3</sup>+/Fe2<sup>+</sup> ratio of 4, compared with the Fe3O4 stoichiometry, the elemental and structural analyses of the higher surface area precipitates are consistent with a major presence of the mixed iron oxide bulk phase (just like the solids from stoichiometric Fe<sup>3</sup>+/Fe2+ ratio synthesis). On the one hand, an averaged Fe:O stoichiometry of 3.0:4.1 (± 0.3), especially consistent with magnetite, was experimentally determined by EDS microanalyses on nanospheres, with 16.2 nm average diameter (Figure 2), to make up the ca. 140 m2·g−<sup>1</sup> solid (Table 2). On the other hand, the 2<sup>θ</sup> peaks at 30.2, 35.7, 43.4, 53.7, 57.3, and 62.9◦ present in the corresponding XRD patterns (Figure 1d–f) are consistent with the (2 2 0), (3 1 1), (4 0 0), (4 2 2), (5 1 1), and (4 4 0) planes of standard magnetite structure (Joint Committee on Powder Diffraction Standards, JCPDS, file No.: 19-0629), respectively. These diffractions are also common to the γ-Fe2O3 structure (JCPDS file No.: 04-0755) whose assignment as major bulk phase was discarded by the absence of its additionally associated peaks from (210), (300), and (320) planes of crystalline maghemite (γ-Fe2O3) [39]. Nevertheless, a minor presence of maghemite cannot be completely ruled out. In fact, many studies claim that it is not straightforward to observe pure magnetite or maghemite, but an intermediate defective structure, Fe3-xO4, between both [50–53]. This has been especially proven for particles within the nanometric scale [52–56], for which a magnetite core with a maghemite shell structure is widely reported [48,54–56]. For the most oxidized iron oxide layer, in the cited core-shell structures, a varying thickness has been found depending on different factors such as the accurate nanoparticle size [54] and/or the reactants used and other preparation conditions [48,57].

Due to the inherent short-range periodic order that maghemite would exhibit in such a scenario, our XRD data would not be useful to directly reveal or deny its minor coexistence together with the magnetite structure. In this respect, Raman spectroscopy was complementarily employed to analyze some representative samples whose spectra are displayed in Figure 7. Thus, the Raman spectrum of the highest surface Fe3O4-based substrate used to incorporate the noble metals (Figure 7, spectrum a1) does not present a typical pure magnetite profile which is considered to be characterized by a single narrow and intense band reported to be centered within the 660–700 cm−<sup>1</sup> range, together with a weak one between 534 and 560 cm−<sup>1</sup> [33,34,58–63]. Instead, the Raman spectrum for the metal-free Fe3O4-based substrate is mainly characterized by the features assigned to maghemite (γ-Fe2O3) in the scientific literature [33,61–65]; i.e., a broad and intense asymmetric band from ca. 600 to 800 cm−1, assigned to *A*1g vibrational mode, and two more bands also asymmetric but weaker: one with two maxima reported to be centered at 345–365 cm−<sup>1</sup> (*E*<sup>g</sup> mode) and 380–395 cm−1, and another one centered at 505–515 cm−<sup>1</sup> (*T*1g mode) [62,65,66]. In addition, two narrow peaks with relatively low intensity appear

at 236 and 302 cm<sup>−</sup>1, matching with the strongest vibrational bands (*A1g* and *Eg* modes, respectively) related to hematite [35,62–65], along with the broadband centered at 1320 cm−<sup>1</sup> [35,67,68]. In this sense, it must be said that Raman spectroscopy is extremely sensitive to the presence of hematite. Even traces of hematite impurities contained within other major iron oxides (as it is known in our case from XRD results) are able to provide narrow peaks with spurious intensity due to the large scattering power for Raman radiation inherent to hematite [60], inducing wrong conclusions about its relative content unless complementary characterization techniques are also employed.

