Lanthanum chromite, LaCrO₃, is a refractory oxide with melting point of 2430 °C, taking on a perovskite-related structure with orthorhombic symmetry (Pnma, space group 62) [15]. It exhibits a small range of cation non-stoichiometry, as indicated in the phase diagram [16], and it can be substituted with a number of homo- and heterovalent cations on both the A (La) and B (Cr) sites. Of greatest interest is the effect of acceptor substitution by alkaline earth cations—notably Ca²⁺ or Sr²⁺—for La³⁺ [17,18], causing charge compensation by oxygen vacancies at low pO₂ and electron holes at high pO₂ [19,20,21]. For this reason, Ca- and Sr-substituted LaCrO₃—hereafter LCC and LSC, respectively—exhibit high p-type electronic conductivity under oxidizing and moderately reducing conditions, falling off under more reducing conditions, but still with dominating p-type electronic conductivity well above 1 S/cm [19,22,23]. This, along with its great stability towards oxidation as well as reduction, has made it of interest as an interconnect material in SOFCs, where it must exhibit electronic conductivity, be gas tight in order to separate the fuel and oxidant, and preferably not conduct oxide ions [19,24].

It was pointed out early on that the presence of oxygen vacancies would give considerable oxide ion conductivity and permeability of oxygen in LCC and LSC [22], but it appeared that the thickness of the interconnect and the rather slow surface kinetics of LaCrO₃-based materials made this a minor problem [30].

It is very difficult to sinter LaCrO₃-based ceramics to high density, and a lot of research has been invested into overcoming this bottleneck, using, for instance, sintering aids [25]. Interconnects of these materials also suffer from Cr evaporation, notably by hexavalent chromium oxohydroxides [26], which poisons SOFC cathodes by the deposition of Cr oxides at the electrocatalytic triple-phase boundary sites between strontium-substituted LaMnO₃ (LSM) cathodes and YSZ electrolytes [27]. Additional weaknesses comprise moderate mechanical properties, chemical expansion by reduction, limited thermal conductivity, and high costs of machining. For these reasons, LaCrO₃-based interconnects are now for the most part being replaced by metallic interconnects—enabled by the reduced temperatures of operation of modern SOFCs [27,28,29].

LaCrO₃ compensates acceptor-dopants with electron holes under oxidizing conditions and oxygen vacancies under reducing conditions, accompanied by corresponding increases the oxygen ion and p-type electronic conductivities. Studies, where oxygen non-stoichiometry has been measured as a function of oxygen partial pressure (pO₂) by means of thermogravimetry, have been conducted for LaCrO₃ with various dopants, doping levels and temperatures [20,21]. Mizusaki et al. [20] reported that the oxygen non-stoichiometry δ for La_1−xSr_xCrO_3−δ (x = 0.1–0.3) was close to zero for pO₂ > 10⁻⁵ and eventually reached x/2 with decreasing pO₂. These pO₂ regimes represent two limiting conditions where the acceptor is predominately charge-compensated by electron holes (i.e., ) and oxygen vacancies (i.e., ), respectively. Moreover, the limiting pO₂, at which oxygen vacancies and electron holes are equally dominant, shifts to higher pO₂ with increasing doping levels and temperatures. The formation of oxygen vacancies at the cost of electron holes can, in Kröger-Vink notation, be written:

The equilibrium constant for this reaction is given by:

and Mizusaki et al. estimated the standard formation enthalpy of oxygen vacancies according to this reaction to 303 kJ/mol for x = 0.1. A similar behavior was observed for LaCr_(1−x)Mg_xO_3−δ [21], where the corresponding enthalpy was seemingly independent of dopant concentration and estimated at 272 kJ/mol for x = 0.02–0.1.

From the limiting electroneutrality condition, the concentration of electron holes will be independent of pO₂ when the acceptor is charge-compensated by electron holes, since it follows from Equation (2) that it exhibits a pO₂^1/4-dependecy when the acceptor is charge-compensated by oxygen vacancies. The defect concentrations as a function of pO₂ are illustrated in Figure 2. Due to higher mobility of electron holes in general compared to ionic defects, the total conductivity will also exhibit the pO₂-dependencies of the electron holes [22,23].

