Synthesis of a nanosilica supported CO2 sorbent in a fluidized bed reactor

Synthesis of a nanosilica supported CO2sorbent by Atomic Layer Deposition in a non-assisted Fluidized Bed Reactor at atmospheric pressure C. Soria–Hoyoa,∗ , J. M. Valverdea, J. R. van Ommenb, P. E. S´anchez–Jim´enezc, L. A. P´erez–Maquedac, M. J. Sayag¨u´esc aFacultad de F´ısica, Universidad de Sevilla, Avda. Reina Mercedes s/n, 41012 Sevilla, SPAIN bDepartment of Chemical Engineering, Delft University of Technology, Product and Process Engineering, Julianalaan 136, 2628 BL Delft, The Netherlands cInstituto de Ciencia de Materiales (CSIC - Universidad de Sevilla), Americo Vespucio 49, 41092 Sevilla, SPAIN Abstract CaO has been deposited on a nanosilica powder matrix by a procedure based on Atomic Layer Deposition (ALD) in a Fluidized Bed Reactor at atmospheric pressure following a potentially scalable process. In previous works ALD in gas fluidized bed has been mostly performed under reduced pressure, which hampers scaling-up the production technology. The material synthesized in the present work is tested as CO2solid sorbent at Calcium Looping conditions. Multicyclic Thermogravimetric analysis (TGA) shows that the nanosilica support stabilizes the capture capacity of CaO. EDX–STEM analysis illustrates the presence of Ca well distributed on the surface of the SiO2nanoparticles. Keywords: Atomic Layer Deposition, Nanosilica, CO2sorbent, atmospheric pressure, Calcium looping, Fluidization. 1. Introduction Calcium-looping (CaL) is a technology based on the multicyclic carbonation of CaO at high temperature and calcination to regenerate the sorbent. The CaL technology is being currently investigated for relevant applications in the energy sector such as postcombustion CO2capture[1] and thermochemical storage of5 concentrated solar power (CSP) using as CaO precursor the widely available and low cost natural limestone (CaCO3)[2, 3]. It is well known that CO2chemisorption on CaO derived from natural limestone proceeds through two different phases [4]. A first kinetically–controlled ∗Corresponding author. Tel: +34 954559898, fax: +34 954239434 Email address: [email protected] (C. Soria–Hoyo) Preprint submitted to Applied Surface Science December 10, 2014 fast reaction phase takes place at the surface of the particle. After a shallow10 layer (of thickness around 50 nm) is built up on the surface of the particles the reaction continues by a much slower process of diffussion of CO2into the bulk of the solid[5]. Only the fast reaction phase is thought to take place in postcombustion capture at practice since the concentration of CO2in the flue gases is low and the velocity of the flow is high (on the order of ms−1) lead-15 ing to short gas–solid residence times. A critical issue of the CaL technology is that the sorbent capacity decays progressively as the number of carb/calc cycles is increased. The drop of capture capacity is particularly marked at postcombustion CO2capture conditions[1]. Sorbent deactivation is attributed to the reduction of active surface area by pore sintering at the tipically high20 calcination temperatures (above 900oC) and high CO2partial pressures in the calciner environment [1]. Silica nanoparticles (NPs) have been coated with CaO in a previous work by capillary impregnation with an aqueous solution of calcium nitrate [6]. The resulting composite showed a stable behaviour as CO2 sorbent at typical CaL conditions and higher mechanical strength, which was25 attributed to the stabilizing effect of the silica support. The present work explores the suitability of Atomic Layer Deposition (ALD) carried out in a Fluidized Bed Reactor (FBR) for coating silica NPs with CaO in order to provide it with a mechanically and thermally stable support. A further purpose of our work will be to demonstrate the feasability of ALD at30 atmospheric pressure in a non–assisted FBR, which would facilitate the scaling up of this technology with many potential uses in a wide and growing number of applications. ALD is a technique to deposit thin films of functional materials on substrates [7, 8]. It is based on sequences of self–terminating gas–solid reactions,35 allowing atomic control of the film thickness [7]. This chemical vapor deposition technique does not require line–of–sight access to the surface, achieving high conformality even on complex shapes [9]. Miniaturization of semiconductor electronic devices has been the main motivation for recent developments in ALD. Yet, it has been also applied successfully to different kinds of substrates40 [8], such as particles [10], nanomembranes and nanowires [11] and even biological templates [12]. Controlled ALD on particles was developed at the University of Colorado [13, 14, 15]. In regards to particle ALD in a FB it was firstly achieved by Wank et al. [16]. Hakim et al. [17, 18] extended45 the technique to NPs-FBR. The first successful application of ALD on particles was achieved by Ferguson et al.