**Figure 7.** (**a**) Raman spectra of pristine Fe3O4-based substrate (a1), 3% Pd-Fe3O4 (a2), and 3% AgPd-Fe3O4 (a3) nanocomposite catalysts, collected at 0.6 mW. (**b**) High-resolution Raman spectrum and its second derivative by the Savitzky–Golay algorithm, using 81 convolution points of the Fe3O4-based substrate submitted to identical reduction treatment (250 ◦C/2 h in H2) as the noble metal-containing catalysts. Acronyms from (a): *Mh* (maghemite), *H* (hematite), and *M* (magnetite) label the dotted lines according to the assignment ranges found in the literature (references at the end of the footnote). Symbols from (b): numerical ranges together with the black and grey labels define the regions where maxima have been reported in the literature for maghemite [33,61–65] and magnetite [33,34,58–63].

Regarding maghemite, its strongest band is usually described as consisting of two components whose maxima are reported within 645–665 cm−<sup>1</sup> and 700–740 cm−<sup>1</sup> ranges [48]. Additionally, a magnon band centered at ca. 1440 cm−<sup>1</sup> and two low-intensity peaks at 125 and 210 cm−<sup>1</sup> are also typical of maghemite structure [66,69]. Unfortunately, the characteristic Raman features linked to magnetite (at ca. 547 and 680 cm<sup>−</sup>1) overlap with the three major bands cited for maghemite, which make difficult unambiguous statements about magnetite just based on the Raman spectrum when maghemite is present as well. A tool that might help to clarify this is the second derivative of the Raman spectrum. For this purpose, a high-resolution Raman spectrum within the region of interest was collected for the Fe3O4-based substrate, this time after being submitted to the same reduction treatment (250 ◦C/2 h in H2) as the noble metal-containing samples. The spectrum obtained and its second derivative are displayed in Figure 7b. Regardless of the higher resolution, the Raman spectrum is very similar to that obtained from the pristine substrate within the same region (Figure 7a1) which leads to identical conclusions. Nevertheless, the second derivative shows a resolved accurate position for the maxima associated to the vibrational Raman modes among which it is possible to distinguish a maximum within the 534–560 cm−<sup>1</sup> region (Figure 7b), where no maximum but the one from a weak vibrational mode of magnetite is reported [63,66].

Therefore, Raman results represented by the spectra in Figure 7, provide unambiguous evidence for the presence of maghemite. Unfortunately, further conclusions about either the structural configuration or the phases distribution of the given iron oxide mixture cannot be done from Raman data, due to the intimate overlapping between the main Raman band of magnetite and one of the strongest bands of maghemite (Figure 7). Furthermore, the ultrathin maghemite layer derived from a minor magnetite surface oxidation cannot be completely ruled out, although laser power was strictly kept far below the limits reported for the laser-induced magnetite oxidation [33,34]. In this respect, transformation into maghemite at temperature values as low as 109 ◦C has been reported for magnetite nanoparticles presenting 13–16 nm diameters (i.e., within the range measured for our Fe3O4 substrate) (Figure 2).

Moving forward to the discussion from the second experimental stage, the incorporation of noble metal nanoparticles to the magnetite-based substrate (the one with the highest surface area), followed by the given selective reduction treatment, leads to the corresponding mono- or bi-metallic (Ag)Pd-Fe3O4 nanocomposites (Table 2). Raman spectra of the representative catalysts 3% Pd-Fe3O4 and 3% AgPd-Fe3O4 are displayed as spectrum a2 and a3, respectively, in Figure 7a. These spectra show meaningless differences for the mixed iron oxide vibrational features compared with the pristine metal-free substrate, apart from the greater or lesser intensity of the narrow peaks at 236 and 302 cm−<sup>1</sup> from the hematite impurity traces. Therefore, identical conclusions as those discussed above apply here.

