Fluid dynamic modeling of oxygen permeation through mixed ionic–electronic conducting membranes

https://doi.org/10.1016/j.memsci.2011.05.016Get rights and content

Abstract

The oxygen transport in a lab-scale experimental set-up for permeation testing of oxygen transport membranes has been modeled using computational fluid dynamics using Finite Element Analysis. The modeling considered gas hydrodynamics and oxygen diffusion in the gas phase and vacancy diffusion of oxygen in a perovskite disc-shaped membrane at 1273 K. In a first step, the model allowed obtaining the coefficient diffusion of oxygen. The parametric study showed that the set-up geometry and flow rate in the air compartment did not have major influence in the oxygen transport. However, very important polarization effects in the sweep-gas (argon) compartment were identified. The highest oxygen permeation flux and the lowest oxygen concentration on the membrane surface were obtained for the following conditions (in increasing order of importance): (1) a large gas inlet radius; (2) short gas inlet distance; and (3) a high gas flow rate.

Highlights

Computational fluid dynamics allowed obtaining the coefficient diffusion of oxygen. ► Concentration polarization effect in permeate compartment. ► Sweep-gas flow rate and inlet distance to membrane are crucial.

Introduction

High-purity oxygen production through ceramic membranes at high temperature [1] is an interesting cost-effective alternative to the cryogenic methods for several industrial applications. Membrane separation will make possible to reduce energy requirements and investment costs. Among the different applications, the use in fossil fuel power plants in the so-called Oxyfuel process [2], [3] would allow minimizing CO2 emissions since the final flue gas stream consists principally of moist CO2, which can be readily liquefied and transported. In this case, the ceramic oxygen transport membrane modules can be thermally integrated and make possible to match the desired oxygen production due to the intrinsic modularity of these systems. Apart from usual applications of oxygen in several industries, e.g. steel industry or petrochemistry, oxygen transport membranes are able to achieve high grade purity (theoretical selectivity of 100% for defect-free membranes) extending its possible applications. Additionally, another important field of application of this kind of membranes is their implementation in high-temperature catalytic membrane reactors while typically reactions are synthesis gas production from methane [4], [5], oxidative coupling of methane to yield ethane and ethylene [6], [7], [8], pocket selective ammonia oxidation [9], among others. The high selectivity can be achieved because oxygen separation is based on the transport of oxygen-ion vacancies through the lattice of a crystalline mixed oxide material. The most usual oxygen-ion conducting materials are based on the perovskite structure (ABO3) or related structures [5], [10] and comprise Fe/Co/Ni and mixtures of lanthanide and alkali-earth metals in suitable proportions.

Fundamental research on this topic makes use of bench-scale testing units, where the oxygen permeation through a small disc-shaped membrane sample is measured as a function of different operation variables such as temperature, oxygen partial pressure at the inlet of each chamber, and inlet gas flow rate in each chamber. A widely used permeation set-up uses a disc membrane (15–25 in diameter and 0.2–2 mm in thickness) with the geometry shown in Fig. 1. Nevertheless, the determination of the true oxygen flux through the material is not always straightforward. Indeed, there are several lurking experimental variables that introduce some noise/deviation to the permeation results. This is especially patent when comparing experimental data from different groups, which apparently employ similar permeation units. Among the different possible causes for this, it can be highlighted (i) temperature gradients in the membrane; (ii) oxygen partial pressure gradients on the membrane surface (at each membrane side); and (iii) gas transport limitations from/toward the membrane surface (polarization) which results in a reduction of the net driving force (gradient of oxygen chemical potential). The last aspect can be particularly critical for thin-supported membranes on porous substrates [11]. The importance of these effects associated to the fluid dynamics in the testing unit depends on the chosen experimental set-up configuration and operating conditions.

