Advancing Green Hydrogen Purity with Iron-Based Self-Cleaning Oxygen Carriers in Chemical Looping Hydrogen

Abstract

Green hydrogen is central to the energy transition, but its production often requires expensive materials and poses environmental risks due to the perfluorinated substances used in electrolysis. This study introduces a transformative approach to green hydrogen production via chemical looping, utilizing an iron-based oxygen carrier with yttrium-stabilized zirconium oxide (YSZ). A significant innovation is the replacement of Al₂O₃ with SiO₂ as an inert support pellet, enhancing process efficiency and reducing CO₂ contamination by minimizing carbon deposition by up to 700%. The major findings include achieving a remarkable hydrogen purity of 99.994% without the need for additional purification methods. The Fe-YSZ oxygen carrier possesses a significantly higher pore volume of 323 mm³/g and pore surface area of 18.3 m²/g, increasing the pore volume in the iron matrix by up to 50%, further improving efficiency. The catalytic system exhibits a unique self-cleaning effect, substantially reducing CO₂ contamination. Fe-YSZ-SiO₂ demonstrated CO₂ contamination levels below 100 ppm, which is particularly noteworthy. This research advances our understanding of chemical looping mechanisms and offers practical, sustainable solutions for green hydrogen production, highlighting the crucial synergy between support pellets and oxygen carriers. These findings underscore the potential of chemical looping hydrogen (CLH) technology for use in efficient and environmentally friendly hydrogen production, contributing to the transition to cleaner energy sources.

1. Introduction

Green hydrogen, which is crucial for our energy future, faces production challenges despite its extensive industrial use [1,2]. At present, 99% of the hydrogen that is essential to the production of fertilizers and semiconductors is obtained from fossil fuels [3]. Shifting to green hydrogen from renewables could cut CO₂ emissions by up to 620 Mt/year [4] and help to achieve the net-zero greenhouse gas emission goals [5]. Electrolysis, particularly using a Polymer Electrolyte Membrane (PEM), is often considered as a game-changer in producing green hydrogen to decarbonize the industry [6,7,8]. Its advantages include high-purity hydrogen production (99.999 vol%) suitable for fuel cells [9]. However, the technology faces challenges, such as the use of Pt/Ir catalysts, which have issues with stability, cost, and efficiency, and the high energy input required for water splitting [10,11]. A significant yet under-discussed issue is the use of perfluorinated and polyfluorinated substances (PFAS) in electrolysis membranes, known as “Forever Chemicals”, which pose risks of cancer, immune responses, and metabolic syndromes [12,13,14,15].

Hydrogen is not only an ideal carbon-free energy carrier with an extremely large energy density (about 142 MJ kg⁻¹) used to replace fossil fuels, but also a critical raw material for hydrogenation, petroleum refining, and fertilizer production. More importantly, in this electrochemical water-splitting route for hydrogen production, H₂ is produced from H₂O without CO₂ emission, which is very green, environment-friendly, sustainable, and highly aligned with the global goal of carbon neutrality [16]. Despite these challenges, other cutting-edge technologies have to be developed to advance hydrogen production alternatives. Chemical looping hydrogen (CLH) is a promising technology that can be used to address these issues. CLH, an innovative approach for green hydrogen production and storage from renewable resources, aims to meet climate targets [17,18]. Utilizing cost-effective iron oxide oxygen carriers (OC) and gasifying biomass through CLH can enable us to overcome challenges related to hydrogen production costs, centralized systems, and long transport distances [19,20]. By using environmentally friendly materials, CLH presents a preferable alternative to fluorinated polymers and rare metals like platinum used in electrolysis.

Studies have highlighted the potential utility of novel materials and methods in enhancing the efficiency and sustainability of hydrogen production [21]. For instance, new catalysts and support materials have shown promising results in reducing CO₂ emissions and improving hydrogen purity [22]. Efforts in utilizing biomass for high-purity green hydrogen production via chemical looping have shown promising results [23]. A novel oxygen carrier material, based on the combination of Fe₂O₃ with a mixed ionic-electronic conductor (MIEC) like yttria stabilized zirconia (YSZ), has demonstrated remarkable performance [24,25,26]. In particular, the Fe₂O₃-YSZ oxygen carrier system has demonstrated exceptional hydrogen production and storage capacity, achieving 12 molH₂/kgOC and maintaining stability for 100 cycles [27]. However, the efficiency of this combination under real gas conditions with CO/H₂ is still untested. Biomass gasification faces challenges such as carbon deposition during the reduction step, potentially contaminating hydrogen based on the CO₂ formed in the oxidation mechanism [28,29,30]. This can be seen in the two side reactions 3a and 5a in the process cycle:

Reduction step:

3 Fe₂O₃ + H₂/CO → 2Fe₃O₄ + H₂O/CO₂

(1)

Fe₃O₄ + H₂/CO → 3FeO + H₂O/CO₂

(2)

3 FeO + 3H₂/CO → 3Fe + 3H₂O/CO₂

(3)

Reduction side reaction:

2 CO → C_(s) + CO₂

(3a)

Oxidation step:

3 Fe + 3H₂O → 3FeO + 3 H₂

(4)