With respect to the crystalline phases from the XRD results, new features appear for metal Pd or AgPd alloy structures, along with those remaining from the Fe3O4 substrate (Figure 3). On the other hand, no intensity is detected for the main diffraction of either metal Fe [70] or PdO [71–73] at 44.7 and 33.9◦ 2θ, respectively, which is consistent with the selective and complete noble metal reduction. The average diameter for the metal clusters was found to increase from ca. 3.3–3.4 nm for 1% Pd- and 2% Pd-Fe3O4 samples, to ca. 4.3–4.4 nm for the 3% Pd- and 3% AgPd-Fe3O4 ones. In the specific case of the bimetallic nanocomposite sample, the stoichiometry determined in the solid by structural and elemental analysis (i.e., by XRD (Figure 3) and by SA-EDS from HAADF-STEM imaging (Figure 5 and Table 2), respectively) was consistent with the nominal Ag:Pd stoichiometry used (0.2:0.8) during the impregnation treatment. Although we did not employ surfactant or any other sophisticated method to control the homogeneous size of the noble metal particles during the incorporation treatment, the standard deviation determined for the size distribution of (Ag)Pd nanoparticles on the magnetite-based substrate was acceptable (ca. 33%) in both mono- and bi-metallic nanocomposite samples (Table 2).

Finally, with respect to the stage of catalytic reactivity evaluation, promising results were obtained for the gas-phase methane partial oxidation into formaldehyde (Figure 6), for which the catalytic system (Ag)Pd-Fe3O4 was never reported, to the best of our knowledge. The highest formaldehyde productivity was shown by the 3% AgPd-Fe3O4 catalyst, yielding ca. 45 gCH2O kgcat−<sup>1</sup> h−<sup>1</sup> with a turnover frequency (TOF) of 24.7 h−<sup>1</sup> (calculated for a 26% Pd dispersion) at 250 ◦C, far below the reaction temperatures reported so far [8,12]. In this sense, oxidation of methane has been hardly reported to occur below 600 ◦C [8,10,12]. Thus, the catalytic data provided by most of the studies on methane selective oxidation to formaldehyde correspond to reaction temperatures within the 600–800 ◦C range [8,12]. Ironically, formaldehyde decomposing into methanol and carbon monoxide starts at temperatures above 150 ◦C, although the kinetics for the uncatalyzed reaction is slow below 300 ◦C [74]. As a matter of fact, 450 ◦C appears as the lowest temperature reported so far, providing heterogeneous catalysis data of formaldehyde yielded by methane oxidation using molecular oxygen [75]. Nonetheless, the use of N2O as a more reactive oxygen donor has been reported to run methane partial oxidation at 350 ◦C on several Fe-doped zeolites, although no selectivity to formaldehyde above 8% is claimed in that work [9]. Instead, a dispersion of partial oxidation products (CH3OH, CH2O, and C2H4) with low total selectivity was obtained, close to 40% in the best case [9]. On the contrary, selectivity to formaldehyde above 95% is achieved at 200–250 ◦C with the 3% AgPd-Fe3O4 catalyst developed in the current work (Figure 6). In addition to the major formation of formaldehyde, CO2 was the only byproduct obtained in our catalytic tests for methane partial oxidation. No trace of methanol or CO was detected as

byproducts, suggesting no formaldehyde decomposition. Moreover, reported pure radical mechanisms for aldehyde formation via methanol intermediate and/or consecutive decomposition involving CO formation, before or along with CO2 production [8], can be discarded as plausible mechanisms being held on our (Ag)Pd-Fe3O4 catalysts. Instead, the so-called ion-radical mechanism, reported to involve electron transfers from the catalyst lattice oxygens, appears the most consistent with our catalytic results. Thus, this mechanism justifies the exclusive formation of formaldehyde and/or CO2, by single-electron transfer from O– centers and/or two-electron transfer from O2 2– centers, respectively, both at the catalyst surface [8]. The ion-radical mechanism has been reported for catalysts based on metal oxides containing ambivalent metals that are able to easily change their oxidation states, such as MoO3 or V2O5. As a matter of fact, these two oxides supported on SiO2 are the catalytic systems grabbing a majority of studies concerning the oxidation of methane into formaldehyde [4–7]. Similarly, the cited ambivalence is also present in the mixed iron oxide from magnetite, the major component in our catalytic system. In fact, meaningful results in methane partial oxidation to formaldehyde have been reported for several catalysts containing iron oxide species [3,8–14]. The most encouraging finding to keep deploying resources in CH4 conversion technology is, indeed, nature's iron-containing active center located in the methane monooxygenase enzyme able to selectively transform CH4 into methanol, formaldehyde, and formic acid at ambient conditions [76,77]. Moreover, along with Mo and V, Fe was in the six elements list whose SiO2-supported oxides were concluded to be the most suitable active phases for methane partial oxidation based on the screening of 43 elements [78]. However, no specific mention to the use of Fe3O4 in methane direct conversion reactions has been found in the literature, neither as an active phase nor support. With respect to the latter, in addition to SiO2, other high surface substrates such as Al2O3, TiO2, SnO2, and zeolites are among the most used for supporting the active phases for partial oxidation of methane.

