Partial oxidation of methane using silver/gadolinia-doped ceria composite membranes

a r t i c l e i n f o


Introduction
Syngas, a mixture of carbon monoxide and hydrogen, is used for the production of synthetic fuels via the Fischer-Tropsch process or transformed into other chemicals, for example dimethyl ether (DME) (Sousa-Aguiar et al., 2011). Syngas is commonly produced on a large scale from methane by steam reforming, a method that requires large amounts of energy as the reaction is endothermic (Eq. (1)). The reaction also leads to a H 2 /CO ratio of 3, higher than the ideal H 2 /CO ratio of 2 necessary for further conversion to synthetic fuels (Sousa-Aguiar et al., 2011;Rostrup-Nielsen et al., 2002). Another method for the conversion of CH 4 is dry or CO 2 reforming, also an energy intensive process that yields a lower H 2 / CO ratio of 1 (Eq. (2)) [2].
An ideal process to achieve the preferred H 2 /CO ratio of 2 is the partial oxidation of methane. One way of achieving this is autothermal reforming, where a mixture of methane and oxygen is raised to a temperature of 1300 1C at which thermodynamic equilibrium favours the formation of syngas. The main disadvantages of this method are the high temperature and the need for high purity oxygen to hinder side reactions between nitrogen and the other components.
Catalytic partial oxidation on a mixed ionic-electronic membrane has been proposed to solve this problem (Yu et al., 2014;Balachandran et al., 1995). In practice, rather than direct partial oxidation, it is a two-step process. The first step is the complete oxidation of methane to CO 2 (Eq. (3)) that provides the energy necessary for the second step, dry-steam reforming in the presence of nickel and excess methane. To be thermally efficient the energy released in the full oxidation step must be consumed in the reforming reactions. By combining Eqs. (1), (2) and (3), partial oxidation of methane to syngas produces the ideal H 2 /CO ratio of 2 as shown in Eq. (4).

Steam reforming CH 4 þH
Previous literature reports have shown that methane can also be partially oxidised using the permeated oxygen from air in a gas separation membrane operating above 800 1C in the presence of a reforming catalyst (Balachandran et al., 1995;Dyer et al., 2000;Shao et al., 2001), as illustrated schematically in Fig. 1.
BSCF, one of the materials that has attracted more interest, is nonetheless strictly limited to operation at pO 2 above 10 À 5 bar, therefore limiting its use in hydrocarbon atmospheres (Vente et al., 2006); it is also known that oxygen permeation is greatly reduced in the presence of CO 2 , due to carbonate formation, and even though the carbonation process might be reversible, use in carboncontaining atmospheres is further restricted by the complexity of operation (Arnold et al., 2007;Bucher et al., 2008). Although a variety of strategies have been implemented to solve these problems, the stability of BSCF is the main challenge for its long term operation under highly reducing and carbon containing atmospheres.
Herein, we present a metal-ceramic composite for in-situ oxygen production or partial oxidation of methane operating in the temperature range: 500-700 1C where economic heat sources can be used and with good tolerance to carbon containing atmospheres and good mechanical stability. The two main components are Ce 0.9 Gd 0.1 O 1.95 (CGO) as the ionic conductor and silver as the electronic conductor, labelled hereafter as Ag/CGO. CGO has been shown to be catalytically active for both steam and dry reforming (Ramirez-Cabrera et al., 2002;Ramı́rez-Cabrera et al., 2004) and it exhibits high oxygen ion conductivity. Doped ceria-based materials are stable in carbon containing atmospheres and their high oxygen storage capacity is useful in ceria-zirconia solid solutions for exhaust gas cleaning (Ozawa et al., 1993).
Silver is an excellent catalyst for the reduction of oxygen and its additon to CGO increases the oxygen surface exchange coefficient by orders of magnitude (Ruiz-Trejo et al., 2014;Seeharaj and Atkinson, 2011). Silver is also considerably cheaper than Pt, Pd or other metals used in similar membranes (Sunarso et al., 2008). Furthermore, silver is also catalytically active for partial oxidation reactions (Nagy and Mestl, 1999).
