Effect of Zirconium Doping on Hydrogen Permeation through

2768

Ind. Eng. Chem. Res. 2010, 49, 2768–2774

Effect of Zirconium Doping on Hydrogen Permeation through Strontium Cerate Membranes Jay Kniep and Y. S. Lin* Department of Chemical Engineering, Arizona State UniVersity, Tempe, Arizona 85287-6006

SrCe0.95Tm0.05O3-δ perovskite-type ceramic membranes offer high hydrogen selectivity, thermal stability, mixed protonic-electronic conductivity, and mechanical strength at temperatures above 600 °C. However, in order for the SrCeO3-based membranes to be used in industrial applications, the chemical stability of the membranes in various environments must be improved. The effect of doping zirconium on the chemical stability, lattice structure, protonic and electronic conductivity, and hydrogen permeation properties of SrCe0.95-xZrxTm0.05O3-δ (0 e x e 0.40) was studied. X-ray diffraction analysis verifies that all samples consist of a single perovskite phase. Doping zirconium in SrCe0.95Tm0.05O3-δ results in a decrease in both the protonic and electronic conductivity of the materials under reducing conditions, and a more significant decrease in hydrogen permeability of the membrane in CO2 free gas streams. In a CO2-containing environment SrCe0.75Zr0.20Tm0.05O3-δ membranes have a larger steady-state H2 flux and superior chemical stability over SrCe0.95Tm0.05O3-δ membranes. 1. Introduction Perovskite-type materials based on SrCeO3 and BaCeO3 have been studied extensively since they were found to have appreciable proton conduction at high temperatures in hydrogencontaining atmospheres.1,2 Applications of these materials include fuel cell electrolyte, hydrogen or steam sensor, hydrogen separation membranes, and membrane reactors.3-6 Both pure SrCeO3 and BaCeO3 exhibit low electronic conductivity; therefore, doping is crucial to enhancing proton and electron transport through the membrane. To maintain electroneutrality, the partial substitution of a trivalent cation for cerium causes charge compensating oxygen vacancies to form in the material. Doped proton conducting perovskite-type materials with improved electronic conductivity include BaCe0.90Nd0.10O3-δ, BaCe0.95Y0.05O3-δ, SrCe0.95Yb0.05O3-δ, and SrCe0.95Tm0.05O3-δ.7-10 For hydrogen permeation through dense, proton conducting perovskite structured ceramic membranes, protons can be incorporated into the membrane from either water vapor or a hydrogen-containing gas. The equilibrium between the membrane and either wet or dry hydrogen containing environments can be illustrated by the following equations: H2O + VO•• + OOX T 2OHO•

(A)

1 OOX + H2 T OHO• + e′ 2

(B)

In the equations, VO•• is an oxygen vacancy with a +2 charge, OOX is neutral lattice oxygen, and OHO• is a hydroxyl ion, which represents an interstitial proton associated with a lattice oxygen. The protons migrate through the membrane by hopping between adjacent lattice oxygen with the driving force for migration being the hydrogen chemical potential gradient across the membrane. Although SrCeO3- and BaCeO3-based mixed proton-electron conducting perovskite-type ceramic membranes can offer extremely high hydrogen selectivity, thermal stability, and mechanical strength at high temperatures, the chemical stability of these materials in various environments must be improved * To whom correspondence should be addressed. E-mail: [email protected].