Noble metals such as Pd or Pt have also been briefly studied as active phases for methane activation, compared with transition metal oxides. Particularly, nanosized Pd has been reported to be highly reactive at temperatures as low as 200 ◦C, although toward complete methane oxidation [79]. Despite some interesting proposals for inhibiting the uncontrolled Pd activity leading to total combustion products [12,79], noble metals were practically left out in research on methane partial oxidation. Within this context, the nanocomposite material addressed in the current work, based on Fe3O4-supported (Ag)Pd, appears as a relatively known system, but a novel type of catalyst for the methane direct conversion into oxygenated partial oxidation products. The initial working hypothesis was to comparatively incorporate Ag with the purpose to downgrade the expected overreactive nanostructured Pd. In this regard, Ag was reported to selectively segregate at the surface of nanostructured Pd, suppressing the sites with the strongest oxidizing reactivity [30,31]. Indeed, the highest catalytic performance is achieved with the Fe3O4-based catalyst containing the bimetal AgPd alloy (Figure 6). Thus, the results settle that Ag improves selectivity toward the partial oxidation product at higher catalytic activity, which suggests confirmation of the working hypothesis about a softening effect on Pd reactivity. However, in general, the catalysts consisting of pure Pd supported over Fe3O4 have surprisingly shown high formaldehyde selectivity as well, rather than the total combustion products expected according to the published literature on Pd predisposition toward complete methane oxidation [79]. Even more, the selectivity to formaldehyde rises with increasing Pd load (from 1 to 3 wt.%). Regarding the assessment of catalytic activity from the pristine Fe3O4-based substrate itself, it could be dismissed by getting no conversion at all for the methane oxidation at identical reaction conditions used for the (Ag)Pd-containing catalysts. The latter statements might suggest a kind of synergy effect between Pd and the magnetite substrate. In this sense, by means of thermogravimetry analyses under reducing or oxidizing conditions, a relationship was found between the redox properties of the iron cations and the Pd content added to the mixed iron oxide. Thus, Figure 8b displays the derivative of thermogravimetry curves (DTG) collected under a H2 stream (i.e., reducing conditions), with the positive peak maximums in the graph representing the greatest rates for each weight loss step which correspond to each dehydration accompanying a reduction. For the metal-free mixed iron

oxide substrate, two separated reduction stages can be distinguished, namely, a small peak at low temperatures and a large one at a higher temperature range, below and above 360 ◦C, respectively. The first region between 200 and 360 ◦C corresponds to the reduction of minority pure Fe (III) oxides to Fe3O4. This reduction is clearly comprised of at least two overlapped peaks with maxima at 284 ◦C and 334 ◦C which, according to the reported literature [80], should correspond to the reduction of maghemite (γ-Fe2O3) to Fe3O4 and hematite (α-Fe2O3) to Fe3O4, respectively. This assignment is also consistent with the hematite impurity observed by XRD (Figure 1) and the homogeneous minor presence of maghemite proved to coexist with magnetite by comparing Raman spectroscopy (Figure 7) and XRD results (Figure 1). According to the following stoichiometry: 3Fe2O3 + H2 → 2Fe3O4 + H2O; a Fe2O3 content (γ-Fe2O3 + α-Fe2O3) of ca. 2% was calculated from the weight loss within the 200–360 ◦C region during thermogravimetry (TG) analysis in H2 of the metal-free Fe3O4-based substrate (Figure 8a). On the other hand, the second region, from 360 ◦C to 550 ◦C, reflects major reduction consisting of a two-step Fe3O4 reduction sequence to Fe<sup>0</sup> through wüstite (FeO) intermediate formation [81].

