This study utilized three commercially available adsorbents, namely KMG30 and MG30 obtained from SASOL (Germany), and KSORB obtained from BASF (Germany). Table 1 provides a summary of the material names and compositions (Coenen et al., 2017; Maroño et al., 2013). Taking KMG30 as an example, it is a potassium-promoted hydrotalcite with a Mg/Al molar ratio of 0.54 (weight ratio of 30:70 for MgO:Al₂O₃) and a potassium carbonate loading of approximately 17 wt%. KMG30 has been extensively investigated in the literature (Cobden et al., 2007; Boon et al., 2015) and shows promise for industrial applications due to its high mechanical stability and decent CO₂ adsorption capacity (Coenen et al., 2016). The adsorbent, a potassium-promoted hydrotalcite-based material characterized by high MgO content, specifically KMG70 (weight ratio of 70:30 for MgO:Al₂O₃) from SASOL, was not included in our current study. Compared with KMG30, KMG70 showed a higher cyclic working capacity due to a higher MgO content. However, slow formation of MgCO₃ in the bulk phase was observed for KMG70, which leads to poor mechanical stability and increased CO₂ slip during long-term CO₂/H₂O adsorption and desorption cycles (Jansen et al., 2013; Coenen et al., 2018). Except for anhydrous potassium carbonate (K₂CO₃ ≥99%, Sigma Aldrich), all other materials were crushed and sieved to obtain a particle size fraction ranging from 32 to 90 μm. All materials underwent pre-calcination at 523.15 K for 1 h and 723.15 K for 24 h, and were subsequently stored in a glove box under a N₂ atmosphere prior to use or characterization. The pre-calcination procedure was suggested by the manufacturer (Maroño et al., 2014).

The MG30 adsorbent with a K₂CO₃ loading of 20 wt% was prepared in-house and referred to as 20K-MG30. Two methods, namely the solid-solid mixing method and the impregnation-evaporation method, were employed for its preparation and subsequent characterization. In the first method, 0.8 g of pre-calcined MG30 and 0.2 g of anhydrous K₂CO₃ were mechanically stirred together. As for the latter method, 0.4 g of pre-calcined MG30, 0.1 g of anhydrous K₂CO₃, and 5.0 g of water were magnetically stirred at 353.15 K until a paste was formed. The solid material prepared using both methods underwent consecutive calcination steps at 363.15 K for 6 h, 473.15 K for 6 h, and 873.15 K for 3 h.

To investigate the long-term effects of different operating cycles, particularly those involving steam, on K₂CO₃-promoted hydrotalcite, a series of experiments was conducted using a custom-built thermogravimetric analysis (TGA) setup under atmospheric pressure. Approximately 100 mg of solid sample was placed in a porous ceramic basket within a quartz reactor (inner diameter approximately 15 mm). The sample was subjected to different gases with a total flow rate of 500 Nml/min. A thermocouple was positioned near the sample to accurately control the reaction temperature. The weight change of the sample was measured using a microbalance (CI-Precision MK2-5M) with a sensitivity of 0.1 μg. The balance head temperature was kept constant, and it was continuously purged with a constant flow of N₂ to protect the balance and ensure accurate mass readings.

The experimental setup, including the use of a Bronkhorst controlled evaporator mixer (CEM) for generating N₂-containing steam, sample lines wrapped with heat wires and insulation materials to prevent steam condensation, and the absence of mass transfer limitations for CO₂ adsorption by KMG30, has been previously published by our group (Coenen et al., 2016). The experiments for KMG30 are summarized in Table 2. Before each experiment, a pre-calcined sample underwent a pretreatment step with N₂ flushing at 873.15 K for 2 h. Each experiment consisted of multiple steps, indicated by hollow arrows, such as Step 1 ⇨ heat treatment ⇨ Step 2 ⇨ Step 3 ⇨ heat treatment ⇨ Step 4, and so on. A heat treatment process involved N₂ flushing to regenerate the sorbent at 873.15 K for 2 h but was not considered when numbering the steps. The sorbent after the heat treatment process is referred to as the heat-treated sorbent.

