Influence of material composition on the CO2 and H2O adsorption capacities and kinetics of potassium-promoted sorbents

A R T I C L E I N F O


Introduction
Layered double hydroxides (LDHs), also often called hydrotalcitelike materials (HTC's), belong to a large group of anionic or basic clays. These materials showed a wide range of applications as catalyst, precursors and adsorbents [1]. Several thousands of tons of HTC's are produced annually by several chemical companies, such as BASF, SASOL, Clariant, Kisuma Chemicals, Sakai Chemical, etc. [2]. A general formula for this type of materials is [M 2+ 1−x M 3+ x (OH) 2  , Cl − or OH − . A schematic representation of a LDH can be seen in Fig. 1.
More recently hydrotalcite-based adsorbents have been investigated for the use in sorption-enhanced water-gas shift (SEWGS) processes. SEWGS is a promising concept for pre-combustion CO 2 capture, which combines the water-gas shift (WGS) reaction with in situ CO 2 removal. In situ CO 2 removal increases the CO conversion attainable at elevated temperatures (300-500°C) due to the shift of the WGS equilibrium (Le Chatelier's principle). Additionally, the operational costs in comparison to alternative processes, like the conventional Selexol process, can be reduced due to a lower steam/CO 2 ratio [4,5]. To deal with the periodic behavior of the adsorption/desorption cycles usually multiple columns are operated in parallel to obtain a continuous process.
In addition to their high stability [6,7] and fast adsorption/desorption kinetics, hydrotalcites show a high selectivity towards CO 2 compared to CO and H 2 [8,9]. At elevated temperatures, the typical layered structure is lost and the LDH is converted into a mixed metal oxide (LDO) with a high surface area and basicity [10]. To increase the basicity and the reversible adsorption capacity of hydrotalcite-based adsorbents, they are promoted with alkaline cations [7]. K 2 CO 3 promotion has been reported to be very beneficial in terms of CO 2 adsorption capacity [8,9,[11][12][13][14][15]. In addition, promotion with also other alkaline cations, like Sr and Cs, and their effect on the CO 2 adsorption capacity were investigated [13]. Not only hydrotalcite-based adsorbents with different Mg/Al ratios and promotion with different alkali-metals were investigated, but also potassium-promoted alumina and MgO were studied as a possible adsorbent for their use in a SEWGS process [3,12,16]. However, the formation of MgCO 3 can lead to mechanical instability for some HTC-based adsorbents and MgO, which can cause major problems for the operation in packed-bed reactors [17]. Usually, the cyclic working capacity of CO 2 is reported for the different adsorbents at various operating conditions. Since the sorbents will be exposed to high partial pressures of CO 2 and H 2 O, not only the CO 2 adsorption capacity, but also the capacity for H 2 O and the influence of H 2 O on the sorption of CO 2 are important, especially for a more accurate process optimization and cost analysis of SEWGS processes with these sorbents. Capacity and adsorption kinetics of a potassiumpromoted hydrotalcite with a Mg/Al ratio of 0.54 for both CO 2 and H 2 O were investigated in an earlier work. It has been shown that the adsorption capacity of CO 2 and H 2 O are highly dependent on the desorption time due to the relatively slow desorption kinetics [18].
In order to study the influence of the material composition on the adsorption and desorption kinetics of CO 2 and H 2 O, a hydrotalcitebased adsorbent with a higher MgO content and a potassium-promoted alumina were compared in this study at the same operating conditions. In the first part of this paper we will show how the MgO content can influence the adsorption kinetics of CO 2 and we will demonstrate the positive effect of a higher MgO content on the CO 2 cyclic working capacity, while the adsorption kinetics of H 2 O are not significantly affected by the material composition. Finally, to explain the decomposition of potassium carbonate on the sorbents due to exposure to H 2 O at high temperature TGA-MS and TGA-TPD experiments were conducted. XRD and BET analysis were used to characterize the sorbent before and after steam treatment.