in order to be used in industrial applications. Specifically, exposure to a CO2- or H2O-containing atmosphere at high temperatures degrades the performance of the materials. In CO2containing atmospheres, the materials form carbonates and metal oxides.11,12 With respect to these materials as hydrogen separation membranes, this has a detrimental effect on the membrane’s mechanical strength and hydrogen flux through the membrane as the carbonates and metal oxides inhibit the surface reactions. Cerium-containing perovskite-type ceramic membranes have also been found to be reactive in water vapor at elevated temperatures forming hydroxides, metal oxides, and other compounds on the membrane surface.13,14 Doped SrZrO3 and BaZrO3 have been shown to be mixed protonic-electronic conductors and have excellent chemical stability in both CO2- and H2O-containing atmospheres at high temperatures.13,15,16 However, these materials have lower electrical conductivities than SrCeO3- or BaCeO3-based materials as well as very high sintering temperatures. To produce a material with both high protonic conductivity and chemical stability, various researchers have investigated solid solutions containing alkaline earth metal cerates and zirconates.17-23 All of the resulting solid solutions exhibited enhanced chemical stability and a decrease in electrical conductivity with increasing Zr content. However, the Zr content needed for a compromise between chemical stability and electrical conductivity for industrial application varies depending on the alkaline earth metal and trivalent dopant present in the material. The effect of Zr doping on the hydrogen permeation through dense protonic-electronic conducting membranes is still unclear and has yet to be studied. This paper reports the synthesis, protonic and electronic conductivity, and hydrogen permeation properties of SrCe0.95-xZrxTm0.05O3-δ (0 e x e 0.40) dense membranes. The main objective is to determine the effect that Zr doping has on the hydrogen permeation through dense membranes in various reducing atmospheres. Electrical conductivities were measured under different atmospheres at high temperatures. Hydrogen permeation tests were conducted in the presence of CO2 to help determine the chemical stability of the materials. This work identifies a composition, which is a good compromise between

10.1021/ie9015182  2010 American Chemical Society Published on Web 02/19/2010

Ind. Eng. Chem. Res., Vol. 49, No. 6, 2010

chemical stability and conductivity, for use in industrial applications. 2. Experimental Methods 2.1. Preparation and Characterization of Membranes. SrCe0.95-xZrxTm0.05O3-δ (0 e x e 0.40) samples were prepared using the liquid citrate method. In this method, stoichiometric amounts of the corresponding metal nitrates Sr(NO3)2 (Alpha Aesar, 99.0%), Ce(NO3)3 · 6H2O (Alpha Aesar, 99.5%), ZrO(NO3)2 · xH2O (Sigma-Aldrich, 99.0%), and Tm(NO3)3 · 5H2O (Alpha Aesar, 99.9%) were mixed with citric acid in distilled water. The amount of citric acid was three times the total molar amount of metal ions present. The transparent liquid was heated to 95-100 °C and remained there under reflux and stirring for 4 h for the polymerization reaction to occur. The lid was then removed, and excess water was evaporated off at 100 °C leaving the solution a viscous gel. The gel was dried for 24 h at 110 °C, and the resulting brittle, porous material was heated to 400 °C for self-ignition to burn out the organics that were present. The material was then ground with a mortar and pestle for 15 min followed by calcination at 850 °C for 8 h (ramp rate ) 5 °C/min). Samples of the calcined powders were put into a die with a diameter of 2.30 cm and pressed with a hydraulic press (Carver, Model No. 3853) to 180 MPa. The resulting green disks were sintered in air at 1495 °C (x ) 0.0) or 1525 °C (for samples doped with Zr) for 24 h in a furnace (Thermolyne, 46100) with a ramp rate of 2 °C/min. The gas tightness of each membrane was verified using a room-temperature un-steady-state permeation system with helium. The membrane was considered gastight if the He permeance was less than 10-10 mol/(m2 · Pa · s). X-ray diffraction (XRD; Bruker; Cu KR) was used to characterize the phase structure of each composition. Characterization of the powders and membranes was evaluated in the 2θ range of 20-70° with a step size of 0.02°/s. Scanning electron microscopy (SEM) was used to characterize the morphology and grain size of the membrane surface. 2.2. Partial Conductivity Measurements. The four-point direct current (DC) method was used in the measurement of the total electrical conductivity of SrCe0.95-xZrxTm0.05O3-δ membranes in different temperature and oxygen partial pressure atmospheres. Previously sintered dense SrCe0.95-xZrxTm0.05O3-δ membranes were sanded into bars (approximately 2.0 × 10.0 × 16.0 mm3). Four lines were painted on a bar using silver conductive paste, and four silver wires were wrapped around the bar onto the lines. Each bar was then placed in a quartz tube with the two outer wires being connected to a potentiostat (Radiometer A/S, PG201). The potentiostat applied a stable direct current to the bar, and the voltage drop along the inner section of the bar was measured by connecting a digital multimeter (Protek, B-845) to the two inner silver wires. A detailed schematic of the conductivity setup can be found elsewhere.4 The conductivity was calculated using the following equation: σ)