**Figure 8.** Thermogravimetry analysis curves collected in H2 for the different (Ag)Pd-Fe3O4 catalysts prepared and for the pristine metal-free Fe3O4 substrate (**a**), and their first derivative (**b**).

The Pd incorporation on Fe3O4 leads to nanocomposite materials presenting a reduction temperature lower than the pristine magnetite-based substrate, either for the one-step reduction from minority Fe2O3 phases to Fe3O4 or the two-step reduction from Fe3O4 to FeO and Fe0. Thus, the reduction temperature for the minor Fe2O3 phases moves bellow 200 ◦C, in agreement with previous publications employing Pd or other noble metals, such as Au, on Fe2O3 [80,82]. In Figure 8a, it can be observed that the corresponding weight loss for (Ag)Pd-Fe3O4 catalysts starts already at 100 ◦C, while the beginning of the equivalent loss for the pristine substrate delays until 200 ◦C.

The derivative TG curve, within the major reduction region, is displayed in Figure 8b, where the maximum of the major peaks correspond to the temperature for the maximum rate of Fe3O4 reduction which decreases following the next trend: Fe3O4 (468 ◦C) > 1% Pd-Fe3O4 (412 ◦C) > 2% Pd-Fe3O4 (376 ◦C) > 3% Pd-Fe3O4 (370 ◦C) > 3% AgPd-Fe3O4 (358 ◦C). According to this sequence, the Fe3O4 reducibility seems to be enhanced as the Pd load increases (from 1 to 3 wt.%) in the monometallic catalysts. Nevertheless, the tendency breaks for the 2.4 wt.% Pd load in the bimetallic 3% AgPd-Fe3O4 catalyst for which the doping with Ag forming the Ag0.2Pd0.8 alloy appears to favor the lowest temperature for Fe3O4 reduction at 358 ◦C (i.e., the highest reducibility).

A thermogravimetry (TG) analysis under oxidizing conditions (dry air stream) was also performed in three representative samples, in order to distinguish whether the effect on the redox properties of iron species, observed by the drop in their reduction temperature, goes further than the spillover effect of atomic hydrogen produced by catalyzed H2 dissociative adsorption on Pd [83,84]. Thus, Figure 9 displays the first derivative of the TG curve collected under synthetic air flow for 3% Pd-Fe3O4 and 3% AgPd-Fe3O4 catalysts, and their pristine Fe3O4 substrate. The first negative peaks observed in Figure 9 within the 150–200 ◦C range correspond to the maximum rate of weight gain produced by oxygen insertion causing Fe2<sup>+</sup> oxidation to Fe3<sup>+</sup>, for which the corresponding temperature, appearing at 164 ◦C in the metal-free Fe3O4 increases in the noble metal-containing catalysts, especially in the bimetallic 3% AgPd-Fe3O4 with the maximum at 191 ◦C. Therefore, the redox properties of iron species undergo generalized changes induced by the Ag and/or Pd nanostructures supported, whereby the Fe3<sup>+</sup> species become easier to reduce (i.e., present higher oxidizing power), while the Fe2<sup>+</sup> ones become harder to oxidize. The strong correlation, existing between the trend in the change of the mixed iron oxide redox properties and the increased productivity to formaldehyde, points to a significant metal–support interaction that seems to be responsible for the tendency observed in the catalytic behavior of the different nanocomposite catalysts.

**Figure 9.** The first derivative of the thermogravimetry (TG) curves collected under synthetic air flow for the monometallic 3% Pd-Fe3O4 and the bimetallic 3% AgPd-Fe3O4 catalysts, as well as for the pristine metal-free Fe3O4-based substrate.

