Introduction

Oxygen-transporting membranes (OTMs) based on mixed ionic-electronic conductors (MIECs) have attracted considerable attention in promising applications, such as in the production of oxygen-enriched air [1], in the selective oxidation of methane and ethane [2], as cathode material in solid oxide fuel cells [3], and also in rechargeable lithium-air batteries [4]. Currently, perovskite-type MIECs containing Ba2+ and/or Sr2+ cations exhibit immense oxygen permeability due to the high amount of oxygen vacancies [5,6,7]. Unfortunately, perovskite materials such as Ba0.5Sr0.5Co0.8Fe0.2O3−δ (BSCF) or doped SrCo0.8Fe0.2O3−δ are vulnerable when operated in CO2-containing atmosphere because alkaline earth carbonate layers can form on the surface of the membranes, blocking the oxygen flux and thus restrict their potential uses [8,9,10]. In contrast, La2NiO4+δ (LNO) possesses a high long-term chemical stability in CO2 atmosphere [11, 12]. However, the oxygen permeability of the polycrystalline LNO ceramic is approximately 4–6 times lower than that of the BSCF perovskite, which can be explained by the lack of exploitation of its anisotropic oxygen transport properties [11, 13, 14].

LNO belongs to the Ruddlesden-Popper (RP) phase series with the general formula Lan+1NinO3n+1 (n = 1 − ∞) and represents its first member (n = 1). This RP oxide features a K2NiF4 structure composed of alternating cubic perovskite LaNiO3 layers and LaO rock-salt layers arranged along its c-axis [15,16,17]. Oxygen hyperstoichiometry can be observed in this material due to the incorporation of interstitial oxygen ions into the rock-salt layers, which are involved in the oxygen transport mechanism [18, 19]. The additional oxygen content taken up in LNO is described by the oxygen excess δ, varying in value between 0 and 0.3 [20]. Depending on δ and the temperature, LNO adopts either an orthorhombic or a tetragonal structure [16, 20,21,22,23,24,25]. The oxygen diffusion through LNO is highly anisotropic, primarily occurring in the rock-salt layers along the a,b-plane via a two-dimensional (2D) interstitialcy migration mechanism. This is reflected in the higher values of the oxygen bulk diffusion coefficient D* and oxygen surface exchange coefficient k* compared to those in the c-axis for LNO single crystals and oriented thin films. Oxygen diffusion along the c-axis is also feasible but to a lesser extent and results from the migration of oxygen vacancies through the perovskite layers [26,27,28,29,30,31,32].

To comprehend and improve the oxygen transport properties of LNO ceramic membranes with polycrystalline character, various factors have been extensively studied. These include the membrane thickness [33], grain size [11, 33, 34], doping, and/or substitution of the lanthanum or nickel site [11, 35,36,37,38,39,40,41,42,43] as well as different sintering techniques [34]. Nevertheless, in polycrystalline membranes with randomly oriented grains, the anisotropic properties of LNO are not exploited. The recent development of LNO-based single crystals [27, 28, 32] or epitaxial thin films [30, 44] has opened up the possibility of utilizing the anisotropic nature of this material for controlling oxygen transport along the a,b-plane or c-axis. Similarly, the anisotropic properties of LNO can also be employed in polycrystalline ceramic membranes by texturing the material to precisely regulate the oxygen ion migration. Well-established methods for the preparation of highly textured polycrystalline ceramics represent the templated grain growth (TGG) or the magnetic orientation process [45,46,47,48,49,50,51,52]. In the TGG, commonly used for the production of textured polycrystalline piezoelectric ceramics [46,47,48,49,50,51], large anisotropic template particles (TPs), e.g., with a plate-like morphology, are initially mixed with small, fine-grained equiaxed matrix particles (MPs), solvents, binder, and plasticizer to form a slurry. Since the TPs are randomly oriented in the slurry, they are typically aligned by tape casting before heat treatment. Here, the orientation of the template particles is induced by shear forces. During sintering, the TPs grow by the consumption of the MPs, leading to the formation of a textured polycrystalline ceramic, which can exhibit comparable properties to those of the corresponding single crystal [45, 46, 53]. Another approach to manufacturing textured polycrystalline membranes, which eliminates the need for a slurry, requires only a minimal quantity of TPs, and is straightforward to implement, involves uniaxial pressing. Due to the anisotropic shape of the TPs, they can be properly aligned in the matrix by the pressing force, resulting in a textured membrane after sintering. Regarding LNO, both particle types can be produced using established synthetic methods. Anisotropic LNO TPs with plate-like crystal shape are obtained via molten-flux synthesis (MFS) [54], while LNO MPs are prepared by a sol-gel route [55] or industrially through soft chemistry approaches.

