Spray‐Dried Sodium Zirconate: A Rapid Absorption Powder for CO2 Capture with Enhanced Cyclic Stability

Abstract Improved powders for capturing CO2 at high temperatures are required for H2 production using sorption‐enhanced steam reforming. Here, we examine the relationship between particle structure and carbonation rate for two types of Na2ZrO3 powders. Hollow spray‐dried microgranules with a wall thickness of 100–300 nm corresponding to the dimensions of the primary acetate‐derived particles gave about 75 wt % theoretical CO2 conversion after a process‐relevant 5 min exposure to 15 vol % CO2. A conventional powder prepared by solid‐state reaction carbonated more slowly, achieving only 50 % conversion owing to a greater proportion of the reaction requiring bulk diffusion through the densely agglomerated particles. The hollow granular structure of the spray‐dried powder was retained postcarbonation but chemical segregation resulted in islands of an amorphous Na‐rich phase (Na2CO3) within a crystalline ZrO2 particle matrix. Despite this phase separation, the reverse reaction to re‐form Na2ZrO3 could be achieved by heating each powder to 900 °C in N2 (no dwell time). This resulted in a very stable multicycle performance in 40 cycle tests using thermogravimetric analysis for both powders. Kinetic analysis of thermogravimetric data showed the carbonation process fits an Avrami–Erofeyev 2 D nucleation and nuclei growth model, consistent with microstructural evidence of a surface‐driven transformation. Thus, we demonstrate that spray drying is a viable processing route to enhance the carbon capture performance of Na2ZrO3 powder.


Introduction
Powder sorbentsf or CO 2 at high temperatures are of interest for an umber of applications,i ncluding the production of H 2 by steam reforming, in which removal of CO 2 shifts the chemical equilibrium in favor of greaterH 2 yield and purity.S orptionenhanced steam reforming (SESR) based on aC aO sorbent (CaO (s) + CO 2(g) QCaCO 3(s) )h as been demonstrated at the research level. [1] Calcium oxide sorbents work best at approximately 600-700 8C, andh ence, coupled to steam reforming reactions;t he sorbent may be regenerated by calcination in air at around 800 8Co ra bove.T his type of calcium looping technology has been considered widely for post-combustion capture (PCC) from fossil-fuel-fired power plants (notably,c oalfired) and other single-point industrial emitters. [2] The technolo-gy could be implemented using two parallel, fluidized beds operating as carbonator and regenerator,o ru sing fixed-bed reactors with alternating carbonation/calcination reactions by feed-flow control. [2] For the proposed implementation in PCC, the decarbonation step would be performed in an ear pure CO 2 stream, necessitating calcination temperatures ! 950 8C. [2] In SESR applications,i nw hich oxygen looping is employed to exchangeo xygen with the metal catalyst, air (oxygen depleted) would be the sorbent regeneration stream at temperatures ! 800 8C. [1] An acceptable sorbent for PCC or SESR should have ah igh CO 2 -uptakec apacity per unit mass and remain closetoits original CO 2 -capture capacity over repeated carbonation/regeneration cycles. [3] Materialc osts should be low and the sorbent should be mechanically robust, as in the case of calcium oxide (CaO) and other inorganic oxides. CaO from limestone is the most inexpensive and readily availableo ptions. CaO however shows serious loss of CO 2 capacity after repeated calcination cycles at 800 8Co wing to the effects of partial sintering and loss of surface area and porosity. [4] An umber of additive powders (e.g.,S iO 2 ,A l 2 O 3 ,Z rO 2 )h ave been investigated as meansofimproving the multicyclestability of CO 2 -capture performance of the active CaOc omponent. [3] The greater the volume fraction of the refractory additive the more durable the sorbent, but there is at rade-off in the dilution of the active component that leads to loss of initial capture capacity:2 0-30 wt %i sacommon compromise loading. The added oxide component often reacts with CaO to form ab inary compound, and it is this compound, for example, Improved powders for capturing CO 2 at high temperatures