[19]. Earlier theoretical studies related to ALD can be found in the Eastern literature as well as an experimental work of vanadium layering on coarse silica gel particles in a fixed bed [20]. ALD has been applied to NPs in a FBR at50 pressures between 1 and 10 Torr achieving completely conformal and pinhole– free coatings [21, 10]. The interested reader may see table II in Longrie et al.[22] for a detailed review on ALD in NP–FBRs at reduced pressure. Most of ALD- FBRs currently employed carry precursors into the FBR by an inert gas carrier at low pressures (around 1 Torr). This reduced pressure yields an optimum55 2 balance between the diffusion and entrainment of reactives into the main flow of the inert gas [8, 23]. Application of ALD at atmospheric pressure would usually require long reaction times [8]. The development of efficient ALD in FBRs at atmospheric pressure would be however desirable for scaling-up the functional coating of60 nano-structured particles from laboratory to industrial applications. Beetstra et al.[24] have successfully coated LiMnO4particles (with typical size from 200 to 500 nm) with Al2O3. The precursors were tri-methyl aluminium and water, which is usually considered as close to ideal materials for the ALD process[7]. The resulting coatings were amorphous and their widths showed a dependence65 with the lattice orientations of the base particles. In a very recent work, ALD in FBRs at atmospheric pressure has been applied to manufacture a relatively large batch size (120 g) of functional aggregates of 200–500 nm LiMnO4particles coated with Al2O3. ALD was carried out in that work by using a FBR assisted by external vibration in order to improve the gas–solid contacting efficiency in70 the otherwise heterogeneously fluidized bed [25]. In the present paper we show an experimental work aimed at coating aggregates of ∼10 nm amorphous SiO2NPs with CaO in non–assisted FBRs at atmospheric pressure. Nanosilica fluidization properties have been analyzed extensively elsewhere[26]. In contrast with the heterogeneous fluidization be-75 havior exhibited by most nanopowders, nanosilica can be fluidized uniformly by a gas yielding an optimum gas-solid contacting efficency. As will be shown, this allows carrying out ALD the coating in the FBR without the need of external assistance methods, which would be necessary otherwise in order to avoid the likely formation of gas channels and macroscopic bubbles serving as a bypass for80 the gas flow and thus hampering the ALD coating process. Moreover, low-cost and high purity nanosilica may be easily produced using cheap precursors such as rice husks [27]. It must be remarked, however, that analyzing the chemistry of the ALD process in the NP-FBR is out of the scopus of the present work, which is mainly focused on demostrating the85 production of a stable CO2sorbent by a potentially scalable technique. Since the self–terminating character of the reactions on the surface of particles will not be assessed it can not be strictly assured that a purely ALD process is taking place. Nevertheless, the self– terminating character of ALD reactions on particles in a FBR has90 been already demonstrated by King et al.[21] by means of in-situ mass–spectrometry. 2. Materials and methods Two different kinds of nanostructured amorphous silica (commercially available from Evonik) have been used for this study. AerosilrR974 is an hydropho-95 bic nanosilica powder with average BET area of 150 m2g−1[28]. Aerosilr300 has also been employed, which is hydrophilic and has an average BET area of 300 m2g−1[28]. The precursors used for the ALD coating process have been Ca(thd)2(Bis(2,2,6,6-tetramethyl-3,5-heptanedionato)Ca, from Sigma-Aldrich), 3 which is a β-diketonate, and O3. The reactions taking place during the ALD100 coating process would be the following[32] || − O−H + Ca(thd)2→ || − O−Ca(thd) + H(thd) (1) || − O−Ca(thd) + 20 O3→ || − Ca −OH + 22 CO2+ 19 H2O (2) where || denotes the surface of the silica NPs. The net stoichiometry for this set of reactions is thus Ca(thd)2+ 20 O3→CaO + 22 CO2+ 19 H2O.(3) At the temperature of the reactor CaO and CO2will rapidly react to CaCO3. According to the manufacturer [28], Aerosil 300 has a total density of SiOH105 groups of about 2.5 groups per nm2on its surface with around 1.5 groups per nm2isolated and the rest bridged [29]. In the manufacturing of Aerosil R974 the silanol groups have been replaced by dimethylsilyl to make the surface of the NPs hydrophobic [28]. However, a small population of sylanol groups (0.54 groups/nm2total, 0.1 isolated groups/nm2) still remains [29] on the surface.110 The area [30] of the Ca(thd)2molecule (≃0.87 nm2) imposes a lower limit to the attainable density of a monolayer of this precursor on the surface of the SiO2NPs. Taking into account that the curvature of the particles allows for some additional space, the amount of Ca(thd)2needed to form a monolayer on the surface of a SiO2NP would be around 1 molecule/nm2. This gives115 3×1020 molecules, i.e. 5×10−4mol or 0.2 g of