**2. Results**

#### *2.1. Preparation and Characterization of the Supported Catalysts*

The CeZr and Zr oxide supports were synthetized by *w*/*o* microemulsion technique [28] followed by calcination at 750 ◦C, and were characterized by XRD (X-Ray Di ffraction) analyses (Figure 1). The di ffractogram of the former showed reflections ascribable to the Ce0.5Zr0.5 O2 (CZOm) phase, which is not easy to obtain with other synthetic methods in terms of composition homogeneity and structure. This specific phase possesses higher oxygen mobility and storage properties that enhance the activity in reforming reactions [28]. Debye Scherrer calculations on the peak at lower 2θ showed an average crystallite dimension of 8 nm.

The XRD analysis on the zirconium oxide sample, ZrO2 (ZOm), showed the formation of the sole tetragonal phase of zirconia, while the pattern related to the monoclinic phase, associated with a lower surface area [32–35], was not present. Debye Scherrer calculations gave an average crystallite dimension of 10 nm.

Table 1 shows the surface area, volume of the pores, and average pore diameter of the CZOm and ZOm samples, determined through Barrett-Joyner-Helenda (BJH) analysis.

**Figure 1.** X-Ray Diffraction (XRD) spectra of the Ce0.5Zr0.5O2 (CZOm) and ZrO2 (ZOm) samples synthetized by microemulsion and followed by calcination at 750 ◦C.

**Table 1.** Surface area, pore volume, and average pore diameter of the CZOm and ZOm.


The Ce-containing sample showed a higher surface area and a smaller average pore diameter. This is related to the smaller average particle dimension that was obtained for the Ce-containing sample (8 nm) compared to the zirconium oxide one. In any case, high surface and small particles were obtained thanks to the microemulsion technique, in which micelles, used as microreactors, helped to module the crystallite dimensions during the synthesis. Nevertheless, the total pore volume was similar for the two samples, suggesting that a higher number of smaller pores was found in the CZOm sample.

The two supports were also analyzed through IR (Infrared) spectroscopy in the solid state, in nujol mull between NaCl plates, both before and after the deposition with the Rh4(CO)12 cluster, in order to verify that this actually occurred. Initially, both spectra showed peaks at ca. 2100 (w), 2031 (s) and 1892 (w) cm<sup>−</sup>1, typically associated to the sole supports. The cluster compound was then deposited on CZOm and ZOm (Rh loading of 0.6% wt/wt). The preparation was carried out by mixing a n-hexane solution of Rh4(CO)12 with a slurry of the desired support in the same solvent for 24 hours, under CO atmosphere because of the instability of the neutral cluster in air. In the end, the IR spectra of the n-hexane solutions showed a complete absence of any residual Rh4(CO)12, as indicated by the disappearance of its typical νCO absorptions. Moreover, both supports changed their colors from white to light burgundy after the treatment with the cluster and their IR spectra showed significant changes with respect to the initial ones. More specifically, the spectrum recorded on CZOm exhibited νCO peaks at 2078 (sh), 2051 (s), 2043 (m), 2019 (sh), 1803 (sh) and 1797 (w) cm<sup>−</sup>1, while the one of ZOm showed νCO peaks at 2068 (w), 2038 (s), 1990 (sh) and 1774 (w) cm<sup>−</sup>1. These experimental evidences consistently indicate that the cluster deposition on the CZOm and ZOm supports occurred. Figures 2 and 3 show the solid-state IR spectra of the bare and Rh-deposited CZOm and ZOm supports, respectively.

**Figure 2.** Solid-state infrared (IR) spectra of the bare (**left**) and Rh-deposited (**right**) CZOm support.

**Figure 3.** Solid-state IR spectra of the bare (**left**) and Rh-deposited (**right**) ZOm support.

Subsequently, both catalysts were pelleted and treated under N2 at 500 ◦C, and new nujol mull spectra were recorded. As illustrated in Figure 4, the ZOm-supported catalyst exhibited residual νCO peaks at 2023 (w) and 1874 (w) cm<sup>−</sup>1, while the CZOm-supported one displayed the same peaks shown by the support before the deposition [26].