Hydration of oxygen vacancies in oxides occurs according to:

with the corresponding equilibrium constant:

When protons are included, the overall electroneutrality condition of LSC can be approximated by:

As can be deduced from Equations (1) and (3) and seen in Figure 2, the concentration of protons increases with decreasing pO₂. From Equation (4), it follows that the concentration of protonic defects as a function of water vapor partial pressure (pH₂O) is proportional to pH₂O^1/2 if electron holes or oxygen vacancies are the dominating charge compensating defects (and thus independent of pH₂O). If protons become the dominating charge compensator, the minor concentration of oxygen vacancies becomes proportional to pH₂O⁻¹, while the concentration of electron holes, via Equation 2, becomes proportional to pH₂O^−1/2.

To the best of our knowledge, there have been no significant reports of protons in LSC until this study, and, as mentioned above, the correlation between the enthalpy of the hydration of oxygen vacancies and the stability of the perovskite does not predict that hydration is favorable at all, while the three d-electrons of Cr³⁺ suggests that LSC may be an exception. The defect concentrations as a function of water vapor partial pressure are illustrated in Figure 3, based on a moderately favorable hydration enthalpy. One may note that at high pH₂O, hydration and the presence of protons should be detectable as a decrease in the p-type electronic conductivity as a function of pH₂O.

The electron holes in LaCrO₃ appear to behave as small polarons localized to the Cr-site, and the hole mobility exhibits a small activation energy. Consequently, the p-type conductivity is also activated (), and in the region where the holes are dominating and their concentration given by the constant concentration of acceptors, E_a, is given only by that of the mobility. Karim and Aldred [31] reported that the electrical conductivity of La_1-xSr_xCrO₃ increased from the order of 10⁻² S/cm at temperatures above 1000 °C and pO₂ = 1 atm for x = 0, to values two orders of magnitude higher for high values of x up to the maximum level (x = 0.4). The electron hole activation energy was estimated at 0.192 eV for the undoped chromite (x = 0), and decreased with increasing acceptor-concentrations to 0.113 eV for x = 0.4.

LaCrO₃ is also an oxide ion conductor when acceptor doped. Oxygen vacancy diffusion coefficients in the order of 10⁻⁶–10⁻⁵ cm²/s were calculated from electrical measurements of LaCrO₃ with varying dopants and doping levels (0.1–0.35 mol%) at temperatures 900–1100 °C [22]. Based on electrochemical methods, Yokokawa et al. [32] estimated a lower vacancy diffusion coefficient in the order of 10⁻⁷ cm²/s at 1000 °C, in agreement with calculations by Van Hassel et al., based on oxygen permeation [33]. Later on, Kawada et al. [34] revisited this method and estimated an oxygen vacancy diffusion coefficient in the order of 10⁻⁵ cm²/s. The lower values of the oxygen vacancy diffusion coefficients reported earlier were mainly explained by an underestimation of the oxygen non-stoichiometry in the high pO₂ region.

Figure 4 presents the total a.c. conductivity of La_0.9Sr_0.1CrO_3−δ as a function of pH₂O in reducing atmospheres at temperatures from 300 to 800 °C. Each isotherm is made by diluting a mixture with fixed pH₂/pH₂O ratio with dry Ar, thus keeping pO₂ essentially constant, while changing pH₂O (and pH₂). The values of pO₂ vary from 3 × 10⁻¹⁷ and 4 × 10⁻²¹ atm (depending of pH₂/pH₂O ratio) at 800 °C to 2 × 10⁻⁴² atm at 300 °C, and under these conditions we may assume that the acceptor dopant is, to a first approximation, charge-compensated by oxygen vacancies. This assumption is in agreement with the higher p-type electronic conductivity observed at pO₂ = 3 × 10⁻¹⁷ than at pO₂ = 4 × 10⁻²¹ atm at 800 °C, as the concentration of electron holes will increase with increasing pO₂ when in the minority. Still, electron holes dominate the total conductivity, due to higher mobility than for ionic defects.

Below approximately 600 °C, the conductivity decreases with increasing pH₂O, reflecting a decrease in the concentration of electron holes as protonic defects are introduced in significant numbers compared to the dominating oxygen vacancies.

By assuming that oxygen vacancies and protons are the dominating charge compensators (), the equilibrium constant of Equation (3), K_hydr, was extracted from the fit of the conductivity curves for each temperature. The results are presented in a van 't Hoff plot (Figure 5), and linear regression of lnK_hydr as a function of 1/T yields standard enthalpy and entropy changes of the hydration reaction of −70 kJ/mol and −90 KJ/mol, respectively. The equilibrium constant extracted from conductivity measurements at 300 °C was excluded in this fitting, due to significant deviation from the values obtained at higher temperatures, likely a consequence of slow kinetics. The exothermic hydration enthalpy of −70 kJ/mol for La_0.9Sr_0.1CrO_3−δ is far from the endothermic expected, based on the already mentioned empirical correlation [7,8,9,10], and the aforementioned crystal field stabilization of the d-electrons in fully vs. only partially occupied octahedra of the perovskites is an interesting input to understanding this and the many other variations from the correlation.