Materials and methods
Two different calcined potassium-promoted hydrotalcite-based adsorbents (SASOL), with different Mg/Al ratio were used in the experiments and compared to a potassium-promoted alumina (BASF). The material names and compositions have been summarized in Table 1. The materials were pre-calcined by the manufacturer at 250°C and 450°C for 24 h [19].
KMG30 and KMG70 adsorbent pellets (4.7 × 4.7 mm) and KSORB (spheres with a diameter of 2 mm) were crushed to powders with similar particle size, which have subsequently been used in TGA-measurements.
The materials were characterized using a helium Pycnometer (Quantachrome Upyc 1200e), BET (Thermo Fischer Surfer), XRD (Rigaku Miniflex 600) and SEM-EDX (FEI-Quanta) to study the morphology of the used material. Particle size distribution was measured in a Fritsch Analysette  Sorption experiments were carried out in two similar in-house developed TGA setups, which were already described elsewhere [18]. Prior to the adsorption experiments all materials were pretreated in the same way (heating the samples to 600°C in N 2 with a dwelling at 600°C for 120 min) as reported in an earlier publication [18]. In this publication we will distinguish between different experiments that have been conducted with TGA.
1. In order to investigate the adsorption mechanism and kinetics of the different materials, adsorption of CO 2 and H 2 O were conducted according to the conditions listed in

Results and discussion
3.1. Characterization of different materials N 2 adsorption isotherms were conducted for the three different materials, in order to determine their surface area. Fig. 2 shows the adsorption isotherms and the resulting surface area (BET method). All materials show an isotherm typical for a mesoporous material. The determined BET surface area decreased for the three materials in the following order: KMG30 > KSORB > KMG70. It has been reported earlier in the literature that the surface area of the original HTC prior to K 2 CO 3 promotion for KMG30 is higher than for KMG70 [9,20,21]. Since K 2 CO 3 leads to a significant reduction in surface area, one can expect the same order for potassium-promoted materials. However, the measured surface areas reported in the literature do not always show the same trend. This may be caused by different impregnation or calcination methods used. A pore size analysis shows that the pore size of all sorbents is below 40 nm as determined with BHJ method analysis (Fig. 2b). KMG30 has the broadest pore size distribution compared to the other sorbents. From the isotherms it is evident that the pore volume for KMG70 is smallest and about 5 times smaller than for KMG30.

CO 2. Adsorption experiments
The CO 2 cyclic working capacity was determined for the different sorbents using different half-cycle times. It can be seen that for all sorbents the cyclic working capacity can be increased by increasing the half-cycle time (Fig. 3). In general the increase in cyclic working capacity for an increased half-cycle time is a result of more CO 2 being desorbed during the regeneration step which was reported earlier for KMG30 [18] and is also valid for the other sorbents, independent of their material composition and difference in morphology. For all sorbents, the cyclic working capacity is nearly doubled when increasing the half-cycle-time up to 600 min. Although 600 min is not a realistic regeneration time for actual applications, the results give an indication on how much the cyclic working capacity could be increased if the regeneration of the sorbent were improved.
During all experiments the sorbents can be sorted according to their cyclic working capacity in the order of KMG70 > KGM30 > KSORB. According to this, it can be stated that, even if the surface is lowest for KMG70, the highest cyclic working capacity was measured for this sorbent. It is known that the MgO content in the sorbent has a significant influence on the basicity of the material [22], which results in more active adsorption sites for CO 2 on a material with a higher MgO content, explaining the obtained results.
We have reported before that usually a higher operating temperature can have a positive effect on the cyclic working capacity for hydrotalcite-based adsorbents due to the increase in the desorption rates at higher operating temperatures [18,23]. We plotted the cyclic working capacity determined for CO 2 at different temperatures in Fig. 4 for the different sorbents. It can be seen that for KMG70 the already described trend for KMG30 is confirmed. However, it seems that the cyclic working capacity of CO 2 for the potassium-promoted alumina is not increased significantly when increasing the operating temperature.