In this work the composite membrane was fabricated by coating ultrafine CGO particles (5 nm) produced by continuous hydrothermal flow synthesis (CHFS), with silver followed by careful sintering. We demonstrate the selective permeation of oxygen through such a membrane, confirming our earlier report (Ruiz-Trejo et al., 2014) and then further demonstrate the effectiveness of such a membrane for the partial oxidation of methane using nickel as a reforming catalyst.

Experimental
The nanoparticles from which the membranes were fabricated, were produced by CHFS. A single aqueous solution of ammonium cerium nitrate (0.45 M) and gadolinium nitrate (0.05 M) (pumped at a flow rate of 200 mL min À 1 ) was mixed in line with an aqueous solution of KOH (0.5 M) at a flow rate of 200 mL min À 1 before being mixed with a stream of water at 450 1C and 24.1 MPa flowing at 400 mL min À 1 in a co-current mixer as described elsewhere (Tighe et al., 2013). The CGO nanoparticles were precipitated instantly, and the stream was then cooled and the pressure reduced to ambient by passing the products through a back pressure regulator. The particles were then separated by centrifuging the initial solids and cleaning the concentrated sludge by placing it in a dialysis bag in a large plastic vessel and replacing with clean water several times in the vessel. Once cleaned, the wet sludge was freeze dried overnight.
The dried particles were then silver coated using Tollens' reagent (Ruiz-Trejo et al., 2014;Tollens, 1882). In a typical experiment concentrated NH 4 OH was gradually added to a solution of AgNO 3 (0.1 M), the addition was stopped when the black precipitate formed initially disappeared. The solution was then basified with KOH (0.8 M). The nanoparticles were suspended in this solution with the aid of an ultrasonic bath for 20 min. The initial ratio was 10 w% Ag and 90 w% CGO as this guarantees percolation in the final sintered sample. Finally, to initiate the reaction of reduction, 3 mL of dextrose solution (0.25 M) were added drop-wise to the mixture and left under stirring at room temperature for five minutes. The suspension was centrifuged and rinsed at least five times to eliminate all remaining reagents. The dry, clean, coated powders were pelletized (2.5 cm diameter) uni-axially and then coldisostatically pressed (250 MPa) before being sintered at 1100 1C for 4 h. The pellets were polished at both surfaces with metallographic paper down to the final 1 mm thickness. The density and open porosity was measured with Archimedes' method using deionised water as the immersion fluid. The samples were boiled for 1 h to ensure all open porosity was filled with water. Fig. 1. Using a mixed oxygen ionic-electronic membrane, oxygen from air is incorporated into the ionic crystal lattice. Oxygen diffuses across the gas-impermeable membrane driven by a gradient in the chemical potential of oxygen. In the case of a composite membrane, oxygen migrates as oxygen ions through the ionic material and electrons are transported in the opposite direction mainly through the electronic conductor network. On the reducing side of the membrane the permeating oxygen then reacts with methane in the presence of a reforming catalyst to form syngas.
A Field Emission Gun Scanning Electron Microscope (FEG-SEM Gemini 1525) was used for imaging of the nano-composite materials. Images of fracture surfaces of the membranes were collected to check the dense nature of the samples and identify the presence of any cracks or large pores. The distribution of silver, doped-ceria and pores were analysed, and this has been discussed in our previous report (Ruiz-Trejo et al., 2014). The crystallinity of the powders and consolidated samples was analysed with an X'Pert PRO MRD X-ray diffractometer and the composition was estimated using the Chung method (Chung, 1974).
A 1 mm thick pellet with surface area of 3.1 cm 2 was placed in an alumina tube and sealed around its periphery with an alumina-based ceramic sealant (Aron Ceramics, Japan). The sealed sample was kept at 500 1C overnight and then held and measured at each test temperature for one hour. One side of the pellet was fed with air at 130 mL min À 1 at NTP (Normal Temperature and Pressure) conditions while the other side was swept with zero-grade argon or diluted methane at a rate of 100 mL min À 1 at NTP conditions. The gases were of commercial purity (BOC, UK). The mass spectrometer was calibrated with separate certified mixtures of N 2 (1%), O 2 (2%) and CO 2 (1%) each diluted in argon; CH 4 was calibrated using mixtures made with mass flow controllers (Bronkhorst). The mass spectrometer was calibrated at the beginning of the each permeation experiment. Two permeation experiments were carried out both on the same disc.