3FeO + H₂O → Fe₃O₄ + H₂

(5)

Oxidation side reaction:

C_(s) + 2 H₂O → 2 H₂ + CO₂

(5a)

Therefore, the process requires additional purification steps or the use of energy-intensive vacuum pumping [31,32]. Large-scale tests using Fe₂O₃-Al₂O₃ as an oxygen carrier system have primarily focused on process parameters, neglecting in-depth investigations of the material’s microscopic properties and behavior [33,34]. Notably, the potential deactivation of iron by Al₂O₃ over extended cycles, forming FeAl₂O₄ [35,36], and its prevention using Fe₂O₃-YSZ, is an area that warrants further investigations. Additionally, we want to investigate the role of the inert support pellets like SiO₂ and Al₂O₃ within the reactor bed, which has not yet been thoroughly taken into account in the literature [32,37,38]. The SiO₂ and Al₂O₃ pellets are essential for the large-scale CLH application in preheating the gases in the fixed bed and keeping the oxygen carrier in the heating zone of the fixed bed (see Figure 1 and Figure S1) to ensure an effective reduction reaction [23,39]. Currently, it is assumed that hydrogen purity is only influenced by the oxygen carrier or the process parameter [28,40,41,42]. Unfortunately, this is often because tests are typically conducted on a milligram or gram scale, while larger scales and the use of support pellets, as found in real industrial applications, are overlooked. Therefore, the interplay between the oxygen carrier and support pellet has currently never been considered or examined in more detail. This represents an unexplored area that needs to be addressed to increase the overall efficiency of the process and provide alternative green hydrogen methods for energy production. Our research aims to analyze the impacts of these materials on hydrogen purity, focusing on both micro- and macroscopic phenomena. Specifically, we seek to understand the role of stabilization materials, such as YSZ or Al₂O₃, in influencing the behavior of the oxygen carrier by using a real syngas mixture CO/H₂. Additionally, we will also explore how the support pellets SiO₂ or Al₂O₃ affect carbon deposition during the chemical looping process, and if they have an influence on the process efficiency. This study seeks to provide a novel perspective on green hydrogen production at a larger scale in the context of chemical looping hydrogen.

This research article unveils a novel approach for generating ultra-pure green hydrogen, achieving a remarkable 99.994% purity while utilizing an eco-friendly iron-based oxygen carrier. It emphasizes a unique application of self-cleaning materials, ingeniously tackling the challenge of carbon deposits, and thereby reducing CO₂ contamination to below 100 ppm across multiple cycles. The study finds that SiO₂, in contrast to Al₂O₃ support pellets, facilitates carbon flake formation on the pellet’s exterior, enabling rapid oxidation during the oxidation phase. This leads to a swift CO₂ release while maintaining high hydrogen quality, a self-cleaning effect of sorts. A pivotal discovery is the synergistic interaction between support pellets and oxygen carriers, significantly influencing the hydrogen production process. The strategic integration of Fe-YSZ-SiO₂ as both oxygen carrier and support pellet emerges as a key innovation, which works by suppressing the formation of side compounds like FeAl₂O₄ observed in the Fe-Al₂O₃-SiO₂ system. The article goes on to propose comprehensive design strategies aimed at enhancing the production and storage of green energy carriers, spotlighting the advantages of a meticulously optimized oxygen carrier system, thus marking a significant leap in sustainable energy technology.

2. Results and Discussion

2.1. Green Hydrogen Generation in Fixed-Bed CLH

The influence of both microscopic and macroscopic phenomena on hydrogen production was investigated using the experimental setup illustrated with the used cycle sequence of reduction and oxidation (see Figure 1). This fixed-bed CLH system was designed to examine the interactions between various materials. Three fixed-bed material systems were tested, as described in the Materials and Methods section. The abbreviations Fe-YSZ and Fe-Al₂O₃ refer to the oxygen carrier mixtures, where the iron oxide is stabilized with the respective support material (YSZ or Al₂O₃). The third parts in Fe-YSZ-SiO₂, Fe-YSZ-Al₂O₃ and Fe-Al₂O₃-SiO₂ denote the support pellets (SiO₂ or Al₂O₃), which were used to stabilize the oxygen carrier in the heating zone (see Figure 1 and Figure S1a). In Figure 2 and Supplementary Figures S3–S8, the amounts of hydrogen produced in respective oxidation cycles, depending on the purity of the hydrogen, for all oxygen carrier systems are shown. Only CO₂ was detected as a contaminant. Notably, from the first cycle of oxidation, it was evident that the support pellets significantly affected the process efficiency (Figure 2). When the inert support pellet was switched from an Al₂O₃- to a SiO₂- support pellet, there was an increase in hydrogen purity from 99.98% to 99.994%, using an yttrium-stabilized zirconium oxide (YSZ)–iron oxygen carrier (OC) (Figure 2). Further studies involving the replacement of YSZ with Al₂O₃ as the stabilization material in the OC revealed its effect on hydrogen purity. Hydrogen with impurities below 100 ppm was achievable only when the combination of SiO₂ pellets with an OC composed of Fe-YSZ was used in the fixed-bed (Figure 2 and Supplementary Figures S3 and S4).