A standard adsorption step, such as Step 1/2/3/4, is denoted as [Procedure 1→Procedure 2] × (number of repetitions), where each procedure lasted for 30 min and the procedures in the square brackets were cyclically repeated at 673.15 K. For the procedures in the square brackets, the CO₂ partial pressure in all CO₂-containing gases was fixed at 0.15 bar, while the H₂O partial pressure in all H₂O-containing gases was kept at 0.30 bar. The ratio of steam to CO in a water gas shift reactor is normally between 2:1 and 3:1 (Ebrahimi et al., 2020). The remaining gas consisted of N₂. The adsorption/desorption capacity for a single procedure was defined by Eq. (1), whereas the cyclic working capacity for an adsorption-desorption cycle, such as [CO₂→N₂], was defined by Eq. (2). The average cyclic working capacity of the last 5 measurements ([CO₂→N₂] × 5) for each step was reported as the sorption capacity for that step. The sorbent cyclic capacity was based on the sample mass after the pretreatment step. It is worth noting that in Step 7 of Exp. 6, that consisted of four steps ([CO₂→N₂→H₂O→N₂] × 6), but only the sorption capacity for a single procedure was considered. A schematic diagram of Exp. 3 is shown in Figure 1. q C O 2 h a l f   c y c l e = Δ m h a l f   c y c l e M C O 2 * m p r e t r e a t e d _ s a m p l e (1) q C O 2 c y c l i c = Δ m a d s + Δ m d e s 2 * M C O 2 * m p r e t r e a t e d _ s a m p l e (2)

To gain a better understanding of the sorption performance of KMG30, some experiments in Table 2 were conducted for MG30, KSORB, and K₂CO₃. One experiment, denoted as [CO₂→N₂→H₂O→N₂]×6, was carried out to determine the adsorption capacities of CO₂ on different sites. In this experiment, part of the adsorbed CO₂ could be easily removed by reducing the CO₂ partial pressure in the gas phase through N₂ flushing (second step), while the remaining adsorbed CO₂ required reaction with steam (H₂O purge in step 3) for removal. Experiments of [CO₂→H₂O]×40 and [CO₂→N₂]×50 were performed to assess the potential loss in sorption capacity over time. Similar to the experiments conducted with KMG30, the CO₂ partial pressure in all CO₂-containing gases in these experiments was fixed at 0.15 bar, and the H₂O partial pressure in all H₂O-containing gases was fixed at 0.30 bar.

The hydrotalcite-based materials were subjected to multiple adsorption/desorption cycles within the TGA setup. After each cycle, the samples were promptly transferred for characterization. The morphological changes of the sorbents were investigated using various techniques. XRD analysis was performed using a Rigaku Miniflex 600 instrument with an air-sensitive sample holder to determine characteristic reflections. BET analysis was conducted using a Thermo Fischer Surfer instrument to determine the BET surface area and BJH (Barrett-Joyner-Halenda) pore size distribution. SEM-EDX analysis was carried out using a Thermo Scientific Phenom ProX instrument, and elemental compositions were determined using an Agilent 4200 MP-AES (microwave plasma atomic emission spectroscopy) instrument. Working cycles of H₂O adsorption→N₂ flushing, CO₂ adsorption→H₂O purge, CO₂/H₂O co-adsorption→N₂ flushing, CO₂/H₂O co-adsorption→H₂O purge and CO₂ adsorption→N₂ flushing were involved for treatments. The preparation procedures for the hydrotalcite-based material (KMG30) and the characterization techniques used are summarized in Supplementary Table S1.

