Upgrading morphological properties of ceramic-carbonate coating for high-temperature CO 2 capture

1 Introduction

Over the past decades, human activities have intensified significantly, resulting in a corresponding increase in fossil fuel consumption. This rise has led to a significant growth in carbon dioxide emissions, widely recognized as the primary driver of global warming. To address this concern, several strategies have been developed to mitigate CO₂ emissions, including pre-combustion capture, oxy-fuel combustion, and post-combustion capture (Seyyedattar and Zendehboudi, 2026). Post-combustion capture is the most practical for integration with existing power plants. Several post-combustion technologies can be applied; key methods include absorption, adsorption, hydrate-based techniques, and membrane-based separation. Within them, membrane technology provides benefits in energy efficiency, modularity, and environmental performance. In contrast to absorption or adsorption technologies, membranes separate CO₂ without requiring regeneration stages, thereby simplifying the process and reducing environmental impact (Park et al., 2025).

High-temperature CO₂-selective membranes show strong potential for both pre- and post-combustion applications. Among the strategies explored, ceramic-carbonate dual-phase membranes have emerged as a promising technology for CO₂ separation, particularly in temperature regimes where other membrane materials or processes exhibit significant limitations (Fu et al., 2024). These membranes consist of a molten carbonate phase infiltrated into the pores of a porous support, resulting in a dense dual-phase membrane. CO₂ permeation through these materials occurs via an ionic mechanism that involves the following steps: i) CO₂ in the gas phase combines with O^2- or with O₂ and e⁻ to form the carbonate ionic species CO₃ ^2-; ii) the CO₃ ^2- diffuses through the molten carbonate phase toward the opposite side of the membrane as a result of the CO₂ partial-pressure gradient; iii) the permeating ionic species CO₃ ^2- is converted back into CO₂ and O^2-, or into CO₂, O₂, and e⁻, on the permeate side (the downstream stream). Throughout the permeation process, the support and the molten carbonate phase provide pathways for the transport of e⁻ and/or O^2-, as well as for carbonate ions, respectively. The support materials may be selected from electron-conducting metals, oxide-ion-conducting ceramics, or mixed oxygen-electron-conducting ceramics. Since the initial demonstration of a dense dual-phase inorganic membrane capable of separating CO₂ at intermediate and high temperatures (Chung et al., 2005; Anderson and Lin, 2010; Zhuang et al., 2020), the field has advanced significantly. Multiple dual-phase systems have been studied, with particular attention given to solid oxide electrolytes such as yttrium-doped zirconia (YSZ) (Wade et al., 2011), gadolinium-doped ceria (CGO) (Wade et al., 2011), and samarium-doped ceria (SDC) (Ovalle-Encinia and Lin, 2022). These materials have shown both good membrane stability and relatively high CO₂ permeation rates.

Experimental studies suggest that the CO₂ flux through ceramic-carbonate membranes can be enhanced in two main ways: by reducing membrane thickness or by increasing the oxygen-ion conductivity of the ceramic phase. The transport of CO₂ across dual-phase membranes is primarily controlled by oxygen-ion conduction; therefore, substantially reducing the thickness of the dual-phase membrane is a practical approach to achieve higher CO₂ flux. In this way, ongoing studies focus on reducing the thickness of composite membranes and enhancing the interaction between the carbonate and ceramic phases to further optimize material performance by modifying their microstructure (Lu and Lin, 2013; Dong et al., 2013; Chen et al., 2020; Tsochataridou et al., 2025; Shen et al., 2025). In this line, an approach proposed is to implement asymmetric support with a graded ceramic layer (Dong et al., 2013). Additionally, sacrificial templates have been used to prepare dual-phase membranes with a high degree of interconnected porosity (Ortiz-Landeros et al., 2013), improving CO₂ separation performance. One practical approach to tuning the porosity in a ceramic coating is to include pore-forming agents. In this way, YSZ porous coatings have been successfully prepared using pore-forming agents such as polymethyl methacrylate (PMMA) and starch as templates, thereby increasing porosity and enabling precise control over pore size and distribution (Albano et al., 2009; Liu et al., 2025).