The reason for this difference is the generally different adsorption and desorption behavior of hydrotalcite-based adsorbents compared to potassium-promoted alumina. From the TGA curves shown in Fig. 5a, it can be seen that for HTC-based adsorbents during the first adsorption cycle, the initially adsorbed amount of CO 2 is nearly independent of the operating temperature. However, the amount of CO 2 desorbed during the regeneration step is increased at a higher operating temperature, as already indicated. Fig. 5b shows that this is not the case for potassiumpromoted alumina. At a lower operating temperature more CO 2 is being adsorbed initially, where the adsorption kinetics seem to remain the same, even if the CO 2 loading is different (see the small figure inside Fig. 5b). After the first cycle the cyclic working capacity for potassium-promoted alumina remains basically unchanged during the following 10 cycles (Fig. 6), where the cyclic working capacity is slightly higher at lower temperature. One would expect that at higher temperatures less CO 2 can be adsorbed due to increasing desorption kinetics (high kinetic energy of gas molecules), which can be seen for the potassium-promoted alumina. It has been reported in the literature that the adsorption of CO 2 on potassium-promoted HTC can be described with a bi-Langmuir adsorption isotherm, where both physisorption and chemisorption lead to a similar adsorption capacity at different operating temperature [24].
It is remarkable that during the first adsorption cycle the CO 2 adsorption capacity is higher for KSORB than for both HTC based materials at a lower operating temperature. Considering both surface area (adsorption) and bulk basicity one would expect that the HTC based adsorbent would show a higher adsorption capacity. One explanation could be that at lower temperature condensation of CO 2 could occur in the very small pores of the potassium-promoted alumina. The theory of a possible condensation of CO 2 in nanopores was reported earlier in the literature [25] and could be an explanation for the obtained behavior, since KSORB has the lowest a mean pore diameter of around 10 nm (see Fig. 2b) compared to KMG30 and KMG70. This could explain why pore condensation only occurs for this material or contributes more to the adsorption mechanism. Another reason can be that the integration of K 2 CO 3 species in the HTC-based adsorbents is different. Adsorption sites on KSORB, created by K 2 CO 3 impregnation, are available on the surface and therefore strongly dependent on the surface coverage, whereas for the HTC-based adsorbent CO 2 is also bond by a reaction. It has been   reported that needle shaped crystals on the surface of an adsorbent have been identified as K 2 CO 3 using SEM-EDX and have been found on potassium-promoted alumina to a higher extent than on a potassiumpromoted HTC [12]. Adsorption and desorption kinetics are very important for the effectivity of the used sorbents especially when using them in cyclic experiments. A comparison of the normalized kinetics of the different sorbents with respect to the total adsorbed mass during the adsorption cycle of 300 min are plotted in Fig. 7. It can be seen that the potassiumpromoted alumina has a very fast initial adsorption rate, where equilibrium seems to be established after about 8000 s. The HTC-based sorbents also exhibit a very fast initial adsorption rate, however at approximately 80% normalized weight change a very marked decrease in adsorption rate is observed, followed by a relatively constant but slow adsorption rate. For HTC-based materials this slow adsorption rate contributes to a much larger extent to the total CO 2 adsorption capacity. A higher MgO content seems to enhance this effect, which can be explained by a slow reaction of MgO sites with CO 2 . It has been reported that potassium-promoted HTC can form bulk MgCO 3 under higher partial pressures of CO 2 and H 2 O. At low pressure and dry adsorption conditions a possible explanation of the observed behavior could be the migration of adsorbed CO 2 on the surface to highly basic adsorption sites in the bulk of the material. It was reported in the literature for a similar material as KMG70 that different adsorption steps (from fast adsorption to diffusion of gaseous CO 2 to finally chemisorption in basic sites) can explain the adsorption kinetics on this type of sorbents [26]. During the desorption step this slowly adsorbed CO 2 seem to be desorbed quite fast, indicated by the faster relative desorption kinetics for KMG70 compared to KMG30 and the potassium-promoted alumina (KSORB). With two independent reaction mechanisms involved in the HTC-based materials compared to one single adsorption site for KSORB the obtained results for both adsorption and desorption can be described. It was already reported that promotion with potassium increases the amount of basic sites available for HTC and alumina, where alumina itself does not show any significant CO 2 adsorption capacity [12]. It was concluded that especially the interactions between K 2 CO 3 and the aluminum centers play a crucial role in the creation of basic adsorption sites [12].

