**1. Introduction**

The reduction of catalytic activity over time is an issue of considerable and continuing concern in industrial practices of catalytic processes. Catalyst replacement and process shutdown could cost the industry very large financial resources every year. The catalyst deactivation changes extensively; for example, in the case of catalytic methane decomposition, catalyst deterioration may be on the scale of seconds, whereas in NH3 synthesis the Fe-catalyst may stay for 5–10 years. Nevertheless, it is unavoidable that all catalysts drop their activities.

Hydrogen is a pure fuel source which can replace fossil fuels. It can be utilized to run numerous devices like fuel cells, engines, vehicles, and electric devices [1]. Subsequently, the attention given

to hydrogen production has been progressively growing these past years. H2 can be obtained from several sources and methods, for example, reforming of hydrocarbons, biomass, electrolysis, and photo-splitting of water, in addition to water-gas shift reaction [2–6]; however, the steam reforming reaction is the main conventional method for hydrogen production. The high H2/CO ratio of the saturated hydrocarbons, particularly CH4, makes them the major feedstock for H2 production by reforming reactions. For example, partial oxidation, if properly controlled, is a more suitable method to produce hydrogen than the dry-reforming process because the H2/CO ratio is 2 and the reaction is mildly exothermic, whereas dry-reforming being endothermic is energy-intensive with a low ratio: H2/CO = 1 [7]. The graphene hydro/dehydrogenation process might be used as an active and eco-friendly device to yield and store hydrogen from water. Its mechanisms involve water decomposition at the graphene/metal interface at room temperature to hydrogenate the graphene sheet, which is buckled and decoupled from the metal substrate. Likewise, thermal programmed reaction experiments demonstrate that molecular hydrogen can be released upon heating the water-exposed graphene/metal interface above 400 K [8].

The endothermic catalytic methane decomposition (CMD) reaction [9] could provide a promising substitute for the conventional processes, like steam reforming, partial oxidation, and autocatalysis of methane, used for hydrogen production [10–12]. The CMD process produces pure hydrogen and valuable carbon (i.e., free carbon nanotubes) [13–16]. The formation of carbon nanotubes (CNT) can be of high economic interest; it enjoys many applications, such as catalysts, catalyst-support, H2 storage, electronic components, and polymer additives [17–20]. Transition metals such as Ni, Co, and Fe supported on different oxides, such as MgO, Al2O3, and SiO2, are often used for CMD reaction [21–24]. Furthermore, it has been stated that the properties of a single metal catalyst can be improved by introducing a second metal to form a bimetallic catalyst [24–28]. Numerous articles have been lately published on bimetallic catalytic systems for CMD reaction [29–34].

The utilization of supported bimetallic Ni–Fe, Ni–Co, and Fe–Co catalysts was examined for catalytic methane decomposition by several researchers such as Awadallah et al. [35]. The catalytic data showed that the bimetallic catalyst exhibited remarkably higher activity and stability and a higher yield of multi-walled carbon nanotubes. The Fe and Co play the role of active metal in the catalyst; they are cheap and abundant compared to noble metals and Ni. The catalytic activity of 30 wt % Fe and 15 wt % Co bimetallic catalyst, reduced at 500 ◦C and operated at 700 ◦C, showed excellent performance among all the tested catalysts which comprised Fe/Al2O3, Co–Fe/Al2O3, Ni–Fe/Al2O3 with different amounts of Co and Ni loading [13]. The present work underscores the features of catalyst regeneration performance in activity and stability.