With the parameters obtained by curve-fitting of the hydration equilibrium constant, we have estimated the concentrations of the defects in La_0.9Sr_0.1CrO_3−δ in wet reducing atmospheres (i.e., pH₂O = 0.025 atm, pH₂ = 1 atm). All charge carriers were assumed to follow an activated transport process, i.e., . For oxygen vacancies, the pre-exponential factor of mobility, u_0,vo, and the enthalpy of mobility, ΔH_mob,vo, were set to 200 cm² K/Vs and 90 kJ/mol, respectively, in accordance with oxygen vacancy diffusion coefficients obtained at temperatures from 950–1100 °C for La_0.8Sr_0.2Cr_0.95Ni_0.05O_3−δ [30]. For protons, the pre-exponential factor of mobility, u_0,OH, was set to 13 cm²K/Vs, in accordance with values obtained for similar perovskite oxides from [9] and the data set of [8], and the enthalpy of mobility, ΔH_mob,OH, was set to 60 kJ/mol, from the empirical 2/3 of the value for oxygen vacancies [8]. The pre-exponential factor of hole mobility, u_0,h, and the enthalpy of hole mobility, ΔH_mob,h, were set in accordance with Weber et al. [35] to 200 cm²K/Vs and 12.5 kJ/mol, respectively. Furthermore, it was assumed that the nominal Sr acceptor level was effective.

The estimated defect concentrations and partial conductivities are presented as a function of inverse temperature in Figure 6 and Figure 7, respectively.

The estimated proton conductivity reaches approximately 5 × 10⁻⁴ S/cm at 800 °C, while the oxygen vacancies reach a conductivity in the order of 10⁻³ S/cm. Figure 7 also illustrates that the electron hole conductivity dominates the total conductivity, even when electron holes are minority defects under reducing conditions. The bends in the conductivity curves are the result of the changeover to a proton-dominated defect structure at lower temperatures.

Based on the estimated partial conductivities, we may estimate the ambipolar proton-hole conductivity of La_0.9Sr_0.1CrO_3−δ from

and it is limited by the lower of the two partial conductivities. In LSC, proton conductivity is orders of magnitude lower than that of electron holes, thereby limiting this ambipolar conductivity.

Hydrogen production measurements—interpreted as apparent hydrogen flux—were performed on disk-shaped LSC membranes of approximately 0.5 mm thickness, prepared with various sintering agents (see Table 1). They showed an initial apparent flux of 0.04–0.02 mL min⁻¹cm⁻², which decreased over some 50–200 h at 1000 °C to final values of 0.02–0.01 mL min⁻¹cm⁻² at 1000 °C in 50% humidified H₂ on the feed side and humidified argon at the permeate side (see Figure 8). The initially high apparent flux may be related to an initially high concentration of intrinsic defects created under the very high sintering temperature. The decrease in apparent flux may alternatively be related to the relaxation and decrease in surface catalytic activity of the polished surfaces during initial operation at elevated temperatures.

The effect of different types of gradients was further investigated in order to separate situations with only hydrogen (proton-hole) transport and cases where oxygen transport, and hence water splitting, contribute to the hydrogen production. Four different gradients are shown in Figure 9, where all have 50% or 10% hydrogen on the feed side and argon on the permeate side. The only difference is that the gases are either wet or dry. Accordingly, the following effects can be seen:

Dry feed/Wet permeate (DF/WP) gives the highest hydrogen production. This can be attributed to a high level of water splitting on the permeate side, combined with a high driving force for transport of oxygen from permeate to feed. Wet feed/wet permeate (WF/WP) gives a slightly lower hydrogen production. One can expect the driving force for water splitting to be lower, since the difference in p(O₂) is less. Wet feed/dry permeate (WF/DP) gives yet lower hydrogen production. In this case we have no possibility of water splitting and the hydrogen must come from ambipolar proton transport. Dry feed/dry permeate (DF/DP) gives the lowest hydrogen production when comparing cases with equally amounts of hydrogen on the feed side. Here, we are probably additionally limited by the hydration and concentration of protons in the oxide.

