Dry Reforming of Methane over a Ruthenium/Carbon Nanotube Catalyst

In this study, CH4 dry reforming was demonstrated on a novel microwave-synthesized ruthenium (Ru)/carbon nanotube (CNT) catalyst. The catalyst was tested in an isothermal laboratory-packed bed reactor, with gas analysis by gas chromatography/thermal conductivity detection. The catalyst demonstrated excellent dry-reforming activity at modest temperatures (773–973 K) and pressure (3.03 × 105 Pa). Higher reaction temperatures favored increased conversion of CH4 and CO2, and increased H2/CO product ratios. Slight coke deposition, estimated by carbon balance, was observed at higher temperatures and higher feed CH4/CO2. A robust global kinetic model composed of three reversible reactions—dry reforming, reverse water gas shift, and CH4 decomposition—simulates observed outlet species concentrations and reactant conversions using this Ru/CNT catalyst over the temperature range of this study. This engineering kinetic model for the Ru/CNT catalyst predicts a somewhat higher selectivity and yield for H2, and less for CO, in comparison to previously published results for a similarly prepared Pt_Pd/CNT catalyst from our group.


Introduction
Because of enhanced petroleum production by hydraulic fracturing, vast quantities of co-produced natural gas are flared annually worldwide [1,2]. In Texas, for example, a lack of sufficient gas pipeline capacity has severely depressed local natural gas market prices, resulting in negative prices and more flaring [3]. In addition to flaring, there is also concern for fugitive CH 4 emissions associated with petroleum production [4]. Methane has a global warming potential 30 times higher than CO 2 [5].
Methane dry reforming (DR) (overall: CH 4 + CO 2 → 2CO + 2H 2 , ∆H 298K = 2.473 × 10 5 J) would remove two greenhouse gases while generating useful synthesis gas (H 2 , CO) as a chemical feedstock [6][7][8], especially for important chemicals such as ammonia and methanol, and for Fischer-Tropsch liquids [9,10]. The energy needed for the reforming can be supplied by burning some of the CH 4 . The CO 2 can be trucked in, especially from sequestration sources. While this scheme does not ultimately eliminate the greenhouse gases, it does provide some useful return in the form of fuels and/or chemicals. Alternatively, the DR could be solar powered [11].
Typical heterogeneous DR catalysts use inorganic oxide supports; e.g., Al 2 O 3 , SiO 2 , TiO 2 [6,[12][13][14]. The active sites fall into two groups. First are base-metals, including Fe, Co, and Ni [6,12,13]. The Ni is widely studied since it is catalytically active and cheap. However, the Ni catalyst sinters at common DR reaction temperatures (800-1100 K), and has a relatively weak coking resistance [6,13]. Cobalt offers an improved coking resistance compared to Ni [15]. In a recent development [16], a highly stable and coke-resistant catalyst composed of Mo-doped Ni nanoparticles on a MgO support was demonstrated for high DR activity.
The second active site group consists of noble metals, commonly Rh, Ru, Pt, Pd, and Ir [6,14,17]. Although expensive, noble metal catalysts resist coking, sinter less, and are more active for higher-temperature applications.
A potentially useful support for DR catalysts is carbon nanotubes (CNTs). The CNTs are one-dimensional cylindrical structures consisting of wrapped single or multi-layer graphene sheets [18]. Its special structure results in excellent thermal and electrical conductivities, high mechanical strength, large surface area, relatively high oxidation stability, surface chemical flexibility and porous structure [19].
The advantages of CNTs as catalyst supports for a number of heterogeneous catalysis processes have been identified. Dry reforming of CH 4 over Ni/CNT and Ni/SiO 2 catalysts produced approximately the same conversions, but Ni/CNT showed much better stability. This was attributed to a redirection of carbon deposits away from the Ni sites to growth along the CNT tips [20]. Dry reforming of CH 4 over Co/CNT and Co/MgO showed higher CH 4 conversions from Co/CNT, which also had lower carbon deposition rates and less catalyst deactivation [21].
In this work, Ru supported on a CNT catalyst was synthesized by a versatile microwave technique, and then tested for methane DR activity. There is no current literature on the use of Ru (as the sole metal) on CNT used as DR catalyst for CH 4 . Observed data are compared against equilibrium. An engineering model, useful for calculations such as reactor design, and based on three global reactions, adequately represents the DR data.

