Hydrogen Purification through a Highly Stable Dual‐Phase Oxygen‐Permeable Membrane

Abstract Using oxygen permeable membranes (OPMs) to upgrade low‐purity hydrogen is a promising concept for high‐purity H2 production. At high temperatures, water dissociates into hydrogen and oxygen. The oxygen permeates through OPM and oxidizes hydrogen in a waste stream on the other side of the membrane. Pure hydrogen can be obtained on the water‐splitting side after condensation. However, the existing Co‐ and Fe‐based OPMs are chemically instable as a result of the over‐reduction of Co and Fe ions under reducing atmospheres. Herein, a dual‐phase membrane Ce0.9Pr0.1O2−δ‐Pr0.1Sr0.9Mg0.1Ti0.9O3−δ (CPO‐PSM‐Ti) with excellent chemical stability and mixed oxygen ionic‐electronic conductivity under reducing atmospheres was developed for H2 purification. An acceptable H2 production rate of 0.52 mL min−1 cm−2 is achieved at 940 °C. No obvious degradation during 180 h of operation indicates the robust stability of CPO‐PSM‐Ti membrane. The proven mixed conductivity and excellent stability of CPO‐PSM‐Ti give prospective advantages over existing OPMs for upgrading low‐purity hydrogen.

ratios. The mole ratio of total metal cation: EDTA: citric acid was 1:1:1.5 and the pH value of mixed solution was adjusted to around 9 by adding NH 4 OH solution. In order to avoid tetrabutyl titanate hydrolysis, it is necessary to dissolve it in an ethanol solution containing lactic acid and acetic acid firstly (weight ratio of tetrabutyl titanate: lactic acid: acetic acid is 1: 1: 1). Then the prepared solution was slowly added into the previous precursor solution followed by stirring for 24 h until the solution became transparent. After heat and further combustion, precursor ashes were obtained and then calcined at 950 °C for 10 h to remove the residual organics. The final gas-tight CPO-PSM-Ti and CPO-PSM-Fe membranes with diameter of 15 mm and thickness of 0.7 mm were prepared by compressing the above powders into disks under a pressure of 10 MPa and sintering at 1450 °C in ambient air for 10 h. The CPO-PSM-Ti coating paste was printed on the membrane surface with a brush. The coated membrane was fired at 1200 °C for 1 h to improve the adhesion between coating layer and membrane.

Characterizations
The crystal structures of different powders and membranes were studied using X-ray diffraction (XRD, D8 Advance, Bruker-AXS, with Cu Kα radiation) at 2θ through the range of 20°-80° with intervals of 0.02°. In situ XRD for CPO-PSM-Ti membrane was performed using a high temperature cell (Buehler HDK 2.4 with REP 2000), with steps of 100 °C at 12 °C min -1 heating rate in the temperature range 25-900 °C. To observe the surface morphology, grain boundary and cross section of fresh and spent membranes were investigated by scanning electron microscopy (SEM, Hitachi S-4800).
The backscattered electron micrographs and energy-dispersive X-ray (EDXS) analyses were characterized by a Thermo Scientific Prisma E scanning electron microscope (SEM). To compare the chemical stability, the CPO-PSM-Ti and CPO-PSM-Fe membranes were treated under 50 vol. % H 2 in dry or wet N 2 (total flow rate was 60 mL min -1 ) at 900 °C for 10 h.
The XPS of CPO-PSM-Ti powder was tested by a X-ray photoelectron spectroscopy equipped with The electrical conductivity measurement for the sintered membrane was carried out using Interface 5000 E potentiostat / galvanostat (GAMRY Instruments) at 600-940 °C. Both sides of membrane were grounded and polished to have the desired thickness and flat surfaces. The gold paste was firstly

SUPPORTING INFORMATION
4 brushed on the both surfaces of the membrane as the electrodes and dried at 200 °C -10 min, followed by calcination at 800 °C for 2 h. The data was acquired using a two-electrode cell configuration ensuring equilibrium conditions at each point. Electrical conductivities were tested under pure argon and hydrogen with different concentrations, respectively. The system was equilibrated under each condition until reaching the steady state, and then the first data point was taken.