In this work, uniaxial pressing was used to align micrometer-sized plate-like LNO particles obtained from a NaOH melt via MFS in the LNO matrix. Following the sintering process, ceramic membranes textured along the c-axis of LNO were manufactured. The degree of texturing was thereby increased by adjusting the fraction of TPs. Several characterization methods were applied to analyze the microstructure of the textured ceramics. To conclude, the impact of the template particle content on oxygen transport in the textured LNO on oxygen transport was assessed.

Experimental

Starting materials

Two distinct powder sources were used in the fabrication of textured LNO membranes. The LNO matrix powder was purchased from Marion Technologies (Verniolle, France), which was obtained by a soft chemistry technique. In contrast, the LNO TPs were synthesized chemically using the molten-flux method. For this purpose, an ultrafine powder consisting of La2O2CO3 and NiO from a sol-gel process was mixed with NaOH beads (Alfa Aesar, 99.9%) and distilled water in a nickel crucible. The mixture was then heated at 673 K for 10 h. After purification of the crude product, plate-shaped LNO crystals were acquired. A detailed description of the MFS can be found in a previous publication of our research group [54].

Preparation of textured La2NiO4+δ ceramic membranes

The LNO TPs were blended with the well-ground powder of LNO MPs in weight ratios of 1:99, 5:95, 10:90, 20:80, and 40:60. The powder mixtures were then placed in a 16 mm die and uniaxially pressed for 15 min at 150 MPa. This resulted in green bodies, which were pressurelessly sintered at 1673 K for 15 h in ambient air, with a heating and cooling rate of 2 K/min, to produce ceramic membranes with a diameter and thickness of about 14 mm and 1.2 mm, respectively. Two reference samples containing only LNO TPs or LNO MPs were also prepared using the same procedure as the ceramics composed of both types of particles.

Microstructural characterization of the materials

Phase identification of the powders and the ceramic membranes was performed by X-ray diffraction (XRD, Bruker AXS GmbH, Bruker D8 Advance) using monochromatic CuKα radiation at 40 kV and 40 mA. XRD data were collected in the 2θ range from 10° to 50° with a step size of 0.01° and a duration of 1s per step. The following powder diffraction files (PDFs) were used to evaluate the recorded XRD patterns: La2NiO4.18 (PDF: [01-089-3589], tetragonal, a = 3.866 Å, c = 12.678 Å), and LaNiO3 (PDF: [00-033-0710], cubic, a = 3.861 Å). Moreover, Eq. (1) is applied to determine the Lotgering orientation factor f from the measured XRD patterns, which provides information about the degree of (00l) texture in the LNO ceramic membranes [56, 57].

$$f_{{(00}l)}=\frac{P_{{(00}l)}-P_0}{1-P_0}$$
(1)

The orientation factor f (Eq. (1)) is given by Eqs. (2) and (3) as follows:

$$P_{{(00}\mathrm l)}=\frac{\sum I_{{(00}l)}}{\sum I_{(hkl)}}$$
(2)
$$P_0=\frac{\sum I_{{(00}l)}^{\,0}}{\sum I_{(hkl)}^{\,0}}$$
(3)

where P(00l) is the ratio of the sum of the relative intensities of all (00l) reflections (= \(\sum I_{{(00}l)}\)) to the sum of the relative intensities of all (hkl) reflections (= \(\sum{\textit{I}}_{{(hk}l)}\)) in the sintered textured LNO membrane. P0 is defined as the proportion of the summation of the relative intensities of all the (00l) reflections (= \(\sum I_{{(00}l)}^{\,0}\)) to the summation of the relative intensities of all the (hkl) reflections (= \(\sum I_{(hkl)}^{\,0}\)) for a non-textured LNO sample. The f factor was calculated for the 2θ range between 10–50°.