For oxygen permeation measurements argon was used as sweep gas and this stream was analysed in a mass spectrometer (Prolab Benchtop) connected downstream. The oxygen leaks into the sweep gas were taken into account by measuring the amount of nitrogen in the permeate side and assuming it leaks in gaseous form with the same composition as air. The oxygen flux reported is the oxygen flux measured minus the oxygen leaks; the corrected value is then the oxygen permeating through the membrane solely by solid state transport. The temperature range investigated was 550-800 1C.
A diagram of the set-up used for partial oxidation is shown in Fig. 2. A piece of nickel foam (Goodfellow, 1.6 mm thick, external area 2.8 cm 2 , porosity 95%, and purity 95%) was placed directly in contact with the membrane on the oxygen-lean side as a catalyst. Four compositions of methane were used: 1.8%, 2.5%, 3% and 4% diluted in argon and with a total flow rate of 100 mL min À 1 at NTP. The gas analysis was performed on the methane stream. To check for carbon deposition, Raman spectroscopy was used after the experiments (Horiba Jobin Yvon LabRAM 800 h Raman Spectrometer). The temperature range tested was 500-700 1C unless otherwise indicated; the composite tested was run at least 48 h at T4500 1C in a pure methane environment. Leaks from air to the permeate side were assumed to be similar to the pure oxygen permeation experiments where N 2 can be measured unambiguously; leaks of methane to the air side are considered to be minimal as demonstrated in the case of H 2 leaking to the air side in Ag/CSO membranes (Ruiz-Trejo et al., 2014).

Results
The XRD data showed no signs of reaction between CGO and silver before or after sintering. The large surface area of the ultrafine Fig. 2. Schematic of membrane partial oxidation reactor. The membrane was sealed to an alumina tube with an alumina based sealant. Air was blown on one side and methane diluted in argon was blown on the other side. A nickel mesh was placed close to the surface of the membrane to catalyse the reforming of methane.  particles of CGO are used as the driving force to sinter the material at the relatively low temperatures needed to avoid excessive loss of silver. The distribution of the different phases has been analysed in our previous work using FIB-SEM tomography (Ruiz-Trejo et al., 2014). The silver content of the membrane was 7 vol% from XRD analysis and its density was 7.02 g cm À 3 by Archimedes' method, which corresponds to 95% of the expected theoretical density of 7.37 g cm À 3 for this Ag/CGO composition. The open porosity obtained by water displacement was 2% and was measured in the sample used for the permeation experiments. It should be noted that the detection of open porosity does not imply that the pores percolate the thickness of the membrane. Fig. 3 shows the measured conductivity of the composite Ag/CGO in air and the values for silver metal conductivity and CGO bulk ionic conductivity in air (ASM International Handbook Committee, 1990;Steele, 2000). The bulk membranes were metallic despite the low level of silver content; however the lowest silver content required to achieve percolation has not yet been determined. Nonetheless, our fabrication technique led to percolation of the silver well below the value of 30 vol% usually taken as the minimum. At room temperature the conductivity was ca 100 s cm À 1 and at 600 1C the electronic conductivity was ca 44 s cm À 1 , at least three orders of magnitude larger than the bulk ionic conductivity of the CGO, indicating that the latter is the rate limiting process for oxygen transport in the composite. Fig. 4a shows the SEM microstructure of a typical membrane. Fig. 4b shows the two regions that are produced during sintering: a very dense silver-rich CGO and one that is silver-poor with more porosity visible. A more detailed analysis of this characteristic was carried out in our previous work (Ruiz-Trejo et al., 2014). Although there may be variations between Ag/CSO and Ag/CGO two main points obtained in that work were: in the silver-rich regions (11.8 vol%) most of the silver was percolating with a low level of porosity (0.3 vol%) and in the silver-poor regions (4.5 vol%) the silver was not percolating and showed a high level of non-percolating porosity (4 vol%). An ideal membrane should be more homogenous and be similar to the silverrich regions: dense and percolating. The presence of two regions is probably an indication that the nanoparticles were not coated individually but rather agglomerates of CSO were coated leading to an uneven distribution of silver. To achieve a single region it might be necessary to percipitate silver directly into the suspension of the nanoparticles as obtained after cleaning the suspension obtained via CHFS. An optimisation of this process is underway.