Ca(thd)2per gram of SiO2 NPs. Assuming an ideal gas behaviour for the O3, and taking into account the stoichiometry of the reaction (eq. 3), the amount of O3needed for that purpose would be about 220 cm3. It must be taken into account that the reaction rate rfor O3decomposition depends on temperature Taccording to an Arrhenius120 type law r∝exp(−E/kT ), where E > 0 is the activation energy and Ris the ideal gas constant. Thus, at 100oC the O3half life is about 12 minutes, while at 250oC the half life drops to just 1 s [31]. Higher temperatures would lead to a very quick decomposition thus severely hampering the ALD coating process. Therefore, it is convenient to perform the ALD coating cycles at not125 too high temperature to hinder O3decomposition but sufficiently high to favor the vaporization of the solid precursor. For the solid precursor Ca(thd)2, the manufacturer reports a sublimation vapor pressure of 0.1 Torr at 205oC. Thus, a working temperature of 250oC has been chosen as a compromise to avoid O3 rapid decomposition while, at the same time, achieving a rate of vaporization130 fast enough to carry out each ALD coating cycle in a reasonably short period of time. On the other hand, the temperature of the process could influence the growth–per–cycle. However, at least for a reduced pressure (1.8 mbar), it has been shown that the growth-per-cycle of CaO on SiO2is insensitive to temperature within a range of temperatures from 200oC to 400oC [32].135 Ten ALD consecutive coating cycles have been applied on two samples of nanosilica, one of them consisting of 0.5 g of Aerosil 300 and the other one being 1.0 g for Aerosil R974 both resulting in a surface area of 150 m2. As 4 a first step, both Aerosil 300 and R974 samples have been sieved to remove large cohesive agglomerates that may be built up during transport and storage.140 In each ALD coating cycle, the solid precursor if fed from above the FBR (see figure 1) for five minutes after which the bed is purged for 2 minutes by flowing N2and subsequently subjected for 5 minutes to a controlled flow of O2(previously passed through an ozonizer) followed by other 2 minutes N2 purging. The flow rates of both N2and O2were 200cm3/min. For this value of145 the gas flow rate, the superficial gas velocity is about 2cm/s which is above the minumum fludization velocity reported for the nanosilica powders used in our study [26, 33]. The dosing and delivery of a solid precursor to the reactor is a critical issue determining in general the efficiency of the ALD coating. With this purpose, di-150 verse techniques have been proposed in the literature [34, 35]. Storing the solid precursor at a high enough temperature to attain a convenient vapor pressure and delivering the vapor to the reactor must be avoided because heating the precursor for long periods may result in its decomposition[35]. The solid precursor used in our work, Ca(thd)2, is a cohesive fine powder. In order to achive155 a uniform Ca(thd)2dosing, a mass of 1 g of this cohesive powder has been mixed for 15 minutes in a 20 mm diameter drum at 2 rpm with one thousand 2 mm diameter glass beads. In this process, the Ca(thd)2fine particles become uniformly adhered to the surface of the carrier glass beads, which allowed us to accurately select of 1 mg doses for ALD precursor direct delivery into the FBR.160 Enthaly change ∆H0 Tfor vaporization of many metal β-diketonates (as reviewed by Tsyganova et al.[36]) range from 55 to 124 kJ mol−1. Although to our knowledge there are not reported values of ∆H0 Tfor Ca based diketonates we may assume that it should be of the order of 100 kJ mol−1. On the other hand, the flow of heat to Ca(thd)2on the surface of a carrier glass bead can be165 calculated as the heat flow to a sphere in a N2flow. Expressions for the average Nusselt number NuDfor an isolated sphere submerged in a gas flow are given by Whitaker[37, 38] in terms of the Reynolds (ReD=uD/ν, where uis the superficial gas velocity, Dis the diameter of the sphere and νis the kinematic gas viscosity) and Prandtl (Pr =ν/α where αis the thermal diffusivity) numbers.170 For fluidization of Aerosil, the gas flow velocity is around 1 cm s−1resulting in Re=0.22 . For these values, the Nusselt number can be estimated[37, 38] as NuD≃2. The ratio of heat flow ˙ Qto the difference of temperature between the sphere, T, and the surrounding medium, Tm, can be thus calculated from NuD≡hD k≃2 (4) ˙ Q T−Tm ≃hπD2≃5×10−4WK−1(5) where his the heat transfer coefficient and kis the gas thermal conductivity175 (k= 41.7×10−3Wm−1K−1). On the other hand, using ∆H0 T∼100kJ mol−1, the enthalpy change for 1 mg of Ca(thd)2is 0.24 J. The difference of temperature T−Tmduring the evaporation process varies in time as the Ca(thd)2evaporates. 