Carbon Nanotubes
The multi-walled carbon nanotubes (Cheap Tubes Inc., Cambridgeport, VT, USA) used were 10-30 × 10 −6 m length by 20-30 × 10 −9 m outer diameter. Other compounds used in the synthesis (Sigma-Aldrich) were 95%+ purity. The synthesis strategy begins with CNT functionalization (becoming F-CNTs) with carboxylic acid groups, followed by the addition of Ru.
The Ru/CNT synthesis is described in detail elsewhere [22]. The synthesis begins with CNT carboxylation to enhance dispersibility for the subsequent Ru addition. A known mass of CNTs are dispersed into concentrated H 2 SO 4 + HNO 3 (aqueous). The suspension is radiated with microwaves to 413 K for 1200 s, and then cooled (room temperature). Vacuum filtering (10 −5 m pore size) follows, with Milli Q water washing to pH ≈ 7. These F-CNTs are then vacuum-dried at 343 K.
The dried F-CNTs are then dispersed in diethylene glycol by sonication. A RuCl 3 .nH 2 O, sodium acetate, and Milli Q water mixture is added, followed by more sonication. The new dispersion is processed in the microwave reactor at 473 K for 600 s. It is estimated that a roughly 20% conversion of the Ru starting material occurs. Upon cooling, the mass is vacuum filtered and ethanol-washed. The Ru/CNT are finally vacuum-dried at 343 K to a constant mass.

Catalyst Testing System
The apparatus used for the Ru/CNT catalysis testing is described elsewhere [23,24]. The CH 4 , CO 2 , and diluent He flow rates were set by calibrated mass flow controllers. Catalyst was uploaded into the fritted metal cup of a 6.35 × 10 −3 m stainless steel Swagelok (Nupro) in-line filter. The stainless steel tube assembly was placed in a 3-zone electric furnace. Results presented elsewhere [23] show that the steel reactor vessel was inert during these DR experiments.
The on-line gas analysis was performed with a model 5890 Hewlett-Packard gas chromatograph/ thermal conductivity detector (GC/TCD) fed through a gas sample loop and valve. The pressure in the reactor flow system was monitored and manually controlled. The operator alternately directed feed (bypass) or reactor effluent to the on-line GC/TCD. In this DR research, the inlet CH 4 and CO 2 were diluted by 85% He, resulting in approximately constant total molar flow rate. The experimental CH 4 , CO, CO 2 mole fractions allowed atomic carbon balances, and estimates of any carbon deposits. These data also facilitated H and O balances for H 2 and H 2 O concentrations. Where feasible, H 2 levels determined this way were verified by the GC/TCD data for H 2 .

Catalyst Characterization
The Ru/CNT was characterized using a scanning electron microscope (SEM, JEM 2800) and a transmission electron microscope (TEM, F2000). Figure 1 (left) shows SEM images of the synthesized Ru/CNT. After Ru deposition, the CNTs are intact with no apparent damage. Close inspection of the TEM image in Figure 1 (right) reveals Ru nanoparticles. Their approximate particle size range is 1-5 × 10 −9 m. Subsequent SEM testing of the catalyst after all the DR runs showed little if any sintering of the Ru. The thermogravimetric analysis (TGA, in air) of our Ru/CNTs is shown in Figure 2. Weight loss in the 403-873 K range is due to Ru/CNT decomposition in air. The 10% remaining mass is likely RuO 2 . Elemental analysis from energy-dispersive X-ray spectroscopy (EDX) ( Table 1) is consistent with this hypothesis.  To extend the physical mass of the solids for ease of handling, the Ru/CNT were mixed with Y-zeolite of similar particle size. The Y-zeolite was the inert Na form, used as delivered (Alfa Aesar, Haverhill, MA, USA) with a 5.5:1 molar ratio SiO 2 :Al 2 O 3 . The solids mixture was 20 wt.% Ru/CNT. Results presented elsewhere [23] show that the added zeolite was effectively inert during our DR experiments. The particles (~3 × 10 −5 m diameter) were small enough to avoid significant external and internal mass transfer resistances (Mears and Weisz-Prater criteria, respectively, as discussed in [23]). Based on CO adsorption and BET surface area tests of the Ru/CNT-zeolite catalyst, the Ru site density was~9 × 10 −8 moles/m 2 [23].