Ghidossi et al. [12] made a review of computational dynamics applied to membranes, emphasizing the possible contribution of computational fluid dynamics in the development of new membrane processes. Some other studies have been devoted to the validation and the application of computational simulators to reliably predict the fluid dynamic and the separation performances in inorganic membranes modules for gas separations. Koukou et al. [13] demonstrated the validity of two-dimensional mathematical model to predict the influence of non-ideal flow effects on membrane separator performance. Furthermore, Takaba and Nakao [14] tested a CFD simulator to evaluate the influence of the concentration polarization on the membrane performance. They model a bi-channel and a tubular geometry when treating a H2/CO mixture. These results showed that the CFD simulation is capable of evaluating the concentration polarization effect in a membrane module for gas separation and they concluded that the CFD simulation can be used to design a membrane module involving prediction of selectivity and cut. More recently, Coroneo et al. [15] used a CFD simulation without introducing any simplified hypothesis on the velocity field and the concentration distribution of the species. They obtained good agreement between the experimental and the predicted data and concluded that CFD is a useful and reliable tool for design of new membrane modules.

Siegel [16] made a thorough review of the application of computational modeling of heat and mass transfer in polymer electrolyte membrane fuel cells where it is shown an overview of models in literature according to their dimensionality. For three-dimensional models, the modeling efforts of the fluid dynamics of the gas phase inside the cells have conducted to a more realistic approach of the cell performance. Moreover, this review evaluates the CFD available software. In particular, the potential of COMSOL Multiphysics is highlighted as an efficient tool for modeling complex systems for which fluid dynamics and mass and heat transfer are relevant. Sousa et al. [17] used COMSOL to obtain the concentration profile of the components in direct ethanol fuel cells. COMSOL and other software [18], [19] have also been used to model solid oxide fuel cells.

Oxygen diffusion has been also modeled using computational techniques in diverse industrial processes like oxyfuel combustion [20], [21] or fuel rods [22]. The oxygen diffusion in ceramic membrane modules has been studied by our group [23], [24] and by others researchers [2], [25]. Recently, [26] the diffusion in porous mixed ionic–electronic layers, e.g. solid oxide fuel cell cathodes, has been modeled successfully using COMSOL, and CFD combined with impedance spectroscopy made it possible the determination of intrinsic materials properties as the diffusion coefficient.

In this work, it is intended to understand the effect of several experimental factors on the measured permeation rate by means of computational fluid dynamics modeling using COMSOL Multiphysics®. The aims of this contribution are (i) to assist the researchers in the selection of the most suitable setup configuration and operating conditions and/or to prevent the use of inadequate operating conditions/set-up geometries; and (ii) to calculate the intrinsic material properties by considering the real fluid dynamics in both membrane chambers.

The strategy followed in this work is the next. Firstly, flux experimental results were obtained for an experimental rig of specified geometry. This geometry is taken as a reference case. Secondly, a model of the system using COMSOL Multiphysics was developed taking into account the different involved phenomena. Thirdly, a diffusion coefficient of oxygen vacancies was obtained as an intrinsic material parameter by an iterative method to match flux experimental results to those given by the model for the reference case. Finally, using the vacancy diffusion coefficient obtained, the effects of changes in the inlet geometry with respect to the reference case (inlet radius and distance to the membrane) and inlet gas flow rate were studied for both permeate and feed chamber.

The manuscript is organized accordingly to the mentioned procedure:

  • -

    In the experimental and methods section, the experimental set-up is described. The modeling using COMSOL Multiphysics and the procedure of determination of the vacancy diffusion coefficient for a reference case are also shown.

  • -

    The results and discussion section is organized in the following parts: study of the reference geometry with determination of the diffusion coefficient of oxygen vacancies, parametric study of the geometry of the inlet of the feed compartment, parametric study of the geometry of the sweep-gas inlet of the permeate compartment, and parametric study of the entering flows of gas in both compartments.