IL VA

(1)

in which I is the applied current, L is the distance between the two inner wires, V is the voltage drop between the two inner wires, and A is the cross-sectional area of the bar. The quartz tube was then loaded into a tubular furnace and heated to 900 °C for 15 min, with a ramp rate of 5 °C/min. At 900 °C, the silver conductive paste forms a continuous layer between the bar and the silver wires. The tubular furnace was then ramped

2769

down to a specific temperature for measurements. The error in determining the conductivity using this procedure is approximately (3.0%. For perovskite-type ceramic membranes, the total conductivity is the summation of the proton, oxygen ion, and electrical (electron or electron-hole) conductivity. In the case of SrCe0.95-xZrxTm0.05O3-δ membranes, the oxygen ion conductivity can be neglected as both SrCe0.95Tm0.05O3-δ and SrZrO3-δ have been shown in literature to have negligible oxygen ion conduction in the experimental temperature range.10,15 Therefore, the protonic and electronic conductivity can be calculated from the total conductivity data measured under different atmospheres. The electronic conductivity of the samples was measured in different oxygen partial pressure environments by using O2/N2 gas mixtures in the temperature range of 700-900 °C. For total conductivity experiments in dry (PH2O ) 10-4 atm) O2/N2 gas mixtures, the protonic conductivity can also be neglected due to the absence of H2 in the atmosphere. The protonic conductivity was found by the difference in the total conductivity of the samples in either a 10% H2/He or CO/ He mixture of the same oxygen partial pressure in the temperature range of 700-900 °C. The partial pressure of CO was varied at different temperatures in order to have the same oxygen partial pressure as the 10% H2/He experiments had at the respective temperatures. The oxygen partial pressure for either the H2- or CO-containing atmosphere was determined using the thermodynamic equilibrium of the following reactions: 1 H2 + O2 T H2O 2

(C)

1 CO + O2 T CO2 2

(D)

The oxygen impurity in the He gas cylinder (which was found to be 185 ppm using an oxygen sensor (Illinois Instruments, 6000 Oxygen Analyzer)) was also taken into consideration, so the equilibrium oxygen partial pressure for the system was calculated using PO2 )

(

2PO2(He) PrKr

)

2

(2)

in which PO2(He) is the oxygen partial pressure in the helium gas, Pr is the partial pressure of the reducing gas (H2 or CO) in the atmosphere, and Kr is the reaction equilibrium constant for either the H2 or CO reaction. The assumption that all of the oxygen present in the He gas reacted with either H2 or CO to form CO2 or steam was used with eq 2. This method has been used in our research group before and is explained in greater detail in a previous publication.4 For all cases, the total gas flow rate into the quartz tube was 100 mL/min, which was monitored using a mass flow readout (MKS, Type 247) and mass flow controllers (MKS). To ensure steady-state conditions in each measurement, data were taken at least 3 h or 30 min after a gas concentration change or temperature change for N2/O2 atmospheres and 5 or 1 h for the 10% H2/He and CO/He mixtures. 2.3. Hydrogen Permeation Measurements. The H2 flux through dense SrCe0.95-xZrxTm0.05O3-δ (0 e x e 0.40) membranes was measured under various conditions in the temperature range of 700-900 °C using a high-temperature gas permeation system. A schematic of the system is shown in Figure 1. The gas permeation module (Probostat, Norwegian Eletro Ceramics AS) utilizes spring force and metal seals to ensure sealing for

2770

Ind. Eng. Chem. Res., Vol. 49, No. 6, 2010

Figure 1. Experimental high-temperature hydrogen permeation setup.