H 2 O adsorption experiments
Similar to CO 2 , the influence of the half-cycle-time on the cyclic working capacity of H 2 O on the different adsorbents was investigated at 400°C and P H2O = 0.34 bar during the adsorption step. It can be seen from Fig. 8 that KMG30 has the highest cyclic working capacity for H 2 O followed by KMG70 and KSORB. Considering only the available surface area, one would expect KSORB to show a higher cyclic working capacity for H 2 O than KMG70. However, when considering the surface Fig. 5. a) mass change during the first cycle for HTC-based adsorbents; b) mass change for the potassium-promoted alumina during the first cycle with a close-up of the desorption step if the initial weight is set to 0 after the adsorption step to compare the desorption kinetics.  basicity and the difference in the materials one can understand that the HTC-based adsorbents show a higher cyclic working capacity for H 2 O, since stronger sites are available. Surface basicity becomes very important at higher operating temperatures during adsorption (greater ratio of gas phase molecules and shorter residence time of gas molecules on the surface). In this case a simple physisorption mechanism for the adsorption is not able to explain the experimental results. However, the fact that the order in highest cyclic working capacity is changed for H 2 O confirms that the mechanism for adsorption is different from the mechanism for CO 2 [18]. Similar to CO 2 , the cyclic working capacity for H 2 O is determined by the relatively slow desorption kinetics. Therefore, an increase in half-cycle-time leads to an increase in cyclic working capacity. However, this increase is less significant for H 2 O than for CO 2 . One reason for this observed behavior can be found in the normalized adsorption and desorption kinetics of H 2 O with respect to the total adsorbed/desorbed amount of H 2 O within one step. It can be seen from Fig. 9 that the adsorption of H 2 O is extremely fast and reaches adsorption equilibrium even within a very short time. It seems that the adsorption kinetics are the same for the different materials. This suggests that only surface adsorption is taking place on different types of basic sites without any bulk diffusion or reaction like it was observed for CO 2 . The same observation can be made for the desorption step (Fig. 9b), where the normalized adsorption kinetics are similar for the different materials. Comparing the desorption kinetics of H 2 O (Fig. 9b) to CO 2 (Fig. 7b) it becomes clear that H 2 O desorbs much faster and thus an increase in half-cycle-time leads to a less pronounced increase in cyclic working capacity.
Increasing the operating temperature leads to a decrease in cyclic working capacity for H 2 O (see Fig. 10) contrary to the observations for CO 2 . The differences in cyclic working capacity regarding the different sorbents decrease at higher operating temperatures. It was observed for experiments at higher operating temperatures that during the first few adsorption cycles, the sorbents KMG70 and especially KSORB show a weight decrease in mass after the first rapid increase which was attributed to the desorption of CO 2 due to the presence of steam on the sorbent.
