Introduction

The using of mixed, oxygen ion and electron, conductivity oxides (MIECs) as ceramic membranes for the integrated process of oxygen separation and partial oxidation of methane to syngas is currently considered as one of the most advantageous and cost-efficient MIEC applications. The driving force for the transport of oxygen is a difference in oxygen activity between the feed and permeate sides of the membrane. As a result, oxygen permeates from the side with high oxygen activity to the side with low activity. On the feed side surface of the membrane, the adsorbed oxygen molecules reduce to two oxide ions \( {O_2} \rightleftarrows 2{O^{2 - }} + 4{e^{-} } \). This oxide ions migrate through a thermally activated hopping mechanism to the permeate side surface of the membrane where they can either directly interact with adsorbed methane molecules or recombine and desorb, thus supporting reactions with methane in the gas phase. Simultaneously, electrons drift from the permeate side to the feed side, thus serving to short circuit the membrane and ensure its electrical neutrality. Depending on the prevailing reaction mechanism (surface or gas phase), different arrangements of the catalyst can be used at the membrane permeate side. Therefore, the respective device can be characterized as a combined electrochemical–catalytic membrane reactor. More details concerning scientific foundations of the relevant transport phenomena are considered in work [1]. Authors [2] report on economic aspects of the emerging technology. However, particular emphasis in recent years has been given to development [3–10] and testing of membrane materials [10–18]. Still, the achieved lifetime of MIEC membranes is not sufficient for their wide introduction in practice. It is worth noticing that, though the durability of membranes is influenced to a degree by structural features and defect architecture, it seems to depend most considerably on their chemical composition that often includes large amounts of easily reducible oxides of 3d row end metals such as Co or Cu [19–21]. On the other hand, ferrous materials, i.e., based on more stable iron oxide, have not only appreciable mixed conductivity but also exhibit a rather more suitable combination of service properties (thermodynamic stability, moderate thermal and chemical expansion, etc.). We have shown earlier that lanthanum–strontium ferrite La0.5Sr0.5FeO3−δ (LSF) can be successfully utilized as a membrane material for oxygen separation in the POM process [22]. In continuation of these studies, the focus in the present work is on the thermal regime and long-term stability of the POM reactor.

Experimental

The powder of LSF was prepared by glycine nitrate synthesis [23]. The typical procedure was as follows: The high-purity nitrates taken in desirable proportions were dissolved in distilled water at moderate heating. Then, amino acetic acid was added to the nitrates in double excess. The gradual heating resulted in thickening and ignition of the desiccated material. The combustion product was ground and subjected to calcination at 1000–1200°C for 10 h. X-ray powder diffraction was used to confirm single-phase final products. The synthesized powders with the average grain size of about 1 μm were mixed with commercial paraffin at 70–80°C and cooled down to room temperature. The batch was separated in small pieces to be fed in a Haake Polidrive extruder with tubing die. The extruded green body was dried at 200°C in order to remove the paraffin, and then temperature was slowly increased to 1250–1300°C. The firing time was 5 h. The wall thickness and external diameter of the obtained tubular ceramic membranes were about 1 and 10 mm, respectively. The density was 95–97% of the theoretical value. In order to verify the absence of microcracks, the obtained ceramic tube was attached to the pressurized air line and checked on air leaks by submerging in ethanol.

Alumina ceramic cylinders with a diameter and length of 2 and 6–10 mm, respectively, were used as catalytic support. The support was soaked in a water solution of nickel nitrate, dried, and fired at 1000°C to form the catalysts with the nickel metal load of about 10% wt.

The experimental setup for POM tests is shown in Fig. 1. The reactor encasement was made of quartz tube with the diameter and length of 40 and 700 mm, respectively. The butt ends of the encasement were covered with flanges and equipped with alumina tubes for air passage. The oxygen membrane under test was sealed in between upper and lower supporting auxiliary alumina tubes. The typical length of the membrane for regular tests was 50 mm. In order to reduce feed gas consumption, 20-mm membranes were used in long-term stability tests. The spacing between encasement and membrane was filled with the catalyst. The Bronkhorst mass flow controllers were used for the preparation and measurement of the inlet feed gas. The outlet stream was passed through the condenser in order to remove water vapors, and then a gas chromatograph was used for the analysis of volume concentrations (x i) of different gases in the outlet stream. In the preliminary work [22], the furnace temperature was controlled only. In order to gain a deeper understanding of temperature variations across the reactor in this study, we used S-type thermocouples Tc1, Tc2, and Tc3 set at different check points where the respective temperature values, T 1, T 2, and T 3, were read with the help of a Keithley 2000 digital multimeter. Yokogawa UT-155 controller was used for temperature maintenance of the furnace. A personal computer with specially developed software was used for running the experiments and data processing. The catalyst at the membrane outside was flashed with argon (purge gas) during 5 h upon achieving experimental temperature, and then methane was gradually added in the stream so that in 20 h the argon feed was completely replaced with methane. Based on the data of the outlet gas analysis, the POM process was characterized with conversion (\( {X_{{\rm{C}}{{\rm{H}}_4}}} \)), selectivity (S CO), hydrogen-to-carbon monoxide ratio (H2/CO), and oxygen separation flux density (\( {J_{{{\rm{O}}_2}}} \))