The objective of the present study was to optimize the dip-coating procedure to obtain thin YSZ coatings on stainless steel substrates, which were subsequently infiltrated with a Na/Li/K carbonate mixture. The molten carbonate was confined to the YSZ layer by vacuum-assisted infiltration. Two pore-forming agents, PMMA and starch, were evaluated for their ability to modify the porosity of ceramic coatings. This work reports the physicochemical properties of dual-phase ceramic-carbonate samples and preliminary CO₂ perm-selective performance of dual-phase membranes under actual operating conditions.

2 Experimental

2.1 Synthesis of the ceramic-carbonate dual-phase membranes

2.1.1 Deposition of the ceramic support

Porous and non-porous 316L stainless steel discs (1.27 cm diameter) were used as substrates for the development of ceramic-carbonate dual-phase membranes. To reduce the ceramic layer thickness and, consequently, the overall thickness of the dual-phase membrane, this work proposes using porous SS discs as substrates for depositing the dual-phase membrane. Figure 1 shows a schematic of the proposed membrane configuration. Non-porous 316L stainless steel substrates were initially used to optimize YSZ coating deposition. After establishing the experimental parameters required to obtain a homogeneous coating, these conditions were subsequently applied to deposit YSZ onto porous substrates, thereby producing composite dual-phase membranes for CO₂ permeation experiments. The ceramic phase was deposited by dip-coating using a colloidal suspension of yttria-stabilized zirconia and polyvinyl alcohol (PVA). For deposition onto porous SS substrates, vacuum-assisted dip coating was used.

To develop a differential porosity suitable for infiltrating carbonate suspensions into the ceramic matrix, the ceramic phase was modified using a YSZ-PVA suspension with commercial starch or polymethylmethacrylate (PMMA). The supports were subjected to sequential deposition-calcination cycles; each cycle consisted of five YSZ immersions (five minutes each), followed by drying at room temperature. A detailed description of the suspensions and the procedure is provided in our previous publication (Dalla Fontana et al., 2021). After each cycle, calcination was performed at 723 K for 12 h (heating rate: 1 °C min^-1).

2.1.2 Preparation and infiltration of the carbonate phase

Li₂CO₃, Na₂CO₃, and K₂CO₃ salts were separately subjected to thermal treatment. Such treatment for the Li₂CO₃ and Na₂CO₃ salts included heating up to 673 K (3° min^-1) and holding for 24 h at that temperature. For the K₂CO₃ salt, the treatment was performed at 473 K for 24 h. Before preparing the carbonate suspension for infiltration, the salts were ground in a mortar to achieve a homogeneous particle size. The carbonate mixture was prepared by mixing powders of Li₂CO₃, Na₂CO₃, and K₂CO₃.

Infiltration was performed using a suspension of Li, Na, and K carbonates at the eutectic composition (43.5:31.5:25 mol%) in anhydrous ethanol (0.025 g/L). This carbonate mixture was selected as the ion-conducting phase due to its relatively high ionic conductivity and the lowest melting point among the carbonate mixtures (Mutch et al., 2019). Carbonate infiltration was performed under vacuum. This technique enabled uniform, controlled infiltration of the carbonate phase into the porous YSZ support. Subsequently, the carbonates melt in situ at 873 K, and their permeation properties are then evaluated.

Table 1 presents the primary characteristics of the samples studied. Note that two different numbers of depositions were evaluated, and, in addition, two types of substrates were employed, porous (PSS) and non-porous (NPSS) 316 L stainless steel discs. The samples prepared on non-porous stainless-steel substrates were designated according to the nomenclature NPSS-YSZ + x, where NPSS denotes the non-porous stainless-steel substrate, YSZ refers to the deposited yttria-stabilized zirconia layer, and x identifies the pore-forming agent used during preparation. On the other hand, the samples prepared on porous stainless-steel substrates were labeled as PSS-YSZ-y, where y indicates the number of deposition cycles.

TABLE 1. Main properties of the samples studied.

Dual-phase membrane prepared for CO₂ evaluation.

Samples prepared by applying ultrasound during the deposition.

Total weight gain after final calcination.

2.2 Samples characterization

2.2.1 X-ray diffraction (XRD)

Diffractograms of the modified substrates and molten carbonate membranes were obtained using an Empyrean PANalytical diffractometer with Cu Kα radiation (λ = 1.542 Å), operated at 40 kV and 45 mA, with a scan speed of 1° min-1. The analyzed range was 2θ = 10° to 90°.

