*2.2. Surface Characterization*

The Brunauer–Emmett–Teller BET surface areas were determined by nitrogen adsorption-desorption at 77 K. All nitrogen adsorption–desorption isotherm curves and variations of the specific surface areas are depicted in Figure 4. According to the International Union of Pure and Applied Chemistry (IUPAC) standard, the catalysts exhibited the IV type isotherm showing that the materials were essentially mesoporous. The fresh sample exhibited H4 hysteresis loop which corresponded to the narrow slit-like pores, whereas the hysteresis loops of spent (90 min) and spent/regenerated (SP-180, 360, and 720 min) samples had the characteristic of a type H3 loop. Compared to fresh and not regenerated samples, the spent/regenerated samples showed a lower amount of nitrogen adsorbed at the high relative pressure (P/P0 = 1), which suggests a strong decrease in the mesoporosity because of the aggregated pores between the CNTs. In addition, as is shown in Figure 4, the displacement of the hysteresis loops toward the lower partial pressures is an indication of the porosity's evolution.

**Figure 4.** N2 adsorption-desorption isotherms and variation of specific surface areas for fresh, spent, and spent/regenerated catalysts at different reaction times.

BET surface areas of spent and spent/regenerated samples were relatively small compared to that of the fresh sample. They decreased linearly with the time of CMD reaction in the following order: fresh-sample (122 m2/g) > SP-90 min (109 m2/g) > SP-180 min (74 m2/g) > SP-360 min (64 m2/g) > SP-720 min (61 m2/g). During the CMD reaction or oxidative regeneration process, the formation of new phases (such as Co and Fe metallic species, CoCx, carbon, and θ-Al2O3, etc.) caused the blocking of the pores with the subsequent loss of surface area as evidenced by XRD analysis. This result was consistent with AAS findings, which showed an increase in the mass of carbon deposited with the number of reaction cycles. The decline in catalyst activity might be related to the reduction in porosity which occurred during the decomposition process.

#### *2.3. H2-TPR and TG Analysis*

Hydrogen temperature programmed reduction experiments were carried out in 100 to 1000 ◦C temperature range under a hydrogen atmosphere. H2-TPR profiles of fresh, spent, and spent/regenerated 15Co–30Fe/Al2O3 samples are shown in Figure 5a,b. Figure 5a is for fresh catalyst, whereas Figure 5b is for the spent catalyst. The reduction profiles can be classified into three regions. In region I (200–450 ◦C), the reduction of hematite (Fe2O3) to Fe3O4 and the reduction of Co3O4 to CoO take place. In region II (450–710 ◦C), the reduction of Fe3O4 to FeO and the reduction of CoO to Co occur. In region III (710–900 ◦C), the reduction FeO to Fe happens. Similar results were observed by other investigators [49–51].

**Figure 5.** (**a**) H2-TPR for fresh catalyst, (**b**) H2-TPR for used catalyst, and (**c**) thermogravimetric (TG) profiles, of 30Fe-15Co/Al2O3 catalysts at different regeneration temperatures.

From the XRD analysis, the possible reducible phases in fresh sample were the following: Fe3O4, CoAl2O4 spinel, and probably some amorphous (or highly dispersed) Fe2+ and Co2+ species which could not be detected easily by XRD. Al2O3 was non-reducible under this work's conditions due to the high binding energy between Al and O. Large differences in H2-TPR profiles were observed that depended on the time of reaction and the number of decomposition-oxidation cycles. However, for all samples, the reduction peaks observed below 450 ◦C were ascribed to the reduction of Co2+ species to nanoparticles of metallic cobalt, whereas the peaks observed between 450 and 900 ◦C were ascribed to the reduction of Fe3+ species of metallic iron. In the case of the SP-360 min and SP-720 min samples compared to other samples, several reduction peaks were observed at low and high temperatures; they could be ascribed to successive steps of the reduction to cobalt and iron species and/or to the presence of various types of Co and Fe species of different interactions with the support or with different particle size. Indeed, it is well known that the increase in the degree of the agglomeration and consequently the size of metal oxide particles could cause a decrease in the metal-support interaction and make reduction easier. From H2-TPR results, it should be pointed out that the H2 consumption decreased with the increasing reaction time (or the number of decomposition-regeneration cycles), probably because of the increase in the encapsulation rate of metallic particles in a zero-valence state that makes their oxidative regeneration difficult due to their inaccessibility to hydrogen molecules. This could justify the lower reducibility of both SP-360 min and SP-720 min samples. During the H2-TPR process, a hydrogasification of carbon deposits could also happen at temperatures above 700 ◦C by releasing methane (C(s) + 2H2 → CH4). An analysis of gaseous effluents (using, for example, mass spectrometry) is however necessary to assess this hypothesis.