This observed weight loss can be seen in Fig. 11 for the sorbent KSORB and KMG70. For KSORB it can be seen, that the during the first adsorption cycle a strong decrease in mass is detected at 500°C. For KMG70 this weight decrease is less prominent at 500°C, but it is visible during the following cycles. Both sorbents only show the described behavior if the temperature is increased (> 400°C). It can be seen that this effect was not observed in the experiments at 300°C. Since the desorption kinetics (for the experiments at 500°C) remain unchanged during the cycles it can be concluded that most probably the sorbent releases CO 2 due to the exposure to H 2 O at a high operating temperature. We have reported that for potassium-promoted HTC (KMG30) an adsorption site exists which is can be occupied by either CO 2 or H 2 O, where one component replaces the other [23]. The existence of this type of site on KMG70 and KSORB could explain the described results. In order to confirm this hypothesis TGA-MS and TGA-TPD were carried out to study this effect and the influence of H 2 O on the cyclic working capacity of CO 2 .   The normalized weight change for the different sorbents and reactor temperature during the TGA-TPD experiment are plotted in Fig. 12a (for a detailed description of the procedure see Section 2), while Fig. 12b shows the mass change during the steam adsorption and TPD during steps 2 and 3 in more detail. In order to compare the weight change in the different sorbents, the initial mass was set to 0 after the pretreatment and both x -and y-axes were set to 0 at the start of the adsorption at 300°C. After H 2 O adsorption at 300°C, all sorbents are losing weight during the increase in temperature to 850°C. Where KMG30 shows a more or less linear weight loss during the linear increase in temperature, for KSORB at about 500°C a strong decrease in weight is observed, while at even higher temperatures the weight loss is linear with the increase in operating temperature. For KMG70 the strong weight loss occurs at a higher temperature (about 650°C, Fig. 12b) and continues until the final temperature of 850°C is reached. At the end of the adsorption step only a small amount of H 2 O is left on the sorbent, which is basically the same for all sorbents. This can be seen in the following desorption step, where for all sorbents the same mass decrease is detected upon changing the feed gas from H 2 O/N 2 to N 2 . It is expected that some of the previous adsorbed H 2 O desorbs when the temperature is raised due to an increase in the desorption rate shifting the equilibrium constant for H 2 O. Regarding the linear weight loss of KMG30 one would expect that only water is desorbing form the sorbent during this step. Comparing the weight before the adsorption of H 2 O in Fig. 12b and after desorption, KMG30 has lost about 14 mg/g and KSORB and KMG70 about 40 mg/g which can be attributed to additional CO 2 being desorbed.
It has been reported in the literature that K 2 CO 3 decomposes releasing CO 2 at higher operating temperatures on potassium-promoted HTC and potassium-promoted alumina [12]. Fig. 13 shows the weight loss rate of the different sorbents together with the MS signal for CO 2 and H 2 O being released by the sorbent during the temperature treatment in a gas mixture of N 2 /He. The weight loss at low temperature < 300°C can be mainly attributed to the desorption of H 2 O. For KMG30 the largest amount of released CO 2 is detected at about 568°C.
For KSORB CO 2 seems to be released mainly at 340°C together with H 2 O and at 776°C (only CO 2 ). KMG70 shows the highest CO 2 release rate at 870°C. These results support our theory that the treatment of H 2 O at elevated temperatures leads to a decomposition of carbonate species. H 2 O seems to lower the activation energy for carbonate species decomposition resulting in a decomposition at much lower temperatures than under dry conditions. The activation energy for K 2 CO 3 decomposition on KGM70 seems to be higher compared to KSORB resulting in the decomposition of these species at a higher temperature. The fact that KMG30 does not show K 2 CO 3 decomposition at higher temperature could be explained with a different incorporation of K 2 CO 3 into the structure of KMG30. It was also reported in the literature, that a very high surface area of the sorbent before impregnation can lead to a better incorporation of K 2 CO 3 into the sorbent structure [27]. Our BET measurements are in line with this hypothesis showing that KMG30 indeed has a high surface area compared to the other two materials after K 2 CO 3 impregnation.
After the treatment with steam, the CO 2 cyclic working capacity of the different materials is significantly reduced (Fig. 14). Where the cyclic working capacity for KMG30 is reduced by 50%, KMG70 has only 21% of its capacity compared to the normal pretreatment. A possible explanation could be a significant loss of surface area due to the high temperature (agglomeration of metal species), which could however be excluded for KMG70. It can be seen from Fig. 15a that the N 2 adsorption isotherms are quite similar for KMG70 before and after TGA-TPD. The surface area of KGM70 is higher after the experiment in the TGA. It has been reported that CO 2 desorption is creating "vent holes" in the surface of potassium-promoted HTC, which results in an increase in surface area [28]. Al-spinel formation after the steam treatment at higher temperatures which could explain the reduction in available surface sites for KMG70, could not be detected using XRD with a closed sample holder (to avoid exposure to humidity and CO 2 from the lab atmosphere) (see Fig. 15b). It was reported that steam aging leads to the formation of MgO and Al spinel which reduces the number of basic sites for a HTC similar to KGM70 (without K 2 CO 3 promotion) [29]. However, both recorded XRD spectra show only MgO as detectable crystalline phase of KMG70 before and after the TPD experiment. Fig. 11. Normalized weight change during pretreatment, adsorption and desorption of H 2 O for at 300 and 500°C for P H2O = 0.34 bar during adsorption and P N2 = 1 bar during regeneration a) KSORB b) KMG70. Fig. 12. a) TGA-TPD for different sorbents 1) Pretreatment at 600°C 2) Adsorption of H 2 O at 300°C and P H2O = 0.34 bar and increase of operating temperature to 850°C 3) Regeneration of sorbent at 850°C with N 2 and cooldown to 400°C 4) determination of CO 2 cyclic working capacity at 400°C and 30 min half-cycle-time b) close up of mass change setting the initial mass to 0 after the pretreatment in order to direct compare the mass change of the different materials during steps 2 and 3.