**Figure 4.** Solid-state IR spectra of the Rh-deposited CZOm (**left**) and ZOm (**right**) supports after treatment under N2 at 500 ◦C.

A possible explanation for this different behavior could be that, in the CZOm support, the high mobility of oxygen atoms linked to the Ce(IV)/Ce(III) couple favored the complete oxidation of the Rh carbonyls [36]. After treatment under hydrogen flow at 500 ◦C, both catalysts only showed IR bands associated to the sole supports.

#### *2.2. Catalytic Tests on Cluster-Derived Catalysts*

#### 2.2.1. Cluster-Based Catalyst with High Rh Loading

The cluster-based catalyst with a metal loading of 0.6% deposited on Ce0.5Zr0.5O2 (Rh0.6-CL-R-CZOm) was tested in different conditions, and the effects of the S/C (Steam/Carbon) ratio, temperature, pressure and GHSV (Gas Hourly Space Velocity) were studied. The precursor was charged in the reactor and heated under N2 at 500 ◦C, and after cooling down to 200 ◦C, the sample was heated at 500 ◦C under H2 to obtain the final active catalyst. The analyses of the reaction parameters have been carried out with the aim of studying the conditions with which the catalyst could maximize both syngas and hydrogen productions.

At 500 ◦C, the increase of the S/C ratio had a positive effect on the methane conversion (X CH4), as shown in Figure 5. In particular, at low S/C ratios, the methane conversion was 19% at S/C 1 and 22% at S/C 1.5, very close to the equilibrium values (19% and 24%, respectively). A further increase in S/C up to 2 and 3 resulted in higher conversions of 23% and 26%, though not enough to meet the increase of the equilibrium values (29% and 36%, respectively), deriving from higher steam concentration, showing a kinetic limitation of the steam reforming reaction. The H2/CO ratio was also considered to analyze the contribution of the water gas shift reaction (WGSR). In all conditions, the ratio was higher than the stoichiometric value of 3, due to the occurrence of the WGSR. Thus, an advantage of low temperature steam reforming is the possibility of carrying out the two reactions together, as opposed of conducting them in three different reactors (one for steam reforming, a high-temperature shift reactor and a low-temperature shift one), which is the usual procedure. For instance, a H2/CO ratio of 28 was obtained at S/C 1. When the S/C ratio increased, the H2/CO raised thanks to an augmented contribution of the WGSR, reaching the values of 35, 43, and 63 at S/C 1.5, 2 and 3, respectively. Indeed, the high amount of steam reacted with CO and produced a major amount of hydrogen. This was also confirmed by the CO selectivity (sCO), which decreased when increasing S/C, falling from 13 (at S/C 1) down to 6 at S/C 3.

Figure 6 highlights the effect of the temperature on the catalytic conversion of methane at 30,000 <sup>h</sup>−1, S/C 3 and atmospheric pressure. In all conditions, the methane conversions were far from the equilibrium values (indicated by the empty bars). In general, experimental and equilibrium methane conversions increase with temperature as both steam reforming thermodynamics and kinetics are favored at high temperature. In fact, a conversion of 4% was observed at 350 ◦C, far from the equilibrium value of 11%, which increased to 9% and 17% at 400 and 450 ◦C, but they were both still far from the equilibrium conversions (18% and 26%, respectively). The equilibrium value was not reached even at 500 ◦C, where the observed experimental conversion was 26% and the equilibrium one 36%. However, a sharply decreasing trend was observed for the H2/CO ratio. The WGS and SR reactions are favored in opposite conditions; therefore, at 350 ◦C water gas shift was predominant, and the as-produced CO was directly converted into hydrogen, resulting in a final gas composition with CO concentration of 75 ppm. Conversely, at 500 ◦C the water gas shift is less favored, and an increased rate of steam reforming consumed water, further depressing the WGSR. Overall, this resulted in a decrease of the H2/CO ratio from 1700 at 350 ◦C to 63 at 500 ◦C, and in an increase of the CO selectivity with temperature up to 6%.