To further investigate the texturing in selected green bodies and sintered membranes, pole figures of the (101), (004), and (110) lattice planes (polar angle θ = 0–70°, in 5° steps, azimuthal angle φ = 0–360°, in 5° steps, integrating β-measurement 10 s/5°) were recorded using a 5-circle X-ray diffractometer (3003 ETA, Seifert Analytical X-Ray) with CoKα radiation, an X-ray tube power of 30 kV and 40 mA, a 2 mm point collimator, and a silicon drift detector (AXAS M, Ketek GmbH). The analysis of the pole figure data was carried out with the LaboTex software from LaboSoft.

The morphology of the samples was studied with a field-emission scanning electron microscope (FE-SEM, JEOL JSM-6700F) using a secondary electron detector at 2 kV. Crystal and grain sizes were determined from the recorded SEM micrographs using ImageJ [58], with over 100 particles or grains evaluated. To better understand the microstructure of the sintered disk membranes, fractured and vibration-polished cross-sections were prepared for SEM by breaking off or cutting a small slice from the sintered samples. For the polished cross-section, the slice was embedded in epoxy resin and then subjected to a multiple polishing program with diamond lapping films followed by vibratory polishing. The elemental composition of the specimens was examined by an energy-dispersive X-ray spectrometer (EDXS, Oxford Instruments INCA-300) with an ultrathin window at 15 kV placed on the SEM. The density of the sintered disks was estimated using the Archimedes method (ISO 5018:1983) with propan-2-ol as the fluid.

Analysis of the oxygen permeation properties

Oxygen permeation tests were carried out between 1023 and 1223 K using a homemade high-temperature permeation cell, as previously described in detail elsewhere [59, 60]. Prior to the measurements, both sides of the disks were polished to a thickness of 1 mm with a 120-grit SiC sandpaper. The membranes were then sealed onto an alumina tube using ceramic paste (Huitian 2767). Synthetic air (20 vol.% O2 and 80 vol.% N2) was used as feed gas at a flow rate of 150 mL/min and He as the sweep gas at a flow rate of 29 mL/min. An online-coupled Agilent 7890A gas chromatograph equipped with a Carboxen® 1000 column was employed to analyze the gas mixture. The absolute flow rate was determined using Ne as an internal standard with a flow rate of 1 mL/min in the sweep side. The relative leakage of O2 due to insufficient sealing was calculated by measuring the N2 concentration and subtracting it from the total O2 permeation flux. Assuming that the leakage of O2 and N2 follows the Knudsen diffusion, the fluxes of leaked N2 (= \(J_{{\mathrm{N}}_{2}}^{\mathrm{leak}}\)) and O2 (= \(J_{{\mathrm{O}}_{2}}^{\mathrm{leak}}\)) are related by Eq. (4):

$$J_{{\mathrm{N}}_{2}}^{\mathrm{leak}}: J_{{\mathrm{O}}_{2}}^{\mathrm{leak}}= \sqrt{\frac{32}{{28}}} \times \frac{{80\, \%}}{{20 \,\%}}{ = 4.28}$$
(4)

where 32 and 28 are the molar masses of N2 and O2, respectively, and 80% und 20% represent the fraction of N2 and O2 in the synthetic air, respectively. By considering Eq. (4), the oxygen permeation flux J(O2) through the ceramic membranes can be calculated using the following Eq. (5):

$$J(\mathrm{O}_{2})=\left[{{c}}(\mathrm{O}_{2}) \, - \, \frac{c(\mathrm{N}_{2})}{4.28}\right]\times\frac{F}{S}$$
(5)

Here, c(O2) and c(N2) are the measured oxygen and nitrogen concentration on the sweep side, F is the total flow rate of the sweep stream, and S is the effective membrane area [61, 62].