It is noteworthy that what we believe is water splitting from ambipolar oxygen transport and what we believe is ambipolar hydrogen transport both appear to have similar and rather high activation energies. As we shall see later, this is possibly related to both being limited by surface kinetics of the polished disk samples used.

Hydrogen production measurements were performed up to 5 bars on both feed and permeate sides, using high steam contents on both sides, in order to increase the proton concentration and reduce the vacancy concentration in LSC by water incorporation, according to Equation (3). Results for 1000 °C are plotted on Figure 10. The apparent hydrogen flux is largely independent of total pressure and of steam content, see Figure 10a, c, and instead the amount of hydrogen in the feed plays a somewhat larger role—see Figure 10b. These observations do not immediately lend support to either proton or oxygen transport; the hydrogen production can well come from both hydrogen flux and water splitting. We note, however, that the values of apparent flux are qualitatively similar to those done at atmospheric pressure and normal wet gases above.

The hydrogen production of an LSC monolith module was also tested at 1000–800 °C and 20 bars pressure in moist hydrogen. The monolith had 37 active channels: 21 for feed gas and 16 for permeate sweep gas. The dense membrane was applied on the porous walls in the monolith on the sweep channel side and had a thickness of 50 μm. The concentration of hydrogen in the permeate gas was 19%–22%. Figure 11 shows the area specific hydrogen leakage and apparent hydrogen flux.

Accordingly, there is a considerable apparent flux through the membrane, and the leakage is low compared to the calculated apparent hydrogen flux. The measurements were performed with co-current flow of feed and sweep gas, rapidly decreasing the driving force for hydrogen permeation, and the gas feeding rate (1000 mL/min) was also low compared to the high apparent flux and membrane area (86 cm²). Besides, the performance of the module was not tested with air on the sweep side, which would have served to maintain a high driving force throughout the monolith and increase the permeation of hydrogen, as mentioned in the introduction for the selected PCDC process. It is therefore expected that the module would perform even better under realistic conditions, e.g., counter-current flow regime with air at the sweep side. There is also room for decreasing the thickness of the membrane well below 50 µm. A hydrogen production rate of the order of 10 mL cm⁻² min⁻¹ under real operating conditions seems therefore to be within reach with this material.

It may still be added that this experiment may also have had a large portion of the hydrogen produced coming from water splitting. The leakages were large enough to prevent a detailed analysis of the mass balances that could have answered this question.

We note that the apparent flux densities are now some two orders of magnitude larger than in the case of the polished disk samples and that the temperature dependency, i.e., activation energy, is now small. We speculate that the surface of the membrane applied inside the monolith channels is catalytically active compared to the polished disk surfaces, and that the results of the monolith reflect bulk limited diffusion, while those of the disk samples reflect surface kinetics limitations.

In order to unambiguously prove diffusion of hydrogen and perhaps quantify the possible contribution from water splitting on the permeate side of the membrane to the produced hydrogen, permeation measurements were carried out using D₂ and D₂O on the feed side of the membranes. Mass spectrometry was used to analyze the permeate gas exiting the cell. Results are plotted in Figure 12.

With dry permeate gas, the presence of only D₂O on the feed side gives rise to very little D-containing species in the permeate, showing that there is little D₂O flux from ambipolar proton (here deuteron) and oxide ion transport, see Figure 12 (first hour). Introduction of D₂ on the feed side gives a driving force for ambipolar deuteron-electron hole transport, and a large and dominating increase in the D₂ content on the permeate side results. This proves unambiguously the presence of hydrogen flux due to ambipolar proton (here deuteron) and electron hole transport.

Experiments were also made with H₂O-wetted permeate gas. In this case, D₂ in the permeate, originated from flux, will isotopically mix with H₂O, resulting in an increase in the concentration of species HD, HDO, H₂ and D₂O, and a decrease in the concentration of H₂O, see Figure 13.

A mass balance analysis of the permeate flue gas for this experiment at 1040 °C results in around 80% of the hydrogen (here hydrogen and deuterium) produced comes from hydrogen (here deuterium) flux and that 20% comes from water splitting.

It may be noted that the sample used here was again a polished disk membrane. The high level of hydrogen flux compared to water splitting suggests that although both processes may be surface kinetics limited, the water splitting is more limited by surface kinetics than the hydrogen flux is.