#### **5. Conclusions**

From the systematic study on the synthesis of Fe3O4-based solid with an improved specific surface area, an enhancement from 80 to 140 m2 g−<sup>1</sup> was achieved by doubling the nominal Fe3+/Fe2<sup>+</sup> ratio from 2 to 4 and decreasing the synthesis temperature from 110 to 80 ◦C. (Ag)Pd-Fe3O4 nanocomposite materials were developed through solvent-assisted (Ag)Pd impregnation on the Fe3O4-based substrate with optimized surface area, followed by a selective reduction treatment. The resulting nanocomposites can be defined as Pd metal or AgPd alloy nanoparticles with ca. 3.3–4.4 nm diameter supported on Fe3O4-based nanoparticles of 16.2 nm average diameter. The coexistence of maghemite was proven by Raman, although in a minor amount (<3 wt.%) according to XRD and TG results. Further research must be done in order to establish the accurate structural configuration linking the major Fe3O4 and the minor γ-Fe2O3 phases within the nanostructures developed.

Catalytic activity for methane conversion appears in all (Ag)Pd-Fe3O4 nanocomposite catalysts tested, at temperatures as low as 200 ◦C. The reported propensity toward methane combustion by nanostructured Pd has shown to be suppressed within the Pd-Fe3O4 nanocomposite systems (1–3 wt.% Pd load), which present a major selectivity to formaldehyde. The higher Pd load, within the studied range, increases the catalyst selectivity to formaldehyde which reaches above 95% for the catalysts containing above 2 wt.% Pd. On the other hand, a correlation was also observed between the increased Pd load and a gradually enhanced reducibility of the Fe3<sup>+</sup> species in magnetite, which confirms a significant metal–support interaction. This reducibility in the Fe3O4 phase improves even further

with the Ag incorporation forming the Ag0.2Pd0.8 alloy. These results suggest that the iron oxide substrate is most likely involved in the catalytic partial oxidation of methane into formaldehyde. This assumption is also supported by the CO2 formation as the only secondary product which has been reported to occur by the ion-radical mechanism involving electron transfers from the catalyst lattice oxygens in catalysts containing oxides from ambivalent metals.

The nanocomposites have proven worthy for further research in gas-phase methane partial oxidation at relatively low reaction temperatures, compared with those reported until now using other catalysts. Forthcoming works will increase the low catalytic activity observed and shed more light on the mechanisms, active sites involved, and the role of the metal–support interaction.

**Author Contributions:** Conceptualization, methodology, writing—original draft preparation, project administration, F.I.-B.; validation, formal analysis, investigation, writing—review and editing, B.M.-N., E.A.-N., F.I.-B., R.S., and B.S.; resources, F.I.-B. and B.S.; data curation, B.M.-N., E.A.-N., F.I.-B., and R.S.; visualization, B.M.-N., E.A.-N., and F.I.-B.; supervision and funding acquisition, F.I.-B. and B.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research and the APC were funded by *Comunidad de Madrid*, grant number 2017-T1/IND-6025 project within the program "*Atracción y Retención de Talento Investigador*" of the V PRICIT. The catalytic test experiments were funded by *Ministerio de Ciencia, Innovación y Universidades de España*, grant number MAT2017-84118-C2-1-R.

**Acknowledgments:** The authors gratefully acknowledge José María Gavira Vallejo (Dept. Ciencias y Técnicas Fisicoquímicas, UNED) for the Raman measurements; Inmaculada Rodríguez Ramos (Instituto de Catálisis y Petroleoquímica, ICP-CSIC), and Antonio R. Guerrero Ruiz (Dept. Química Inorgánica y Química Técnica, UNED) for their support in infrastructure and technical resources; as well as the administrative support from Mª Dolores Fernández, Cristina Ramos, and Mª Elisa Estébanez (PAS, UNED) and Rosa Mª Martín Aranda (Vicerrectora Investigación, UNED).

**Conflicts of Interest:** The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Nanomaterials* Editorial Office E-mail: nanomaterials@mdpi.com www.mdpi.com/journal/nanomaterials

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18