Results and discussion

Microstructural characterization of the starting materials

Two types of LNO powders, differing in crystal size and morphology, were utilized to obtain textured ceramic membranes. Commercial fine-sized LNO powder served as MPs, while large plate-like LNO crystals from MFS were used as TPs. XRD, SEM, and EDXS were applied to analyze both starting materials. The XRD patterns in Fig. 1(a, b) display body-centered tetragonal LNO with a K2NiF4 structure in the I4/mmm space group as the main phase in both powders. The tetragonal phase was expected since the synthesis of both components took place in air and at high temperatures [20, 63, 64]. In the case of LNO TPs (Fig. 1(b)), cubic LaNiO3 was also registered as a minor impurity. Its formation is attributed to the chosen reaction parameters, as discussed in a previous work [54]. The matrix consists of nanometer-sized equiaxed particles (~ 90 nm) that form agglomerates (Fig. 1(c)). In contrast, the LNO TPs are characterized by a well-defined plate-like shape and a size in the low micrometer range (Fig. 1(d)). The average length and width of the crystals measure about 10 µm and 4 µm, respectively. In a prior study [65], selected-area electron diffraction patterns were utilized to prove that the (001) facets dominate compared to the (100) and (010) facets of a plate-like LNO crystal from MFS. A schematic view of an LNO plate with relevant crystallographic information is depicted in Fig. 1(e). Crystal growth along the crystallographic c-axis is strongly impaired by the high surface energy of the (001) facets and steric constraints [65]. Therefore, the plate-like particles preferentially grow along the a- and b-axes. The EDXS analysis of the LNO MPs (Fig. S1) shows a homogeneous distribution of La, Ni, and O in the powder. The La/Ni ratio here is almost 2:1. Furthermore, a small amount of Si was found (0.6 at.%), which was evenly dispersed across the entire sample area. This impurity was likely introduced during the manufacturing process. As demonstrated elsewhere [34], the Si contamination does not have a noticeable effect on the properties of the ceramic materials due to its minor amount. The EDXS elemental maps and spectrum in Fig. S2 affirm that the LNO TPs are homogeneously composed of the elements La, Ni, and O, with a La/Ni ratio of approximately 2:1. Some Na (3.07 at.%) was found as an impurity originating from the NaOH flux, which may occupy the La site in LNO because of its similar ionic radius to lanthanum [54, 65]. Partial doping of LNO with Na has been observed and described in other work [66]. Sodium is a chemical element that easily evaporates at elevated temperatures [67]. As will be indicated later, sodium is no longer present in the LNO membranes, likely due to the high sintering temperature of 1673 K employed. While the small particle size of commercial LNO makes it a suitable matrix powder for embedding the LNO TPs, the plate-like nature of the micrometer-sized TPs facilitates their preferential alignment during uniaxial pressing, resulting in the formation of textured ceramic membranes after the sintering step.

Fig. 1
figure 1

a, b XRD patterns and c, d SEM micrographs of LNO MPs and LNO TPs, respectively; e schematic representation of a plate-like LNO TP with selected surfaces and coordinate system