As the cyclization process progresses, experimental results indicate that the choice of support pellet significantly affects hydrogen purity for chemical looping (Supplementary Figures S3–S8). The combination of Fe-YSZ-SiO₂ achieves an impressive hydrogen purity of up to 99.994% throughout the cycles. In contrast, Fe-Al₂O₃-SiO₂ manages to produce a hydrogen purity of 99.98%. A notable change in hydrogen production purity is observed when switching the support pellet to Al₂O₃ in the Fe-YSZ-Al₂O₃ system. Gas chromatographic investigations suggest an enrichment of carbon in the Al₂O₃ pellets over numerous cycles. The total hydrogen production values over the number of cycles are comparable in all materials, but only Fe-Al₂O₃-SiO₂ shows a lack of production in the first four cycles (Supplementary Figure S2). Comparing the images of the support pellets after 30 cycles (Supplementary Figure S1c), it is evident that a different mechanism of carbon deposition occurs when changing the support pellets from Al₂O₃ to SiO₂. Carbon formation can occur as a result of the side reaction of CO with itself [43,44]. In a case of using SiO₂, the carbon forms flakes outside of the pellets, and for the Al₂O₃ pellets, it appears that carbon is deposited within the pellet structure (Supplementary Figure S1b,c).

These findings are particularly significant, as they have not been previously described in the literature. Moreover, they challenge the prevailing assumption that the oxygen carrier is the sole influencer of hydrogen purity. Consequently, these microscopic and macroscopic effects require further investigation and understanding in order to fully comprehend their impacts on hydrogen production processes.

2.2. Investigation of the Carbon Deposition on the Support Pellets

Our fixed-bed investigations have unequivocally demonstrated that the hydrogen purity of internal support pellets is significantly influenced by factors other than the oxygen carrier, contrary to the common assumption set out in the literature. Scanning Electron Microscopy (SEM) images reinforce this finding. The Al₂O₃ samples exhibit a substance deposited on the surfaces of the pellets, enveloping them in a darker coating indicative of an element with a lower atomic number (Figure 3(1a–4a), Supplementary Figure S11). Energy-dispersive X-ray (EDX) mapping reveals that this substance is carbon, preferentially deposited on Al₂O₃ with a weight percentage of 46.65 wt. % (Supplementary Figure S12 and Table S2). In contrast, the SiO₂ samples show a consistent surface structure remaining intact even after 30 cycles (Figure 3(5a–8a) and Supplementary Figure S13), with the formation of few carbon spots, which are visible on the surface (Supplementary Figure S13a). The EDX mapping of SiO₂ shows that carbon deposits are scarce, and the flat surfaces show a carbon concentration of just 5.28 wt.% (Supplementary Table S3 and Figure S14a). A closer investigation of the SEM/EDX analysis with a line scan shows that they are small carbon islands (Supplementary Figure S14b). A close EDX analysis of the individual surface’s area (Supplementary Figure S15a and Table S4) proves that the carbon spots consist of 82.92 wt.% carbon. But this is only the case in a few places, so almost no carbon is deposited on the SiO₂ surface.

Thermogravimetric analysis (TGA) further confirms these observations (Figure 3b and Supplementary Figure S10) of the SEM/EDX and GC investigations. Carbon at 0.13–0.20 wt. % was detected in the Al₂O₃ pellets and at 0.01–0.04 wt. % in the SiO₂ pellets (Supplementary Table S1) using water vapor as the oxidation agent at 800 °C, similar to in the reactor. This suggests that 700% more carbon is deposited in the Al₂O₃ pellet structure during the reduction process compared to SiO₂ pellets, explaining the disparity in hydrogen purity between Fe-YSZ-SiO₂ and Fe-YSZ-Al₂O₃ systems.

2.3. Investigation of the Pore Structure Network in the Support Pellets

Building on our investigations, we aimed to determine whether the carbon deposition within the pores was the sole factor affecting hydrogen purity, or if there were other contributing factors. To this end, X-ray fluorescence (XRF) analyses were conducted on the support pellets to assess any atomic composition changes resulting from high temperatures and the harsh conditions during the reduction and oxidation cycles at 800 °C. The XRF analysis revealed that both SiO₂ and Al₂O₃ materials underwent no significant alterations during cyclization, as evidenced in Supplementary Figures S26 and S27. IR analysis further supported this, indicating no formation of new bonds such as -OH in either Al₂O₃ or SiO₂ pellet samples (Supplementary Figure S9) [45,46,47]. X-ray diffraction (XRD) analysis also showed no changes in crystal properties or the formation of additional side compounds (Supplementary Figures S28 and S29). Interestingly, the SiO₂ samples exhibited an amorphous silica structure, while in the Al₂O₃ pellets, the SiO₂ showed a crystalline structure. Therefore, we conclude that the carbon deposition is attributable to a CO side reaction, leading to the formation of carbon without any carbide formation when using the SiO₂ or Al₂O₃ support pellets or other intermediate compounds.