Additionally, the potassium promoted Alumina did not show a high loss in sorption capacity, where Mg-Al spinel formation can be excluded due to the absence of Mg in the material, spinel formation does not explain the obtained results during this study. Irreversible decomposition of the potassium carbonate species and possible interaction with Al and Mg under the presence of steam at high temperature resulting in a significant loss of basic sites available for cyclic CO 2 adsorption is therefore one explanation for the obtained results.
It is very important regarding the use of potassium-promoted sorbent for SEWGS and other applications at elevated temperatures, that the temperature for decomposition of potassium carbonate can be significantly reduced by the presence of steam and this should be avoided in order to prohibit the deactivation of the sorbent. This shows the importance of not only studying the CO 2 adsorption capacity at ideal conditions, but to also consider conditions outside the direct process window. Considering the findings during this study, KMG30 was found to be the best candidate for SEWGS applications regarding the stability of the sorbent and the relatively high cyclic working capacity for CO 2 .

Conclusions
The CO 2 and H 2 O cyclic working capacity was determined for three sorbents with different material compositions for their use in sorptionenhanced water-gas shift processes at a temperature range between 300 and 500°C. The potassium-promoted hydrotalcite based adsorbent with a high MgO content shows a higher cyclic working capacity for CO 2 at different operating temperatures. More and stronger bulk basic sites are responsible for this increase in cyclic working capacity. For all sorbents the desorption step is limiting the cyclic working capacity due to slower desorption kinetics, which determines the amount of CO 2 that can be adsorbed in the subsequent adsorption step. HTC-based adsorbents exhibited different adsorption kinetics compared to potassium-promoted alumina. After the first fast initial adsorption of CO 2 a sorbent with a higher MgO content adsorbs a larger amount of CO 2 slowly which could be caused by slow formation of bulk carbonates. During a desorption step with N 2 the additionally adsorbed CO 2 can be desorbed relatively fast compared to the slow adsorption. Probably two different independent surface reactions are responsible for this behavior caused by MgO in the sorbent structure. The adsorption and desorption kinetics of H 2 O are similar for all sorbents independent of their material composition. A combination of a high surface area and the strength of basic sites could explain that the HTC-based materials show the highest cyclic working capacity for H 2 O at three different operating conditions. The presence of steam at high operating temperature can lead to the irreversible decomposition of surface carbonates resulting in a strong decrease in CO 2 cyclic working capacity of the sorbent. It was found that steam seems to be able to reduce the temperature when the Fig. 13. a) Weight loss rate of different adsorbents as function of temperature (heating rate 20°C/min in 50% N 2 and 50% He b) MS signal of CO 2 and H 2 O for the different experiments as function of operating temperature.
Fig. 14. CO 2 cyclic working capacity after steam treatment at 850°C (TPD) compared to CO 2 cyclic working capacity for sorbents pretreated at 600°C for 2 h in N 2 (usual procedure). The relative reduction in CO 2 cyclic working capacity after steam treatment is indicated in percentages. decomposition of these carbonates occurs. For the use of these sorbents in sorption-enhanced water-gas shift processes this observation is of major importance regarding the selection of the sorbent and optimization of the process conditions. The potassium-promoted HTC KMG30 thus seems to be the best candidate for sorption-enhanced water-gas shift processes.