Section snippets

Experimental set-up and conditions

A detailed scheme of the experimental set-up is shown in Fig. 1b. A membrane disc is placed between the feed and permeate compartments. The membrane disc has a diameter of 15 mm and thickness of 0.75 mm. The membrane material was an oxygen deficient perovskite (La1−ySryFeO3−δ with y = 0.1). Due to the necessary seal, the area exposed of the membrane is 1.039 cm2. Dry air enters into the feed chamber from an inlet placed in the axis of the module. The flow enters perpendicularly to the membrane and

Results and discussion

In this section, three magnitudes (flow streamlines, oxygen concentration and oxygen flow rate) has been analyzed in the different set-up volumes in order to determine the influence of the different geometric and operational factors on fluid dynamics and therefore on the final permeation. Moreover, two figures of merit, i.e., average oxygen flux through the ceramic membrane and average oxygen concentration on the corresponding membrane surface, have been selected to show the goodness and

Conclusions

The oxygen transport in a lab-scale experimental rig for characterization of ceramic membranes has been modeled taking into account gas hydrodynamics and oxygen diffusion in the gas phase and vacancy diffusion of oxygen in a perovskite membrane. The rig geometry and feed flow rate had a significant effect on the oxygen transport through the membrane. The model was useful to obtain the coefficient diffusion of oxygen vacancies which is a material property. The parametric study showed that the

Acknowledgements

The Spanish Ministry for Science and Innovation (JAE-Pre 08-0058 grant and ENE2008-06302 project) and through FP7 NASA-OTM Project (NMP3-SL-2009-228701) is kindly acknowledged.

References (31)

Cited by (35)

  • Design and performance of asymmetric supported membranes for oxygen and hydrogen separation

    2021, International Journal of Hydrogen Energy
    Citation Excerpt :

    Electronic transport depiction is omitted here for brevity. Modern ceramic oxygen separation membranes are built using complex Sr-doped perovskites (Sr-doped lanthanum ferrites-nickelates (LSFN), nickelates-cobaltites (LSNC) etc.) since they have a high oxygen diffusivity (DO ~ 10−7 cm2/s at 700 °C) and mixed ionic-electronic conductivity (σtotal ~ 102–103 S/cm at 700 °C), which provide a high permeability and, thus, fine characteristics in catalytic reactions [8–10,22,33,38–42]. Oxygen separation membranes based on Sr-doped La ferrite [43], Sr-doped Ba ferrite-cobaltite [42,44] are used also.

  • Characterization of oxygen transport phenomena on BSCF membranes assisted by fluid dynamic simulations including surface exchange

    2020, Chemical Engineering Journal
    Citation Excerpt :

    The equations of the multidimensional model (Table 1) were implemented in the software COMSOL Multiphysics v4.4. As the axial geometry permits using 2D computational domains, then the model complexity can be greatly reduced from a computational point of view if only radial and z-components of the equations are considered [47]. Therefore, three interconnected domains were considered: permeate, membrane, and feed compartments.

  • Towards improved partial oxidation product yield in mixed ionic-electronic membrane reactors using CSTR and CFD modelling

    2019, Chemical Engineering Science
    Citation Excerpt :

    Methane oxidation in the sweep stream of a MIEC membrane reactor has been modelled using a range of chemical kinetic mechanisms: a three-step mechanism for POX (Jin et al., 2016), a four-step mechanism for OCM (Tung and Lobban, 1992; Wang and Lin, 1995), and the 325 reaction GRI MECH mechanism (Hong et al., 2012, 2013; Mancini and Mitsos, 2011a,b), where the latter has only been applied to simplified geometries. Computational fluid dynamic (CFD) analysis of a 2-D parallel flow reactor (Habib et al., 2013), a typical button cell geometry (Nemitallah et al., 2015; Gozálvez-Zafrilla et al., 2011), as well as a modified geometry (Nemitallah, 2016) have been performed, but considered either non-reacting or simplified kinetic mechanisms (e.g., 1–5 steps). Using a reactor network analysis based on a 1-D model, an optimal reactor network configuration and operating parameters were numerically identified (Godini et al., 2012).

View all citing articles on Scopus
View full text