TGA results of spent and spent/regenerated catalysts are shown in Figure 5c that shows a weight change in the temperature range 100–850 ◦C. All samples displayed weight loss (WL) from 500 to 850 ◦C which was related to the combustion of different kinds of carbon (cobalt carbide amorphous carbon and carbon nanotubes). The TGA curves showed that the weight loss decreased in the order WL(SP-90 min) > WL(SP-720 min) > WL(SP-360 min) > WL(SP-180 min). As expected, the highest weight loss was observed in the spent and not regenerated sample (SP-90 min). Comparing the regenerated samples, it was found that SP-180 min exhibited the lowest weight loss, indicating that the accumulation of encapsulating carbon deposition (difficult to remove) increased with the increasing number of decomposition-oxidation cycles. This result corresponded to the XRD finding which exhibited a more intense peak of carbon for samples having undergone several decomposition–regeneration cycles. In the case of SP-720 min, a change was noticed in the slope of the TGA curve, which suggested two kinds of coke deposit may be related to the more abundant presence of cobalt carbide over this catalyst.

#### *2.4. Catalysts Behavior under CMD-Regeneration Cycles*

In most cases, during CMD reaction, carbon encapsulates the active sites and this carbon accumulation is responsible for catalyst deactivation. To increase the catalyst lifetime, regeneration cycles by gasification using oxygen as an oxidizing agen<sup>t</sup> was introduced as has been described before. The regeneration step obviously required the interruption of the production process for 10 min. The catalytic activity of catalysts was determined by evaluating methane conversion and hydrogen yield obtained during the CMD process. In the present work, Weight hourly space velocity (WHSV) is calculated from the feed gas consisting of 15 mL/min of CH4 and 10 mL/min of N2 and weight of catalyst 0.3 g, as follows:

$$\text{WHSV} = \frac{(15+10)\text{ mL/min} \times 60\text{ min/h}}{0.3\text{ g-cat}} = \frac{25 \times 60\text{ mL/(h)}}{0.3\text{ (g-cat)}} = 5000\text{ mL/(h\cdot g\text{-cat.})}$$

The methane conversion obtained here for regenerated catalysts is above 60%. Zhou et al. conducted an Fe catalyst for CMD and investigated the effect of space velocity on the methane conversion. When a space velocity (3750 mL/g-cat/h), close to the one used in the present work (5000 mL/g-cat/h), was considered, they found a methane conversion of about 38% [52].

Figure 6 compared the evolution, with time of stream (TOS), of the methane conversion obtained over the fresh catalyst and the spent catalysts subjected to successive cycles of CMD/regeneration. As can be seen, the results were not the same after the regeneration cycles. Only the points on the curve at 90, 180, 360, and 720 min corresponding to the samples SP-(90, 180, 360, 720 min) were considered. As is shown, the methane conversion decreased, as the reaction time was increased, in the order of 68% (90 min, zero cycle) > 63% (180 min, after two cycles) > 62% (360 min, after four cycles) > 56% (720 min, after eight cycles). This degradation of catalytic activity and loss of stability of the catalysts (about Δ% = 12% from 90 to 720 min of TOS) was due probably to irreversible deactivation caused by covering and encapsulating of active sites by deposited carbon. The same trend was observed in the BET surface area which decreased when the number of decomposition/regeneration cycles was increased, as is shown in Figure 3. On the other hand, it was shown that the reducibility of catalysts decreased when the number of decomposition/oxidation cycles was increased; this was attributed to the blocking of metallic particles (in the oxidation state zero) by the formation of encapsulating carbon species, consequently leading to the loss of active sites and catalyst deactivation. From these results, it could be concluded that the degradation of physicochemical and catalytic properties was essentially related to carbon formation that was not completely removed after repetitive oxidative regeneration steps. Therefore, the higher the amount of carbon deposited, the lower the catalytic activity.

