Effects of Catalysts and Membranes on the Performance of Membrane Reactors in Steam Reforming of Ethanol at Moderate Temperature

Abstract

Steam reforming of ethanol in the membrane reactor using the Pd₇₇Ag₂₃ membrane was evaluated in Ni/CeO₂ and Co/CeO₂ at atmospheric pressure. At 673 K, the H₂ yield in the Pd₇₇Ag₂₃ membrane reactor over Co/CeO₂ was found to be higher than that over Ni/CeO₂, although the H₂ yield over Ni/CeO₂ exceeded that over Co/CeO₂ at 773 K. This difference was owing to their reaction mechanism. At 773 K, the effect of H₂ removal could be understood as the equilibrium shift. In contrast, the H₂ removal kinetically inhibited the reverse methane steam reforming at low temperature. Thus, the low methane-forming reaction rate of Co/CeO₂ was favorable at 673 K. The addition of a trace amount of Ru increased the H₂ yield effectively in the membrane reactor, indicating that a reverse H₂ spill over mechanism of Ru would enhance the kinetical effect of H₂ separation. Finally, the effect of membrane performance on the reactor performance by using amorphous alloy membranes with different compositions was evaluated. The H₂ yield was set in the order of H₂ permeation flux regardless of the membrane composition.

1. Introduction

Hydrogen has been considered as one of the most promising clean energies because its combustion emits only water. Most of the hydrogen produced currently comes from catalytic steam reforming of natural gas [1]. Steam reforming of natural gas is a mature technology as a practical application, and it has been employed for hydrogen production from various hydrogen sources such as liquefied petroleum gas [2,3] iso-octane [4] and kerosene [5,6,7,8]. Considering the sustainable society, hydrogen production from fossil fuels is undesirable, and it would shift to renewable sources such as biomass-derived fuels. In particular, biomass-derived liquid fuels are preferable to direct hydrogen storage for on-site hydrogen production owing to their convenience for storage and transport. Among biomass-derived liquid fuels, bioethanol has been frequently studied for hydrogen production because it is easy to handle and distribute and it is readily available [9]. This process can be realized under far milder conditions than those of methane steam reforming. Therefore, the steam reforming of bioethanol is more attractive from practical and environmental view points. In the last decade, many researchers have made efforts to develop catalysts for steam reforming of ethanol aimed at hydrogen production [10,11,12]. Noble metal catalysts (Pd, Pt, Rh, Ru) exhibited high catalytic activity and stability [13,14,15]. Inexpensive catalysts involving Ni or Co have been also proposed as appropriate candidates in terms of activity and selectivity [16,17,18,19]. However, the process efficiency for hydrogen production must be further improved when hydrogen can be used as an energy source for fuel cells. Additionally, extremely purified hydrogen is required for the low-temperature fuel cells such as proton exchange membrane fuel cells because of irreversible catalyst poisoning even by a trace amount of CO [20,21].

Membrane reactors using hydrogen selective membranes such as Pd and Pd alloy membranes are expected to be one of the promising technologies to achieve high hydrogen production efficiency with high hydrogen purity, because extremely purified hydrogen can be obtained in one step with high hydrogen yield over the thermodynamic equilibrium, owing to simultaneous hydrogen separation from the reaction zone. Therefore, hydrogen production from several fuels using membrane reactors have been extensively studied so far [22,23,24]. Recently, steam reforming of ethanol has been also investigated by several research groups [25,26,27,28,29,30]. Basile and co-workers systematically investigated the performance of a membrane reactor using a Pd-Ag membrane packed with Co/Al₂O₃ catalysts in steam reforming of ethanol [25,26], and achieved 100% ethanol conversion, 95.0% CO-free hydrogen recovery and up to 60% CO-free hydrogen yield at 673 K and 3.0 bar. They also investigated the effect of by-products such as acetic acid and glycerol as well [27]. Llorca and co-workers investigated the effect of reactor configurations over Co talc [28] and Co hydrotalcite [29] and demonstrated the higher performance of the catalytic membrane reactors using Pd-Ag membranes compared to the staged membrane reactor, where the catalyst has been placed in-series with the membrane. The enhancement of reactor performance by simultaneous hydrogen separation was also reported by Oyama’s group. They evaluated the effect of membrane performance on the reactor performance by comparison between Pd-Cu and SiO₂-Al₂O₃ membranes and found that both permeance and selectivity had a favorable effect on steam reforming of ethanol in membrane reactors [30]. These studies clearly demonstrated the positive effect of membrane reactors in steam reforming of ethanol. However, there are only a few comparison studies on the effect of catalysts and membranes, and, in this study, we evaluated the Pd-Ag membrane reactor packed with Ni or Co based catalysts and then carried out the comparison study of the effect of membranes on the reactor performance using Pd-Ag and amorphous alloy membranes.