To validate the accuracy of the TGA setup, the cyclic capacity of CO₂ (at 66.6 kPa) was compared with literature values. The reported literature values for KMG30 range between 12 and 13 mg/g (Coenen et al., 2017), whereas our experimental value was found to be 12.2 mg/g (Supplementary Figure S1). In Experiment 1, CO₂ was adsorbed from a CO₂/N₂ mixture at 400°C, followed by N₂ flushing at the same temperature. Thermal regeneration was optionally employed between adjacent steps to further regenerate the sorbent. As depicted in Figure 2, although the sample was regenerated at 600 °C before Steps 2 and 3, the sorption capacity exhibited a relatively rapid drop in the first five steps. Subsequently, it decreased slowly at a rate of approximately 0.018 mg/g per step, as indicated by the dashed lines in the figure. Regarding the two adjacent blue rhombus-shaped markers (indicating steps with prior heat treatment) after Step 5, the first blue rhombus was higher than the preceding black point, while the second rhombus was lower than the first one in most cases. The heat treatment (at 600 °C) only partially restored a sorption capacity of 0.25 mg/g (the distance between the two dashed lines). In the adsorption-desorption process, a continuous loss in sorption capacity was observed when only N₂ flushing and/or thermal swing was employed for sorbent regeneration.

This continuous loss in sorption capacity can be attributed to differences in the kinetics of sorption and desorption. As shown in Figures 2B,C, rapid mass changes occurred at the beginning of the procedures, where approximately 90% and 60% of CO₂ were adsorbed or desorbed within the first 200 s, respectively. Subsequently, CO₂ was only slowly adsorbed by KMG30 (Figure 2B), probably related to the formation of bulk carbonates, while the desorption from the sorbent still occurred at faster rates (Figure 2C). During the last 10 min (1200–1800 s), weight increases of 0.1–0.2 mg/g were observed for the adsorption curves, while weight decreases of 0.4–0.7 mg/g were observed for the desorption curves. Thus, the slow kinetics of the desorption is limiting the cyclic working capacity of KMG30. The reduced half-cycle capacity during the desorption procedures (from Steps 1 to 27 in Figure 2C) accounted for the loss in sorption capacity during multiple cycles of CO₂ adsorption followed by N₂ purge. Similar phenomena were also observed for the steps without heat treatment beforehand (black points in Figure 2). The adsorption and desorption kinetics of these steps are summarized in Supplementary Figure S2.

Exp. 2–5 were conducted to examine the impact of [H₂O→N₂] cycles. Exp. 2 was similar to Exp. 1, where the CO₂ is adsorbed at 400 °C and desorbed with N₂ flushing at the same temperature. Thermal regeneration was implemented before Steps 1, 2, 3, and 8, 9. Exp. 3 was a modification of Exp. 2, where Steps 3, 5, and 9, 11 consisted of 10 cycles of H₂O adsorption/N₂ flushing. The results for Exp. 2 and 3 are presented in Figure 3. The findings of Exp. 2 were consistent with Exp. 1, with a rapid decrease in sorption capacity during the first 5 steps, and the increased capacity in Step 8 was attributed to the pre-treatment. For Exp. 2 and 3, comparable sorption capacities were observed in Steps 1 and 2. The focus was on the difference between the two experiments. Following 10 cycles of [H₂O/N₂→N₂] in Step 3 of Exp. 3, the sorption capacity decreased from 7.7 mg/g in Step 2–6.8 mg/g in Step 4. In Exp. 2, the capacity decreased from 7.7 mg/g in Step 2–7.3 mg/g in Step 3. This finding is intriguing as it is commonly believed that steam can further regenerate the sorbent through reaction with the carbonate (Coenen et al., 2017). This holds true when considering the capacities of Steps 8, 10, and 12 in Exp. 3, where similar values were obtained, indicating that the [H₂O→N₂]×10 step helped maintaining the sorption capacity. Similar findings were observed in the additional tests following Step 12 in Exp. 2. As shown in Figure 3, the sorption capacity in Step 14 was higher than that in Step 12, and comparable values were found for Steps 16 and 14. Therefore, early exposure to steam purge/N₂ flushing cycles led to a capacity loss of 0.5 mg/g for KMG30. However, if the sorbent had already undergone 80 cycles of CO₂ adsorption/N₂ flushing, cycles of steam purge/N₂ flushing would help restoring the sorbent capacity.