the H2 flux measurements. The feed gas mixture contained 10% H2 with the balance He while the sweep gas was a 20% O2/Ar gas mixture. Long-term H2 flux tests were conducted at 900 °C with the feed gas containing a mixture of CO2, H2, and He and a 20% O2/Ar sweep gas mixture. For H2 permeation measurements, a gastight membrane and silver seal (Alfa Aesar, 99.9%) were mounted on the inner alumina tube and held in place by spring pressure applied by the alumina spacer on top of the membrane. The sealing procedure consisted initially of heating the setup from ambient conditions to about 950 °C to soften the silver ring. Next, He and Ar were introduced on the feed and sweep side, respectively. The flow rate of the inert gases on either side of the membrane was 30 mL/min, which was regulated by mass flow controllers (MKS, Model 1179) and a four-channel readout (MKS, Type 247). The amount of He in the Ar stream (and therefore, the leakage rate through the seal) was determined by running gas samples through a gas chromatographer (Agilent, 6890N) with a packed column (2836PC, Alltech) and a TCD detector. Once the He content in the sweep stream was minimized, the system was ramped down (1 °C/min) to experimental conditions. For H2 permeation flux measurements with a 20% O2/Ar sweep, the relative humidity in the effluent of the sweep gas was measured by a thermohygrometer (Cole Palmer, 37950) and used to calculate the flux of H2 through the membrane. The system was allowed to equilibrate for 5 h after a gas mixture change or 1 h after a temperature change for all experiments. The error in determining the H2 permeation flux using this procedure is approximately (8%. The standard temperature and pressure used for determining the hydrogen permeation are 0 °C and 1 atm.

Figure 2. XRD patterns of SrCe0.95-xZrxTm0.05 O3-δ. Table 1. Unit Cell Parameters for SrCe0.95-xZrxTm0.05O3-δ Zr content

a (Å)

b (Å)

c (Å)

volume (Å3)

0.0 0.1 0.2 0.3 0.4

8.574 8.551 8.519 8.497 8.455

5.986 5.979 5.969 5.959 5.942

6.139 6.099 6.074 6.047 6.012

315.1 311.8 308.9 306.2 302.0

structures can be described by the tolerance factor (t),25 which quantifies the extent to which the perovskite structure varies from ideal cubic structure (where t ) 1) and is defined as t)

(RA + RO)

√2(RB + RO)

where RA, RB, and RO are the respective ionic radii in ABO3. The tolerance factor increases with increasing Zr content, with t ) 0.888 for SrCe0.95Tm0.05O3-δ and t ) 0.908 for SrCe0.55Zr0.40Tm0.05O3-δ. Figure 3 is an SEM image of the surface of a SrCe0.75Zr0.20Tm0.05O3-δ membrane. The membrane grain size is 10-20 µm with no surface defects. The membrane was determined to be impermeable to He (He permeance < 10-10 mol/(m2 · Pa · s)) using a room-temperature un-steady-state permeation system. Therefore, the gas tightness and quality of the membrane was confirmed on the basis of these results. 3.2. Protonic and Electronic Conductivity of SrCe0.95-xZrxTm0.05O3-δ. Dense SrCe0.95-xZrxTm0.05O3-δ (0 e x e 0.40) membranes were cut and polished into rectangular bars for total

3. Results and Discussion 3.1. Membrane Characteristics. Figure 2 shows the XRD patterns of sintered SrCe0.95-xZrxTm0.05O3-δ (0 e x e 0.40) samples. All compositions exhibit a single perovskite phase with orthorhombic distortion. The split peaks in the XRD patterns are not from impurities, but show the distortion of the perovskite structure.4,10 The XRD patterns can be indexed by an orthorhombic lattice with the lattice parameters, determined by Rietveld analysis, given in Table 1. From the patterns in Figure 2, there is a noticeable peak shift to higher angles with increasing Zr content. This was expected as the ionic radius of six coordinated Zr4+ (0.72 Å) is smaller than that of Ce4+ (0.87 Å).24 This indicates uniform distribution of Zr and Ce in the lattice. Besides a linear decrease in the unit cell volume, another evident trend with increasing Zr content is the decrease in the degree of orthorhombic distortion. The symmetry of the

(3)

Figure 3. SEM image of SrCe0.75Zr0.20Tm0.05O3-δ membrane surface.

Ind. Eng. Chem. Res., Vol. 49, No. 6, 2010

Figure 4. Total conductivity of SrCe0.95-xZrxTm0.05O3-δ in O2/N2 gas mixtures at 900 °C.