Pudukudy et al. [34] investigated the direct decomposition of CH4 over SBA-15-supported Ni-, Co- and Fe-based bimetallic catalysts. Their results specified that all of the bimetallic catalysts were highly active and stable for the reaction at 700 ◦C even after 300 min of time on stream. Co–Fe/SBA-15 catalyst revealed high catalytic stability [34]. Bimetallic Ni–Fe, Ni–Co, and Fe–Co supported on MgO catalysts with a total metal content of 50 wt % were examined for CMD by Awadallah et al. [35]. The catalytic data exhibited that the bimetallic 25%Fe–25%Co/MgO catalyst displayed remarkably higher activity and stability up to ~ 10 h of time on stream and a higher yield of multi-walled carbon nanotubes [35]. The deactivation of heterogeneous catalysts due to carbon deposition is a global issue that causes a reduction in catalytic activity with time. Various types of carbon and coke that change in morphology and reactivity are created. The more reactive, amorphous forms of carbon created at low temperatures are transformed into less reactive, graphitic forms at high temperatures over a period of time [36]. Normally, catalyst regeneration is considered to restore catalytic activity by removing carbon, poisons, and site blockage [37]. Hazzim et al.; studied the regeneration for CMD using an activated carbon catalyst [38]. They found that the activity at the start and the final mass gain of the catalyst increased as the reaction temperatures rose from 850 to 950 ◦C. However, at 850 and 950 ◦C reaction temperatures, the activity and mass gain declined after each regeneration step. The decrease was slower under severe regenerating conditions [38].

In this work, a cobalt–iron supported on alumina catalyst (noted 15%Co–30%Fe) was synthesized for catalytic methane decomposition (CMD) reaction (CH4 → C + 2H2:ΔH◦ = 75:6 KJ/mol). The effect of the catalyst regeneration was evaluated by using different oxidizing forced periodic cycling. The regenerated catalysts were tested again in CMD reaction for coproduction of hydrogen and carbon nanomaterials. This study focused on identifying the nature of the carbon deposits formed after each CMD testing that precedes the regeneration cycles; also in addition, the obtained yields of hydrogen were quantified to understand the reaction's mechanism.

#### **2. Results and Discussion**

Because of the materials' complexity, structural and textural properties and morphology of spent/regenerated catalysts (referred to as SP-180 min, SP-360 min, and SP-720 min) were examined using X-ray diffraction (XRD), transmission electron microscopy (TEM), laser Raman spectroscopy (LRS), N2-physisorption, hydrogen temperature programmed reduction (H2-TPR), and thermogravimetric analysis (TGA), and the carbon amount was evaluated by atomic absorption spectrometry (AAS). For each of these three samples, the oxidative regeneration process was performed every 90 min of the test. For comparison, the un-regenerated SP-90 min was used as a reference.

#### *2.1. Structure and Morphology*

XRD patterns of fresh catalyst (15Co–30Fe/Al2O3), SP-90 min, and three spent/regenerated samples are shown in Figure 1. As mentioned previously [39], the diffractogram of the fresh solid showed the presence of γ-Al2O3 phase at 2θ ca. 40◦, 49◦, and 64◦ in accordance with (ICDD# 004-0875) lines. In this sample, additional peaks related to Fe3O4 magnetite were found at 2θ ca. 32◦, 35.5◦, 45◦, 55◦, and 60◦ according to (ICDD# 04-006-6550), whereas Fe2O3 hematite was not observed at 2θ: 32◦, 35◦ (most intense) and 25◦, 50◦, 54◦ (less intense) in accordance with (ICDD# 089-0598). Peaks observed at 2θ ca. 32◦, 37◦, 45◦, 56◦, 59◦, and 66◦ correspond to CoAl2O4 spinel (ICDD#44-0160). The formation of CoAl2O4 was ascribed to the strong interactions between cobalt species and γ-Al2O3 lattice during the synthesis process. No peak corresponding to Co–Fe mixed oxides like CoFe2O4 spinel was observed; this was plausibly due to a stronger interaction between cobalt and alumina which have more pronounced acidic properties than that of iron oxides.

The XRD pattern of spent SP-90 min was completely different from that of the fresh catalyst; it was dominated by the reflection of (200) planes observed in 2θ = 20◦–30◦ range, attributed to carbon species with a graphite-like structure. No reflections corresponding to CoAl2O4 spinel, Fe3O4 magnetite, or Fe2O3 hematite phases were detected. The presence of γ-Al2O3 phase was not clearly ascertained. However, a close inspection of this diffractogram revealed the presence of the common reflection of cobalt and iron chemically in the oxidation state zero, indicating the reduction under the reaction mixture. Part of CoAl2O4 and Fe3O4 directly forms Co and Fe metallic species or probably Co–Fe alloy. The authors supposed that the reduction of CoAl2O4 and Fe3O4 was incomplete under CH4 atmosphere. The presence of unreduced species CoAl2O4 spinel and Fe3O4 could not be clearly checked because their lines are probably overlapped by those of the Co and Fe metallic species.