An increase of the GHSV caused a decrease in the methane conversion from 23% at 30,000 <sup>h</sup>−1, to 17% at 50,000 h−<sup>1</sup> and 13% at 100,000 <sup>h</sup>−1, which also resulted in lower hydrogen production (Figure 7). Moreover, the conversion values were far from the equilibrium ones, which were set at 29%, 28%, and 25%, respectively. This was due to the minor contact time between the gas stream and the

catalyst. In these conditions, in fact, the reaction appeared to be strongly limited by kinetics, and the catalyst gave conversion results that were far from the thermodynamic equilibrium. The lower methane conversion is accompanied by a higher water partial pressure, which decreased the CO amount due to WGSR and, overall, increased the H2/CO ratio from 43 at 30,000 <sup>h</sup>−1, to 66 at 50,000 h−<sup>1</sup> and 91 at 100,000 h−1. This catalyst behavior showed the predominance of the WGSR in determining the H2/CO ratio, suggesting a faster kinetics for the WGS reaction over the steam reforming one. However, it is difficult to offer a complete comparison among the different tests as lower catalytic bed temperatures are developed at higher GHSV, due to decreased conversions. In fact, lower temperatures decrease the equilibrium conversion and steam reforming activities.

**Figure 5.** Comparison among tests carried out on Rh0.6-CL-R-CZOm at 30,000 <sup>h</sup>−1, 1 atm and 500 ◦C. The graphs report the methane conversion (blue) with respect to the equilibrium values (white bars with black frames), H2/CO ratio (red), and CO selectivity (yellow) versus different S/C ratios.

**Figure 6.** Comparison among tests carried out on Rh0.6-CL-R-CZOm at 30,000 <sup>h</sup>−1, 1 atm and S/C 3. The graphs report the methane conversion (blue) with respect to the equilibrium values (white bars with black frames), H2/CO ratio (red), and CO selectivity (yellow) versus different T. The Y scale related to the H2/CO-ratio graph is very different from the others due to the large values.

**Figure 7.** Comparison among tests carried out on Rh0.6-CL-R-CZOm at 1 atm, S/C 2 and 500 ◦C. The graphs report the methane conversion (blue) with respect to the equilibrium values (white bars with black frames), H2/CO ratio (red) and CO selectivity (yellow) versus different Gas Hourly Space Velocity (GHSV).

At higher pressure, lower methane conversions were observed (from 29% at 1 atm, to 20%, 17%, and 13% at 3, 5, and 10 atm, respectively) (Figure 8), owing to a decreased thermodynamic equilibrium of the SR reaction. However, the equilibrium conversion was close to the experimental conversion at 5 atm (17%) and reached it at 10 atm (13%). In analogy with what observed above, a lower methane conversion is associated with a higher residual amount of water, which enhances the WGSR and, in turn, the H2/CO ratio, which raised from 43 at 1 atm to 108 at 10 atm.

**Figure 8.** Comparison among tests carried out on Rh0.6-CL-R-CZOm at 30,000 <sup>h</sup>−1, S/C 2 and 500 ◦C. The graphs report the methane conversion (blue) with respect to the equilibrium values (white bars with black frames), H2/CO ratio (red) and CO selectivity (yellow) versus different GHSV. The Y scale related to the H2/CO-ratio graph is slightly different from the others due to higher values.

The effect of the S/C ratio was investigated at 500 ◦C and 10 atm (Figure 9). In these conditions, the methane conversion increased with S/C, reaching the equilibrium values up to S/C 2 (9% at S/C 1, 11% at S/C 1.5 and 13% at S/C 2), and the experimental conversion (15%) close to equilibrium value (17%) at S/C 3. The highest conversion was observed at S/C 3 thanks to the increased equilibrium values at higher S/C ratios.

**Figure 9.** Comparison among tests carried out on Rh0.6-CL-R-CZOm at 30,000 <sup>h</sup>−1, 10 atm and 500 ◦C. The graph reports the methane conversion (blue) with respect to the equilibrium values (empty bars with black frame) versus different S/C ratios.