Figure 14 presents the ambipolar proton-electron hole and oxygen vacancy-electron hole conductivities as a function of inverse temperature estimated from electrical measurements, the ambipolar proton-electron hole conductivity extracted from the measurements of hydrogen production with the disk membrane (wet and dry permeate gas), and the ambipolar proton-electron hole conductivity extracted from the measurements of hydrogen production with the monolith module (wet permeate gas).

Two different models were used to describe the hydrogen flux and relate it to ambipolar conductivity for wet and dry permeate gas. Both assume that the transport number for electron holes is unity, so that ambipolar conductivity equals proton conductivity. For wet permeate gas (and wet feed gas) we assume that the hydration level and thus proton concentration is constant throughout the membrane. The difference in pH₂ is the only driving force for hydrogen (proton) flux, and the following model is applied:

Here, the proton conductivity is the one valid for the pH₂O used in the measurement, L is the membrane thickness, and the other symbols have their usual meanings.

From the dry permeate case, we assume that the pH₂/pH₂O ratio is constant throughout the membrane, which is attained to a first approximation through ambipolar water flux accompanying the hydrogen flux. The proton conductivity is thus not constant throughout the membrane, and the following model is applied:

Here the pH₂/pH₂O used. It may be noted that both models yield proton conductivities that are valid for a given water vapor content, which may be recalculated for another one assuming that the protons are minority defects and their concentration proportional to pH₂O^1/2, and this has been applied in plotting Figure 14.

While the models are uncertain, the choice of model does not dramatically alter the resulting proton conductivity.

Based on the knowledge of the defect structure and transport properties of LSC, the conductivity measurements, the D₂ flux measurements and isotope exchange data, and the hydrogen production measurements from disk and monolith membranes, we may now tentatively conclude the following from our studies so far:

From the changes in p-type electronic conductivity with pH₂O, we know that LSC hydrates significantly and contains considerable concentrations of protons. The thermodynamics of hydration can be extracted, and we can make a first guess of the proton conductivity in LSC in wet atmospheres as a function of temperature and compare this with an estimate of the oxide ion conductivity which is based on more solid data (Figure 14). From this and the dominating p-type electronic conductivity we can predict that LSC will be permeable to hydrogen, oxygen, and water in gradients of pH₂, pO₂, and pH₂O. The two first can be used respectively for hydrogen production by hydrogen flux and water splitting.

The D₂ flux measurements prove unambiguously that there is in fact hydrogen flux in the material. Moreover, they suggest that hydrogen flux and water splitting contribute within the same order of magnitude to the hydrogen production (80% and 20%, respectively, for a disk sample at 1040 °C in wet gases).

Hydrogen production measurements using a monolithic membrane module with large area at high pressures and high steam contents yield a high output, one order of magnitude higher than estimated from the guessed proton conductivity, and with a small temperature dependency as predicted for hydrogen flux limited by bulk transport, see Figure 14. All in all, we suggest that this is to a large extent hydrogen flux, and that the protons therefore may have a higher mobility than guessed.

Hydrogen production measurements using polished disk samples show variations with atmosphere composition that support the simultaneous presence of hydrogen flux and water splitting. The activation energies are, however, consistently higher and the apparent flux lower than expected from the ambipolar conductivities (see Figure 14). This suggests that these samples are severely limited by surface kinetics. Based on this, the activation energy for surface kinetics is similar for both hydrogen flux and water splitting, although details in the results from the various experiments can be interpreted to reflect that water splitting is more limited by surface kinetics than hydrogen permeation.

The monolith module demonstrated promising flux densities and hydrogen production rates, and appears to have been produced with relatively active surfaces.

Future studies of LSC membranes for hydrogen production must focus on the kinetics and engineering of its surfaces.

The LSC powders were prepared by spray pyrolysis at 750 °C of an aqueous nitrate solution with chromium, strontium and lanthanum concentrations according to the perovskite stoichiometry. The solution contained 1 mole metal ions per liter, corresponding to approximately 115 g perovskite per liter. To retain maximum sintering activity and suppress the concentration and evaporation of carcinogenic Cr(VI), the powder was calcined in vacuum (1 mbar) at 850 °C. The powder was finally milled in a steel attrition mill for 120 min at 300 rpm with 3 kg distilled water to 2 kg LSC. To increase sinterability, small additions (2–4 mol%) of various partly-proprietary sintering aids were added during milling. Dilatometer studies of a powder batch with 2 mol% CaMnO₃ demonstrated densities of >95% by sintering in air at 1700 °C or in 4% H₂ in N₂ at 1600 °C. The latter condition was used for sample preparation, where the LSC powder with various sintering aids (Table 1) was pressed into green bodies and sintered to dense pellets.