are required for H 2 production using sorption-enhanced steam reforming.H ere, we examine the relationship between particle structurea nd carbonation rate for two types of Na 2 ZrO 3 powders. Hollow spray-dried microgranules with aw all thickness of 100-300nmc orrespondingt ot he dimensions of the primary acetate-derived particles gave about 75 wt %t heoretical CO 2 conversion after ap rocess-relevant 5min exposure to 15 vol % CO 2 .Ac onventional powder prepared by solid-state reaction carbonated more slowly,achievingo nly 50 %c onversiono wing to ag reater proportion of the reactionr equiring bulk diffusion through the denselyagglomerated particles. The hollow granular structure of the spray-dried powder wasr etained post-carbonation but chemical segregation resulted in islands of an amorphous Na-rich phase (Na 2 CO 3 )w ithin ac rystalline ZrO 2 particlem atrix. Despite this phase separation, the reverser eaction to re-form Na 2 ZrO 3 could be achieved by heatinge ach powdert o9 00 8Ci nN 2 (no dwell time). This resulted in av ery stable multicycle performance in 40 cycle tests using thermogravimetric analysis for both powders. Kinetic analysis of thermogravimetric data showed the carbonation process fits an Avrami-Erofeyev 2D nucleation and nucleig rowth model,c onsistent with microstructural evidence of as urface-driven transformation.T hus, we demonstrate that spray drying is av iable processing route to enhance the carbon capture performance of Na 2 ZrO 3 powder. Ca 12 Al 14 O 33 (mayenite), which acts as the "refractory spacer" second phase designed to inhibitC aO particle sintering and densification. [3g-k] Au niform distribution of the second phase is essential to minimized ensification of the active CaO phase and suppress multicycle degradation.T he performance of ar ange of CaO-based sorbents is summarized in the study of Zhao et al. [3m] The more complex( and costly)t he processing technique, for example, sol-gel or chemical templating, the finer the particle size, and the more uniform the dispersion. Consequently,s olution-derived composite powders generally have the bestm ulticycle performance relative to the base CaO sorbentmaterial.
Another approach to avoid multicyclep owder densification problemsh as been to use alternative sorbent materials to CaO, such as Li 4 SiO 4 and Li 2 ZrO 3 . [5] The latter has received considerable attention for both post-and pre-combustion capture, and for SESR applications. Li 2 ZrO 3 absorbs CO 2 according to the reversible reaction: Li 2 ZrO 3(s) + CO 2(g) !Li 2 CO 3(s) + ZrO 2(s) (giving am aximum increase in sorbent mass of 28 %). It also acts as ab asic catalyst that has the advantage of promoting tar degradation in SESR processes. However, its utilizationh as been inhibited by poor reaction kinetics at low CO 2 partial pressures (< 0.02 MPa)a nd high temperatures (> 500 8C). The more active, metastable, tetragonal crystal structure-the major contributor to CO 2 chemisorption-is potentially transformed to al ess reactive monoclinic form during high-temperature cycling. The Li 2 ZrO 3 -based sorbents are best suited to processes operating at temperatures < 550 8Cs uch as steam reformingo fs imple compounds such as methane, ethanol, or glycerol. Solid solutionso fL i 2 ZrO 3 with Na 2 ZrO 3 have also received attention. [6] There are also reports on the use of Na 2 ZrO 3 and K 2 ZrO 3 as CO 2 sorbents. [7] From thermodynamic considerations, Na 2 ZrO 3 and K 2 ZrO 3 absorb CO 2 at lower CO 2 partial pressures and higher temperatures than Li 2 ZrO 3 .H owever K 2 ZrO 3 sorbents are more difficult to regenerate. To reach ag ood balance between ease of capture and regenerationa th igh temperatures ( % 650-750 8C), Na 2 ZrO 3 is more promising than either Li 2 ZrO 3 or K 2 ZrO 3 .
To reduce the particle size of Na 2 ZrO 3 sorbents,anumber of solution-baseds ynthesis routes were developed. [7a, 8] These result in faster carbonation rates since ag reater proportion of the CO 2 uptakeo ccurs through interfacial solid-gas reactions, and the diffusionl engths for ion migration in the later stages of the reaction (in which the rate of mass transfer is controlled by SS diffusion) are reduced.