2. Experimental Section

2.1. Preparation of Catalysts

For the experiment, 15 wt% Ni/CeO₂ and 15 wt% Co/CeO₂ were prepared as the literature reported [31]. For Ni/CeO₂, nickel acetate tetrahydrate was dissolved in deionized water and stirred at 343 K. After adding CeO₂ (mean particle size: 1 μm), the pH was adjusted to 9 by adding 0.25 M Na₂CO₃ aqueous solution. Then, the water was slowly vaporized at 373 K to obtain the precipitation, and the precipitation was calcined at 673 K for 5 h. For preparation of Co/CeO₂, the preparation procedure was the same to that for Ni/CeO₂. Cobalt acetate tetrahydrate was used as the cobalt source.

Preparation of Ru-Co/CeO₂ and Pd-Co/CeO₂ was as follows. Ruthenium chloride or palladium chloride was dissolved in deionized water and stirred. Then, Co/CeO₂ was added in the solution (M/Co = 0.003 w/w, M = Ru or Pd) and the solvent was slowly vaporized at 373 K. The obtained solid material was calcinced at 823 K under an N₂ flow for 5 h.

2.2. Preparation of Amorphous Alloy Membranes

Alloy ingots were prepared by arc melting a mixture of pure metals with the appropriate composition. After remelting the alloys several times to make better homogeneity, a ribbon sample was obtained by a single roller melt-spinning method. Both surfaces of the ribbon were polished and sputtered with Pd coating with the thickness of approximately 100 nm. The appearance of obtained membrane was shown in Figure 1.

2.3. Characterization

Hydrogen permeation tests. The amorphous alloy membranes (size: 10 mm × 10 mm) were placed in the separator sealed by Cu gaskets. The membrane was pre-heated in vacuum up to 673 K. Then, the membrane was cooled down to 623 K. After keeping the membrane at 673 K, H₂ was introduced at the appropriate pressure (0.05, 0.10 and 0.15 MPa-G). The flow rate at the permeation side was measured by the soup-film flow meter.

Catalytic tests. The steam reforming of ethanol was carried out in a conventional reactor and membrane reactor. In the conventional reactor, the 10 g of catalysts was placed in the reactor and heated to 773 K under an N₂ flow. Then, the catalyst was reduced under an H₂ flow at 773 K for 1 h. After controlling the reaction temperature (623–773 K), a mixture of water and ethanol with the steam to carbon ratio (S/C) of two was fed into the reactor at atmospheric pressure. The products were analyzed with a GC-8A gas chromatograph (Shimadzu Corporation, Kyoto, Japan) and the outlet gas flow rate was measured by the soup-film flow meter.

In the membrane reactor, the catalysts and membrane were placed in the reactor. The reactor was heated to 773 K under an N₂ flow, and catalyst was reduced for 1 h under an H₂ flow. The Ar sweep gas was used at the permeation side in the reactor. A mixture of water and ethanol with the S/C of two was fed into the reactor after staying at the reaction temperature. The total pressure at both feed and permeate sides was maintained at atmospheric pressure. The gas composition in the retentate and permeate side was analyzed with the gas chromatograph and the outlet gas flow rate in both sides was measured by the soup flow meter. A Pd₇₇Ag₂₃ membrane (thickness: 20 μm, purchased from Tanaka Kikinzoku Kogyo K.K., Tokyo, Japan) was used as reference.