Modeling
Modeling begins with reaction equilibrium calculations. Then we propose a three reaction global engineering kinetic model calibrated against experimental data.

Equilibrium Calculations
In this study, all experimental runs were tested with the Chemkin-Pro ® equilibrium application [25]. The calculation is based on the element-potential method used in Stanjan [26]. The equilibrium composition minimizes total Gibbs Free Energy.
The equilibrium calculation is run with specified temperature (constant), pressure (constant), and feed composition. The allowed equilibrium species are H 2 , H 2 O, CO, CO 2 , CH 4 , He, and solid carbon (Cs, when allowed-assumed graphite). The Chemkin-Pro database provides the thermodynamic properties of the species. Table 2 compares the equilibrium simulation results with and without the solid carbon Cs for an experimental case. For a feed where CH 4 /CO 2 = 1, the amount of H 2 is roughly about the same in the two cases. However, allowing Cs results in much less equilibrium CH 4 and CO. Indeed, Cs is the second largest quantity species after H 2 , and the equilibrium H 2 /CO ≥ 1. Without Cs, equilibrium H 2 /CO < 1. A similarly large impact of Cs in equilibrium was reported elsewhere [27]. Table 2. Impact of solid carbon (Cs) on equilibrium; T = 973 K, P = 3.03 × 10 5 Pa, initial CH 4 /CO 2 = 1.

Global Kinetic Model
An engineering global kinetic model has the advantage of offering relatively easy calculations for an otherwise complex reaction system. It is not meant to substitute for a detailed elementary reaction mechanism that requires a complex software package such as Chemkin-Pro to evaluate. The global model is a screening tool, and is limited to the region over which it is calibrated.
Analysis began with an experimental carbon balance. Measured feed CH 4 + CO 2 mole fractions were compared to measured outlet CH 4 + CO 2 + CO. Heavy He dilution minimized the impact of any changes in total molar rate by reactions. Figure 3 presents a parity plot of the experimental carbon balance. Points below the diagonal suggest carbon deposition occurred during those runs. A closer look of the marker distribution in the graphic reveals:

1.
Circle I shows the runs at 773, 823, and 873 K. These are very close or even at the parity line suggesting little carbon deposition at these lower temperatures.

2.
Circle II and III present the runs at 923 K and 973 K. Compared to circle I, higher temperature favors coke formation.

3.
Circle III shows two cases at 973 K. They are even further from the parity line than the other runs at 973K (Circle II, feed CH 4 /CO 2 range 0.51-1.01) due to higher feed CH 4 /CO 2 (1.53, 2.08). The same observation is obtained at the other temperatures-higher CH 4 /CO 2 at comparable temperatures favors coke formation.

4.
For all runs, the further below the parity line, the higher H 2 /CO is observed. It is implied that the formation of higher H 2 is coincident with coke formation.
In view of Figure 3, our global kinetic model uses the following three reactions: Equation (1) presents the ideal DR reaction, and is chosen naturally. The reverse water gas shift (RWGS, Equation (2)) occurs during reforming [28]. Reactions 1 and 2 suggest H 2 /CO <1, which is coincident with most of our cases. However, numerous runs showed H 2 /CO >1 at higher temperatures and feed CH 4 /CO 2 . This might be explained by either the CH 4 decomposition (MD, Equation (3)) generating more H 2 , or Boudouard reaction (2 CO → CO 2 + Cs) consuming more CO. However, higher temperatures favor MD equilibrium but discourage Boudouard. It has been claimed that DR occurs through a catalytic decomposition of CH 4 to adsorbed C and H atoms, thus facilitating carbon deposits [28]. Finally, Reactions 2 and 3 are known to occur during CH 4 DR [21].
The selected reversible reactions for the global model and their rate expressions are summarized in Table 3. If the approach to equilibrium η i > 1, the reaction goes left to right; if η i < 1, the reaction goes in reverse; if η i = 1, the reaction is at equilibrium. The equilibrium constants K pi (Table 3), are regressed vs. temperature from an on-line calculator [29]. For Reactions 1 and 3, the first order CH 4 kinetics are suggested elsewhere [28]. The Reaction 2 first-order CO 2 kinetics are also suggested elsewhere [30]. A regression strategy was used to estimate the Arrhenius parameter pairs (A i , E i ), Assuming a packed-bed reactor (PBR) model (Table 4), our DR experiments were simulated. The experimental CH 4 and CO 2 conversions are sufficiently high (>10%) that an integral PBR model is used rather than a simple differential reactor. Table 3. Global kinetic engineering model with primary dry reforming (DR) and secondary reverse water gas shift (RWGS), methane decomposition (MD) reactions.