2.2.2 Scanning electron microscopy (SEM-EDS)

Top surface images of the YSZ-modified supports, the molten carbonate membranes, and the bulk atomic composition of the samples were obtained by SEM-EDS using a Carl Zeiss Sigma microscope, equipped with an EDS detector (OXFORD-AZTEC XMAX 80).

2.2.3 Confocal microscopy (CM)

Roughness and pore size were determined with an OLYMPUS LEXT OLS4000 confocal microscope. This confocal microscope uses a 405 nm laser. The laser wavelength defines its spatial resolution in the (x,y) plane, and its resolution in the z-axis is 10 nm.

2.2.4 Dynamic light scattering (DLS)

The particle size distribution of the commercial YSZ stock suspension (YSZ-Nyacol) and that prepared in our laboratory from this suspension with PVA (YSZ-PVA) were determined using DLS. The determination was also performed for the ZrO₂-PVA suspension. A Zetasizer Nano ZS (Malvern, United Kingdom) equipped with a light source (λ = 532 nm) and a scattering angle of 90° was used. The isoelectric point of the suspensions was determined using the same equipment but with a different cuvette.

2.2.5 Preliminary permeation experiments

CO₂:N₂ gas permeation measurements were performed using a device developed by our research group (Tarditi et al., 2011). All gases were fed to the permeator via calibrated mass-flow controllers, and a back-pressure regulator controlled the pressure differences across the membranes. The effective permeation area of the membranes was 0.5 cm². The temperature during the experiments was controlled using a type K thermocouple connected to a temperature controller. CO₂ permeation measurements were performed at 873 K and a pressure difference (ΔP) of 50 kPa, using a CO₂/N2 mixture with a 1:1 ratio as the feed. Ar was used as a sweeping gas on the permeate side during the permeation experiments. Two bubble flow meters, calibrated at room temperature and atmospheric pressure, were used to measure the permeation flow rates. The feed, retentate, and permeate streams were analyzed online using a Shimadzu GC-8A equipped with a Porapak column and a thermal conductivity detector (TCD).

3 Results and discussion

3.1 Optimization of the YSZ deposition

To achieve a thinner ceramic-carbonate film, this work proposes using porous stainless-steel disks as substrates for depositing a dual-phase ceramic-carbonate selective film. By employing these porous substrates, the resulting film is expected to be thinner and more uniform, thereby enhancing the functional properties required for its intended application. To optimize the coating of the ceramic support surface for subsequent carbonate-phase infiltration, several samples were prepared. The first synthesized samples consisted of porous stainless-steel discs coated with YSZ after 1 and 3 deposition-calcination cycles, designated PSS-YSZ-1 and PSS-YSZ-3, respectively. Figures 2A,B show the morphology of the PSS-YSZ-3 sample after coating, followed by calcination. Note that after three cycles, homogeneous surface coverage of the stainless-steel substrate was observed. Using this coating procedure, a uniform, continuous, and smooth film was observed by SEM, with a thickness of approximately 15 μm, showing neither uncovered zones nor detachment after calcination at 673 K. On the other hand, larger pores are observed in the top view of the PSS-YSZ-1 sample (Figures 2C,D), suggesting that, after one deposition cycle, the coverage of the porous support was inhomogeneous. Although the YSZ coverage of the substrate is homogeneously distributed across the surface, the presence of pores up to approximately 10 µm in size prevents a homogeneous, high level of contact between the carbonate and the ceramic phase.

Table 2 summarizes the bulk atomic composition of the PSS-YSZ-1 and PSS-YSZ-3 samples, as determined by EDS, indicating homogeneous elemental distribution in both samples. A higher concentration of Zr and Y was observed for PSS-YSZ-3, confirming the increase in YSZ coverage after three deposition-calcination cycles. This trend is also supported by the weight-gain data reported in Table 1, which was normalized by the geometric area of the discs to estimate the coating mass loading. The calculated loading increased from 0.002 g cm^-2 for PSS-YSZ-1 to 0.009 g cm^-2 for PSS-YSZ-3, in agreement with the higher Zr and Y contents detected by EDS and with the more complete and compact surface coverage observed by SEM.

TABLE 2. Volumetric composition of the sample determined by EDS.