The same tests were also conducted at lower temperature, namely 450 ◦C (Figure 10). In these operative conditions, the experimental methane conversion was close to the equilibrium values up to a S/C ratio of 2, however, it remained distant from it a S/C 3 due to the lower catalytic activity at this temperature. In particular, conversions of 5%, 6%, 7%, and 8% were, respectively, observed by increasing the S/C, while equilibrium conversions concurrently increased from 5% to 7%, then 9% and 12%.

**Figure 10.** Comparison among tests carried out on Rh0.6-CL-R-CZOm at 30,000 <sup>h</sup>−1, 10 atm and 450 ◦C. The graph reports the methane conversion (blue) with respect to the equilibrium values (empty bars with black frames) versus different S/C ratios.

Finally, in order to evaluate the stability of the catalysts in the operative conditions, a test with the same parameters as the first one (500 ◦C, 30,000 <sup>h</sup>−1, 1 atm and S/C 1.5, Figure 11) was periodically conducted after every three tests carried out in different conditions, giving rise to the so-called return tests. The results indicate a rather high catalyst stability towards methane conversion. However, a conversion drop of 4% was observed in the last return test, which was carried out after those conducted at low temperature and high pressure.

**Figure 11.** Methane conversion in deactivation tests carried out at 500 ◦C, 1 atm, 30,000 h−<sup>1</sup> and S/C 1.5.

This deactivation could be due to the formation of carbon on the active site, owing to Boudouard reaction, which in fact is favored at low temperature and high pressure. The Raman analysis on the spent Rh0.6-CL-R-CZOm catalyst (Figure 12) showed two very small peaks related to the presence of carbon. In particular, the bands detected at 1350 and 1580 cm<sup>−</sup><sup>1</sup> could be attributed to the D-band and G-band of carbon, respectively [37–39]. In general, the G-band derives from the stretching of sp<sup>2</sup> carbon bonds in planes of graphene type [28], while the D-band is caused by the vibrations of carbon atoms in disordered species, such as amorphous carbon [40]. Both bands are rather broad and weak even after a long time on stream and different reaction test conditions, demonstrating the properties of the Ce0.5Zr0.5O2 support synthetized by microemulsion in retarding carbon formation, which was also reported previously [28]. Considering the weakness of the bands, other phenomena, such as the re-oxidation of the catalyst during high-pressure tests, where high water partial pressure is present, need to be taken into account. In those conditions, especially in the first part of the catalytic bed where the presence of hydrogen is still very low, the gas mixture had a higher oxidative potential that could affect the oxidation state of the Rh.

**Figure 12.** Raman analysis carried out on the used Rh0.6-CL-R-CZOm catalyst. The two bars indicate the presence of carbon.

#### 2.2.2. Catalytic Tests with Low Rh Loading

The catalysts with a lower Rh loading were prepared diluting the 0.6% sample down to 0.05% wt/wt for further tests, to allow a better evaluation of the activity in a kinetic regime and a better understanding of the deactivation phenomena and metal support effects. Furthermore, the study of catalysts with low amounts of active phase is interesting for some specific applications, such as structured catalysts, membrane reactors, and microreactors, where the active phase may be deposited or coated in thin layers or on low-surface-area supports.

Catalysts with a low % of active phase (0.05% wt/wt) on the two different supports (CZOm and ZOm) were prepared and treated in two different ways: either just heated under N2 or treated firstly under inert gas and then hydrogen. The four produced catalysts were named as follows:


Tests were carried out at fixed experimental conditions (GHSV = 30,000 <sup>h</sup>−1, P = 1 atm, T = 500 ◦C, S/C 1.5) to compare the different catalysts' performances.

Before the catalytic tests, a preliminary screening was carried out to compare the Rh-particle size and distribution on the not-reduced (only heated under nitrogen) CZOm- and ZOm-supported catalysts. The TEM (Transmission Electron Microscopy) analyses performed on both supports evidenced the presence of small Rh particles (1-2 nm) with narrow size distribution (Figure 13). However, it is important to note that the identification of particles below 1 nm is difficult and may be strongly underestimated.

**Figure 13.** Rh particle-size distribution of the Rh cluster deposited on CZOm support determined by Transmission Electron Microscopy (TEM) analysis.