In Exp. 4, the sorbent was initially subjected to 10 cycles of steam purge/N₂ flushing to validate the findings. As depicted in Figure 4, the sorption capacity in Step 2 of Exp. 4 was 7.3 mg/g, compared to a value of 8.2 mg/g for Step 1 in Exp. 3. This indicates a decrease of 0.9 mg/g in the sorption capacity of CO₂ on KMG30 when cycles of steam purge/N₂ flushing are applied at the beginning. However, a gradual decline in sorption capacities from Step 2 to Step 16 was observed, and it was evident that the blue stars mostly exceeded the black points after Step 3, suggesting that cycles of steam purge/N₂ flushing could restore the capacity subsequently. A comparison diagram illustrating the adsorption kinetics of Step 2 in Exp. 4 with the kinetics of Steps 1 and 2 in Exp. 2 is presented in Supplementary Figure S3A. Apart from the smaller half-cycle capacity for Step 2 in Exp. 4, the shapes of the curves were similar. The same trend was observed for the desorption kinetics (Supplementary Figure S3B). Therefore, after undergoing cycles of steam purge/N₂ flushing, the CO₂ adsorption mechanism remains unchanged while the number of adsorption sites decreases, resulting in a smaller half-cycle capacity for both the adsorption and desorption procedures.

In Exp. 5, the sorbent was subjected to 40 cycles of steam purge/N₂ flushing, followed by 20 cycles of CO₂ adsorption/steam purge. Surprisingly, the cyclic capacity continued to increase throughout the cycles until reaching a constant value of approximately 12.3 mg/g. This trend is in stark contrast to the decreasing trend observed in cycles of CO₂ adsorption/N₂ flushing (Figure 2A). The final stable value is significantly higher than the CO₂ sorption capacity observed in Exp. 4 (around 7.0 mg/g). This indicates the presence of different CO₂ adsorption sites, some of which can only be regenerated using steam (Coenen et al., 2017). The descriptions of the adsorption sites (Sites A and C for H₂O, Sites B and C for CO₂) following the work of Coenen et al. are presented in Supplementary Figure S4. The cyclic capacities of last five cycles in Figure 5 are consistent with the sorption capacities for the CO₂ adsorption/steam purge cycles in Section 3.3, showing that the loss of sorption capacity of KMG30 could be restored by subjecting it to cycles of CO₂ adsorption/steam purge. A detailed calculation procedure for the capacities of CO₂ and H₂O on different sites is presented in the next section.

In Exp. 6, the sorbent stability under cycles of CO₂ adsorption/steam purge was tested. Cycles of [CO₂→N₂→H₂O→N₂] were employed to determine the adsorption capacities of CO₂ and H₂O in different sites. Following multiple cycles of CO₂ adsorption/steam purge, cycles of [CO₂→N₂] were performed to compare the sorption capacities of CO₂ with those obtained in Exp. 1.

From Figure 6, it can be observed that the sorption capacities of Steps 1–6 and 11–15 exhibit variations around 12.2 mg/g. KMG30 appears to demonstrate remarkable stability during cycles of CO₂ adsorption/steam purge. This stability can be attributed to the rapid kinetics of both the adsorption and desorption processes. Over the last 200 s, the weight increases in the adsorption curves and weight decreases in the desorption curves are approximately 0.04 mg/g and 0.08 mg/g, respectively. These weight changes are insignificant when compared to the overall mass change of approximately 12 mg/g throughout the entire adsorption/desorption period (1800 s). Furthermore, unlike the declining trend in half-cycle capacities observed from Step 1 to Step 27 in Exp. 1 (Figure 2), there is no consistent decreasing or increasing trend in half-cycle capacities from Step 1 to Step 15 in Exp. 6 (Figures 6B,C). For instance, the curves of Step 1 overlap with the curves of Step 11.