3.1.1 Effect of the addition of a pore-forming agent

Achieving intimate contact between the ceramic and the molten carbonate phases is a key step in developing this type of membrane, as it is essential for enhancing CO₂ permeation (Fu et al., 2024). A variation of the dip-coating protocol was investigated to improve contact between the carbonate and the ceramic phase (YSZ). Note that in this stage, non-porous stainless steel substrates were used. The strategy included adding a pore-forming agent (e.g., PMMA or starch) to the YSZ-PVA colloidal suspension to modify the microstructure of the ceramic phase. These agents were selected for their ease of handling and, in principle, low cost. On the other hand, they do not require extreme conditions to be eliminated. The particle size distributions of the original PMMA (in latex), the prepared YSZ + PMMA-PVA, and YSZ-PVA suspensions were determined by DLS; the data are summarized in Figure 3.

The suspension of PMMA nanospheres in latex (blue line, Figure 3) exhibited two peaks: a broad peak with an average size of ca. 4,801 nm and a particle-size distribution ranging from 1719 to 5,560 nm, and a second peak with an average size close to 59 nm and a distribution between 8 and 825 nm. For clarity, Figure 3 shows only the region up to 1,600 nm, allowing a better visual comparison with the other suspensions. On the other hand, a single peak with an average particle size of approximately 24 nm and a particle-size distribution ranging from 18 to 295 nm was observed for the YSZ + PMMA-PVA suspension (green curve). When the DLS distribution for the starch-containing sample was analyzed, two peaks were observed: a narrower peak with a maximum at 112 nm and a broader distribution corresponding to larger sizes (up to 500 nm, red line). For comparison, the DLS data of the YSZ-PVA suspension is presented. For this sample, two peaks were observed: a broad peak with an average particle size of approximately 396 nm and a narrow peak with an average size of about 59 nm (red curve). As inferred from the data, the YSZ + PMMA-PVA suspension exhibits the narrowest particle-size distribution. Ruiping and Wang (2013) proposed that by decreasing the diameter of PMMA micro-balls, the mean pore size and thermal conductivity of porous YSZ ceramics decreased. Additionally, the flexural strength of the support with the same porosity initially increases and then decreases.

The thermal behavior of the PMMA (in latex) suspension and the starch samples was studied by TGA and DTG; the obtained data are summarized in Figure 1S. In the TGA profile, a significant mass loss is observed between 523 K and 673 K, corresponding to two peaks in the DTG: a more intense peak at 615 K and a well-defined shoulder at 552 K. It has been reported that the thermal degradation of PMMA proceeds in three steps, with the least stable step probably initiated by H–H bond cleavage (Arisawa and Brill, 1997). Secondly, a step around 543 K was perhaps initiated by scissions at unsaturated ends. Finally, random scission within the polymer chain initiates the last stage of PMMA decomposition (ca. 623 K). On the other hand, for the starch sample, two peaks located at 353 K and 589 K were evidenced in the derivative profile (Supplementary Figure S1b). Based on these results, the samples were calcined at 773 K to ensure complete decomposition of the pore-forming agent.

Figure 4 compares the morphology of the NPSS-YSZ + STARCH and NPSS-YSZ + PMMA samples deposited by dip-coating with two deposition cycles. For comparison, an image of the NPSS-YSZ sample is also presented (Figure 4, top view).

As shown, incorporating a pore-forming agent into the YSZ suspension affects the coating morphology. The sample synthesized using starch as a pore-forming agent (NPSS-YSZ + STARCH) exhibited defects and cracks (indicated by red circles). In contrast, the surface of the NPSS-YSZ + PMMA sample exhibited regions resembling burst bubbles (yellow circles). During deposition, some PMMA agglomerates may settle and, during calcination, generate cavitation. The high-magnification images of the samples clearly show a significant change in the coating microstructure upon the addition of starch and PMMA. The addition of starch or PMMA produces a more granular coating, with PMMA exhibiting a more pronounced effect.

To improve deposition and prevent the formation of these features, samples NPSS-YSZ + STARCH_wu and NPSS-YSZ + PMMA_wu, were prepared with ultrasound applied during deposition, (Figure 5). It is important to note that, in this sample, cavitation was homogeneously distributed across the surface, with no evidence of collapsed bubbles, unlike the behavior observed in the absence of ultrasound (Figure 4). Nevertheless, a more compact top layer was observed, which appeared to be detached in the regions associated with these cavitation features.