Through this study's experiments, the authors were able to verify that significant catalyst deactivation occurred toward higher regeneration times (involving a succession of cycles) despite a very interesting catalytic formulation. The formation of cobalt carbide CoC2 may then complicate the regeneration of metal catalysts (Figure 1), as reported in the literature [52].

**Figure 6.** Time on stream (TOS) yield of hydrogen versus different regeneration times with 15%Co–30%Fe/Al2O3 spent-catalyst. TOS conditions: Tregeneration-activation = 500 ◦C; Treaction = 700 ◦C; (WHSV = 5000 mL/H·g-cat), total flow rate = 25 mL/min and CH4/N2 = 1.5.

Figures 6–8 present, respectively, the evolution with time on stream (TOS) of H2 yield and the coproduced carbon on SP-(90, 180, 360, 720 min) samples. The hydrogen production by 67.8% (90 min) > 66.8% (180 min) > 63.7% (320 min) > 57.7% (720 min) followed the same evolution as the methane conversion. Similar to the conversion of methane, the hydrogen yield was affected by the number of regeneration cycles. It decreased with the increase in the regeneration number, but to a lesser extent. On the other hand, in the case of the carbon deposit, an increasing trend of carbon with the increase in the regeneration number was observed: 1.2 (180 min) < 1.5 (320 min) < 3.2 (720 min).

As reported by some authors [53], the catalyst deactivation mechanism rests on progressive pore blocking by carbon deposition. The carbon formation involved in the first step is the dissociative adsorption of methane on the metal surface (iron and cobalt) producing hydrogen and carbon atoms. The adsorbed carbon atoms could dissolve and diffuse through the metal. Those on the surface form, thereby encapsulating carbon, blocking the access of methane to the active sites, and causing a reduction of activity, as proposed by the following steps. During CMD reaction, the carbon may chemisorb firmly as a monolayer or physically adsorb in multilayers while releasing hydrogen [54]. In either case, this blocks the reactants from reaching the metal surface. The behavior is complex, because the carbon threads may grow from the top surface of the metal particles or the carbon may spread into the metal and form bulk carbides (e.g., formation of CoC2), as observed in the study's XRD. According the H2-TPR conclusions, it can be stated that methane molecules may be formed through the reaction between carbon encapsulating particles and chemisorbed hydrogen gases.

The formation of encapsulating carbon on the metallic surface could be limited by Cads + H2 → CH4 reverse reaction (with H2 produced by reaction or added to the feed). This step was favored in the presence of a high hydrogen concentration in the feed which depended on the methane conversion. Besides coke deposition, the sintering of metallic active phase (Co, Fe) occurred during CMD reaction and oxidative regeneration process and was the second factor responsible for catalyst deactivation. Similar to carbon deposition, sintering brings about the loss of surface area and porosity. Despite the coke deposits and many oxidative treatments, the present catalysts remained active with a methane conversion and hydrogen yield exceeding 56% and 58%, respectively (after eight reaction cycles).

Further tests were needed to optimize the regeneration step to eliminate quantitatively the carbon deposits and improve the catalytic performance.

**Figure 7.** Comparison of different H2 yields (%) obtained from 90, 180, 360, and 720 min of reaction (the white bar is a limit between each regeneration cycle).

**Figure 8.** Carbon yields obtained after 90, 180, 360, and 720 min of reaction.