Step 7 in Exp. 6 involves six cycles of [CO₂→N₂→H₂O→N₂], which are conducted to measure the adsorption capacities of CO₂ and H₂O on different sites. In this process, the H₂O adsorbed on Site A and CO₂ adsorbed on Site B can be regenerated when the partial pressures of the adsorbed species in the gas phase decrease. On the other hand, H₂O and CO₂ adsorbed in Site C can only be replaced by each other. Specifically, CO₂ will desorb from Site C only if steam is present in the gas phase and the partial pressure of CO₂ decreases. The interchange between CO₂ and H₂O in Site C can be represented by Eqs (3), (4). The final step, Step 6, involves steam purge. The first cycle of Step 7 is depicted in Figure 7. During this cycle with CO₂ feed, H₂O is desorbed from both Sites A and C, while CO₂ is adsorbed on Sites B and C. The specific weight changes observed in Figure 7 are used to determine the sorption capacities of Sites A, B, and C.

During the steam purge, although some H₂O (4.4 mg/g) is adsorbed on Site C, there is a simultaneous exchange (resulting in a net weight loss of (3.1 mg/g)) of CO₂ and H₂O on Site C. The sorption capacities of CO₂ and H₂O on Site C can be determined from this net loss using Eqs (3), (4). The final results are presented in Figure 7, which accurately describe the entire figure. For more detailed information on the adsorption sites of a hydrotalcite-based adsorbent for CO₂ and H₂O adsorption, please refer to the work by Coenen et al. (Coenen et al., 2017). S i t e   a c t i v a t i o n : C • O + H 2 O → C O H 2 (3) S i t e   e x c h a n g e : C O H 2 + C O 2 ↔ C C O 3 + H 2 O (4)

The sorption capacities of Steps 8 and 10 in Exp. 7, indicated by red stars in Figure 2A, are consistent with the sorption capacities of Steps 1 and 3 in Exp. 1. Even after multiple cycles of CO₂ adsorption and steam purge, the sorption capacities of CO₂ remain the same as those observed with fresh KMG30. This further demonstrates the exceptional stability of KMG30 under cycles of CO₂ adsorption and steam purge. Furthermore, the sorption capacity of CO₂ in Site B is determined to be 9.5 mg/g according to Figure 7. In contrast, the sorption capacity of Step 1 in Exp. 1 (Figure 2A) is significantly lower, approximately 8.0 mg/g. Upon examining the cyclic capacities of the ten cycles in Step 1 of Exp. 1 (Supplementary Figure S5), the desorption capacities decrease from 9.1 to 8.0 mg/g, all of which are smaller than the determined CO₂ capacity in Site B. One possible reason for this difference is that, apart from Sites B and C, Site D also contributes to CO₂ adsorption (Coenen et al., 2017). CO₂ can be desorbed from Site D under N₂ flushing. Consequently, the method used to determine the capacity of Site B may not be sufficiently accurate. However, since the primary objective of this study is to investigate the long-term stability of KMG30 under various operational cycles, in which it will remain occupied throughout, Site D is not considered further. Another reason for the disparity could be that the sorption capacity on Site B increases during cycles of CO₂ adsorption and steam purge.

In Exp. 7, the sorbent stability under cycles of co-adsorption of CO₂ and H₂O/N₂ flushing was investigated. The results are summarized in Figure 8. The cyclic capacities exhibit a gradual decrease and eventually stabilize at around 15.1 mg/g from Cycles 1 to 60 (Steps 1–6). The stable performance can be attributed to the rapid desorption kinetics, as observed in Figures 8B,C. Over the last 200 s, the weight increases in the adsorption curves and weight decreases in the desorption curves are approximately 0.10 and 0.15 mg/g, respectively. These weight changes, which are of comparable magnitudes, are insignificant when compared to the overall mass change of around 15 mg/g during the entire adsorption/desorption period (1800 s). The gradual decrease in sorption capacities observed during the initial 30 cycles may be associated with the slightly slower desorption kinetics compared to the adsorption kinetics, as well as some small morphological changes of KMG30.