2771

Figure 5. Total conductivity of SrCe0.95Tm0.05O3-δ and SrCe0.75Zr0.20 Tm0.05O3-δ in either 10% H2/He (fixed composition) or CO/He of various compositions shown in Table 2.

Table 2. CO Concentrations in CO/He Mixture Needed for Same Oxygen Partial Pressure as 10% H2/He at Various Temperatures for given temperature

CO concentration (%) PO2 (atm)

700 °C

800 °C

900 °C

5.1 2.3 × 10-26

7.9 6.9 × 10-24

11.3 7.6 × 10-22

conductivity experiments using the four-point DC method. Figure 4 shows the total conductivity of SrCe0.95-xZrxTm0.05O3-δ (0 e x e 0.40) in dry O2/N2 gas mixtures at 900 °C. In the instance of measuring the total conductivity under dry O2/N2 mixtures, only electronic (electron and electron-hole) conduction plays a role in the total conductivity. As seen in Figure 4, the total conductivity of all samples increases linearly with increasing PO21/4. This indicates that p-type (electron-hole) electronic conduction is dominant in SrCe0.95-xZrxTm0.05O3-δ (0 e x e 0.40) under dry O2/N2 gas mixtures. The other trend evident from Figure 4 is that the electronic conductivity of the samples decreases with increasing Zr content. This was expected due to the lower conductivity of zirconates. Although a decrease in total conductivity with increasing Zr content has been reported for various materials,18-22 those experiments were conducted with H2-containing atmospheres (wet gases and/or H2 gas). Therefore, the decrease in the total conductivity for those systems could be a result of a loss of protonic as well as electronic conductivity. For example, Katahira et al.20 found that BaCe0.9ZrxY0.1O3-δ oxides exhibit decreasing protonic conductivity with increasing Zr content while the electronic conductivity of the oxides was nearly independent of the Zr content. To determine the effect that Zr doping has on the protonic conductivity, the total conductivity of SrCe0.95-xZrxTm0.05O3-δ (x ) 0.0 and 0.20) was measured in 10% H2/He and in a CO/ He gas mixture with the same oxygen partial pressure. Table 2 lists the oxygen partial pressure for 10% H2/He at various temperatures and the CO concentration required on the basis of calculations detailed in Experimental Methods. The total conductivity of SrCe0.95Tm0.05O3-δ and SrCe0.75Zr0.20Tm0.05O3-δ measured in 10% H2/He and CO/He environments in the temperature range of 700-900 °C is shown in Figure 5. The same membrane for either composition was used for the conductivity measurements in both reducing environments. The protonic conductivity of both compositions is the difference

Figure 6. Dependence of protonic conductivity and proton-transfer number on temperature for SrCe0.95Tm0.05O3-δ.

Figure 7. Dependence of protonic conductivity and proton-transfer number on temperature for SrCe0.75Zr0.20Tm0.05O3-δ.

between the total conductivity in 10% H2/He and CO/He atmospheres. Figures 6 and 7 show the protonic conductivity and proton-transfer number (σprotonic/σtotal) for each composition. Consistent with other proton conducting strontium cerate based perovskite-type materials, both SrCe0.95Tm0.05O3-δ and SrCe0.75-

2772

Ind. Eng. Chem. Res., Vol. 49, No. 6, 2010

J1 ) -

σ1σ2 (σ1 + σ2)z1 F

2 2

[

∇µ1 -

z1 ∇µ z2 2

]

(4)

where F is Faraday’s constant and σi, zi, and 3µi are the partial conductivity, charge number, and chemical potential gradient of species i (here 1 is for proton and 2 is for electron-hole). Following the Wagner theory, equilibrium is assumed between the proton (charged species 1), electron-hole (charged species 2), and neutral molecular hydrogen. Thus, 1 ∇µ ) ∇µ1 - ∇µ2 2 H2

(5)

With z1 ) 1 and z2 ) 1, inserting eq 5 into eq 4 gives the flux for neutral hydrogen as JH2 ) Figure 8. Hydrogen permeation flux comparison of 1.6 mm thick SrCe0.95Tm0.05O3-δ and SrCe0.75Zr0.20Tm0.05O3-δ membranes (upstream, 10% H2/He; downstream, 20% O2/Ar).