Reaction
Rate Expression r i Appr. to Equil. η i K pi Dry Reforming CH 4 + CO 2 = 2CO + 2H 2 Reverse Water Gas Shift

PBR Balances Species j Net Rates r j Mole Fractions y j Partial Pressures
Total molar rate includes inert gas P j = y j P P = total pressure An original Matlab program was developed to simultaneously integrate all PBR species balances with the input of all experimental inlet mole fractions and flow rates at a given temperature and pressure, and then to compare the calculated outlet mole fractions with experimental values. The Matlab program repeats the regression process, resulting in optimized rate constants k i at that temperature. The optimized rate constants k i at each temperature were correlated (Figure 4) to obtain quality Arrhenius parameters of each reaction (Table 5).  Table 5 for Reaction (Rxn) 1 (DR), 2 (RWGS), 3 (MD). A similar analysis method [31] was successfully applied to optimize a 3-global reaction set for methane dehydroaromatization (MDA), inspired elsewhere [32]. In this MDA study, a detailed kinetic mechanism for the surface elementary reactions [33] was evaluated in a PBR model to create a simulated experimental database.
Using the best-fit Arrhenius parameters (Table 5) with the 3-reaction global kinetic set, the experimental DR runs were again simulated, with predicted values compared to the observed. Figure 5 shows example simulated and experimental (outlet) mole fractions at 923 K, 3.03 × 10 5 Pa, feed 1.12 × 10 −6 m 3 /s, and feed CH 4 /CO 2 = 0.5. Species H 2 and CO rise steadily, while CO 2 and CH 4 both drop. The experimental outlet mole fractions are well simulated by the global 3-reaction model. Offering further insight, Figure 6 shows model-based η i profiles for the Figure 5 case. The DR and MD are far from equilibrium everywhere, while RWGS moves quickly toward equilibrium [28].

Discussion
The results above illustrate how the 3-reaction global model above was built up based on the observed reactor outlet species concentrations with various experimental temperatures and feed molar CH 4 /CO 2 ratios. In this section, these results are discussed by demonstrating the utility of the 3-reaction global model by comparison against useful quantities such as reactant conversions and product H 2 /CO ratios. The equilibrium values are also presented in these cases for comparison.

Species Concentrations at Reactor Outlet
For all experimental runs, the CH 4 , CO 2 , CO mole fractions were directly measured by GC/TCD. Oxygen atom balances were used, based on measured inlet and effluent CO 2 and CO concentrations, to estimate H 2 O. No trace of O 2 was detected in any case. Combined with H atom balance, H 2 was estimated using measured inlet and outlet CH 4 .
To compare the 3-reaction model with experimental results, Figure 7 shows outlet species for all cases at 873 K at constant feed rate but different feed CH 4 /CO 2 . Figure 8 presents the compositions of all cases at CH 4 /CO 2 = 1.0 with constant feed rate as a function of temperature. Acceptable agreements establish the precision of the 3-reaction model. The outlet concentrations of CO, H 2 O, and H 2 are significantly impacted by reaction temperature, less so by the inlet CH 4 /CO 2 .