In Exp. 8, the sorbent stability was tested under cycles of co-adsorption of CO₂ and H₂O/steam purge. The results are summarized in Figure 9, where the cyclic capacities gradually decrease and stabilize around 14.6 mg/g from Cycles 1 to 60 (Steps 1–6). The stable performance is also attributed to the fast desorption/adsorption kinetics exhibited under the working conditions. Figures 9B,C illustrate the adsorption and desorption kinetics, respectively. During the last 200 s of the process, the weight increases in the adsorption curves and weight decreases in the desorption curves are approximately 0.10 and 0.25 mg/g, respectively. Once again, these weight changes, which are of similar magnitudes, are insignificant when compared to the overall mass change of around 15 mg/g during the entire adsorption/desorption period (1800 s).

KMG30 demonstrates stability under cycles of CO₂ adsorption/steam purge, cycles of CO₂ and H₂O co-adsorption/N₂ flushing, and cycles of CO₂ and H₂O co-adsorption/steam purge. However, a noticeable decrease in sorption capacities is observed during cycles of CO₂ adsorption/N₂ flushing, which can be attributed to the difference in desorption kinetics. The summary of sorption capacities for different adsorption/desorption cycles was presented in Supplementary Table S2. From a microscopic perspective, Site C appears to play a crucial role in maintaining stable sorption capacities for CO₂. To ensure stable performance, it is necessary to remove CO₂ from Site C. In the cycles of CO₂ adsorption/N₂ flushing, only Site B of KMG30 is utilized. However, under cycles of CO₂ adsorption/steam purge, both Sites B and C are involved (as shown in Figure 5). The same explanation applies to the cycles of CO₂ and H₂O co-adsorption/steam purge, where both Sites B and C contribute to the process (as seen in Figure 9A). In the cycles of CO₂ and H₂O co-adsorption/N₂ flushing, it is hypothesized that during N₂ flushing, the steam desorbed from Site A may assist in removing CO₂ from Site C through Eq. (3). However, this hypothesis requires further validation through breakthrough experiments.

Experiment of [CO₂→N₂→H₂O→N₂]×6 ⇨ [CO₂→H₂O]×40 ⇨ [CO₂→N₂]×50 was conducted using MG30, KSORB and K₂CO₃. For K₂CO₃, negligible weight changes were observed throughout the long test due to its very low surface area. In the cycles of [CO₂→N₂→H₂O→N₂], as shown in Supplementary Figure S6A, B, Sites A, B, and C were applicable to KSORB but not to MG30. MG30, on the other hand, exhibited behavior more consistent with being a sorbent for H₂O adsorption rather than CO₂ adsorption, as evidenced by the cycles of [CO₂→H₂O] (see Supplementary Figure S6C). Further breakthrough experiments are necessary to elucidate and quantify the behavior of MG30 during the cycles of [CO₂→N₂→H₂O→N₂].

During the cycles of [CO₂→N₂], KSORB exhibited higher cyclic capacities than MG30, but lower than KMG30 (see Figure 2A). This can be attributed to the ability of K₂CO₃ to create basic sites for CO₂ adsorption through its interaction with MgO and Al₂O₃ on potassium-promoted layered double oxides (Veselovskaya et al., 2013; Zhu et al., 2019). Similar to KMG30, the CO₂ capacities (in Site B) rapidly decreased for KSORB during the first 20 cycles of [CO₂→N₂]. Weight loss was less pronounced for MG30. Thus, the sites created by K₂CO₃ could be a contributing factor to the capacity loss. Within the cycles of [CO₂→H₂O], stable cyclic capacities were observed for both MG30 and KSORB (see Figure 10).