Herein,weuse scanning and transmissione lectron microscopy with energy-dispersive analysiso fX -rays (EDX) to investigate the microstructural differences between Na 2 ZrO 3 particles produced by spray drying am ixed acetate solution,a nd pow-ders prepared by conventionalS Sr eaction. The structural differences we identify account for much faster rates of carbonation in spray-dried (SD) forms.ACO 2 conversion of approximately 0.18 g CO 2 g À1 sorbent ( % 75 %o ft heoretical capacity) is demonstrated for the SD powder after only 5min exposure to 15 vol %C O 2 at 700 8C, namely,u nder carbonation conditions pertinent to SESR. Stable multicycle performance is demonstratedf or both powder types over a4 0cycle thermogravimetric testing program (decarbonation at 900 8C) but because of the slower rate of carbonation for the conventionally prepared Na 2 ZrO 3 ,i ts conversion is only about 50 %o ft he theoretical capacity under these conditions (which are relevant to implementation in SESR). Finally,w el ink our microstructural observations to kinetic modelling of the CO 2 -absorption profiles measured during carbonation to gain mechanistic insights into the surface-driven absorption process.

Results and Discussion
Phase analysis and particle structure:a s-preparedpowders X-ray diffraction (XRD) patterns confirmedt hat both SD and solid-state (SS) powders contained crystalline Na 2 ZrO 3 ,i nt he form of hexagonal and monoclinic polymorphs.F igure 1p resents the XRD pattern fort he SD powder.M inor peaks of ZrO 2 (monoclinic) and very weak peaks of Na 2 CO 3 wered etected, consistentw ith residual intermediate phases from the following reaction [Reaction (R1), only inorganic products are represented]: 2NaðCH 3 COOÞþZrðCH 3 COOÞ 4 ! Na 2 CO 3 þ ZrO 2 ! Na 2 ZrO 3 þ CO 2 ðR1Þ The very weak Na 2 CO 3 peaks relative to the XRD peaks for ZrO 2 are consistentw ith the former being poorly crystallized. The conventionalS Sp owder gave similar diffraction patterns to the SD material ( Figure S1 in the Supporting Information). SEM (Figure2)r evealed the SD powders to be hollow,p erforated, andp artially collapsed spherical granules. These ranged in size from 1-10 mm ( Figure 2a). The walls of the granules were composed of interlocking primary particles (100-300 nm in size, Figure 2b)a nd were as inglep article in thickness (the 100-300nmw all thickness is illustrated in Figure S2 in the Supporting Information). We have observed similarp article structures previously,f or example, in ZrO 2 granules that were spray dried from acetate solution. [9] This type of structure is consistentw ith af ormation mechanism in which liquid atomized droplets, upon enteringt he heatedc hamber of the spray dryer,f irst develop as olid, pliables urfaces kino fs alt particles that surroundsaliquid core. After continued heating, pressure builds up and is released by bursting of the outer solid skin, resulting in characteristics urface rupturing of the hollow granule. If the outer skin remainsp liable at this stage,t he walls collapse to create deformed, hollow spheres. The expelled liquid from the interior of the droplet forms as econdary aerosol, which results in as eries of smaller granules. Aschematic of the proposed SD granule formationm echanism is shown in Figure 3.
The SS powders were composed of denselya gglomerated granules, tenths of mmi ns ize, typical of ac onventionally prepared mixed-oxide ceramic powder;p rimary particle size was approximately 0.05-1 mm (Figure 2c,d).
Carbonation characteristics and effect on particle structure To assess the baselineC O 2 -uptakep erformance of the SD and SS powders, the response to prolonged exposure to 15 %C O 2 at 700 8Cw as analyzed ( Figure 4). The SD powders reached as teady-state increase in mass after about 10 min, equivalent to 0.20 g CO 2 g À1 sorbent uptake and am olar conversion of approxi-mately8 5% of theoretical capacity.A fter 5min, the uptake was about 0.18 g CO 2 g À1 sorbent .The SS powder approached asimilar steady-state level of carbonation but required ad well period of almost2 5min as opposed to only 10 min for the SD powder, indicating am uch slower rate of carbonation in the conventional SS powder.