Performance test of membrane for hydrogen purification
The oxygen permeation tests and hydrogen purification experiments were carried out using a selfmade high temperature device as described in our previous work. [1] The membrane was fixed on an alumina tube and sealed using a glass ring (Schott 8252) at 1053 °C for 1 h. The operation temperature was controlled by a microprocessor temperature controller equipped with a thermocouple. The effective area of the membrane was about 0.6 cm 2 . All gas flows (He, N 2 , H 2 ) were controlled by gas mass flow controllers. The flow rate of steam was controlled by a liquid mass flow controller, and water had been fully evaporated at 180 °C before feeding into the reactor. All gas mass flow controllers were accurately calibrated with a soap flow meter and liquid mass flow controller was calibrated by measured the water weight among given time. The compositions of the outlet gases after condensation were analyzed by a gas chromatograph (GC, Agilent 7890B equipped with Porapak Q and 13X columns) using TCD. When synthetic air and He were used as the feed gas and sweep gas, respectively, the oxygen permeation flux was calculated according to the equation (1). Only when the leakage of oxygen was less than 5 % under air/He atmosphere, additional permeation measurements under air/H 2 and H 2 O/H 2 atmospheres were then carried out.
Where, C O2 and C N2 are the oxygen and nitrogen concentrations on the sweep side. F is the total flow rate of the outlet and A is the effective membrane area. 4.02 is the Knudsen diffusion factor of nitrogen and oxygen leakage through possible pores or cracks, and Knudsen diffusion factor was calculated by following equation (2).
When the diluted H 2 was used as the sweep gas, the oxygen permeation flux was calculated according to the equation (3).
When hydrogen purification experiments were conducted, feed side was mixed gas (35 mL min Where, C H2 is the hydrogen concentrations on the feed side detected by GC. C leak means the leaked hydrogen through possible pores or cracks that was calculated according to equation (6).
M stands for the relative molecular mass, C stands for the gas concentration on the feed side and F stands for the flow rate of the sweep gas. N 2 concentration in the feed side was detected by GC. The H 2 leakage was subtracted for calculating the H 2 production rate. Figure S1. The schematic illustration of hydrogen purification using CPO-PSM-Ti oxygen permeable membrane.       The characteristic peaks of Sr, C, Ti, O and Ce can be obviously observed in the survey spectrum ( Figure S8a). The XPS of O 1s spectrum is shown in Figure S8b from which the deconvolution peaks centred at 529.1 eV and 532.1 eV correspond to the definite oxygen species. [2] The peak at 529.1 eV is ascribed to the lattice oxygen and the one at 532.1 eV is attributed to surface adsorbed oxygen. Figure   S8c displays the high-resolution XPS spectrum of Ti 2p in the CPO-PSM-Ti sample. The fitting results of the experimental data show two peaks at the binding energy of 458.8 eV and 465.1 eV, which point to the Ti 2p3/2 and Ti 2p1/2, respectively. [3] Especially, the Ti 2p3/2 peak is well fitted by the two peaks of Ti 4+ (458.9 eV) and Ti 3+ (457.4 eV) and the Ti 2p1/2 peak is well fitted by the two peaks of Ti 4+ (464.7 eV) and Ti 3+ (463.0 eV), clearly demonstrating that the Ti ions became reduced by H 2 . It is worthy to mention that there are no Ti 2+ , Ti + and metallic Ti species at the surface of sample, indicating that Ti ions are only modestly reduced from Ti 4+ to Ti 3+ without over-reduction, and thus cubic-phase Figure S9. Thermogravimetry analyses of CPO and PSM-Ti powders.

Results and Discussion
As shown in Figure S9, about 2.0 wt. % weight loss was observed for CPO as temperature rise from 300 °C to 900 °C, while the weight loss for PSM-Ti sample was only about 1.5 wt. %. The higher weight loss implied that more available lattice oxygen was released from CPO phase, indicating that the CPO has higher oxygen vacancy concentration under low oxygen partial pressure. [5] Meanwhile, the PSM-Ti also shows oxygen release ability at elevated temperatures, which may be ascribed to the lowvalence Mg 2+ doping [6] or the change in Ti valence states according to Equation: Ti 4+ -(O 2-)-Ti 4+ =Ti 3+ -□-Ti 3+ +1/2O 2 ("□" represents oxygen vacancy) at low O 2 partial pressure.  Considering the membrane surface has been activated by porous layer and the thickness of membrane is large enough, we assume that the bulk diffusion is the rate limiting step for oxygen permeation through the membrane. Therefore, according to Wagner theory, the oxygen permeation flux (J O2 ) during the permeation process can be written as the following Equation (1).
where R is the gas constant, L is the membrane thickness, F is the Faraday's constant, σ t is the total conductivity, σ el and σ ion are the electronic conductivity and oxygen-ionic conductivity, respectively.
P' O2 and P'' O2 are the O 2 partial pressure of feed and sweep sides, respectively. The oxygen ionic conductivity (σ ion ) of a mixed conductor can be calculated from the oxygen permeation flux by integrating Equation (1), thus modifying the final form to Equation (2). [7][8] The oxygen permeation fluxes of CPO-PSM-Ti membrane under a condition of H 2 O//H 2 ( Figure 3) were used to calculate the σ ion . All the equilibrium compositions on the two sides are calculated using Gibbs free energy minimization simulations using HSC Chemistry 5.0. Table S1 shows the thermodynamic equilibrium P O2 on two sides of the membrane and the oxygen ionic conductivity (σ ion ) at different temperatures (860-940 °C) at 1 atm. As shown in Figure S10, the oxygen ionic conductivity of membrane calculated by Equation (2)  As shown in Figure S11, when low-purity hydrogen (H 2 and N 2 ) with different flow rates (20 or 60 mL min -1 ) was used to consume the permeated O 2 under the condition of 40 mL min -1 (H 2 O+He) as feed gas, the H 2 production rate gradually increased with the increase of H 2 concentration on the sweep side.
An increase of H 2 production rate from 0.33 to 0.80 mL min -1 cm -2 was obtained with the increase of H 2 concentration from 20 vol. % to 80 vol. % using 60 mL min -1 H 2 and N 2 as sweep gas. As shown in Table S2, the nominal weight percentages (wt%) of Ce, Pr, Sr, Ti and Mg cations are 42.5%, 7.9%, 17.7%, 9.7% and 0.54%, respectively. In order to quantify the phase composition of fresh and spent membranes, inductively coupled plasma optical emission spectrometry (ICP-OES) analysis was conducted to monitor the chemical composition of the membrane. Table S2 shows the actual elemental compositions of fresh and spent membranes.