For the partial oxidation experiments, the detection of the reactants and products was carried out with a mass spectrometer and it was assumed that the carbon-containing products from methane were only CO and CO 2 since no carbon deposition was detected by Raman spectroscopy. It was further assumed that there were no other processes such as oxidative coupling of methane although this deserves future investigation (Tenelshof et al., 1995). The species unambiguously detected and calibrated were CH 4 and CO 2 and these were used to estimate the amount of CO produced according to Eqs. (5)-(8). The oxygen permeated in the methane oxidation experiments was estimated with Eq. (9), where unwanted oxygen leaks come from air at the same rate as measured in the oxygen permeation experiment.  Table 1 shows the results of the partial oxidation products with the flow of methane feed also tabulated.
After operation for 48 h, the experiments were stopped and the absence of carbon deposition on the composite was first tested by visual inspection and then evaluated with Raman spectroscopy on the surface of the membrane, however no coking was observed by either method. However, longer term stability tests are needed to determine a useful temperature range of operation since at higher temperature silver exudation might be a problem as will be shown later. Fig. 5 shows the percentage CO selectivity for various temperatures and input flows of methane. Although there is some nonlinearity, the high selectivity observed for the production of carbon monoxide under all conditions is significant, proving that partial oxidation of methane was achieved, the main goal of this study. The selectivity seems to generally decrease with increasing temperature and when the input methane flow is low, the reaction is less selective. These high selectivities were obtained despite the use of an un-optimised nickel mesh as a catalyst, and may be improved by employing a different catalyst. Fig. 6 shows that the conversion increases from 500 to 600 1C as expected, however, the fact that the conversion does not increase between 600 and 700 1C suggests that other factors (such as: local temperature variations, microstructural changes, surface limitation, dynamic carbon-deposition/elimination, gas-phase reaction) may be affecting the reaction, and this requires further study. Table 1 Mol percentage of reactants and products of the partial oxidation of methane. The percentage of conversion and selectivity is also shown. The total flow was 100 mL min À 1 (NTP) and the balance was argon.

T/1C
[ The separation membrane was observed to degrade if operated in methane above 700 1C leading to lower conversion rates. One membrane was operated at 800 1C under methane after which droplets of silver were observed on the methane side as seen in Fig. 7a and to a lesser degree on the air side as shown in Fig. 7b. Although nominally both sides of the membrane were at the same temperature, heat released during the complete oxidation of methane has probably exacerbated the exudation of silver droplets. Fig. 8 shows the oxygen flux measured by mass spectrometry using argon as the sweep gas (no methane), and also the lower bound oxygen flux with CH 4 present as calculated from the amounts of CH 4 consumed and CO 2 produced. The real oxygen flux in the latter case was expected to be higher as the oxygen transformed to water was not measured. A recent report for oxygen separation in Ag-CuO-CGO (1 mm thick) composites indicated oxygen fluxes of 0.05 mL cm À 2 min À 1 at 700 1C between N 2 and air. At the same temperature, our values reported here for argon as sweep gas are 0.015 mL cm À 2 min À 1 , but higher when the sweep gas was methane: 0.18 mL cm À 2 min À 1 .
The permeation from air was carried out before the partial oxidation experiment to determine the rate of oxygen flux and estimate unwanted leaks; oxygen flux increased with temperature as expected, but the values obtained remained close to the leakage value. However, with methane the oxygen flux increased by nearly an order of magnitude. Fig. 8 also indicates that a higher methane concentration led to higher permeation of oxygen.

