N₂ adsorption-desorption isotherms and pore size distributions are presented in Figure 1. The associated textural parameters are listed in Table 1. Graphite (G) was characterized by a low surface area and a limited porosity, showing a type IV adsorption/desorption isotherm (Figure 1(Ia)) with a very small volume of adsorbed N₂, which agrees with the low porosity of this material [35]. The pore size distribution showed a narrow peak centered at 33 Å (Figure 1(Ib)). Fishbone type CNF, mesoporous in nature, had a BET surface area and pore volume in the range commonly accepted for these materials (10–300 m² g⁻¹) [34]. The profiles of the N₂ adsorption/desorption isotherms corresponding to these materials (Figure 1(IIa)) can be also assigned to type IV, where the predominant pore size was 256 Å (Figure 1(IIb)) [36].

The carbon structural order was measured by means of temperature-programmed oxidation (TPO) analysis. It is well established that the more structured a carbon material is, the higher the temperature is required for gasification during TPO [37]. Figure 2 and Table 1 show the TPO profiles and the gasification temperatures of the two carbonaceous supports. CNF were oxidized at a lower temperature than G since the former had higher surface area, more defects, and exposed edge planes, and surface C-H/O-H groups, which all together make them much more easily attacked by O₂ [38]. The graphitic character of the supports was also evaluated by XRD. Interlayer spacing (d₀₀₂), average crystalline parameter (L_c) and average number of planes of graphite crystals (npg), shown in Table 1, were completely in agreement with TPO analyses, providing a measure of the structural order of the materials which, increased with both decreasing values of d₀₀₂ and increasing values of both L_c and npg.

Structural features were further assessed by Raman spectroscopy (Figure 3). Raman spectra of the supports exhibited two peaks, commonly denoted as D- and G-bands, at ca. 1354 and 1600 cm⁻¹, respectively. D-band has been attributed to the presence of defects and/or curvature in the carbon structure, while the G-band is associated with well-ordered structures [31,39]. Therefore, the relative intensities of D- and G-bands (I_D/I_G) can be used as an index to assess graphitic character. Again, Raman spectroscopy results corroborated those obtained by XRD and TPO. G support, with the highest graphitic character, generated an intense G-band in the Raman spectrum while CNF supports showed a moderate G-band associated with linearly oriented graphite sheets.

The metal function (Au) was introduced by the impregnation (-IMP) and gold-sol with THPC (Tetrakis(hydroxymethyl)phosphonium chloride) as the surfactant agent (-SGT) methods. Actual metal loading is recorded in Table 2. TPR profiles associated to each catalyst are given in Figure 4, which also includes the response recorded for the corresponding supports. The first hydrogen consumption peak (T_max ranging from 594 to 656 K), common to all the catalysts and absent in the supports, has been attributed to the reduction of the oxidized-Au form to the metallic one (Au⁰) [40]. According to the obtained results, 623 K was chosen as the suitable reduction temperature to ensure metal activation without affecting the surface properties of the support [41,42]. On the other hand, theorical H₂ consumptions values that is, H₂ amount needed to reduce all the Au³⁺ to Au⁰ (47.4 and 27.1 μmol g_cat⁻¹ for Au/CNF-F-IMP and Au/CNF-F-SGT respectively), were always higher than the corresponding experimental ones (12.6 and 9.1 μmol g_cat⁻¹ for Au/CNF-F-IMP and Au/CNF-F-SGT respectively) indicating that, part of the Au was in the metallic form (Au⁰) before TPR reduction. T_max associated with this first hydrogen consumption peak are given in Table 2. It was observed that the reduction temperature was slightly higher for catalysts with lower metal loadings; i.e., catalysts prepared by the gold-sol method. The second hydrogen consumption observed in CNF-based catalysts resulted in a broader peak (T_max ranging from 836 to 894 K) that was attributed to the hydrogen uptake from the support upon decomposition of oxygen surface groups [31,43,44]. Graphite decomposes at very high temperatures (about 1573 K) and, therefore, no response could be observed in the TPR profiles carried out at the maximum temperature of 1273 K. This preliminary assignment was also investigated by temperature-programmed decomposition analyses in He after activation of the Au/C samples at 873 K. It corresponds to the second peak observed in the TPR profile, which also includes the response recorded for both supports. This technique provides qualitative information about the presence of oxygen groups in the carbon structure, as reported elsewhere [45]. The profiles presented in Figure 5 demonstrated that the signal of the support was stronger than that corresponding to the thermal treatment of the catalysts after TPR, due to the decomposition of oxygen surface groups from the support (in He at T ≥ 873 K). Note that the ascribed signal is mainly formed as consequence of the CO and CO₂ release resulting from the decomposition of these groups in the form of water, carboxylic, lactone, anhydride, phenol and carbonyl and quinone groups [45].