The XRD patterns of both powders collected after the end of the isothermal thermogravimetric analysis (TGA) experiment were similar. The pattern for the SD powder is shown in Figure 5, indicating am ixture of Na 2 CO 3 and ZrO 2 ,w ith no evidence of unreacted Na 2 ZrO 3 ( Figure 5). The carbonated SS powderp attern is shown in the Supporting Information. This confirms the carbonation reactiono ft he Na 2 ZrO 3 phase contained in the calcined starting powder [Reaction (R2)] had reachedc ompletion (subjecttoX RD detection limits).
The SEM images of the powders produced after 25 min isothermalc arbonation revealed the carbonated SD granules retained the general structure of the as-prepared material  Figure 2c). The surface of the carbonated SD granules revealed localizedp ockets with as mooth, glass-like appearance.
Close inspection indicated that as imilar phase was also interspersed within the interlocking submicron particles that made up the remainder of the granule surface ( Figure 6a). The SEM/EDX elementalm aps indicated the smooth regions to be Na-rich (Figure 6b), and therefore, we attribute these to be   Na 2 CO 3 ,w hich under the carbonation conditions employed had softened and flowed into isolated islands.T he remainder of the carbonated-granule structure was made up of interlocking faceted particles that were Zr-rich ( Figure 6b)-these are assumed to be the ZrO 2 phase identified by XRD.T here was also some localized glass-like phase interspersed within the (crystalline) ZrO 2 particles. The SS agglomerates showed similar evidenceo faglass-like Na-rich phase surrounding Zr-rich particulatematerial (Figure 6d).

Multicycle carbonation/decarbonation performance
The multicycle performance of the SD powder and the conventional powder over 40 TGA carbonation/decarbonation cycles, is summarized in Figure 7( multicycle TGA plots are shown in the Supporting Information). An increase in the level of CO 2 uptake was observed over the first 3cycles for each powder; this type of self-activation has been observed for other oxide sorbentp owders, for example, CaO, and can be attributedt o the generation of porosity in the powder owing to outgassing in the first few decarbonation cycles. [10] After the initial selfactivation period, the uptake capacity of both the SD and SS powders showed ar emarkable stability,i ndicating high durability to be an intrinsic feature of Na 2 ZrO 3 sorbents (as discussed below). The variationi nm ass conversion of the SD powderw as < 5% between cycles number 3a nd 40. The CO 2uptake level was approximately 0.18 g CO 2 g À1 sorbent (4.1 mmol g À1 ) in cycle 4c orresponding to am olarc onversion efficiency of about 75 %. Because of the slower rate of carbonation of the SS powder (as identified in Figure 4) the level of uptake after the set 5min carbonation within multicycle experimentsw as lower,0 .12 g CO 2 g À1 sorbent (2.7 mmol g À1 )o ra bout5 0% conversion by mass under multicycle conditions. SEM micrographs showed the particle structure of decarbonated SD and SS powders after 10 and 30 TGA cycles, indicating am ore porous structure (Figure 8a,b)t han for the as-    (Figure 2). This is consistent with reports for other oxide sorbents for which an initial increase in porosity owing to self-activation associated with the first few carbonation/decarbonation cycles is shown. [10,11] The cycled SS powders were also more porous than the as-prepared SS samples (Figure 8c,d).

TEM of spray-dried powder
Analysis by TEM of the carbonated SD powder after one TGA cycle and dispersion in heptanei sshown in Figure 9a.O nly fragments of the granules could be imaged as full-size granules are not electron transparent. Thef ragments hows ap olycrystalline substructure (top right image in Figure 9a Or ich whereas EDX spectra of the glassy regions (black) are Na and Cr ich, consistent with Na 2 CO 3 (the background Cu signal is from the support grid). These findings are in agreement with the information inferred by SEM/EDX of full-size SD granulesi maged following extendedc arbonation experiments ( Figure 6) and confirmt hat the walls of the hollow granules are composed of an etwork of interlocking submicrometer, crystalline ZrO 2 particles with regions of partially glassy Na 2 CO 3 phase interspersed between them (only partially glassy because XRD identifies am inor amount of crystalline Na 2 CO 3 ).