The conversion of ethanol to C₁ products and H₂ yield was calculated as follows:

$$Conversion of ethanol to C_1 products=\frac{flow rate of CO, C{O}_2 and C{H}_4 in products}{feed rate of ethanol\times 2},$$

$$H_{2}yield= \frac{flow rate of H_2 in products}{feed rate of ethanol\times 6}.$$

Additionally, we defined H₂ removal ratio as follows to compare the membrane reactor performance:

$$H_2 removal ratio \left[\%\right]= \frac{H_2 permeation flow rate [mol/min]}{Theoritical H_2 production rate [mol/min] }\times 100.$$

3. Results and Discussion

3.1. Membrane Reactor Performance Using Pd₇₇Ag₂₃ Membrane with Co/CeO₂ and Ni/CeO₂ Catalysts

Figure 2 shows the comparison of reactor performance of the conventional reactor and membrane reactor with the Pd membrane over Co/CeO₂ and Ni/CeO₂ catalysts. In the conventional reactor, Ni/CeO₂ exhibited higher conversion of ethanol to C₁ products compared to Co/CeO₂ catalysts. However, the H₂ yield over Ni/CeO₂ did not exceed those over Co/CeO₂ at the reaction temperatures from 673 to 773 K. Regardless of catalysts, the membrane reactor exhibited higher conversion and H₂ yield than the conventional reactor, but the influence of Pd₇₇Ag₂₃ membrane was different between the catalysts. The increase in conversion of ethanol to C₁ products over Ni/CeO₂ by the Pd₇₇Ag₂₃ membrane was much higher than those over Co/CeO₂ catalysts, and it achieved more than 90% at 723 K, whereas the conversion was only 65% at 773 K over Co/CeO₂. The H₂ yield over Co/CeO₂ increased by almost 13% regardless of temperature, e.g., 44.4% and 57.8% at 773 K for the conventional reactor and membrane reactor, respectively. In contrast, it was found that the increase in H₂ yield over Ni/CeO₂ was significantly improved with an increase in the reaction temperature (from 27.4% at 673 K to 64.5% at 773 K). These results indicate that the influence of H₂ removal from the reaction zone through the Pd₇₇Ag₂₃ membrane was higher over Ni/CeO₂ than Co/CeO₂ at the high reaction temperature. However, Co/CeO₂ was suitable for the membrane reactor at low reaction temperature. Figure 3 shows the selectivity of C₁ products. For Ni/CeO₂, the main product at 673 K was methane (58.1%) and a slight decrease in the methane selectivity was observed (51.1%) in the membrane reactor at this reaction temperature. The methane selectivity was slightly decreased to 41.4% at 773 K in the conventional reactor, and the simultaneous H₂ removal by Pd₇₇Ag₂₃ membrane greatly enhanced the CO₂ selectivity with decreasing of the methane selectivity. Furthermore, the membrane reactor showed lower methane selectivity of 25.9% at 773 K. For Co/CeO₂, the main product in the conventional reactor was CO₂ regardless of the reaction temperature and the methane selectivity was very low (28.8% and 18.9% at 673 K and 773 K, respectively) compared to Ni/CeO₂. In the membrane reactor, the methane selectivity was decreased, and it was noting that the decrease in methane selectivity was much higher at 673 K than 773 K in Co/CeO₂. Torres et al. reported the difference in reaction path in the steam reforming of ethanol [16], and their reaction scheme is summarized in Scheme 1. In the literature, at high reaction temperature, the steam reforming of ethanol was dominant over both Ni and Co catalysts. However, the reaction path was largely different at the moderated reaction temperature between Ni and Co catalysts. The initial reaction was ethanol dehydrogenation to acetaldehyde over both catalysts. The product selectivity approached the thermodynamic equilibrium in Ni catalysts because of the methane-forming reaction such as ethanol cracking, acetaldehyde decarbonilation and the reverse methane steam reforming. On the other hand, the Co catalysts did not promote such methane-forming reactions and the steam reforming of acetaldehyde was preferably occurred. Considering their reaction scheme, the highly increased H₂ yield over Ni/CeO₂ in the membrane reactor at 773 K was due to the shift of equilibrium by simultaneous H₂ removal through the Pd₇₇Ag₂₃ membrane. However, at 673 K, it was easy for Ni/CeO₂ to produce methane and the methane steam reforming would be hardly promoted once methane produced even when H₂ was selectively removed from the reaction zone because the methane steam reforming is kinetically and thermodynamically unfavorable at low temperature. Therefore, the low H₂ yield was owing to the preferential production of methane in both conventional and membrane reactors at low temperature. In contrast, because of the low reaction rate of methane-forming reactions in Co/CeO₂, the Pd₇₇Ag₂₃ membrane could remove produced H₂ from the reaction zone before H₂ was consumed by the reverse methane steam reforming. This means the selective H₂ removal through Pd₇₇Ag₂₃ membrane kinetically inhibited the methane production, resulting in high H₂ yield at low reaction temperature. Indeed, Seelam et al. reported Co/Al₂O₂ that showed high membrane performance at 673 K compared to Ni/ZrO₂ because of high methane selectivity in Ni/ZrO₂ [27], which is consistent with our experimental results.