Figure 14 shows the results obtained with the cluster-based catalysts deposited on CZOm in tests carried out at high temperature. An initial conversion around 35% was observed for the catalyst not treated in hydrogen, with a decreasing trend in the first hour and an almost constant value of 30% during the reaction time. This result is in agreemen<sup>t</sup> with the fact that, when using such cluster-derived samples, active species are obtained even without the need of a reduction step (hydrogen treatment). The deactivation of the first part was probably due to the fast sintering of the reactive small particles evidenced by TEM analysis. The sample treated under H2 was able to produce a higher conversion (40%) with stable results over 300 minutes.

**Figure 14.** Methane conversion and hydrogen dry outlet percentage obtained on Rh0.05-CL-N-CZOm (**left**) and Rh0.05-CL-R-CZOm (**right**) tested at 750 ◦C, 30,000 <sup>h</sup>−1, 1 atm and S/C 1.5.

The same test was conducted on an Rh-impregnated catalyst on the same CeZr oxide and reduced at 500 ◦C (Rh0.0.5-IWI-R-CZOm), which represents the best catalyst in oxy-reforming reaction [19] (Figure 15). This test showed an initial methane conversion of 36% with a slightly deactivation trend towards 34%. These results were lower than those obtained with Rh0.05-CL-R-CZO, indicating a higher activity for the cluster-derived catalyst and a readily stable conversion. This could be related to a better dispersion of the Rh cluster on the oxide support.

**Figure 15.** Methane conversion and hydrogen concentration in the dry gas obtained on Rh0.0.5-IWI-R-CZOm tested at 750 ◦C, 30,000 <sup>h</sup>−1, 1 atm and S/C 1.5.

The same test was also conducted on the cluster-based catalyst supported on zirconia (Rh0.05-CL-R-ZOm, Figure 16) and treated under H2, which gave a lower methane conversion (around 33%) with respect to the analogous cluster on CZOm. This is probably related to the e ffect of the Ce(IV)/Ce(III) couple in the sample that acts as steam-dissociating site, and by the higher oxygen mobility in the Ce0.5Zr0.5 O2 phase obtained with the microemulsion synthesis [28,41–44].

**Figure 16.** Methane conversion and hydrogen dry outlet percentage obtained on Rh0.05-CL-R-ZOm tested at 750 ◦C, 30,000 <sup>h</sup>−1, 1 atm and S/C 1.5.

Having assessed the higher conversion of the cluster-based catalysts at 750 ◦C, the tests were also conducted at 500 ◦C. Figure 17 highlights the methane conversion and hydrogen dry outlet concentration versus the reaction time on the two CZOm-supported catalysts.

**Figure 17.** Methane conversion and hydrogen dry outlet percentage obtained on Rh0.05-CL-N-CZOm (**left**) and Rh0.05-CL-R-CZOm (**right**) tested at 500 ◦C, 30,000 <sup>h</sup>−1, 1 atm and S/C 1.5.

The unreduced catalyst provided an initial conversion of circa 9%, while the reduced one gave an initial lower conversion (3%). It has been experimentally demonstrated that hydrogen treatment favors the formation of larger particles [26], and this phenomenon could explain the lower activity of the reduced catalyst with respect to the not-reduced one.

In order to explain the deactivation phenomena that took place in both catalysts, it is important to consider that the reaction environment at low temperature is an oxidizing one due to: i) the low amount of produced hydrogen related to lower kinetics of the catalyst, and also related to the low amount of active phase; ii) the high concentration of water. All this could favor the oxidation of Rh, which is faster when small particles are present, i.e., in the unreduced sample. The oxidation is very likely favored by the presence of Ce(IV) and the high oxygen mobility. The possibility to ascribe the deactivation to sintering of the active phase during reaction can be excluded, since small Rh particles were still found on the unreduced spent catalyst, as evidenced by TEM, where it is clear that the Rh metal particles are only slightly larger (centered between 1.5 and 2.5 nm, Figure 18) than those found on the fresh catalyst.