XRD patterns for the Au-based catalysts are presented in Figure 6. XRD profiles were similar, regardless the method of gold introduction onto the support. Peaks at 2θ ≈ 26° correspond to the (002) graphite plane of carbon (JCPDS-ICDD Card No. 41-1487). Peaks observed at 38°, 44°, 64° and 77° correspond, respectively, to the (111), (200), (220) y (311) planes of metallic gold (JCPDS-ICDD Card No. 01-1172), which was consistent with an exclusive hexagonal geometry of Au⁰ [46]. These results indicated that after reduction the metallic phase was in the Au⁰ form.

Representative TEM micrographs (Figures 7 and 8) and TEM derived particle size distributions (Figure 9) showed the morphology and size of the reduced Au particles. According to the obtained Au particle size (Table 2), Au particles were probably anchored onto the carbon surface since, they could not penetrate between the carbon layers (d₀₀₂ values were much lower than the supported Au particle size). This fact is corroborated by the negligible change in the pore diameter, and by the pore blockage that took place after the Au introduction. In all cases, TEM micrographs showed small and dispersed Au particles, being this observation, as expected, more marked when G was used as the support and when catalysts were prepared by the gold-sol method (THPC stabilized the colloid gold solutions [47]). Regardless the metal introduction method used, the Au particle size distribution was considerably narrower when G was used as the catalytic support.

On the order hand, it could be also observed that the presence of structural defects (low crystallinity) played an important role in the Au deposition and, as consequence, in its catalytic activity. Higher crystallinity and, therefore, higher carbon content in the structures (Table 1), resulted in a greater number of graphite edges exposed in an orderly manner (see L_c values, Table 1), ensuring a strong anchoring of small metal particles. Thus, G, with almost no porosity, generated the smallest surface area-weighted average Au diameter due to its high crystalline nature. In this material a great number of ordered edges were exposed, resulting in a greater anchoring of small, thin, faceted and well dispersed Au particles, revealing a strong metal-support interaction [48]. Nevertheless, Au particles supported over fishbone type CNFs were larger and, consequently, worse dispersed.

Carbon nanofibers were prepared at atmospheric pressure by the catalytic chemical vapour deposition method (CVD) in a fixed-bed reactor consisting of a horizontal quartz tube (100 cm length × 9 cm i.d.). The synthesis was conducted over a Ni/SiO₂ (10% w/w) catalyst at 873 K employing ethylene as the carbon source and hydrogen as the carrier gas (C₂H₄/H₂, 4/1 v/v, 900 cm³ min⁻¹). The growth time was 1 h and the space velocity was 10,000 h⁻¹. The carbon deposit obtained was demineralized using HF (48% v/v) and dried for 12 h at 383 K in air to remove water prior to its characterization. Graphite (supplied by Aldrich) underwent the same treatment with this acid to make their properties comparable. Further details regarding the CNF synthesis are given in a previous work [58].

The gold catalysts were prepared via the gold-sol method with THPC as the surfactant agent (-SGT) [41] and the wet impregnation method (-IMP) [59], using HAuCl₄·3H₂O (Sigma Aldrich) as the metal precursor salt. In the gold-sold method, the metallic sols were first prepared stirring H₂O and NaOH before adding the diluted THPC. After that, the gold aqueous solution was added. The gold-sols were then stirred for 1 h. Aqueous suspensions of the carbon supports were agitated in an ultrasonic bath. Finally, the gold-sols were added into the carbon suspensions and stirred for 1 h for immobilization. In the impregnation method, the sample was placed in a glass vessel and kept under vacuum at room temperature for 2 h in order to remove water and other compounds adsorbed on the carbon material. A known volume of an aqueous metal precursor solution was then poured over the carbon material. Next, the solvent was removed by evaporation under vacuum. The metal content added to the catalyst was controlled by measuring the metal concentration in the impregnating solution.

To finish with the catalyst preparation, the suspension was filtered, washed with deionized water until the filtrate was free of chloride (AgNO₃ test), and dried at 383 K for 24 h. All the catalysts were sieved into a batch of 75 μm average diameter and then reduced at 623 K in a flow of H₂ for 2 h.