For this reason, we investigated the pore structure to understand carbon deposition phenomena in the support pellet by mercury porosimetry. Supplementary Figures S20 and S22 show the volume of intrusion mercury (Hg) per g of the pellet as a function of pressure uncycled, after reduction (upper and lower pellets) and after oxidation in the TGA. This provides the opportunity to see what morphological changes occur in the pellet material during the oxidation and reduction cycle in the fixed-bed. The blue line shows the mercury that migrates into the pellet structure during the pressure increase, and the orange line shows the mercury that migrates out of the pores when the pressure is decreased. Based on the intrusion and extrusion curve, the mathematical pore model was used to calculate the characteristic pore parameter as a function of the distribution over the pore diameter for all samples, such as pore size distribution and the surface area (Figure 4, Supplementary Figures S21 and S23). The mercury porosimetry revealed that the pore structure remained unchanged with the Al₂O₃ support pellet, and the pore volume only slightly changed during the cyclization (Supplementary Figure S22a–c and Table S8). This suggests that the carbon forms a thin layer around the Al₂O₃ pellets. The increase from 571 mm³/g to 585 mm³/g (Supplementary Table S8) may occur because the additional carbon on the surface of the pellet may form fine new pores into which the mercury can penetrate. Despite this, mercury can still penetrate the small pores, and the shapes of the intrusion and extrusion curves remain unchanged. This observation is consistent with the previous SEM/EDX images.

The SiO₂ pellets show a completely different behavior after reduction compared to the upper pellets (Supplementary Figure S20). The Hg- curves show a drastic structural change after the 30th cycle after the reduction cycle and a huge increase in the total pore volume from 1106 mm³/g to 777 mm³/g (Supplementary Figure S20a–c and Table S7). The significant change in pore volume and surface, and the relative pore volume in the SiO₂ samples, indicate that the pores were clogged with carbon, preventing mercury infiltration (Figure 4(1a–1c) and Supplementary Figure S21), because after oxidation in the TGA, the shape (Supplementary Figure S20d) and the pore parameters were nearly the same as before, i.e., in the uncycled sample (Figure 4(1d) and Supplementary Table S7). Please note that we will prove this observation further in the text using µ-CT analysis and SEM/EDX analysis. In addition, the phenomenon only occurred in the upper pellets, which suggests that carbon deposition occurs here. Interestingly, after oxidation in the TGA, the Al₂O₃ volume decreased from 585 mm³/g to 559 mm³/g (Supplementary Table S8). The observed effects could be due to residual carbon within the Al₂O₃ pellets, which reduces the volume of mercury penetration, or to changes in the pellet structure caused by high-temperature processes, e.g., sintering of the pores. It is difficult to determine with Hg porosimetry whether these changes result from sintering under high temperatures and reductive gases, or from carbon deposition.

Therefore, to analyze the pore network more effectively, we employed a three-dimensional imaging method: X-ray microtomography (µ-CT). The pore network of the pellets was scanned with µ-CT, and afterwards, with a deep learning method, the segmentation analysis was performed. The pore size distribution of the network of Al₂O₃ and SiO₂ in the uncycled sample can be seen in the blue line, and that of the sample after 30 cycles is shown in the orange line (Figure 5a,b). The µ-CT analysis has confirmed that the Al₂O₃ did not undergo significant changes in internal morphology over the cyclization period, and the same is shown for the SiO₂ sample (Figure 5 and Supplementary Figure S30). The segmentation analysis indicates that the material becomes a bit denser (Figure 5c–f), which explains the drop in the degree of Hg porosimetry after TGA oxidation (Figure 4(1d,2d)) (Supplementary Tables S7 and S8). The increase in volume and surface area after the reduction in Al₂O₃ suggests a uniform deposition of carbon on the support pellets (Figure 4(2a–2d) and Supplementary Figures S22 and S23), because mercury can now penetrate the additional pores formed on the surface due to the deposited carbon. The µ-CT scanning of SiO₂ and Al₂O₃ illustrated no significant change between the uncycled and cycled samples after the reduction, which is a different observation to that offered by the Hg-porosimetry data (Figure 4(1b,2b)). However, this does not explain the drastic pore alteration from 1106 mm³/g to 777 mm³ and, as TGA oxidation restored the SiO₂ pellet to its initial pore state, 974 mm³/g (Figure 4(1b–1d) and Supplementary Table S7). Given that no changes were observed with µ-CT, XRD, XRF, or IR analyses, it is evident that the change in pore structure is due to carbon deposition, as identified in the Hg-porosimetry. Indeed, carbon itself cannot be detected in the µ-CT measurements. Instead, the observed changes in porosity are solely attributable to the carbon deposition, as the pore structure remains unchanged due to the cyclization process in the reactor. Based on the previously discussed line scanning and the carbon island that was found (Supplementary Figures S14 and S15), it is obvious that this can only be achieved by closing the pores, due to the carbon and making them accessible again through oxidation with steam (Figure 6a). This clearly indicates that almost no carbon reaches the surface or into the pores of SiO₂, which agrees with the analyses of the previous TGA investigation. It seems that that the SiO₂ exhibits a self-cleaning effect, because it shows that the carbon literally rolls off the material (Figure 6b and Supplementary Figure S1) and does not enter the pore network during the cyclization. This allows for its rapid oxidation during the oxidation phase, and the carbon-repellent effect leads to the swift release of CO₂, combined with high hydrogen quality, due to a kind of self-cleaning effect (Figure 6b). In contrast, with Al₂O₃, the carbon accumulates in the pores (Figure 6a), proven by a decrease in the pore volume from 571 mm²/g to 559 mm³/g (Supplementary Table S8), in combination with no changes in the pore network after the 30 cycles in the µ-CT (Figure 5 and Supplementary Figure S30). This results in a gradual release of CO₂, as well as the permanent contamination of the hydrogen, and does not lead to the sudden oxidation of the carbon (Figure 6b). Further investigations reveal that carbon deposition primarily occurred on the upper pellets, as the lower pellets showed no significant change compared to fresh pellets in SiO₂ (Supplementary Table S7) and Al₂O₃ (Supplementary Table S8). SiO₂ demonstrated a minimal accumulation of carbon and no significant agglomeration within the pores. In contrast, Al₂O₃ showed evident storage, correlating with the observed variations in hydrogen purity. The data yielded by the different analysis methods, therefore, explain and prove the observations of the hydrogen purity, as analyzed using GC, and show that the support pellets have a significant impact on green hydrogen production purity in the CLH process.