3.2. Catalyst Development for Improving the Membrane Reactor Performance

Comparing Co/CeO₂ and Ni/CeO₂, Co/CeO₂ was preferable to Ni/CeO₂ for steam reforming of ethanol at low temperature, and the H₂ removal through the Pd₇₇Ag₂₃ membrane was very effective at achieving high H₂ yield because of the kinetic inhibition of methane-forming reactions. To improve the membrane reactor performance, we developed the catalysts with addition of a trace amount (M/Co = 0.003 (w/w)) of precious metals such as Ru and Pd.

Table 1. Performance of conventional and membrane reactors over Ru-Co/CeO₂ and Pd-Co/CeO₂ in steam reforming of ethanol.

Catalysts		Conversion to C1 (%)	H2 Yield (%)	C1 Selectivity (%)
				CO2	CO	CH4
Co/CeO2	CR	63.4	47.6	70.3	6.0	23.6
Ru-Co/CeO2	CR	86.8	42.3	59.0	3.3	37.7
	MR	87.9	70.3	81.8	4.0	14.2
Pd-Co/CeO2	CR	92.7	30.8	51.6	1.9	46.5
	MR	94.2	52.9	65.3	2.4	32.3

Reaction temperature: 623 K, W/F: 1.0 × 10⁴ g-cat min/C-mol. S/C = 2. Membrane: Pd₇₇Ag₂₃, Sweep Ar flow rate: 500 mL/min. CR: conventional reactor, MR: membrane reactor.

shows the performance of Ru-Co/CeO₂ and Pd-Co/CeO₂ in steam reforming of ethanol in the conventional and membrane reactors. It was interestingly found that the addition of a very small amount of Ru or Pd greatly enhanced the conversion of ethanol to C₁ products. However, the H₂ yield in the conventional reactor was not changed in Ru-Co/CeO₂ and significantly decreased in Pd-Co/CeO₂ compared to Co/CeO₂ (here, it should be noted that the H₂ yield and conversion to C₁ product on Co/CeO₂ was higher than those in Figure 2 even at the same reaction condition. This was owing to the refinement of flow system on the reactor in Figure S1). The low H₂ yield can be explained by the thermodynamic equilibrium. The C₁ selectivity at the thermodynamic equilibrium is CO₂:CO:CH₄ = 30.32:0.15:69.53. Indeed, the methane selectivity approached the thermodynamic value with the increased conversion by the addition of Ru and Pd, resulting in low H₂ yield because produced H₂ was consumed by reverse methane steam reforming from CO₂ and CO. In the membrane reactor, the H₂ yield in both Ru-Co/CeO₂ and Pd-Co/CeO₂ was increased by the simultaneous H₂ separation although the conversion was not changed. In addition, the methane selectivity was decreased by the H₂ removal, and it was much lower than that at the thermodynamic equilibrium as shown in Table 1. This indicates that the H₂ removal by the Pd₇₇Ag₂₃ membrane kinetically inhibits the methane-forming reaction but does not shift the equilibrium as mentioned above. The platinum group is well-known to show high H₂ dissociation/association ability and H₂ spillover effect. Otsuka et al. reported that Pt accelerated the formation rates of H₂ and CO in the partial oxidation of methane by the reverse spillover of H₂ [32]. Lei et al. investigated the effect of Rh in the high temperature water-gas shift reaction, and they found that Rh greatly enhances H₂ release during reoxidation by water, presumably by recombining hydrogen atoms transferred from oxide to metal by reverse spillover [33]. Therefore, the Ru and Pd might accelerate the association of hydrogen atom and desorption of H₂ molecules from the catalyst through the reverse spillover mechanism. Comparing Ru-Co/CeO₂ and Pd-Co/CeO₂, Ru-Co/CeO₂ exhibited high H₂ yield and low methane selectivity. From the investigation in C₁ selectivity with the conversions of ethanol as shown in Figure 4, the methane selectivity in Pd-Co/CeO₂ was increased at lower conversions compared to Ru-Co/CeO₂. This clearly indicates that the promotion effect of methane-forming reaction was higher in Pd than Ru, probably caused by high H₂ storage capacity of Pd. Thus, the Pd membrane could effectively remove H₂ through the reverse spillover on Ru before they reacted with CO₂ or CO to methane, resulting in higher H₂ yield in Ru-Co/CeO₂. On the other hand, a certain part of hydrogen would react with CO₂ and CO due to relatively high hydrogen concentration on Pd before desorption of H₂ molecules by the reverse spillover mechanism, although the Pd₇₇Ag₂₃ membrane removed H₂ from the reaction zone.