Comparing the NPSS-YSZ + STARCH_wu and NPSS-YSZ + PMMA_wu samples, it is possible to observe a rougher, more granular microstructure in the sample prepared with PMMA. Ruiz-Morales et al. (2006) reported the synthesis of a material based on YSZ and PVA, incorporating PMMA (Ruiz-Morales et al., 2006). The goal of this work was to improve the microstructural properties of novel materials for the synthesis of SOFT cells. The authors reported a homogeneous pore distribution obtained by using PMMA spheres as templates. They also claimed that there was no evidence of collapse or cracks in the entire sample, mainly due to two factors. On the one hand, the use of PVA improves particle cohesion at the temperature at which PMMA microspheres are removed. On the other hand, the second factor concerns the walls of the microstructure (formed by YSZ grains). Additionally, it was reported that the simplicity of the process enables the use of various deposition methods and different substrates. The differences between the coatings obtained using the two pore-forming additives are evident, with PMMA yielding a more favorable morphology.

To enable a more thorough comparison of surface roughness between the NPSS-YSZ_PMMA and NPSS-YSZ + starch samples, confocal laser microscopy was used to analyze them. Figure 6 summarizes the 3D topography of the NPSS-YSZ + PMMA (A, B, and C) and NPSS-YSZ + starch (D, E, and F) samples, respectively. For each sample, three surface areas were analyzed to evaluate the homogeneity of the coating distribution. For the NPSS-YSZ + PMMA sample, a higher peak to valley depth ranging from 24 to 29 µm was observed at the surface, while on the NPSS-YSZ + starch sample, the maximum depth was between 15 and 20 µm. Although confocal microscopy does not provide a complete pore-size distribution or an open porosity value, the larger Sq, Sa, Sz, and Sv values measured for the PMMA-containing coating indicate more developed surface relief and deeper surface depressions than those obtained with starch. These features suggest a greater availability of surface-accessible cavities for carbonate during the subsequent infiltration step.

Table 3 summarizes the root mean square height (Sq), arithmetic mean height (Sa), maximum height (Sz), maximum pit depth (Sv), and maximum peak height (Sp) at three different areas for each sample. The average values of 2.811 µm and 1.963 µm were determined for the NPSS-YSZ_PMMA and NPSS-YSZ + starch samples, respectively.

TABLE 3. Parameters of the surface roughness analysis performed with confocal microscopy.

Sq (root mean square height) is determined by taking the square root of the average of all square deviations in surface height from the mean plane. It serves as a quantitative measure of the overall surface roughness.

Sa (arithmetic mean height) indicates the average difference between the surface heights and the mean plane, providing a general overview of the surface roughness.

Sz (maximum height) represents the total vertical distance between the highest peak and the lowest valley on the surface, providing insight into the full extent of surface height variation.

Sv (maximum pit depth): This parameter specifies the depth of the deepest valley found on the surface.

Sp (maximum peak height) indicates the height of the tallest peak present on the surface.

Figure 7 shows the diffraction patterns of the NPSS-YSZ + starch, NPSS-YSZ + PMMA samples, and an NPSS substrate after calcination at 773 K. Note that, in addition to the characteristic reflections of the oxidized substrate, associated with the austenitic phase of the stainless steel (JCPDS Number 33–0397), new peaks in the 2θ range between 26° and 39° were also observed. In this region, reflections associated with tetragonal ZrO₂ (PDF: 00421164) and zirconium yttrium oxide (PDF: 010772112) are located. For comparison, the diffraction pattern of a YSZ powder sample is shown. This sample was obtained from the colloidal suspension used after evaporation and calcination. These data are consistent with those reported by Shi et al. (1999), who observed the formation of a tetragonal phase between 673 K and 973 K in an yttria-stabilized zirconia membrane.

3.2 Molten carbonate composite membrane

Following the analysis of YSZ deposition using pore-forming agents, the next step involved the fabrication of a PSS-YSZ-carbonate membrane using PMMA. The membrane was prepared by dip-coating the ceramic-carbonate layer onto a porous stainless-steel substrate to evaluate the development of the composite membrane architecture. The ceramic coating procedure consisted of an initial YSZ deposition cycle without a pore-forming agent, followed by two deposition cycles with PMMA added to the YSZ-PVA suspension. The normalized loading of the PSS–YSZ–carbonate membrane was 0.008 g cm^-2, close to that of PSS–YSZ–3, suggesting good reproducibility of the optimized three-cycle coating protocol. The slightly lower loading in the PMMA-containing membrane may be associated with the higher porosity generated in the ceramic phase after PMMA removal during calcination, since no pore-forming agent was used for PSS–YSZ–3.