Au metal loading was determined by atomic absorption (AA) spectrophotometry using a SPECTRA 220 FS analyzer. Samples (ca. 0.5 g) were treated in 2 cm³ HCl, 3 cm³ HF and 2 cm³ H₂O₂ followed by microwave digestion (523 K). Surface area/porosity measurements were conducted using a Micromeritics ASAP 2010 sorptometer apparatus with N₂ as the sorbate at 77 K. All the samples were outgassed prior to analysis at 453 K under vacuum (5 × 10⁻³ Torr) for 16 h. The total specific surface areas were determined by the multipoint BET method. Microporosity was evaluated by the t-plot and Horwath-Kawazoe (HK) methods and mesoporosity was evaluated by the Barret-Joyner-Halenda (BJH) method. XRD analyses were conducted with a Philips X'Pert instrument using nickel-filtered Cu Kα radiation. Samples were scanned at a rate of 0.02° step⁻¹ over the range 5–90° (scan time = 2 s step⁻¹). The resulting diffractograms were compared with the JCPDS-ICDD references [60]. Micro-Raman spectra of the supports were recorded with a Renishaw Raman Microscope System RM1000 equipped with a Leica microscope, an electrically refrigerated CCD camera and a diode laser at 514 nm as the excitation source, operating at a power level of 3 mW. Temperature-programmed oxidation (TPO) profiles were obtained using a PerkinElmer TGA7 thermogravimetric analyzer, where samples (ca. 10 mg) were ramped from room temperature to 1273 K (5 K min⁻¹) in a 50 cm³ min⁻¹ O₂/He (5% v/v) flow. The carbon content in the solid carbon deposits was determined using a LECO CHNS-932 apparatus. Temperature-programmed reduction (TPR) experiments were conducted in a commercial Micromeritics AutoChem 2950 HP unit with TCD detection. Samples (ca. 0.1 g) were loaded in a U-shaped quartz reactor and ramped from room temperature to 1273 K (5 K min⁻¹) and in a reducing atmosphere (17.5% v/v H₂/Ar, 60 cm³ min⁻¹). Prior to the analysis, samples were treated in an Ar atmosphere (50 cm³ min⁻¹) from room temperature to 523 K (10 min) and from 523 to 773 K (20 min).

Temperature-programmed decomposition profiles were recorded using a Micromeritics TPD/TPR 2900 apparatus. The samples were activated in situ under H₂ (60 cm³ min⁻¹) at 5 K min⁻¹ to 873 K, cooled to room temperature and then heated (5 K min⁻¹) to 1273 K in a flow of He (100 cm³ min⁻¹). Changes in the outlet gas composition were monitored by a TCD device. Acquisition and further treatment of data was performed with the Micromeritics 2900 commercial software.

Transmission electron microscopy (TEM) analysis was made with a JEOL JEM-4000EX unit (an accelerating voltage of 400 kV). Samples were prepared by ultrasonic dispersion in acetone with a drop of the resultant suspension evaporated onto a holey carbon-supported grid. Mean Au particle size evaluated as the surface-area weighted diameter (d̅_s) was computed according to: d ̅ s = ∑ i n i d i 3 n i d i 2where n_i represents the number of particles with diameter d_i (Σ_in_i ≥ 400).

Measurements of the catalytic activity were carried out within a specific tubular quartz reactor. The dimensions of the reactor were 45 cm length and 1 cm of diameter, being the catalyst placed on a fritted quartz located at the end of the reactor. The temperature of the catalyst was measured with a K-type thermocouple (Thermocoax) placed inside the inner quartz tube. The entire reactor was placed in a furnace (Lenton) equipped with a temperature-programmed system. The reaction gases were Praxair certified standards of CO (99.99% purity) and N₂ (99.999% purity), which was used as the carrier gas. The gas flows were controlled by a set of calibrated mass flowmeters (Brooks 5850 E and 5850 S). The content of water in the reaction mixture was controlled using the vapor pressure of H₂O at the temperature of the saturator. All lines placed downstream from the saturator were heated above 100 °C to prevent condensation. The saturation of the feed stream by water at the working temperature was verified by a blank experiment where the amount of water trapped by a condenser was measured for certain time and compared with the theoretical value. Products gases were analyzed with a micro gas-chromatograph (Varian CP-4900).

Prior to the reaction test, the catalyst was reduced with pure H₂ (100 mL/min) at atmospheric pressure and 623 K for 2 h. Catalytic activity was studied in the temperature range of 573–633 K. Gas hourly space velocity (GHSV) was slightly varied in each experiment by adjusting the catalysts weight in order to get always the same Au content in the reactor (2.7 × 10⁻³ g). The total gas flow rate was fixed at 100 NmL/min. Once operating conditions remained constant, steam was added to the preheated feed gas upstream of the reactor. Two different H₂O/CO molar ratios (2 and 4) were used. Effluent gas composition was analyzed online at short intervals of time. A reaction time of 1.5 h was allowed for steady state to be achieved. Taking into account that CO could be also converted to CH₄ (CO methanation) [14], hydrogen selectivity was calculated using the following equation: Hydrogen selectivity ( % ) = CO 2 selectivity ( % ) − 3 ( CH 4 selectivity ( % ) )

The activity of the different catalysts used in this work will be shown as experimental reaction rates, which allowed to establish a better comparison of their catalytic performance.

Finally, supports were checked to be stable at hydrothermal conditions. Experiments performed at 633 K by flowing 100 NmL/min of a feed stream containing 40% H₂O and N₂ balance demonstrated that no traces of CO₂ or CO was detected in the effluent; i.e., no gasification of any of the supports was observed.