In the case of the SP-180 min sample, the pattern performed after the oxidative regeneration process was similar to that of SP-90 min, but the intensity of the peaks was different. The SP-180 min pattern revealed the presence of graphitic carbon with very low-intensity peaks due to the removal of surface carbon in the form of CO2. Cobalt and iron were always observed in the metallic form, that is, no cobalt and iron oxides were observed after oxidative treatment. From these results, it might be inferred that filamentous carbon stabilized the cobalt and iron in oxidation state zero.

In the SP-360 min and SP-720 min samples, the intensity of the carbon lines increased with the time of CMD reaction due to an additional deposit of carbon in well-structured forms that is difficult to eliminate during the regeneration process. The presence of Fe<sup>0</sup> metallic species was always ascertained, but that of Co<sup>0</sup> is unlikely. However, the diffractogram revealed the presence of a new phase attributed to CoC2 cobalt carbide and identified after Rietveld refinement; this identification is observed by a

slight peak splitting situated at 2θ ≈ 45◦ (ICDD#44-0962). No reflections corresponding to iron carbide species (like Fe5C2 or Fe3C) could be observed in these samples.

Nevertheless, with a refinement scan, some traces of Fe3O4 magnetite and Fe2O3 hematite were identified indicating the re-oxidation of iron during the regeneration process. Unlike the spent/regenerated sample SP-180 min, the SP-360 min and SP-720 min samples revealed the presence of a new phase recorded as θ-Al2O3 observed at 2θ positions of 29.8◦, 41.0◦, 44.2◦, 50.81◦, 61.8◦, 67.9◦, and 79.1◦ in accordance of (ICDD#086-1410). From these results, it could be inferred that the θ-Al2O3 would probably help to make and stabilize more iron species at the surface of SP-360 min and SP-720 min samples [40,41].

**Figure 1.** X-ray diffraction (XRD) patterns of fresh, spent, and spent/regenerated catalysts.

Transmission electron microscopy (TEM) was performed in order to ge<sup>t</sup> more information on the carbon morphology detected by XRD. TEM micrographs of the spent and spent/regenerated catalysts are shown in Figure 2. In all cases, the formation of filamentous-type carbon was confirmed. It is known that the fibers of carbon nanotubes (CNTs) are composed of two kinds: (1) single-walled carbon nanotubes (SWCNTs) which consist of a single tube of graphite and (2) multi-walled carbon nanotubes (MWCNTs). Generally, in TEM analysis, it is difficult to establish the length of nanotubes to compare them. However, it is well known that nanotube diameters can range from just a few nanometers for SWCNTs to several tens of nanometers for MWCNTs. In their study, Jorio et al. [42] demonstrated by TEM that diverse SWCNT diameters varied between 0.7 and 3.0 nm. On the other hand, Hou et al. [43], using the same technique, found that MWCNT were generally in the diameter range from 10 to 200 nm. In this work, only MWCNT fibers with 27–53 nm diameters were identified (Figure 2). Similarly, in an earlier report [39] of CMD on SP-90 min, the lengths of the produced carbon (CNT) at 90 min of CMD (observed by TEM) varied between 14.09 and 72.0 nm, which were the characteristics of MWCNTs.

**Figure 2.** Top surface transmission electron microscopy (TEM) micrographs of spent and spent/regenerated 15Co–30Fe/Al2O3 catalysts obtained after different times (90, 180, 320, and 720 min).