Figure 8 presents the microstructural characterization of the PSS-YSZ-carbonate sample both before (a, b) and after (c, d) the deposition of the carbonate mixture, followed by infiltration at high temperature (873 K). The Na/Li/K carbonate mixture was introduced into the porous ceramic phase via vacuum-assisted dip-coating. Carbonates were dispersed in anhydrous ethanol, and the disk bearing the YSZ ceramic phase was immersed in the suspension under vacuum during deposition. Before dispersion, the carbonates were ground in a mortar to achieve a uniform particle size. It was observed that the grinding process performed before preparing the carbonate suspension did not affect the final particle size (see Figure 2S in the Supplementary Material). The infiltration of the Na/Li/K molten carbonate phase with eutectic composition (Li₂CO₃:Na₂CO₃:K₂CO₃ = 43.5:31.5:25.0 mol%) was promoted in situ by heating the membrane to 823 K in a nitrogen atmosphere.

A distinct morphology was observed in the YSZ-modified support before and after carbonate deposition. Note that after carbonate infiltration, the membrane exhibited a smoother surface. This approach facilitated uniform, well-controlled infiltration of the carbonate phase into the porous YSZ layer. EDS analysis across several regions of the sample enabled observation of the atomic composition. The alkaline carbonates exhibited relatively good dispersion and efficient pore filling, as evidenced by the homogeneous distribution of Na and K detected by EDS (Li was undetectable by this method).

Figure 9 displays the elemental mapping obtained by SEM-EDS for a region of the PSS-YSZ-carbonate sample. A uniform distribution of the elements constituting the molten carbonate phase is clearly visible. Furthermore, the elements Zr and Y, components of the ceramic phase, can be identified due to the analytical depth. The values represent the average bulk atomic composition for a selected area of the PSS-YSZ-carbonate sample.

The crystal structure of the PSS-YSZ-carbonate membrane was analyzed by XRD (Figure 10). For comparison, the XRD pattern of a PSS-YSZ sample was included. After infiltration and carbonate melting, the PSS-YSZ-carbonate membrane exhibited reflections corresponding to the carbonate phase. On the other hand, due to the depth of the analysis, signals from the stainless steel are also observed. The peaks at 2θ = 43.6° and 50.8° correspond to the γ-phase of the stainless-steel substrate (JCPDS 33-0397). Small peaks at 2θ = 21.8°–27.1° are consistent with the most intense XRD peaks of (Li/Na)CO3 (hexagonal, JCPDS 34-0933).

3.3 High-temperature CO₂ permeation properties: preliminary results

CO₂ permeation was evaluated using the PSS–YSZ–carbonate membrane prepared with PMMA as a pore-forming agent in the ceramic layer. To check membrane sealing, nitrogen permeation through the membrane was measured before and after infiltration. Once the membrane was placed in the permeator, nitrogen permeation was determined by feeding a pure N₂ stream at a flow rate of 50 mL min^-1. After treatment at 873 K for 1 h under an inert atmosphere, no nitrogen permeation was detected across the membrane under a pressure difference of 50 kPa, indicating that a defect-free membrane was formed via carbonate infiltration into the ceramic phase. Although small cracks were observed on the membrane surface after infiltration (Figures 8C,D), nitrogen was not detected in the permeate. This result suggests that these defects may be confined to the near-surface region of the membrane and do not extend through, or form connected pathways across, the full thickness of the dual-phase layer. Although a numerical CO₂/N₂ selectivity value was not reported, no detectable N₂ permeation was observed either before or after the CO₂ permeation measurements under the experimental conditions used. This result indicates that the membrane was effectively gas-tight to N₂, while CO₂ transport occurred selectively through the carbonate-mediated mechanism. Therefore, the CO₂/N₂ selectivity can be considered very high, or at least higher than the detection limit.