To revealm ore information on the spatial distribution of the component phases, two other TEM samples were prepared: as ample collected after one TGA cycle, the other after 20 cycles.This time powdersw ere dispersed in acetonei nstead of heptane. Acetone is ap olar solvent in which Na 2 CO 3 and any hydroxyl-carbonate phases that may form upon storage in air,o ro ne xposure to moisture presenti nd ispersant liquids (e.g.,b icarbonate), are soluble and leach out of the granule fragments. TEM showedt hat the acetone-dried samples were indeed more porous (Figure 9b), suggesting that the soluble (Na 2 CO 3 )m aterial hado riginally been located between the ZrO 2 nanoparticle networks,c orroborating the interpretations of SEM images (which showed glassy materiala mongst ZrO 2 particlesi na ddition to segregated pocketso fN a 2 CO 3 ). In some areas of the 20 cycle image, the leachedc arbonate phase has re-precipitated in an acicular morphology.

Carbonation reaction: kinetic analysis
As et of isothermal TGA carbonation experiments wered esigned to identify the reaction model that best describes the carbonation process of the SD and SS powders and to derive apparent kinetic parameters.
(1)] was calculated by findingt he minimum and maximum measured TGA masses over the cycle step considered (a cycle consisting of carbonation followed by calcination). For carbonation, the minimum mass is the initial mass at t = 0( m 0 ), whereas the maximum is the final mass at t = t f (m f ).  Conversion versust ime data (a vs. t)c an then be represented using several models of SS (gas) reactions. Hancock and Sharp'sm ethod [12] assigns am odel or af amily of models according to the value of m,asd efined in Equation (2): in which B is ac onstant,t he conversion values (a)r ange typically between 0a nd 0.5, and plottingl n[Àln(1Àa)] versusl nt produces as traight line fit with gradient m. Figure 10 as hows the linear fit for the SD and SS powders carbonating at 700 8Cw ith bestfit values of m and ln B.
According to Hancock andS harp, [12] the SD and SS powders exhibited m values of 1.86 and 1.69 respectively,b oth corresponding to Avrami-Erofeyev( also knowna sJ MAEK) models close to m = 2, termed A2 models. Avrami-Erofeyev AN models, with values of N ! 1, are known as "nucleation and nuclei growth models". In the case of the carbonation of the SD and SS Na 2 ZrO 3 crystals, with fitted valueso fm of 1.9 and 1.7, both close to N = 2, disc-like are the most likely nuclei shapes.
Further confirmation of the Avrami-Erofeyev model being identified as best fitting the SD and SS Na 2 ZrO 3 carbonation reactions is found using the method described by Khawam and Flanagan. [13] In this method, the shape of the plot da/dt versus a is used to determine the mostl ikely reaction model,w ith Avrami-Erofeyev displaying au nique dome-like profile with the apex located at a values between 0.3 and 0.4 for model A2. Figure 10 bs hows that the carbonation of both the SD and SS Na 2 ZrO 3 powders exhibited dome shapes with apices between 0.3 and 0.4, corresponding roughlyt othe A2 model.
Reaction kinetics of the Avrami-Erofeyev AN models can be described by the equation relatingt he integral-conversion function(g(a)) to the reactiont ime followingEquation (3): in which the rate constant k typically follows Arrhenius' law [Eq. (4)]: in which A is the pre-exponential factor, E is the activation energy, R is the universal gas constant, and T is the temperature in K.
Here, the carbonation having been performed at 700 8C, one value of k waso btained fore ach of the materials tested (SD and SS Na 2 ZrO 3 powders).
Inverting Equations (1) and (3) allows the calculation of amodeled value of mass increase (in %) as functionoft ime accordingt oE quations (5) and (6): With a ¼ 1 À exp À kt ðÞ N ðÞ ½ ð6Þ Figure 11 compares the experimentally obtained %m ass increases versus time of the SD andS SN a 2 ZrO 3 powders during carbonation at 700 8Cw ith their modeled counterpart using Equations (5) and (6), and providesafinal test of the suitability of the chosen modelsw ith their derived kinetic rates. It can be seen that an excellent match between experimental and modeled mass increases was obtained for both materials.