Zr0.20Tm0.05O3-δ are primarily proton conductors at lower temperatures and mixed protonic-electronic conductors at higher temperatures.4,15,19 The protonic conductivity has an activation energy of 44 kJ/ mol for SrCe0.95Tm0.05O3-δ and 32 kJ/mol for SrCe0.75Zr0.20Tm0.05O3-δ, respectively. SrCe0.95Tb0.05O3-δ has a reported protonic conductivity activation energy of 39 kJ/mol using the same method,4 which is in good agreement with the experiment results reported here. The protonic conductivity activation energy for other proton conducting perovskite materials ranges from 37 to 77 kJ/mol in wet or dry H2 atmospheres, where such materials are almost pure proton conductors.1,2,15,20 At 900 °C, the protonic and electrionic conductivities are respectively 0.0031 and 0.005 S/cm for SrCe0.95Tm0.05O3-δ, and 0.0026 and 0.0047 S/cm for SrCe0.75Zr0.20Tm0.05O3-δ at an oxygen partial pressure of 7.6 × 10-22. Clearly, Zr doping reduces both the proton conductivity and electronic conductivity of the material, and the extent of the effect is similar in the temperature range of 700-900 °C at such low oxygen partial pressures. 3.3. Hydrogen Permeation and Membrane Stability. The H2 flux through SrCe0.95-xZrxTm0.05O3-δ (x ) 0.0 and 0.20) membranes was measured using various feed gases and a sweep gas of 20% O2/Ar. Figure 8 compares the temperature dependence of the H2 permeation flux through 1.6 mm thick membranes with a 10% H2/He feed gas. To convert the flux values from mL(STP)/(cm2 · min) to SI units of mol/(m2 · s), the values in Figure 8 should be multiplied by 0.00682. The H2 permeation flux values for SrCe0.95Tm0.05O3-δ were previously reported by Qi and Lin.10 SrCe0.95Tm0.05O3-δ has higher H2 permeation flux values than SrCe0.75Zr0.20Tm0.05O3-δ over the temperature range tested. This was expected due to the fact that SrCe0.95Tm0.05O3-δ has higher protonic and electronic conductivities than SrCe0.75Zr0.20Tm0.05O3-δ within experimental conditions. The hydrogen permeation flux for the SrCe0.95Tm0.05O3-δ membrane is about 5 times higher than the SrCe0.75Zr0.20Tm0.05O3-δ membrane. The energy of activation for hydrogen permeation are 36.5 and 78.1 kJ/mol respectively for SrCe0.95Tm0.05O3-δ and SrCe0.75Zr0.20Tm0.05O3-δ under the experimental conditions in Figure 8. For thick membranes, bulk diffusion can be assumed to be the rate-limiting step for hydrogen permeation. The proton flux through a membrane with no external current can be written as27

σ 1 σ2 4(σ1 + σ2)F2

∇µH2

(6)

Since detailed hydrogen pressure dependence of the conductivities is not known, eq 6 cannot be integrated to correlate the hydrogen permeation flux to the permeation conditions.28 However, eq 6 shows that the hydrogen permeation flux is proportional to the constant of [σ1σ2/(σ1 + σ2)], which, for the membrane surface exposed to H2/He at 900 °C, is 0.0019 for SrCe0.95Tm0.05O3-δ and 0.0017 for SrCe0.75Zr0.20Tm0.05O3-δ. The ratio of the constant [σ1σ2/(σ1 + σ2)] using the values for the membrane surface exposed to H2/He mixture is 1.1, clearly not correlated to the ratio of the hydrogen flux for these two membranes at this temperature (about 5). For the membrane surface exposed to the O2/Ar mixture, the electronic conductivity is 0.032 and 0.016 S/cm respectively for SrCe0.95Tm0.05O3-δ, and SrCe0.75Zr0.20Tm0.05O3-δ at 900 °C. The protonic conductivity under such conditions is not known and cannot be measured directly. The hydrogen permeation data suggest that for the membrane surface exposed to the O2/Ar mixture the protonic conductivity is likely to be smaller than that for the surface exposed to the H2/He mixture (