Discussion
The diffusion-controlled oxygen flux across the membrane depends, according to Wagner's theory, is given by Fig. 5. Percentage of CO selectivity at different temperatures and CH 4 concentrations used (in mol%). The total flow fed was 100 mL min À 1 (NTP) and the balance was argon.  where σi is the ionic conductivity, L the thickness of the membrane, R is the gas constant, T the temperature and F Faraday's constant. Although this is a simplification, in that it does not consider the rates of reaction at the two membrane surfaces, this is a starting point to better understand some of the variables that control the flux of oxygen and consequently the partial oxidation of methane. Eq. (10) states that oxygen permeation can be increased by either decreasing the thickness of the membrane or by decreasing the chemical potential of oxygen in the sweep side of the membrane. The latter is achieved by introducing methane as seen in the large increase in oxygen flux observed when the methane concentration was increased (see Fig. 8). Sweeping with hydrogen has also been shown to increase the oxygen flux in similar composites (Ruiz-Trejo et al., 2014). The authors are currently investigating alternative fabrication routes for yet thinner membranes. Eq. (10) does not consider a surface controlled permeation. In previous work we have estimated that for thicknesses above 300-500 μm the flux should be controlled by bulk diffusion (Ruiz-Trejo et al., 2014), therefore, the membrane studied herein probably operates in a bulk-diffusion controlled regime when argon is fed. However, when methane is fed the oxygen flux limitation may be different since the surface rate constant at the oxygen lean side is not known. We can estimate the theoretical value of oxygen flux calculating first the pO 2 in the permeate side using the equilibrium with an equilibrium constant of ca ln K ¼ 26:4 at 700 1C (Ellingham, 1944). For a feed with 4% methane, [CO 2 ] is 0.068% and [CO] is 0.768% (from Table 1), the pO 2 value is 9.2 Â 10 À 26 atm, the ionic conductivity is 0.045 s cm À 1 (Steele, 2000) and the expected flux is then 1.9 mL min À 1 cm 2 , an order of magnitude higher than the measured value 0.18 mL min À 1 cm 2 . In addition to the uncertainty associated with the amount of oxygen converted to H 2 O, the difference might be indicating that the oxygen flux was being limited by surface reaction on the permeate side. At lower temperatures the difference was less dramatic: at 500 1C, 0.43 mL min À 1 cm 2 as expected, while 0.12 mL min À 1 cm 2 was measured as the lower boundary value. As future work will require thinner membranes to increase oxygen permeation rates, spectroscopic in-situ and mechanistic studies are needed to understand the surface evolution/oxidation reactions in CH 4 environments. Although we did not observe any carbon deposition post test, this may still have happened during operation. Work under CH 4 or CO/CO 2 mixtures in fuel cells and electrolysers is often plagued with problems of carbon deposition, however, the capability of ceria-based compounds to oxidise carbon is well-known (Aneggi et al., 2012;Trovarelli, 2002). Furthermore, silver is a well-known catalyst for partial oxidation reactions (Nagy and Mestl, 1999). It is then expected that both silver and ceria profit from each other's capabilities, making them an encouraging combination in membrane reactors for partial oxidation reactions.

Conclusions
Methane has been partially oxidised using novel Ag/CGO composites membranes. These were manufactured by sintering CGO nanoparticles prepared by CHFS and then coated with silver. The oxygen permeation through the membrane from air to argon as sweep gas, was estimated to be 0.02 mL cm À 2 min À 1 (NTP) at 700 1C, while using a CH 4 -containing atmosphere on the low oxygen activity side, the permeation rate was greater than 0.18 mL cm À 2 min À 1 (NTP). At 700 1C, high selectivity for CO (490%) was observed and good methane conversion was measured (approximately 21%), and despite the simplicity of the reactor and the thickness of the membrane. The samples operated at CH 4 for at least 48 h below 700 1C without being mechanically compromised, however if operated at 800 1C, silver exuded to the surface as a consequence of the heat released during methane oxidation. Due to the stability of CGO in carbon atmospheres, the Ag/CGO composites are expected to function well in these reducing atmospheres and under large transmembrane pO 2 , differences but operation at higher temperatures or in the presence of very exothermic reactions might lead to exudation of the silver.