$$ {X_{{\rm{C}}{{\rm{H}}_4}}} = \left( {1 - \frac{{{x_{{\rm{C}}{{\rm{H}}_4}}}}}{{{x_{\rm{CO}}} + {x_{{\rm{C}}{{\rm{O}}_2}}} + {x_{{\rm{C}}{{\rm{H}}_4}}}}}} \right) \times 100\% $$
(1)
$$ {S_{\rm{CO}}} = \left( {\frac{{{x_{\rm{CO}}}}}{{{x_{\rm{CO}}} + {x_{{\rm{C}}{{\rm{O}}_2}}}}}} \right) \times 100\% $$
(2)
$$ {J_{{{\rm{O}}_2}}} = F_{\rm{d}}^{\rm{out}} \times \left( {1.5{x_{\rm{CO}}} + 2{x_{{\rm{C}}{{\rm{O}}_2}}} + {x_{{{\rm{O}}_2}}} - 0.5{x_{{H_2}}} - \frac{{0.21}}{{0.78}}{x_{{{\rm{N}}_2}}}} \right)/s $$
(3)
Fig. 1
figure 1

The sketch of the experimental setup. MFC, mass flow controller; PC, personal computer; DM, digital multimeter; TC, temperature controller; GC, gas chromatograph; C, condenser; AT, alumina tube; SG, sealant glass; Tc1, Tc2, and Tc3, thermocouples; Ct, catalyst; OM, oxygen membrane; QT, quartz tube

Full size image

Here, \( F_{\rm{d}}^{\rm{out}} \) is the flow of the dry outlet gas, 0.21 and 0.78 are volume fractions of oxygen and nitrogen in air, respectively, s is membrane surface, and the last term accounts for nitrogen leaks through microcracks and imperfections in seals. Since conversion exceeded 98% in all experiments, this parameter is not shown in the plots below.

Results and discussion

Ferrite La0.5Sr0.5FeO3−δ is characterized by highest ambipolar conductivity in the series of solid solutions La1−x Sr x FeO3−δ [24, 25], and the calculated oxygen permeation flux density through 1-mm-thick membrane of this oxide can achieve 8 ml × min−1 × cm−2 at 900°C when the permeate side oxygen partial pressure is 10−16 atm while the other side is flushed with air. On the other hand, some limitations on the flux can be imposed by surface reactions [26]. It is interesting, therefore, to observe the oxygen separation efficiency of this material in conditions of a real POM process. The conversion parameters at 900°C versus methane flow are shown in Fig. 2. Notice that temperature was set with the help of thermocouple Tc1 in the spacing between reactor encasement and external heaters. Surprisingly, experimentally achieved flux density of about 10 ml × min−1 × cm−2 exceeded the calculated value. Hence, the temperature of the membrane was probably higher than 900°C.

Fig. 2
figure 2

Performance characteristics of the tubular LSF membrane versus feed flux at T 1 = 900°C. The surface of membrane is 12.5 cm2; the flow rate of air is 1000 ml × min−1

Full size image

In order to better understand temperature regimes in different areas, thermocouples Tc2 and Tc3 were located near membrane inside and outside surfaces, i.e., in air and methane streams, respectively (Fig. 1). The heating power was supplied so that temperature would not exceed 950°C, and when this upper limit was achieved either at Tc2 or at Tc3 respective thermocouple was used for temperature maintenance. The temperature changes at check points and conversion parameters versus methane flow are shown in Fig. 3. It is seen that T 2 is lower than T 3 at \( {F_{{\rm{C}}{{\rm{H}}_4}}} \)< 20 ml × min−1, which is due to colder inlet air and insufficiently intensive heat transfer through the catalytic bed. However, a further increase of methane flow results in faster temperature increase at Tc2, and the air stream temperature approaches 950°C near \( {F_{{\rm{C}}{{\rm{H}}_4}}} \)= 40 ml × min−1 while the temperature in the catalytic bed progressively decreases. These results suggest that the low selectivity at moderate values of the methane flow reflects a considerable development of the total oxidation reaction

$$ C{H_4} + 2{O_2} \rightleftarrows C{O_2} + 2{H_2}O + 802\;kJ $$
(4)

which is characterized with a very large heat effect. On the other hand, the selectivity increase with methane flow (Fig. 3) gives evidence to the increasing impact of the partial oxidation reaction

$$ C{H_4} + \frac{1}{2}{O_2} \rightleftarrows CO + 2{H_2} + 44\;kJ $$
(5)