Full size image

Microstructural characterization of textured La2NiO4+δ ceramic membranes

The powder mixtures were subsequently subjected to uniaxial pressing and sintering to form ceramic disk membranes. XRD patterns were recorded from the reference samples composed of only pristine TPs or MPs, as well as from the membranes consisting of both particle types. The XRD patterns presented in Fig. 2(a) exhibit the body-centered tetragonal LNO structure (space group: I4/mmm) for all specimens. LaNiO3, the by-product of the MFS, was no longer observed (see Fig. 1(b)). This can be interpreted by the fact that the perovskite-type LaNiO3 was converted to LNO due to the high sintering temperature used [68, 69]. In the XRD patterns and the magnified cut-out of the 2θ interval from 42° to 44° in Fig. 2(b), it is particularly noticeable that the (00l) reflections become more intense with an increasing proportion of LNO TPs in the ceramic materials. For the samples with 40 wt.% TPs or 100 wt.% TPs, the intensities of the (004) reflections even exceed those of the (103) reflection, which is typically the main peak for a randomly oriented LNO specimen, such as the reference membrane formed from pure MPs. This implies that during pressing, the randomly oriented plate-shaped LNO crystals in the matrix were predominantly aligned perpendicular to the pressing direction due to their anisotropic morphology (Fig. 2(c)). As a result, all (00l) reflections in the XRD patterns gain more intensity with ascending TP content.

Fig. 2
figure 2

a, b XRD patterns of textured LNO ceramic membranes containing varying amounts of TPs; c the illustration depicts the preferred orientation of the plate-like LNO TPs in the specimens, as determined by XRD

Full size image

By using the XRD patterns in Fig. 2(a) and Eq. (1), the Lotgering orientation factor f was calculated for all membranes to determine the degree of texturing along the c-axis. All (00l) reflections in the 2θ range between 10 and 50° were employed for this purpose. The Lotgering factor f assumes values between 0 and 100%, whereby the latter value represents perfect texturing [70]. The correlation between the Lotgering factor f and the density of the samples as a function of the LNO TPs content in the ceramics is depicted in Fig. 3. The plot indicates that as the mass percentage of the TPs increases in the ceramic materials, the (00l) texturing becomes higher, but concurrently, the density of the samples decreases. For example, in the case of the LNO membrane consisting of 5 wt.% TPs, f is 14% and the density amounts to 91.4% of the theoretical value (theoretical bulk density of LNO according to PDF [01-089-3589]: 7.01 g/cm3). For a ceramic containing only pure LNO TPs, f reaches the maximum value of 55%, but at the same time, it has the lowest density (= 75.4% of the theoretical value). Its f value is moderate but not particularly high, indicating a still-existing misorientation of the LNO plates in the membrane. This statement is reinforced by the corresponding XRD pattern in Fig. 2, which shows additional intense reflections such as (103) or (110) that are not assigned to the crystallographic c-axis of LNO. The highest density is achieved with the membrane consisting only of pure MPs (= 98.4% of the theoretical density) and has no texturing. The fact that the sample density reduces with a growing proportion of TPs in the textured LNO membranes is later supported by the SEM micrographs.

Fig. 3
figure 3

Relationship between the Lotgering orientation factor f and the density of textured LNO membranes as a function of LNO TP content in the samples (circles: Lotgering factor f values, squares: membrane density values)

Full size image

To further quantify the level of (00l) texturing, XRD pole figures were recorded from the (101), (004), and (110) planes of a pressed LNO green body and a textured LNO ceramic membrane. The multiples of a random distribution (MRD) into the c-axis of the materials were calculated based on the estimated orientation density function data using the LaboTex software to determine the strength of texturing. The MRD value can assume values between 0 (= no texturing) and infinite (= single crystal): The higher the MRD, the stronger the texturing in the specimen [71, 72]. The pole figures in Fig. 4 of an LNO sample consisting of 5 wt.% TPs demonstrate pronounced (00l) texturing. In the green body, the MRD is already 4.53. Subsequent conventional sintering further enhances the texturing effect along the c-axis of LNO (MRD = 8.85). In contrast, the (101) and (110) pole figures of the green body and the sintered membrane display significantly weaker intensities compared to that of the (004) reflection (MRD = 1.92–3.48). Using XRD pole figures, it was proved that texturing with the preferential orientation of LNO TPs in the ceramic membrane along the c-axis, was already achieved by uniaxial pressing. The plate-like crystals were predominantly aligned perpendicular to the pressing direction, as indicated on the right side of Fig. 4. The texturing was subsequently reinforced by the sintering process.