3.3. Comparison of Amorphous Alloy Membranes and Pd₇₇Ag₂₃ Membrane in the Membrane Reactor

Figure 5 shows the H₂ permeability of amorphous alloy membranes. The H₂ permeability of Ni-Nb-Zr ternary alloy membrane was increased with increasing Zr content as the literature reported [34,35]. The H₂ permeability of (Ni_0.6Nb_0.4)₇₀Zr₃₀ was approximately 8.8 × 10⁻⁹ mol·m⁻¹·s⁻¹·Pa^−0.5, which is consistent with the value reported by the researchers [36]. Paglieri et al. reported that the addition of Ta lowered the H₂ permeability, although this slightly improves the thermal stability [37]. We have found that the increase in Ta content slightly decreased the H₂ permeability of Ni-Nb-Ta-Zr quaternary alloy membranes [38]. Indeed, Ni-Ta-Zr ternary alloy membranes showed the lower H₂ permeability in our study as well. In contrast, the addition of a small amount of Zr and Ta increased the H₂ permeability in Nb-Ni-Co alloy membranes because the introduction of larger atoms expanded the amorphous structure, resulting in an increase in H₂ diffusivity [39]. Thus, the Ni₄₀Nb₂₀Ta₅Zr₃₀Co₅ alloy membrane exhibited the highest H₂ permeability that was comparable to the Pd₇₇Ag₂₃ membrane.

The steam reforming of ethanol over Ru-Co/CeO₂ was evaluated in the membrane reactors with amorphous alloy membranes.

Table 2. Comparison of the membrane reactor performance using amorphous alloy membranes and Pd₇₇Ag₂₃ membrane in the steam reforming of ethanol.

Membrane	Membrane Thickness (μm)	H2 removal (%)	H2 Yield (%)	C1 Selectivity (%)
				CO2	CO	CH4
(Ni0.5Nb0.5)80Zr20	41.6	40.0	58.1	72.9	3.7	23.4
(Ni0.6Nb0.4)80Zr20	40.6	37.0	57.7	71.8	3.4	24.7
(Ni0.6Nb0.4)70Zr30	27.7	49.2	65.6	75.8	2.8	21.4
(Ni0.67Ta0.33)80Zr20	35.7	10.0	47.9	66.3	5.1	29.0
(Ni0.67Ta0.33)70Zr30	38.7	36.6	55.7	69.3	5.2	25.5
Ni40Nb20Ta5Zr30Co5	30.4	50.9	67.2	79.0	3.8	17.2
Pd78Cu6Si16	28.1	9.6	46.5	65.8	5.4	28.8
Pd77Ag23	20.0	58.0	70.3	81.8	4.0	14.2

Catalyst: Ru-Co/CeO₂, reaction temperature: 623 K, 1.0 × 10⁴ g-cat min/C-mol. S/C = 2, sweep Ar flow rate: 500 mL/min.