2.4. Carbon Deposition and Performance of the Oxygen Carrier

Finally, it is essential to elucidate why the Fe₂O₃-YSZ-SiO₂ oxygen carrier system exhibits better performance than the Fe-Al₂O₃-SiO₂ material system. XRD analysis reveals a notable observation. In the Fe-Al₂O₃ OC-system, a secondary phase of FeAl₂O₄ (hercynite) is formed adjacent to the support material (Figure 7a). It is challenging to reduce this hercynite compound with hydrogen and to oxidize it with steam, leading to the deactivation of iron and, consequently, a reduction in hydrogen release during oxidation [35,36,48]. In contrast, the XRD spectra of Fe-YSZ show only two different crystal structures of ZrO₂ without any intermediate phase forming with iron (Figure 7b). Furthermore, the XRD spectrum does not indicate the formation of carbides like Fe₃C, Al₄C₃, or ZrC. This shows that the carbon is deposited on the oxygen carrier as carbon species, similarly to in the supporting pellets. This absence clearly evidences the effects of CO₂ contamination through substantial carbon deposition and formation on exclusively FeAl₂O₄.

SEM examination of the oxygen carrier offers additional insights. There are black spots on the Fe₂O₃-Al₂O₃ oxygen carrier marked in orange dashed lines, which, upon magnification, show a fine deposit of a compound with a low atomic number (Figure 8b and Supplementary Figure S16). Detailed EDX analysis (Supplementary Figure S17) reveals that the bright spots predominantly contain reduced iron with a low proportion of 8.42 wt. % C (Supplementary Table S5). In the dark spots, the carbon content dramatically increases to 89.19 wt. % (Supplementary Table S5) on the surface. In contrast, using YSZ as a stabilizing material for the oxygen carrier results in a much finer crystalline iron with a uniform distribution, devoid of light and dark spots (Supplementary Figure S18). Since no secondary phases such as FeAl₂O₄ are formed, the iron can exhibit improved oxidizability. This enhancement contributes to a self-cleaning effect and explains the high hydrogen purity. The EDX analyses clearly indicate that the brighter spots are due to an increase in iron content (Supplementary Figure S19). The carbon content at the surface of the Fe-YSZ oxygen carriers is only 3–5 wt. %, accounting for the enhanced hydrogen purity (Supplementary Table S6).

Additionally, mercury porosimetry was also performed on the oxygen carrier to see what influence the formation of a side compound has on the active iron system. It was demonstrated that the Fe-YSZ oxygen carrier possesses a significantly higher pore volume of 323 mm³/g and a pore surface area of 18.3 m²/g (Supplementary Figure S25 and Table S9), facilitating better gas transport within the material. On the other hand, the Fe-Al₂O₃ shows a pore volume of 215 mm³/g and a pore surface area of 14.3 m²/g (Supplementary Figure S24 and Table S9). The side reaction in Fe-Al₂O₃ leads to FeAl₂O₄ formation and creates unreactive “dead spots”, which hamper pore formation and thus diminish reactivity (Figure 8c). By suppressing the side formation by using a chemically inert support material, YSZ, in addition to the active iron component, the pore volume was increased to 50%, and the surface area was increased up to 27%. The observed decline in hydrogen purity (Figure 2) can be scientifically attributed to the progressive carbon deposition on the support material across an increasing number of cycles (Figure 8c). This phenomenon slightly increases the hydrogen generation (Supplementary Figure S2) via the carbon–water reaction, but concurrently intensifies CO₂ contamination, predominantly in the latter stages of the oxidation phase (Supplementary Figures S4–S8). Overall, the Fe-YSZ’s material composition exhibits enhanced performance, characterized by a reduced tendency towards carbon deposition, which results in improved hydrogen purity and cycling stability (Figure 8c). Moreover, an interesting observation is that, with the same flushing time after the reduction in the reactor, Fe-Al₂O₃-SiO₂ demonstrates a CO₂ level of 400–500 ppm, whereas Fe-YSZ-SiO₂ maintains levels below 200 ppm also at high number of cycles (t = 0 min in the diagrams in Figure 2 and Supplementary Figures S4–S8). This suggests that CO₂ is trapped and retained in the pores by the reduced pore network through the formation of FeAl₂O₄. From a technical point of view, there is another advantage to the stabilization of the iron using YSZ, as impurities can be removed from the system more quickly. It is noteworthy that the use of YSZ to stabilize the oxygen carrier results in a rapid release of deposited carbon. This phenomenon occurs regardless of whether Al₂O₃ or SiO₂ pellets are used, elucidating the observed differences between Fe₂O₃-YSZ-SiO₂ and Fe₂O₃-YSZ-Al₂O₃ systems. These studies therefore show how important it is to understand the most suitable stabilizing materials for use in combination with iron to achieve efficient green hydrogen production.