Both modelling methods indicate 2D nucleation and nuclei growth for the carbonate phase, which when combined with the SEM observations (Figures 2a nd 6) suggestasurfacedriven transformation of the Na 2 ZrO 3 granules, consistentw ith ap orous Na 2 CO 3 and ZrO 2 surfacel ayer discussed in another Na 2 ZrO 3 study. [7d] In summary,S DN a 2 ZrO 3 granules exhibit rapid CO 2 uptake reaching0 .18 g CO 2 g À1 sorbent within only 5min (15 %C O 2 at 700 8C), some 50 %g reater conversion within this process-relevant time period than the conventionally preparedS Sp owder. Both powder types are highly durable, showingm inimal decay (< 5%)i nu ptake capacity after the 40 cycles testu nder conditions relevant to steam reforming.
Thus, we demonstrate the intrinsically superior durability of Na 2 ZrO 3 ,a nd that the rate of carbonation may be improved through simple spray drying, which is an industriallys calable process that providesafine primaryp article size within Figure 10. Model identification for carbonationc onversion factors of SD and SS powders using a) Hancock and Sharp method, [12] indicating linear fit with gradient m % 2 [Eq. (2)] corresponding to Avrami-Erofeyev A2 model; b) Khawam and Flannagan method, [13] indicating dome-shape of da/dt vs. a. ap orousg ranular structure. Confirmation of ah igher surface area in the SD powders, as suggested by the SEM images, was obtained from N 2 -adsorption isotherms ( Figure 12). The BET (Brunauer-Emmett-Teller) surface areas of the SD powder were % 20 m 2 g À1 compared to only % 2m 2 g À1 for the SS powder.
Hysteresis in the isotherms indicates mesoporosity.P ore volumes measured by the Barrett-Joyner-Halenda( BJH) method were % 0.039 cm 3 g À1 for the SD powder and 0.007 cm 3 g À1 for the SS powder.T his differencei sc onsistentw ith SEM observations of hollow-perforated microgranules in SD powders, and dense agglomerates in SS powders. The hollow andp erforated microstructure of the SD granulesp rovides easy access of CO 2 to the inner and outer surfaces of the granule walls. This, allied to the thin wall dimensions,r esults in ah igherp roportion of the carbonation processi nvolving ar apid gas-solid reaction (the linear segment of the TGA profile) than is the case for the denselya gglomeratedS SNa 2 ZrO 3 powder.
Crystalline Na 2 ZrO 3 naturally possesses lattice-scale intimate mixing of refractory ZrO 2 and active Na 2 Oc onstituents. The crystal structure of the monoclinicf orm is representedi n Figure 13. The lattice-scale distributionso fe ach component represents ideal mixing of ac omposite metal-oxide sorbent material suited to high-temperature operation, and account for the remarkable durability of Na 2 CO 3 in multicycle operation. This scale of mixing cannotb ea chieved by mechanical mixing or chemical precipitation of two-phase sorbent and refractory spacer powders.
During carbonation, the Na 2 ZrO 3 crystal lattice decomposes in as urface-driven process to at ruly nanoscale composite of ZrO 2 and Na 2 CO 3 .F rom SEM and TEM examination of the walls of the hollow SD granules, ap oorly crystallized/glassy carbonate phase segregates. The reverser eaction to regenerate crystalline Na 2 ZrO 3 occurs readily during the temperature and gasswing decarbonation step, once again creatingasorbentw ith ideal crystal lattice scale distributions of "active" and "spacer" components ready for the next carbonation step. The net result is av ery durable single-phase high-temperature sorbent.
As mentioned in the Introduction, there is aw ide literature on other high-temperature powder sorbents for CO 2 capture, most notably for CaO powders in which refractory additives are introduced, for example, ZrO 2 [3m, 14] to suppress the natural densification (partial sintering) and loss of porosity that degrades cycleo nc ycle the CaO performance, as outlined in the Introduction. Often, very complex chemical solutionp recipitation reactions are employed to promote adequate mixing of the two components. In 2012, we proposed at emperature-induced volume-expanding phase change additive to disrupt densification, [15] others later adopted this concept. [16] However, we found that the volume expansion that occurredb etween regeneration and carbonation wasaccommodated in the residual pore spacesa nd did not induce microcracking to openu p porosityp rior to the next carbonation step. All of these second-phase additives to as orbentp owder require complicated processing to achieves ignificant improvements in durability as performance is only improved if there is intimate mixing of "refractory" additive ands orbent. Even the best chemicalo rm echanical synthetic processes only give mixing of the two particle types on the submicron scale.