Fig. 4
figure 4

XRD pole figures of the (101), (004), and (110) reflections with intensity scale of the MRD in arbitrary units (a.u.) for the LNO-based material consisting of 5 wt.% TPs in both pressed and sintered form

Full size image

To examine the microstructure of the LNO membranes, SEM micrographs were acquired of both the surface and the fractured cross-section, and in some cases, of the vibration-polished cross-section. The SEM analysis of the reference membrane, which does not contain LNO TPs, can be appreciated in Fig. 5. The grains are clearly visible in all scanning electron micrographs. Furthermore, no open pores or cracks were detected on the surface or in the cross-section, consistent with the high calculated relative density value of 98.4% (see Fig. 3). Assuming the grains possess an approximate round geometry, the average grain diameter was estimated to be 4.1 µm. The membrane comprised entirely of MPs was also characterized by EDXS. The results are depicted in Fig. S3. La, Ni, and O are evenly distributed both on the surface and in the vibration-polished cross-section. Notably, the Si contamination originating from the initial powder was absent in the sintered membrane (refer to Fig. S1). Possibly, the Si content is now so negligible that it can no longer be detected by EDXS, suggesting its removal during the high-temperature sintering step.

Fig. 5
figure 5

SEM micrographs of the LNO ceramic membrane composed solely of MPs: a surface; b fractured; and c, d vibration-polished cross-section

Full size image

The results of the SEM microstructure study for a textured sample are given in Figs. 6 and S4, focusing on the LNO membrane with 5 wt.% TPs. Plate-like grains with dimensions of around 20 µm × 7 µm can be recognized on the membrane surface (Figs. 6(a) and S4(b, c)), which are thoroughly embedded in the matrix with a grain size of about 5 µm. During the sintering process, there was significant growth of LNO TPs in the ceramic, as the original particles were 10 µm × 4 µm in size. In the fractured and vibration-polished membrane cross-section (Figs. 6(b) and S4(d−f)), rectangular areas partially surrounded by the matrix were observed. The presence of small closed pores can also be noted, which would explain the lower relative density of 91.4% compared to the reference membrane without TPs (see Figs. 3 and 5). The porosity might have been caused by careless crushing of the sample. On the membrane surface, the LNO TPs stand out due to their well-defined plate-like morphology, as they are aligned with their c-axis parallel to the pressing direction (Figs. 6(i) and S4(a)). By contrast, in the cross-section of the specimen, the width and thickness of a plate-shaped grain can be easily identified (Fig. 6(j)). Additionally, EDXS analysis was performed on the textured LNO membrane. The SEM images of the surface and the crushed cross-section from Fig. 6(a) and (b) were used for this investigation. Both the membrane surface and the cross-section are homogeneously composed of the three elements La, Ni, and O (Fig. 6(c−h)). Impurities such as Si or Na from the starting materials were not identified, likely due to their minor quantities. Therefore, it can be assumed that both elements have been eliminated from the membrane as the result of the high sintering temperature employed. The La/Ni ratio in the sample is almost 2:1, which agrees with the stoichiometry of LNO.

Fig. 6
figure 6

SEM-EDXS characterization of the textured LNO membrane containing 5 wt.% TPs: a, b selected SEM micrographs for EDXS analysis; c–h EDXS elemental maps of the corresponding elements, i, j schematic representation of the surface and cross-section of a textured LNO ceramic membrane