Methane and Carbon Dioxide Conversions
Estimated CH 4 and CO 2 conversions are based on inlet and outlet mole fractions due to the heavy He dilution. Figure 9 shows variations with temperature of experimental, equilibrium, and 3-reaction model predicted conversions at feed CH 4 /CO 2 of 2.0 and 0.5. The 3-reaction model performed very well in predicting observed conversions.  The CH 4 equilibrium conversions are much greater than in the experiment, and are insensitive to temperature, suggesting a kinetic potential to improve conversions. Other experimental temperature runs showed similarly consistent results. Figure 10 shows similar results with variation of feed CH 4 /CO 2 . At 973 K, experimental CO 2 conversions are much closer to equilibrium, suggesting that 973 K is nearly the upper temperature limit for improving CO 2 conversions with this catalyst.

Syngas Molar Ratio H 2 /CO
As a high syngas H 2 /CO ratio is often preferred for many processes, the product mole fraction ratio H 2 /CO is an important index of reforming catalyst effectiveness [34]. Figure 11 shows that higher temperatures and feed CH 4 /CO 2 favor higher H 2 /CO.  The stoichiometric H 2 /CO for the ideal DR reaction is 1.0. At 773 K, the observed H 2 /CO at feed CH 4 /CO 2 = 1 is <1. This is attributed to the RWGS reaction, which is more thermodynamically favored at these relatively low temperatures [35]. At 973 K with CH 4 /CO 2 = 1, the experimental H 2 /CO is ≈1.0.

Comparison of Ru/CNT to Pt_Pd/CNT
A study similar to the one presented here was performed earlier with Pt_Pd/CNT catalyst, also admixed with inert zeolite [24]. Both the Ru/CNT and Pt_Pd/CNT catalysts were prepared in very similar ways. Based on CO adsorption and BET surface area tests of the metal/CNT-zeolite catalysts, the Pt_Pd and Ru catalyst site densities were estimated as 1.3 × 10 −7 and 9.0 × 10 −8 mole/m 2 , respectively [23].   Table 6 presents calculated results from the 3-reaction (DR, RWGS, MD) engineering kinetic models presented in this paper for Ru, and in [24] for Pt_Pd. Each simulation was run with a hypothetical packed-bed reactor with 1 gram of metal/CNT-zeolite catalyst held at a constant 773 K, 3.03 × 10 5 Pa pressure, feeding CH 4 and CO 2 at 5.6 × 10 −5 mole/s total rate (equimolar). While the differences are fairly small, there is a definite trend that the Ru/CNT catalyst, even with a smaller site density, produces a synthesis gas richer in H 2 . The drawback is a slightly larger Cs. Table 6. Comparison of performances from Ru/CNT (model, this study) and Pt_Pd/CNT (model from [21]). T = 773K, P = 3.03 × 10 5 Pa, 0.001 kg catalyst, total feed 5.6 × 10 −5 mole/s, feed CH 4 /CO 2 = 1.

Conclusions
An Ru catalyst supported on CNT, prepared by a novel microwave synthesis technique, was tested for activity on dry reforming (DR) of CH 4 to synthesis gas. The DR studies were undertaken in an isothermal packed bed reactor. The catalyst showed impressive activity and stability at modest temperatures. The outlet species concentrations and conversions were influenced by temperature and feed molar ratio of CH 4 /CO 2 . Based on the analysis of the experiment data, a small amount of carbon deposition was observed during DR, and it was enhanced by higher temperatures and feed CH 4 /CO 2 . Reactant conversions did not reach equilibrium values, suggesting potential to improve conversions and product yield. In order to account for both observed reactant conversions and product species concentrations, a global engineering model consisting of three reversible reactions was developed. The model reactions are dry reforming (ideal), reverse water gas shift, and methane decomposition. The three-reaction model adequately represents the observed species profiles as functions of temperature (773-973 K) and feed CH 4 /CO 2 (0.5-2.0). Linear Arrhenius plots for the forward rate constants are observed for each reaction in the model over the calibration temperature range. The calibrated engineering model for the Ru/CNT catalyst predicts a higher selectivity and yield for H 2 in comparison to previously published results for a similarly prepared Pt_Pd/CNT catalyst from our group, although at the expense of slightly more carbon deposits.