3. Materials and Methods

3.1. Synthesis of the Oxygen Carrier

The oxygen carriers were prepared in a fixed weight ratio of Fe₂O₃ powder (Thermo SCIENTIFIC, Waltham, MA, USA) stabilized with a support material powder in the weight ratio of 80:20 wt. %. First, the powders were mechanically mixed in the Eirich EL1 mixing system. The following support materials were used: yttrium-stabilized zirconium oxide with 8 mol.% Y₂O₃ (IK-Hochrhein) and Al₂O₃ (Alfa Aesar, Ward Hill, MA, USA). The pelletized oxygen carrier samples with a diameter of 2–4 mm were formed in the high-performance mixing system (Eirich EL1) by the fine spraying of distilled water. For this purpose, 400 g Fe₂O₃ and 100 g YSZ/Al₂O₃ powder were mixed for 15 min at 2600 rpm. The mixture was then stirred manually to avoid deposits at the corners. This procedure, with high speed and manual mixing, was carried out four times. The mixture was then accelerated, and ultrapure water was sprayed into the chamber in the form of fine droplets via a steel nozzle. The water was added in doses of 3–5 g, and after each dose the mixture was stirred for 20 s at 2200 rpm until the next burst of water followed, to ensure homogeneous sphere formation. After the last dose, the formed pellets were rotated for a further 4 min. The angle of inclination of the stainless-steel container was 30° for all procedures.

After granulation, the OC was dried for 3 h at 108 °C in a drying cabinet. The pellets were then temperature-treated in a muffle furnace (ELSKLO with Jumo dicon 501 controller), where they were heated from room temperature to 900 °C at a heating rate of 15 K/min and held for 6 h at 900 °C in an air atmosphere. The following abbreviations have been used for the OC mixture composed of Fe₂O₃ and a support material in the weight ratio of 80:20 wt. %: Fe-YSZ for the oxygen carrier pellets with yttrium-stabilized zirconia as the support material, and Fe-Al₂O₃ for the oxygen carrier pellets with alumina as the support material.

3.2. Chemical Looping Fixed-Bed Set-Up

The material screening was performed in a chemical looping fixed-bed reaction set-up built in-house. A steel reactor with a total length of 70 cm and an inner diameter of 5.2 cm was fixed vertically in a holder so that the reactor was situated in the tube furnace. The tip of the thermocouple was positioned at a height of 35 cm inside. The reactor was filled with up to 34 cm of support pellets of Al₂O₃ or SiO₂, 250 g of the corresponding oxygen carrier, and finally support pellets again (see Figure 1 and Figure S1). The different combinations were tested in the reactor set-up, and the following abbreviations were used for the OC mixture in combination with the support pellets: Fe-YSZ-SiO₂, Fe-Al₂O₃-SiO₂ and Fe-YSZ-SiO₂. The thermocouple was therefore located 1 cm inside the OC bed. The experimental reactor was heated to 800 °C under nitrogen flow with a heating ramp of 10 K/min. The reduction cycle was performed with a synthetic gas mixture of 2 NL min⁻¹ H₂ (99.999%) and 0.5 NL min⁻¹ CO (98%) for 60 min. Oxidation cycling was performed with the feeding of 5 g min⁻¹ of H₂O with a pump in a 250 °C evaporator situated at the top of the reactor for 50 min. During the oxidation and reduction, an additional 1 NL min⁻¹ N₂ was added as an internal standard to calculate the volume of hydrogen, using the concentration detected using the GC column A. After each oxidation step, three further additional measuring points were recorded with only 1 NL min⁻¹ N₂. Afterwards, the reactor was flushed with N₂ for 30 min via the purge line. Before each further oxidation or reduction measurement, another three measuring points were recorded with only 1 NL min⁻¹ N₂. Thus, in the diagrams, time [min] = 0 is the point in time from which water or reduction gas was fed into the the chemical looping reactor unit. This also made it possible to show how much residual contamination was present in the reactor after the flushing process. The product gas/steam mixture was separated via a heat exchange unit and water separator in such a way that water droplets or moisture did not enter the GC analysis. After each measurement day, the reactor was kept at 800 °C and an internal gas flow of 0.5 NL min⁻¹ N₂ was passed over the reactor bed. This prevented the unwanted penetration and diffusion of oxygen into the reactor system, the oxidation of the formed carbon, or the entry of contamination into the GC line. After the 30th cycle of the reaction, the reactor was cooled to room temperature with inert gas flow at 1 NL min⁻¹ N₂. All samples were stored for further investigation in sealed sample vessels under an argon atmosphere for further investigation.