Hence, the enthalpy for the combination of reactions (4) and (5) is

$$ \Delta {H_\Sigma } = {S_{\rm{CO}}} \times 44\;{\hbox{kJ}} + \left( {1 - {S_{\rm{CO}}}} \right) \times 802\;{\hbox{kJ}} $$
(6)

Then, the dependence of specific heat power on methane flow in the reactor can be found as

$$ P = \Delta {H_\Sigma }\frac{{{F_{{\rm{C}}{{\rm{H}}_4}}}}}{{22.4{\hbox{ l}} \times {\hbox{mo}}{{\hbox{l}}^{ - 1}}{ }60}}{X_{{\rm{C}}{{\rm{H}}_4}}} $$
(7)

Here, 22.4 mol−1 is molar volume, and 60 is a factor for conversion of minutes to seconds. The results are shown in Fig. 4 as calculated from experimental values of selectivity in Fig. 3. As expected, the heat output increases with methane flow up to 15–18 kJ at 40–60 ml × min−1 and then begins to decline. At the same time, specific heat power calculated at zero selectivity, i.e., when all permeate oxygen is exhausted in the reaction of total oxidation, achieves saturation. The heat power difference clearly demonstrates that the summary reaction in the tubular reactor (Fig. 1) evolves mainly as total methane oxidation, most probably in a thin gas layer near the membrane surface, followed by reactions of combustion products with methane over a catalytic bed where a large amount of heat is consumed. This picture gives a simple explanation to the observed high efficiency of ordinary nickel catalysts that are widely used for steam/carbon dioxide reforming. It follows also from the obtained results that a more precise control of membrane working conditions can be carried out with the help of thermocouple Tc2 located directly at the membrane feed side.

Fig. 3
figure 3

Temperature readings of thermocouples Tc2 and Tc3 (see text) and performance characteristics of the tubular LSF membrane versus feed flux

Full size image
Fig. 4
figure 4

The heat power released at the membrane surface and overall heat power

Full size image

From a practical standpoint, the longevity of the separating membrane in the reforming process is probably the most important issue. The variations of the POM parameters versus time at 950°С are shown in Fig. 5. The selectivity of about 80% was attained at the end of the transitory period of replacement of the purge gas for pure methane while oxygen permeation flux density achieved nearly 6.8 ml × min−1 × cm−2. This value is in agreement with the flux calculated from conductivity [22] when the permeate side oxygen partial pressure equals 10−8.5 atm. A fast increase of nitrogen concentration in the outlet gas after 220 h signals the development of cracks either in seals or in the membrane. The subsequent inspection of the disassembled reactor confirmed membrane degradation. The temperature decrease to 850°С resulted in a respective decrease of the oxygen permeate flux density to 3 ml × min−1 × cm−2. On the other hand, the calculated flux should be near 4 ml × min−1 × cm−2. This difference suggests that temperature decrease from 950 to 850°С results in the development of surface exchange limitations. This conclusion is corroborated with the results of [27] where it was observed that oxygen fluxes through 1- and 1.6-mm membranes of La0.5Sr0.5FeO3−δ did not differ at 800°С and thus evidenced surface exchange control over oxygen transport. However, much better membrane performance was achieved. It is seen from Fig. 6 that conversion parameters remained fairly stable for about 7500 h at selectivity near 92%. The after-test EDX analysis did not reveal any chemical changes across the membrane, and X-ray analysis confirmed the phase integrity of the material.

Fig. 5
figure 5

Performance characteristics of the tubular LSF membrane at 950°C. The surface of membrane is 12.5 cm2; the flow rate of air is 1000 ml × min−1

Full size image
Fig. 6
figure 6

Performance characteristics of the tubular LSF membrane at 850°C. The surface of membrane is 5 cm2; the flow rate of air is 300 ml × min−1

Full size image

The obtained results demonstrate that the optimal working temperature for the ferrite membranes is between 850 and 900°С. The oxygen permeation within these limits depends on the interplay of bulk oxygen chemical diffusion and surface oxygen exchange kinetics. At the same time, kinetic demixing and related phase instabilities [28] do not seem to play any role in the lifetime of the membrane.

Conclusion

In general, we have shown that perovskite-like ferrites are promising materials for oxygen separating membranes to be used in an electrochemical POM reactor. A reasonable compromise of permeability and stability can be achieved by fine-tuning of the chemical composition of the material and precise control of the membrane temperature. Surface-controlled kinetics affects noticeably upon oxygen permeation of the membranes at temperatures below 850°С. Structural instabilities related to kinetic demixing or chemical reactions of the membrane with gaseous components do not develop considerably at this temperature, thus providing longevity of the membrane. The arguments are given in proof of the reforming mechanism that involves total oxidation of the sweep methane with oxygen permeate followed by the interaction of the oxidation products, i.e., water vapor and carbon dioxide, with methane. Therefore, well-developed nickel–alumina catalysts for steam reforming can be used efficiently in a POM reactor.