There are anumber of literature reports of Na 2 ZrO 3 as asorbent for CO 2 :t he conditions used for sorption/desorptionv ary   between the different publications. Martínez-dlCruz and Pfeiffer [7d, e] prepared Na 2 ZrO 3 by as imilarS Sr oute to our SS powderb ut with calcination at 850 8Cf or 6h and found that addition of 20 %e xcess Na 2 ZrO 3 produced ap hase-pure product (by XRD). The surface area of this product was approximately 3m 2 g À1 ,c omparable to the surface area of the SS powderp resented herein. Their 20 cycle sorption/desorption studies were conducted in 100 %C O 2 (as opposed to 15 % herein): temperatures between 550 and 700 8Cw eref ound to give the highest uptakes;d esorption in N 2 was conducted at 800 8C. Sorption-dwell times of 30 minutes were adopted, the samples exhibited CO 2 uptakes corresponding to 18.5-19 mass %. [7d, e] Our SS powder exhibited similar uptake after similar total time periods to these reports (Figure 4) but we adopted as horter (5 minute) carbonation period in multicycle TGA as this replicated more closely the conditions of aw orking sorbent. The same group studied the microstructure of their SS powders and concluded that am esoporouss tructure was formed on the surface of the agglomerates at sorption temperatures of 300-550 8C, but sintering of this shell layer at temperatures above 550 8Ce liminated the porosity and at that stage sorptionk inetics were controlled by diffusion processes throughadense Na 2 CO 3 + ZrO 2 shell. [7d] This is consistentw ith our TEM analysis. The effect of relative humidity on the carbonation and decarbonation processes at low temperatures (30-80 8C) for powders produced by SS reactions hows that high humidity has ap ositive effect, whichw as attributed to bicarbonateformationa tt he surface. [17] Several solution routes have been used to produce Na 2 ZrO 3 sorbentp owders. This includes simple evaporation of sodium acetate and zirconium acetyl acetonate in ethanol and uptakes of CO 2 of about 21 wt %b yT GA over four cycles were recorded involving sorptioni n8 0% CO 2 at 600 8C( for > 100 minutes) and regeneration in argon at 800 8C. [7a, 8] Spray drying of these precursor solutionsw as also investigated. [8] Unlike the SD granules of the present work, their spherical granules disintegrated on calcination to produce an anosized powder of similar particle sizes ( % 50 nm) to the powders produced by simple evaporation drying. Hence, both SD and simple evaporation powders within the study of Zhao et al. [8] exhibited similarC O 2 -capture properties, achieving around1 7.5 wt %m ass increase after 200 si n1 00 %C O 2 at 575 8C. Multicyclep erformancei n5 0% CO 2 up to 11 cycles indicated an uptake of almost 15 wt %. [8] Sodium oxalate and zirconium nitrate, sodiumc itrate and zirconyly nitrate aqueous solutions as well as sodium acetate and zirconyl chloride solutions have been used to produce Na 2 ZrO 3 in an evaporation/drying/calcinationp rocess reported by Ji et al. [18] and Memon et al. [19] The CO 2 -capture kinetics of our SD powdersc ompare favorably to other Na 2 ZrO 3 sorbentp owders, although performance comparisons between different laboratories is complicated by the variability in sorption andd esorption conditions employed. We demonstrate distinctive microstructuralf eaturest hat lead to high surfacea reas, whichexplain the reasonsfor the characteristic rapid rates of carbonation. The direct like-for-like comparison to SS tested under identicalT GA conditions provides an unequivocal demonstration of the superior per-formance of SD. For comparisons with other alkaline metal or alkaline-earth ceramic sorbents, the reader is directed to acomprehensive review article by Memon et al. [19]

Conclusions
The microstructural reasonsf or fasterr ates of CO 2 capture by spray-dried (SD) granules of Na 2 ZrO 3 relative to ap owder prepared by conventionals olid-state (SS) synthesis method have been established using ac ombination of scanning and transmission electron microscopy,s urface area measurements, and kinetic modeling. The hollow and perforated granulars tructure of SD powders presents ah igher surface area than the densely agglomerated conventional powder and promotes the surfacedriven carbonation reaction. This permitted about 75 %o ft heoreticalm ass conversion within 5min exposure to 15 %C O 2 at 700 8C, compared to only around5 0% for the benchmark conventional SS powder.A lthough segregation of Na 2 CO 3 and ZrO 2 occurs duringc arbonation, crystalline Na 2 ZrO 3 is reformed by heating to 900 8Ca nd immediately cooling, ready for the next carbonation step in am ulticycle sorption/desorptionp rocess. High multicycle durability is an intrinsic feature of Na 2 ZrO 3 as the active soda component is held within as table crystal structure. This contrasts to alternative high-temperature sorbentss uch as CaO-based materials in which sintering degrades durability.