Full size image

SEM images of the membranes that consist of 1 wt.% TPs, 10 wt.% TPs, and 20 wt.% TPs are provided in Fig. S5. Plate-like grains are distinctly visible on the surface of the ceramics. Furthermore, the SEM micrographs of the sample cross-sections indicate an increase in porosity with the proportion of TPs in the membranes. In the LNO ceramic membrane made of 20 wt.% TPs, micrometer-sized pores are present on both the surface and the cross-section of the material. These findings are in accordance with the calculated densities of the samples, showing that the higher the content of plate-like TPs in the LNO membrane, the lower the density (Fig. 3). Figure S6 displays the SEM characterization of the reference membrane composed solely of pure LNO TPs. Predominantly, grains with plate-like shape are discernible, accompanied by continuous pores between them. The grains have an average size of 13 µm × 5 µm. Thus, moderate particle growth has taken place, although not as pronounced as in the case of membranes containing both MPs and TPs. Compaction of the membrane by sintering has not yet fully occurred, as the LNO TPs are too large and partially misoriented, preventing effective contact between them. This leads to the formation of large pores in the green body after pressing, which cannot be removed during the sintering step. Consequently, the membrane with 100 wt. % TPs exhibits the lowest density among all the samples (Fig. 3). Similar behavior has been identified in the development of textured piezoelectric materials using TGG. In general, it is difficult to compact TPs with an anisotropic structure into a ceramic material. As the portion of TPs increases, densification becomes more challenging [46, 73, 74]. Attempts to increase the membrane density by extending the sintering time or using a binder, as has been shown elsewhere, did not result in a noticeable increase in sample densification (data are not shown) [50, 75]. Similar to the TGG process, small equiaxial matrix particles seem to play a crucial role in our approach described in this work to promote membrane densification. These particles suppress the contact of TPs by filling the gaps between them, thereby enhancing membrane consolidation. Besides, the consumption of MPs during sintering promotes the growth of TPs, which contributes to the development of texturing [47].

Oxygen permeation measurements

In order to investigate the impact of (00l) texturing on the oxygen permeation properties of the sintered LNO membranes, oxygen permeation measurements were conducted in the temperature range of 1023–1223 K. To this task, textured polycrystalline LNO membranes with a thickness of 1 mm, which corresponds approximately to the characteristic membrane thickness Lc of LNO (Lc ≈ 1 mm), were employed [33]. Since gas-tight membranes with a minimum relative density of 90% are required for effective oxygen separation from air [76,77,78], the permeation experiments were performed on textured LNO membranes of 0 wt.% TPs, 1 wt.% TPs, 5 wt.% TPs, and 10 wt.% TPs.

The principle of such a measurement on a textured sample is schematically explained in Fig. 7(a). While synthetic air was introduced to the feed side of the permeator, He and Ne were supplied as internal standard gas to the sweep side of the permeator. The gas mixtures were then analyzed using a gas chromatograph to determine the effective oxygen transport flux, considering O2 leakage. The oxygen permeation results in Fig. 7(b) show the typical course of such a measurement on an LNO ceramic membrane: The oxygen permeation flux grows with rising temperature [11, 33, 34, 79]. Consequently, the highest permeation fluxes for all membranes are achieved at 1223 K. Additionally, it is evident that the oxygen permeation fluxes decrease with increasing proportion of LNO TPs, which are aligned along their c-axis parallel to the permeation direction in the disk membranes. The reduction in oxygen permeability is more pronounced between 1123–1223 K than between 1023–1073 K. Therefore, the highest permeation performance was observed when using the reference membrane without TPs. As stated in the literature, oxygen transport through LNO is strongly anisotropic, occurring primarily in the a,b-plane of the material. In contrast, oxygen diffusion along the c-axis is unfavorable and thus marginal due to the fully occupied oxygen sites in the perovskite-like LaNiO3 layers [26,27,28,29,30,31,32]. The probable mechanism of oxygen transport along this direction can be described as follows: An interstitial O2− ion moves along the c-axis and displaces an apical O2− ion from the NiO6 octahedron, which subsequently moves downward to a neighboring interstitial site. This process repeats in the next unit cell [80]. Hence, texturing a polycrystalline LNO membrane along its c-axis by increasing the LNO TP content in the samples causes a drop in oxygen permeation flux. The more pronounced the c-direction texturing, the lower the oxygen permeation fluxes become. Using an Arrhenius plot, the activation energies Ea of oxygen permeation were determined for the textured LNO membranes (Fig. 7(c)). For this purpose, the measurement points were fitted by a linear equation, and the value of Ea was obtained from the slope. It was found that the membrane without TPs, exhibits the highest activation energy. As the proportion of LNO TPs in the membranes becomes higher, the value of Ea diminishes.