3.3. Gas Chromatography

All produced gases were analyzed by gas chromatography (GC) via an Agilent 3000A Micro GC with two columns. The GC was implemented after the water separation to enable inline measurements (see Figure 1). A GENIE170 membrane separator was used to prevent any residual moisture from influencing the measurement of the columns. A micro-thermal conductivity detector was used. The gas concentration was calculated by comparing the signals in the chromatograms obtained by the thermal conductivity detector with the calibration curves yielded by reference gases via the integrated peak areas in the EZIQ software from INFICON (version 3.3.2 SP2).

On column A (Molsieve 10 m × 320 µm × 12 µm), H₂, O₂, N₂, CH₄ and CO were analyzed using argon (99.999%) as the carrier gas. On column B (PLOTU 8 m × 320 µm × 30 µm), CO₂, C₂H₄, C₂H₆ and C₂H₂ were analyzed using helium (99.999%) as the carrier gas. Sample injection, analysis, and the rinsing of the column took 3 min in each analysis. The columns were always flushed overnight via a purge program in order to avoid any peak drift on the following measurement days.

3.4. Thermogravimetric Analysis

Thermogravimetric analysis was conducted using a Netzsch STA 449 C Jupiter system. Measurements were taken in a corundum sample cup with a volume capacity of 3.4 mL (Supplementary Figure S10d). Water vapor was generated using an ultra-pure water steam generator (aDROP) and transferred through a heating line maintained at 160 °C. Throughout the procedure, N₂ (99.999%) served as the protective gas, with a flow rate of 100 NmL/min from the upper part and 50 NmL/min from the scale. Initially, the reaction chamber was maintained at 90 °C for a minimum of 20 min to establish moisture and an inert atmosphere, and the TGA balance was zeroed. The sample was then heated to 800 °C at a rate of 10 K/min, followed by a 20 min stabilization period at 800 °C. Subsequently, the oxidation of carbon commenced, lasting 60 min with a steam supply rate of 2.44 g H₂O/min at 800 °C. After oxidation, a gas flow of 100 NmL/min from the upper part and 50 NmL/min from the scale using N₂ was applied for 20 min to evaluate the mass loss. The sample was then cooled under a nitrogen atmosphere and stored under argon in glass vials for further characterization. A blank measurement was performed following the same protocol with an empty corundum cup. Here, the determination limit was influenced by certain sources of error, particularly temperature fluctuations in the purge gas flow rate, notably originating from the steam supply.

3.5. SEM/EDX Analysis

Scanning electron microscopy (SEM) images were obtained with an FEI-XL20 (Philips, Amsterdam, The Netherlands). The energy-dispersive X-ray analysis (EDX) was performed with a Silicon Drift Detector from remX GmbH and analyzed in the IDFix software (2013, Version 1.05). The materials were placed in 0.5″ aluminum specimen stubs from Agar Scientific with conductive double-sided adhesive carbon tabs.

3.6. Mercury Porosimetry

Mercury porosimetry was performed using a PASCAL 140 Series and a PASCAL 440 Series Mercury Porosimeter (Thermo SCIENTIFIC). First, a blank measurement using the CD3 dilatometer from the PASCAL 140 Series and PASCAL 440 was performed. When testing an oxygen carrier sample, approximately 500 mg of sample was weighed into the CD3 dilatometer and transferred into the PASCAL 140 Mercury Porosimeter. After synthesis or cyclization, all samples were always stored under argon in glass vials to prevent oxidation or water retention. The sample amounts of the Al₂O₃ and SiO₂ support pellets were in the range of 300 mg, and for the oxygen carrier an amount of 500 mg was used. The blank and sample measurements were first evacuated to remove physisorbed gases from the interior of the sample’s pores. The experimental run started with outgassing, followed by an air pulse and a filling run in which the dilatometer was filled with mercury. After the low-pressure measurement (400 kPa) in the PASCAL 140 device, the sample was carefully transferred into the PASCAL 440 Mercury Porosimeter, where a high-pressure measurement (400 MPa) was performed. Based on the intrusion and extrusion curves, the mathematical pore model was used to calculate the characteristic pore parameter as a function of the distribution over the pore diameter for all samples. The calculations were performed using the Washburn equation [49] in SOL.I.D software (Version 1.6.6, Thermo SCIENTIFIC).

3.7. Infrared Spectroscopy

FT-IR spectra were recorded with a BRUKER ALPHA II. The samples were crushed with agate mortar to a powder and pressed on a monolithic diamond crystal in the device. Each FTIR spectrum was recorded over 60 co-added scans at a resolution of 4 cm⁻¹ in the region of 400–4000 cm⁻¹. All FTIR spectra were corrected against the atmosphere as a background via 24 scans. For apodization, the Blackman-Harrris 3-Term was used.