**Figure 18.** Particle size distribution of the spent Rh0.05-CL-N-CZOm determined via TEM analysis.

The catalytic performances of the Rh cluster deposited on ZOm, both not reduced and reduced (Rh0.05-CL-N-ZOm and Rh0.05-CL-R-ZOm), are shown in Figure 19.

**Figure 19.** Methane conversion and hydrogen dry outlet percentage obtained on Rh0.05-CL-N-ZOm (**left**) and Rh0.05-CL-R-ZOm (**right**).

The initial conversion of the not-reduced catalyst (10.2%) was higher than the one reached with the CZOm-deposited cluster. Moreover, the fast deactivation that characterized the CZOm-based catalysts was not observed in this case, indicating that the presence of Ce favors the oxidation of the Rh. The reduced catalyst was stable too, but produced a lower conversion than the unreduced one, owing to the larger particles formed during the reduction step [26].

Table 2 reports the turnover frequency (TOF) values of the investigated catalysts at 500 ◦C. In the case of the CZOm samples, they were calculated on the initial conversion as those catalysts showed deactivation. Their comparison with TOF values of some other Rh-containing catalysts reported in the literature [45–47], however, is not always suggestive, because of the different supports, active-phase loadings, and operative conditions of the latter.


**Table 2.** Turnover frequency of the catalysts investigated in this work. \* Calculated on the initial conversion.

#### 2.2.3. Investigation on Deactivation

It is particularly interesting to elucidate the mechanism of catalyst deactivation, since no such investigation on carbonyl cluster-derived catalyst is reported in literature, as very few low-temperature studies have been carried out. Having observed that the deactivation was related to the presence of Ce and the oxidizing environment, its cause was further investigated by TEM and EDS (Energy Dispersive X-ray Spectrometry) analyses. Having also excluded deactivation by sintering, and assessed the absence of carbon formed over the surface, it was hypothesized that the deactivation was caused by the oxidation of the active particles favored by Ce, which is able to provide oxygen to the active phase thanks to its redox properties, enhanced by the oxygen mobility of the microemulsion catalyst [19].

This phenomenon benefits from the oxidizing environment found in the low-temperature SR conditions, in which a relatively low amount of hydrogen is produced together with high concentrations of unreacted steam. SAED (Selected Area Electron Diffraction) analyses were carried out during TEM on the Rh particles of the unreduced CZOm- and ZOm-supported catalysts, after reaction at low (500 ◦C) and high (750 ◦C) temperature with the same Rh loading (0.05% wt/wt). In all samples, the presence of Rh in the analyzed particles was evidenced by the EDS analyses. The spent ZOm-supported samples, both at 500 ◦C and 750 ◦C, showed the presence of metallic Rh, as well as the CZOm-supported catalyst after reaction at high temperature (Figure 20). Experimental d-values of 2.25 and 1.84 Å, in fact, could be attributed to the metallic Rh phase, while those at 3.10 and 2.61 Å are related to the support.

**Figure 20.** Selected Area Electron Diffraction (SAED) and TEM analyses carried out on a Rh containing area of the spent not-reduced CZOm supported catalyst after reaction at high temperature.

However, no such phase was observed in the spent CZOm-supported catalyst after reaction at low temperature (Figure 21); the measured d-values of 2.65 and 3.05 Å were attributed to the tetragonal phase of Ce0.5Zr0.5O2. It was not possible to detect the presence of Rh oxide because its reflexes overlapped those of the support. Nonetheless, all these evidences can be confidently seen as an indirect confirmation that oxidation of Rh only occurred on the CZOm-supported catalyst after reaction at low temperature, which is the only one that showed deactivation.

**Figure 21.** SAED and Energy Dispersive X-ray Spectrometry (EDS) analyses carried out on a Rh containing area of the spent not-reduced CZOm supported catalyst after reaction at low temperature.

The hypothesis of the oxidation of Rh is also substantiated by a computational study on metallic Rh clusters on CeO2, which showed that they are stable in reducing conditions such as those obtained in the high temperature steam reforming, but that the relative oxide is formed in oxidizing ones [36].

#### **3. Materials and Methods**