Fig. 7
figure 7

a Schematic representation of the principle of oxygen permeation measurement; b oxygen permeation fluxes of selected textured LNO membranes consisting of 0−10 wt.% TPs in synthetic air/He gradient and within the temperature range of 1023–1223 K. Measurement conditions: synthetic air (20 vol.% O2 and 80 vol.% N2) as the feed gas (flow rate = 150 mL/min), He as the sweep gas (flow rate = 29 mL/min), and Ne as the internal standard gas (flow rate = 1 mL/min); membrane thickness: 1 mm; c corresponding Arrhenius plots for determining activation energies

Full size image

In Table 1, the activation energies of all LNO membranes manufactured in this work are summarized alongside activation energy data from the literature for oxygen bulk diffusion D* and oxygen surface exchange k* of other LNO materials. Comparison between all values is limited because the measurement conditions for determining Ea are distinct for each LNO sample. The Ea values estimated for the textured LNO membranes in this work can be attributed to both oxygen bulk diffusion and surface oxygen exchange since the membrane thickness roughly corresponds to the Lc of LNO (Lc ≈ 1 mm) [33]. In a previous report, it was calculated that for an LNO single crystal, oxygen diffusion along the c-axis has a lower activation energy (≈ 0.25 eV) than along the a,b-plane (≈ 0.9 eV) [27]. Since the LNO membrane without TPs is a polycrystalline membrane containing grains randomly oriented in the a-, b-, and c- directions, the Ea value is higher than the c-axis textured samples owing to the larger fraction of a- and b-oriented grains. The lower activation energy of oxygen diffusion for the (00l) textured membranes compared to the polycrystalline sample (membrane without TPs) is likely caused by the presence of O− ions. These ions exist in lower concentrations compared to O2− ions, leading to a lower D* along the c-axis. Due to the small size and low charge of the O− ions, their mobility is high, resulting in a reduced diffusion activation energy along the c-direction [16, 27].

Table 1 Comparison of the activation energies Ea for oxygen diffusion of the LNO ceramic membranes prepared in this work with literature values
Full size table

Conclusions

Textured polycrystalline La2NiO4+δ membranes were produced by mixing template particles with matrix particles of the oxide in different mass ratios. Through uniaxial pressing, the template particles in the matrix were preferentially oriented along their c-axis parallel to the pressing direction. Subsequent sintering allowed a further increase in the texturing degree. The higher the amount of the La2NiO4+δ template particles in the ceramic materials, the more accentuated the (00l) texturing. Membranes with a low proportion of template particles exhibit plate-like grains that are well integrated into the matrix. However, with an increasing fraction of template particles, the membrane’s porosity became higher. Texturing La2NiO4+δ ceramics with the crystallographic c-axis preferentially parallel to the oxygen flow direction resulted in a lower oxygen permeation flux and thus lower values of the activation energy for surface exchange and bulk diffusion of oxygen. This is based on the anisotropic properties of La2NiO4+δ. Oxygen diffusion along the c-axis is unfavorable and therefore low due to the fully occupied oxygen sites in the perovskite layers. Because oxygen transport in La2NiO4+δ occurs predominately in the a,b-plane, texturing along this direction may result in improved oxygen permeation compared to a polycrystalline membrane with randomly oriented or c-axis oriented grains. Ceramics with a high loading of template particles, which are beneficial for texturing, and a high density, as required for the membrane application in oxygen separation, may be obtained by using advanced sintering techniques such as hot pressing [82, 83], spark plasma sintering [34, 84, 85] or spark plasma texturing [85,86,87]. These sintering techniques could enable the production of larger sintered bodies, from which rectangular membranes with texturing along the a,b-plane of lanthanum nickelate can be obtained by cutting, thereby improving the oxygen transport properties. The findings from these investigations will be presented in a follow-up study.