3.8. XRF

X-ray florescence (XRF) investigations were performed with a PANalytical Epsilon 1 system using a Ag tube and an acceleration voltage between 10 kV and 40 kV. Al, Cu and Ag filters were used for beam hardening. The fluorescent radiation was collected in an energy-dispersive mode, detecting the characteristic radiation values of elements with atomic numbers larger than 11 (sodium). Qualitative peak assignment to the individual elements and quantification were performed in the software Omnian (2020, Version 2.1) using interaction cross-sections of the individual elements.

3.9. XRD

X-ray diffraction (XRD) was performed with a PANalytical Emprean System using a sealed copper tube with an acceleration voltage of 40 kV. On the primary side, an X-ray mirror was used for monochromatization and for generating a parallel beam. The diffraction signal was collected with a PIXcel3D detector operating in the one-dimensional mode. Qualitative phase analysis was performed by comparing the experimental peak pattern with information in the database PDF2 (Powder Diffraction File 2) from the International Center of Diffraction Data (ICDD).

3.10. µ-CT Measurement

For the scans, samples with an approximate diameter of 1.8 mm were introduced into a polyimide tube with a 1.8 mm inner diameter and a wall thickness of 0.01 mm. The acquisition was performed using the lab-scale UniTOM HR device (TESCAN ORSAY HOLDING, Brno, Czech Republic). The acceleration X-ray voltage was 60 kV, and the target current was 25 μA in nano-focus mode. No filter was used for hardening the X-ray beam. For each scan, 2286 projections were acquired with an angular range of 360 degrees. Each projection was taken with a DEXELA camera with an exposure of 1.5 s, collecting six averages and applying two-fold binning. The camera’s pixel matrix and geometrical magnification of about 374 resulted in an isotropic voxel size of 400 nm. The field of view for each sample was about 0.61 mm × 0.31 mm. The computed tomographic reconstruction of the scans was performed with the Panthera (TESCAN XRE) software (Ghent, Belgium), using a ring filter with a width of 10 and an average filter with a kernel size of three.

3.11. µ-CT Segmentation Analysis

Segmentation of the raw 3D µ-CT images was performed in terms of Al₂O₃ and SiO₂, and the pore space was assessed in the “Al₂O₃ sample” and SiO₂ and in the “SiO₂ sample”. Due to the inherently weak phase contrast between Al₂O₃ and SiO₂ within the “Al₂O₃ sample”, we used a deep learning method for segmentation. Convolutional neural networks have been trained to reliably distinguish the constituents based on characteristic patterns, most prominently their shape (Dragonfly 2020.1, Object Research Systems (ORS) Inc., Montreal, Canada, 2020). For the reliable identification of these patterns, a pseudo-3D UNet architecture was used [50,51] because it is capable of recognizing a three-dimensional context from 2D image stacks without the computational burden of using neural networks operating exclusively with 3D input image data [51]. For the “SiO₂ sample”, conventional threshold segmentation was performed.

For a detailed analysis of the pore space and the subsequent extraction of pore size distribution, a pore network analysis was performed, utilizing the SNOW (Sub-Network of an Over-segmented Watershed) network extraction technique [52,53]. In this analysis, isolated pores, which do not connect to the surface of the sample, were excluded. For this purpose, the Al₂O₃ and SiO₂ phases were combined into one solid phase in the case of the “Al₂O₃ sample”.

4. Conclusions

The summarized work demonstrates that the interaction between support pellets and the active iron oxygen carriers plays a crucial role in CLH when seeking to achieve high hydrogen purity via the self-cleaning effect. By studying both microscopic and macroscopic phenomena, we determined how carbon is deposited in different ways. A strategic switch to SiO₂ instead of Al₂O₃ as a support pellet significantly increases the hydrogen purity to the impressive value of 99.994% without the need to resort to vacuum pumps or other energy-intensive methods. This remarkable improvement is due to the unique behaviors of SiO₂ and YSZ: side reactions of CO cause carbon to precipitate as flakes on the outside of SiO₂, enabling its rapid oxidation via the self-cleaning effect.

Additionally, integrating YSZ into the iron matrix yields remarkable self-cleaning properties in the oxygen carrier. Ensuring the chemical inertness of YSZ within the iron matrix is crucial for optimizing hydrogen production efficiency. Interestingly, carbon appears to roll off of the surfaces of YSZ and SiO₂ effortlessly, enabling its rapid oxidation. A crucial aspect of this process is the inhibition of harmful side reactions with iron as a result of the support material. This was particularly evident in the formation of FeAl₂O₄, which led to the formation of inactive “dead spots” that favored carbon deposition and hindered gas exchange by minimizing pore formation. The interaction underscores a critical aspect of the material’s behavior under cycling conditions, revealing the complex dynamics of carbon’s interaction within the CLH process.

Our studies clearly show that we can produce and store high-purity hydrogen by cleverly control the materials to procure a self-cleaning effect. This breakthrough eliminates the need for high-pressure or vacuum systems, and bypasses other energy-intensive purification processes, representing a significant step towards sustainable green hydrogen production from renewable resources with environmentally friendly materials.