CO₂ permeation through the PSS–YSZ–carbonate membrane was evaluated at 873 K under two driving forces: first, by applying a sweep gas (Ar) while keeping both the feed and permeate streams at atmospheric pressure; second, by increasing the pressure on the feed side to create a pressure gradient across the membrane (total ΔP 50 kPa). The CO₂ permeation flux through the membrane under these conditions was 0.08 and 0.11 mL min^-1 cm^-2 at a feeding pressure of 101.3 and 151.8 kPa, respectively (Figure 11). This increase in permeation flux with increasing feed-side pressure is consistent with data reported in the literature. Studies of a samarium-doped ceria membrane infiltrated with a lithium-sodium carbonate mixture showed that the CO₂ flux through the membrane increases with increasing CO₂ partial pressure (Ruiz-Morales et al., 2006).

Carbonate phases may be susceptible to volatilization, dewetting, or corrosion of stainless-steel components during prolonged high-temperature operation; therefore, long-term membrane stability should be carefully considered. Nevertheless, in the studied configuration, the porous stainless-steel substrate provides mechanical support for the dual-phase membrane without direct contact with the carbonate phase. This is enabled by the ceramic phase with suitable porosity, which helps confine the carbonate phase and reduces both carbonate loss and potential detrimental effects on membrane permeability. Although only short permeation runs were evaluated in this study, extended durability tests are required to confirm membrane performance under realistic operating conditions. These aspects, including long-term operation and start-up/shut-down cycling tests, will be addressed in detail in a forthcoming study.

CO₂ permeation flux and permeabilities are compared with data reported in the literature in Table 4. Using ceramic-carbonate hollow fiber membranes, Zuo et al. (2014) reported CO₂ fluxes of 0.061 and 0.13 mL min⁻¹ cm⁻² at 923 K and 1,123 K, respectively. These data are in the same order as those reported by Wade et al. (2011) for a disc membrane (0.047 and 0.17 mL min⁻¹ cm⁻² at 923 K and 1,123 K, respectively). Note that the CO₂ permeation flux of the PSS–YSZ–carbonate membrane is higher than the data reported in the literature at the same temperature. Although the reported data are preliminary and were obtained from a single membrane, they suggest that the prepared PSS–YSZ–carbonate membrane was gas-tight under the tested conditions and exhibited appreciable CO₂ permeation at high temperature, with flux values within the range reported for related ceramic–carbonate dual-phase membranes. Nevertheless, because cross-sectional SEM analysis was not performed, this conclusion should be regarded as an indirect inference based on gas-tightness measurements rather than as direct microstructural confirmation.

TABLE 4. Comparison of CO₂ permeation flux with some published data.

Sweep gas without ΔP.

ΔP, 50 kPa.

4 Conclusion

In this work, a YSZ ceramic layer was successfully deposited by dip coating onto a stainless-steel substrate and subsequently infiltrated with a eutectic Na/Li/K carbonate mixture, resulting in a uniform and continuous dual-phase coverage that remained stable after high-temperature treatments. The incorporation of pore-forming agents improved the porosity of the ceramic layer and promoted a more effective interaction between the ceramic and carbonate phases. Among the additives tested, PMMA produced a rougher, more granular coating than starch, as shown by SEM and higher roughness/height values from confocal microscopy. This morphology was deemed more suitable for carbonate infiltration and was therefore selected for membrane preparation. A quantitative link between pore-size distribution, permeability, and infiltration efficiency will still require additional porosity/permeability measurements. Preliminary measurements of permeation flux in the presence of an equimolar CO₂:N₂ mixture yielded a CO₂ permeation flux of 0.11 mL min^-1 cm^-2 at 873 K. These preliminary permeation results suggest that the proposed membrane architecture is promising for high-temperature CO₂ separation; however, further experiments using replicate membranes and additional driving-force conditions are required to confirm reproducibility and establish robust performance trends.

In memoriam

In memory of Agustina Dalla Fontana, I would like to express my deepest gratitude. Her dedication to research and her passion for exploring new materials were truly inspiring to all who had the privilege of working with her. She is greatly missed, both as an exceptional scientist and as a dear friend.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frmst.2026.1871627/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Arash Mollahosseini, National Institute for Nanotechnology (NRC-CNRC), Canada

Reviewed by: Najwa Jubier, Wasit University College of Science, Iraq

Dimitar Panayotov, Institute of General and Inorganic Chemistry, BAS, Bulgaria