LSC powder with 2 mol% CaMnO₃ was also used for fabrication of the small asymmetric monolithic membrane module. The support structure was a circular porous monolith, extruded from a mixture of LSC and pore former (20 wt % cellulose fiber) with binder (methyl cellulose), plasticizer (polyethylene glycol) and water. The extruded monoliths were dried at 40 °C, while a fan maintained air circulation through the channels. To remove the organic components, the dry monoliths were heated to 500 °C at a low rate of 0.1°/min to avoid overheating and cracking. After a final pre-sintering of the support monolith at 1200 °C, the LSC membrane was applied by dip coating with access to each second monolith channel in a chess pattern. After sintering in dilute hydrogen at 1700 °C, the dense membrane layer was 50 µm thick, and proved gas impervious at room temperature, while the channel width and wall thickness were 1.7 and 0.4 mm, respectively.

Pictures of the membranes and the monolith have been published elsewhere [3]. The monolith had 37 active channels: 21 for feed gas and 16 for sweep gas. The membrane was applied in the sweep channels. The module was made from a 90 mm long monolith with 86 cm² membrane area. To distribute feed and sweep gas in the appropriate channels, manifolds were sealed to the monolith at each end. Each manifold was made from a dense choke plate and gas distributor in LSC and a hat in yttrium stabilized zirconia. The components were sealed to each other by a glass ceramic with thermal expansion profile matching LSC. The module also had 2 gas baffles in platinum around the monolith, matching the inner diameter of the test reactor. During testing, the hats were sealed to Al₂O₃ support tubes by gold rings. Thus, the feed side of the membrane module was available from the inside of the supply tubes, while the sweep channels were available from the outside through the side holes of the gas distributor.

Electrical conductivity measurements in controlled atmospheres were performed on a La_0.9Sr_0.1CrO₃-sample in a ProboStat^TM cell (NorECs AS, Norway). The sample was rather porous—only 82% dense—in order to get a fast response with changes of atmosphere. An alternating current (a.c.), van der Pauw four point method, was used to measure the conductivity as a function of temperature and concentration of water vapor. The measurements were performed using a Solartron SI 1260 Frequency Response Analyzer operating at 1590 Hz and 1 V amplitude. The conductivity was not corrected for porosity, as this is a minor effect, and has no consequence for the interpretations used here. Different hydrogen-argon mixtures were used and the gas composition was controlled by gas mixing and wetting stages. The conductivity measurements were done under atmospheric pressure.

Measurements of hydrogen production of disk-shaped membranes were performed in a ProboStat^TM cell, using planar polished samples and gold rings as seals. The support tubes were also polished, in order to improve the sealing process. Helium was used to verify successful sealing, indicated by <0.01 mol% detection of He using a gas chromatograph. Measurements were performed by using a humidified He:H₂ mixture (50:50 and 90:10) on the feed side (with a total of 50 mL/min gas flow) and humidified Ar on the permeate side (with a gas flow of 25 mL/min). For pressurized measurements to 5 bars, the normal flow was multiplied by the pressure. Both He and H₂ concentrations were measured with a calibrated Varian 4900 µ-GC. The apparent fluxes were calculated from the measured permeate H₂ concentration and the calibrated flow of Ar sweep gas after correcting for He leak using a Knudsen ratio. For the pressurized flux measurements, a metal mantel was used on the ProboStat^TM cell, which allowed measurement up to 30 bars at 900 °C or 10 bars at 1000 °C. For mass spectrometry measurements, the permeate gas was analyzed with a Pfeiffer PrismaPlus QMG 220, and the spectrometer ion current converted to relative concentrations of species in the gas, based on several background and calibration runs of known gas mixtures.

Measurements of hydrogen production of the module were carried out in a different rig. The module was made from a 9 cm-long monolith, which gave a total of 86 cm² membrane area. It was installed in a reactor of Paralloy CR39W steel alloy, which is suited for test conditions involving high pressure, temperature and hydrogen activity. The feed gas was a mixture of 70% H₂, 20% steam, and 10% N₂, while the sweep gas was 80% Ar and 20% steam. At each side, the total flow rate was 1000 mL/min. After sealing the monolith module to the test reactor at 1000 °C, only 0.5% nitrogen was detected in the sweep side flue gas in the entire temperature range.