summarizes the membrane reactor performance. The H₂ yield was clearly related to H₂ removal ratio. Indeed, the amorphous membranes with the lowest H₂ removal ratio of approximately 10%, such as (Ni_0.67Ta_0.33)₈₀Zr₂₀ and Pd₇₈Cu₆Si₁₆, exhibited the same membrane reactor performance, and a similar H₂ yield was obtained on the (Ni_0.6Nb_0.4)₇₀Zr₃₀ and Ni₄₀Nb₂₀Ta₅Zr₃₀Co₅ with the highest H₂ removal ratio of approximately 50%. However, the H₂ removal ratio was not the same order of the H₂ permeability. For example, the H₂ yield and H₂ removal ratio in the (Ni_0.6Nb_0.4)₇₀Zr₃₀ membrane was higher than that in the (Ni_0.5Nb_0.5)₈₀Zr₂₀ and was almost comparable to Ni₄₀Nb₂₀Ta₅Zr₃₀Co₅, although the H₂ permeability was on the order of Ni₄₀Nb₂₀Ta₅Zr₃₀Co₅ > (Ni_0.5Nb_0.5)₈₀Zr₂₀≅ (Ni_0.6Nb_0.4)₇₀Zr₃₀. This could be owing to their membrane thickness that is in inverse proportion to the H₂ permeation flux when the membrane is thick enough.

Finally, we carried out the steam reforming of ethanol in the Pd₇₇Ag₂₃ membrane reactor with different sweep Ar flow rate to understand the effect of H₂ removal by the membrane on H₂ yield. Figure 6 shows the H₂ yield and C₁ selectivity as a function of H₂ removal ratio. The solid lines were interpolated from the experimental results on the Pd₇₇Ag₂₃ membrane reactor by varying the sweep Ar flow rate to obtain different H₂ removal ratio. Interestingly, the H₂ yield and C₁ selectivity in the amorphous membranes fitted the curve well. This clearly shows that the membrane reactor performance was determined by the H₂ removal ratio. Considering the reaction mechanism mentioned in Section 3.1, at the low reaction temperature, the catalytic performance such as C₁ selectivity was kinetically controlled by the H₂ permeation rate through the membrane. Thus, it is not an unexpected result that the membrane reactor performance using different amorphous membranes fitted those with the Pd₇₇Ag₂₃ membrane. In other words, we can roughly predict the membrane reactor performance based on their actual H₂ permeation flux regardless of the metal composition of the alloy membranes. Indeed, the H₂ removal ratio is clearly related to the H₂ permeation flux of the membrane as shown in Figure 7.

4. Conclusions

We evaluated the Pd membrane performance over Ni/CeO₂ and Co/CeO₂ in steam reforming of ethanol. In the conventional fixed bed reactor, Ni/CeO₂ showed low H₂ yield compared to Co/CeO₂ although the conversion to C₁ products was much higher at the temperatures. In the membrane reactor, the simultaneous H₂ separation improved both conversion of ethanol to C₁ products and H₂ yield. For Ni/CeO₂, the H₂ yield exceeded that in Co/CeO₂ at 773 K. However, at 673 K, although H₂ yield over Ni/CeO₂ was slightly increased by H₂ removal, it was lower than that over Co/CeO₂. The difference of the H₂ removal effect in those catalysts could be due to their reaction mechanism. At high reaction temperature, the higher the reaction rate of steam reforming of ethanol in Ni/CeO₂ compared to Co/CeO₂, the higher the increase rate of H₂ yield that was achieved due to the higher equilibrium shift effect by H₂ removal. However, at low reaction temperature, the methane-forming reaction in Ni/CeO₂ inhibits the H₂ permeation, resulting in a low H₂ yield. From these results, the H₂ separation membrane can improve the H₂ yield thermodynamically at high reaction temperature, but the simultaneous H₂ separation kinetically inhibited methane formation by H₂ removal at a low reaction temperature.

The amorphous alloy membranes with different compositions were employed in the steam reforming of ethanol, and the membrane reactor performance was compared with the Pd₇₇Ag₂₃ membrane. Regardless of membrane composition, the membrane reactor performance could be set in the order of H₂ permeation flux.