For all spent and spent/regenerated samples, TEM micrographs showed large amounts of graphitic carbon layers around the Co metal particles. In fact, these Co particles were mostly encapsulated within CNTs; on the other hand, the Fe particles clung to the CNT surface in accordance with XRD analysis, which showed, after oxidative regeneration, a re-oxidation of iron metallic species in Fe3O4 magnetite or in Fe2O3 hematite (according to XRD observations); it however showed no oxidized form of cobalt which was encapsulated and stabilized with filamentous carbon. According to the literature [44], a metal encapsulated into the CNTs was not very accessible to reactants and therefore resistive to any oxidative regeneration process. Besides metal particles encapsulated (Co) and metallic particles located on top of nanocarbons (Fe), all TEM images (Figure 2) showed agglomerate black particles clearly observed in the tips of MWCNTs and dispersed on the CNT surface. These black particles are more numerous on the unregenerate sample (SP-90 min) and are attributed to condensation of carbon nanoparticles. These black particles could be easily removed in CO2 form ( C → CO2 ) during the regeneration process.

Table 1 summarizes the average size of carbon crystallites determined by XRD and obtained by TEM. The crystallite size (XRD) was estimated using the Debye–Scherrer formula [45] for the most intense (002) peak of carbon and using the equation τ = 0.94 × λ FWHM × Cos( 2θ2 ), where λcu = 1.5406 Å, τ (Å) is the carbon crystallite size, 2θ (radian) is the diffraction peak position, and Full width at half maximum (FWHM) (radian) is full width at half maximum describing the width measurement of carbon peak. In TEM, the calculation of sizes was realized using the ImageJ software package (1.52 h, National Institutes of Health, Madison, WI, USA, 2018). Except for SP-720 min (with instrumental broadening ×50 against ×200), the TEM investigation displays homogeneous distribution of carbon species with a particle size 52.31, 49.28, and 53.23 nm for 90, 180, and 360 min samples, respectively. Similar trends were observed by XRD for these three samples (49.74 and 42.63 Å). In the case of the SP-720 min, the crystallite sizes (XRD or TEM) were the smallest as compared to those of other catalysts because of the removal of amorphous carbon species from the MWCNT surface (loss of carbon C → CO2) [46].


**Table 1.** Estimation of carbon crystallite size using XRD and TEM techniques.

(\*): XRD calculations with Debye–Scherrer formula on the (002) most intense reflection; (\*\*): TEM calculations from 05–08 zones of each micrograph.

As a conclusion, XRD and TEM measurements are complementary techniques, but not comparable: XRD involves some heterogeneity properties such as crystalline structures, types (carbon, carbide, etc.), and amorphous regions, whereas TEM gives information on the particle organization of carbon (homogeneous distribution), dispersion, and distribution of metals.

To confirm the carbon nature and structure observed by XRD, laser Raman spectroscopy (LRS) was applied to provide useful information about material crystallinity, phase transition, and structural disorder. The spectra of the spent and spent/regenerated samples are shown in Figure 3. The single band observed between 1000 and 2000 cm<sup>−</sup><sup>1</sup> could be associated with different carbon bonds; it revealed the presence of a carbon graphite structure on the molecular level. According to previous reports [47], this band with a maximum at 1363.6 cm<sup>−</sup><sup>1</sup> was attributed to the D-band of sp<sup>2</sup> carbon material. Deconvolution of this band revealed two components (Figure 3); the second component, centered at 1580.54 cm<sup>−</sup><sup>1</sup> is associated with the G-band. According to Ferrari [48], the Raman G-band was a characteristic of graphite with high crystallinity and the D-band was attributed to defects and lattice distortions in the carbon structures; an increase in the intensity of the D-band reflected an increase in the disorder of the carbon atoms and thus a restructuring of the nanotubes. Therefore, the ratio of D-band intensity on G-band intensity (ID/IG) gave an indication of defects or graphitic order. In Figure 3, the ratio ID/IG was calculated after deconvolution. The ID/IG ratios, which varied in the order of 1.84 (360 min) > 1.75 (180 min) >1.66 (720 min) > 1.51 (90 min), indicated the formation of new phases detected by XRD in spent/regenerated samples, which could be directly responsible for the higher level of lattice distortion in the carbon graphite structures. However, the values (ID/IG = 1.66–1.84), obtained after regeneration, had slightly increased compared to that of SP-90 min (ID/IG ~ 1.51), which means that with the oxidative treatment, the structure of the nanotubes was disturbed but not totally damaged.

**Figure 3.** Laser Raman spectra of SP samples (90, 180, 360, and 720 min).