Experimental Section
SD powders were prepared from as tarting solution produced by dissolving Na(CH 3 COO)·3 H 2 O) (50 mmol) and Zr(CH 3 COO) 2 (25 mmol) in dilute nitric acid (300 mL) (Sigma-Aldrich reagents) to form ac lear solution. This solution was spray dried using ab enchtop spray dryer (SD-05 Lab-Plant, UK). The operation conditions were:inlet temperature 200 8C, aspirating air flow at 40 m 3 h À1 ,p eristaltic pump speed 0.6 dm 3 h À1 ,a nd compressor pressure of 0.18 MPa. Collected powders were calcined in ab ox furnace at 900 8Cf or 2h to promote formation of Na 2 ZrO 3. The conventional SS powder was prepared by ball-milling Na 2 CO 3 (Acros Organics) and ZrO 2 (Dynamic Ceramics) powders for 16 h, followed by calcination at 900 8Cf or 2h.N itrogen adsorption/desorption isotherms were measured using aQ uantachrome Instruments Nova 2200: surface areas were measured by the BET method and pore volumes by the BJH method. Samples were outgassed under vacuum at 200 8Cf or 3hprior to analysis.
The first assessment of the carbonation characteristics of the SD and SS powders involved isothermal TGA in which a % 15 mg sample was exposed to CO 2 at 700 8C( Mettler To ledo star 1T GA/ DSC). The sample was first heated to 900 8C( 20 8Cmin À1 )i nN 2 to remove any traces of hydrated/carbonated surface phases formed during storage. After cooling (20 8Cmin À1 )t o7 00 8C, the gas was switched to 15 %C O 2 /85 %N 2 and held at this condition for 25 min. Multicycle performance up to 40 cycles was evaluated using 700 8C, 15 %C O 2 /5 min carbonation and regeneration (desorption) achieved by switching to N 2 and heating at 20 8Cmin À1 to 900 8Cand immediately cooling at 20 8Cmin À1 to 700 8C.
XRD data were collected using aB ruker D8 diffractometer (CuK a l = 1.5416 ). Owing to the small quantities of powders generated in the TGA experiments, the powders were deposited on as ilicon sample holder.T he resulting XRD patterns were analyzed using X'Pert HighScore Plus software (Version 3.0e). The diffraction patterns were compared to standard patterns in the ICDD PDF4 database (International Center for Diffraction Data).
The microstructures of as-prepared powders, carbonated powders, and powders after multiple carbonation/decarbonation cycles were characterized by using SEM with energy dispersive EDX elemental analysis (LEO 1530 Gemini field emission gun, FEG-SEM). All samples for SEM were sputter-coated with al ayer of platinum, % 5nm in thickness. TEM was used to analyze an SD sample after 1a nd 20 successive TGA cycles, ending on ac arbonation step (Philips CM200 Field emission gun TEM/STEM with Supertwin Objective lens, and an Oxford Instruments SD 80 mm 2 X-max EDX system running INCA software). Powders were prepared for TEM by dispersing in either acetone or heptane (as detailed) and drop-casting onto standard holey carbon films supported on copper grids (Agar Scientific Ltd).