To investigate the effects and changes in the morphology of KMG30, various characterization techniques were employed, where one should keep in mind that restructuring of the sorbent can take place during cooling down at the end of the experiment. Initially, SEM images were taken to examine the morphological changes of KMG30 after exposure to cycles of [H₂O→N₂] and cycles of [CO₂/H₂O→N₂]. However, as shown in Supplementary Figure S7, there were minimal differences observed in the SEM images. However, this may also indicate that SEM may not be suitable for accurately assessing the morphological changes in KMG30. Based on our assumption, the sites created by K₂CO₃ could be the reason for the capacity loss in Section 3.2.

EDX and ICP-MS analyses were conducted to further understand the distribution of K₂CO₃ on the surface of KMG30. The results obtained from the EDX and ICP-MS analyses were found to be consistent. EDX analysis (details see Supplementary Table S3) is used to investigate the elemental composition of solid surfaces and ICP-MS technique is used to analyze the overall concentration of a sample. The results, with no significant difference between the surface (EDX) and bulk (ICP-MS) concentrations of K, Al, and Mg, suggest that K₂CO₃ is distributed throughout the entire sorbent, including both the pores and surfaces of KMG30.

Notably, EDX point analysis was performed on 10 selected points on the surface of KMG30. The mean composition of these points was summarized in Supplementary Table S4. It was observed that the standard deviations, represented by the signs (±), became larger when KMG30 was exposed to cycles of [H₂O→N₂]. This suggests that the treatment with cycles of [H₂O→N₂] led to more heterogeneous distribution of K₂CO₃ and Al₂O₃ on the surface of KMG30. However, it is important to note that KMG30 was found to be a heterogeneous sorbent, as evident from the varying color depths observed in the EDX analysis (Supplementary Figure S7D). Therefore, drawing conclusive interpretations solely based on the analytical results obtained by EDX may not be convincing. The XRD technique was employed to determine any alterations in the crystal structure of KMG30 and provide valuable insights into the structural changes in KMG30 following different treatment cycles.

The effects of other working cycles including [CO₂→N₂], [CO₂→H₂O], [CO₂/H₂O→N₂] and [CO₂/H₂O→H₂O] on the morphological changes of KMG30 were investigated using XRD. The results, as depicted in Figure 13, show no significant changes in the XRD patterns when the heat-treated KMG30 is subjected to cycles of [CO₂→N₂], heat treatment, cycles of [CO₂/H₂O→N₂], and cycles of [CO₂/H₂O→H₂O]. The effects of cycles of [H₂O→N₂] have already been discussed in Section 4.1, as represented by the green curves in Figure 13A. However, when the heat-treated KMG30 is processed with cycles of [CO₂→H₂O], a small peak is observed at a 2-theta value around 32°, indicating some minor structural changes that warrant further investigation.

The morphological changes of KMG30 were also analyzed using BET surface area analysis. The results, presented in Supplementary Table S5, show that the surface areas of KMG30 exhibit only slight changes when subjected to different working cycles. While surface area is an important property for an adsorbent, these slight changes do not fully explain the observed capacity loss when KMG30 is exposed to cycles of [H₂O→N₂]. To further investigate the pore size distributions, the BJH method was employed, and the results are displayed in Supplementary Figure S12. Overall, it can be observed that the pore sizes tend to increase after different treatments. Specifically, the pore volumes decrease for pores with sizes smaller than 4 nm, while the pore volumes increase for pores with sizes larger than 8 nm. However, when the heat-treated KMG30 is treated with cycles of [H₂O→N₂], as shown in Figure 11A, it is unlikely that the closure of nanopores with sizes less than 4 nm is responsible for the aggregation of K₂CO₃, considering the small pore volumes of these nanopores (<0.008 cm³/g). It should be noted that it is challenging to speculate on the morphological changes of the sorbent under working cycles (at 400 °C) using ex-situ analyses alone. Therefore, it is recommended to use in-situ TEM (transmission electron microscopy) for future studies to gain more insight into the morphological changes of KMG30.