5 Taking into account that at 205oC there is an appreciable value of vapor pressure (0.1 Torr) the difference in temperature during vaporization would be of a few180 tens of oC, resulting in a vaporization time of the order of tens of seconds. The experimental setup used in our ALD coating experiments is sketched in figure 1. Nanosilica aggregates are suspended in the upwards fluidizing flow of gas inside a 300 mm high and 20 mm internal diameter quartz cylinder(1). A porous quartz plate fitted at the bottom of the tube serves to distribute185 uniformly the gas flow entering the bed. The fluidized bed is enclosed within a thermostated clamshell furnace (2) which keeps temperature constant at 250o C. Gases entering the fluidized bed are either pure industrial N2(99.9% purity) or O3carried in the stream of O2, depending on the phase of the ALD coating process. N2is pre-heated at 250oC as it passes through stainless steel tubes190 (3). The N2and O2flow rates are kept at 200 cm3/min by means of mass flow controllers (4, 10) (Alicat). A glass flask (5) connects the quartz tube with the gas conduction pipes. This flask is maintained at a constant temperature of 250oC by a heating mantle (6). The solid precursor is manually dosed from the top of the quartz tube through feedthrough (7) directly into the fluidized bed195 using the carrier glass beads as detailed above. The flow of O3is obtained from industrial grade O2(99.9% purity) passing through 316L stainless steel tube into a specially designed dielectric-barrier-discharge ozonizer driven by a Trek High Voltage amplifier (50Hz AC 5 kV). The ozonizer yields 15% vol. concentration of O3. Finally, the gas exits the system through exhaust (11).200 3. Results and discussion Microstructural characterization of the processed samples was performed by means of an analytical TEM-STEM microscope TECNAI G2 F30 (FEI company) with field emission gun (FEG) and point resolution 0.2 nm. The samples were prepared by dispersion of the powder in acetone and droplets of the suspen-205 sion were deposited onto a coated carbon copper grid. Energy dispersive X-ray spectroscopy (EDX detector, Oxford Instruments) nanoanalysis was performed in STEM mode with a HAADF (High Angle Annular Dark Field) detector and the analysis of the spectra were carried out using a ES Vision software (FEI Company). Images were taken at 300 kV and the micrographs were analyzed210 with the Digital Micrograph software (Gatan Inc.). Microstructural and morphological results of the Aerosil 300 processed sample are presented in figure 2. As shown in figures 2b-c the sample consisted of aggregated NPs with an average size of 12 nm ( the size distribution is shown in figure 2.a). The different contrast observed in the micrographs can be in-215 terpreted in terms of the electronic density, caused either by the difference in thickness of the sample (number of atoms) or by different atoms in different areas (number of electrons per atom). When the electron beam passes through the sample, the contrast is darker where the electron density is higher. At higher magnifications (figure 2.d-e) the micrographs show particles embedded in a ma-220 trix of larger particles. Representative results from the chemical analysis are shown in figure 3. The detected amounts of Ca and Si along the line profile 6 1 2 2 3 4 5 6 7 8 9 10 O2 N2 O3 11 Figure 1: Experimental set-up used for ALD Ca coating in a NP-FBR at atmospheric pressure. (1) Quartz tube (300 mm high, 20 mm internal diameter), (2) clamshell furnace, (3) heated steel tubing, (4) mass flow controller, (5) distributor, (6) heating mantle, (7) feedthrough for solid dosing, (8) tubing, (9) ozonizer, (10) mass flow controller and (11) exhaust. depicted in 3.a are shown in 3.c-d. The relative proportion of Ca is very small, which is consistent with the low number of ALD coating cycles to which the samples were subjected (a work under progress is being carried out to automa-225 tize the process in order to analyze the effect of increasing the number of cycles). The highest Si/Ca atomic ratio value was found to be 95/5 corresponding to the spectrum displayed in figure 3.b and the position in the image profile marked with a cross. To test the capacity of the CaO coated materials as CO2sorbent, the pro-230 cessed samples have been analyzed by thermogravimetric analysis (TGA) at CaL conditions for post-combustion capture. Samples of about 10 mg were carbonated at 650oC for five minutes in a gas mixture of 85% vol air/ 15% vol CO2. Carbonation was followed by calcination for five minutes at 950oC in 30% 7 vol air/ 70% vol CO2. Carbonation/calcination cycles were carried out using235 a Q5000IT TG analyzer (TA Instruments) provided with a infrared halogen furnace allowing for a fast heating/cooling rates (300 oC/min ) and with a high sensitivity microbalance. The thermograms obtained during the carbonation/calcination cycles are shown in figure 4. TGA tests using pure nanosilica showed no gain/loss of weight upon calcination/carbonation cycles. The results240 show that the carbonation activity of the ALD synthesized materials remains quite stable after a few cycles, in constrast with the behaviour of limestone derived CaO, which shows a gradual decline of the capture capacity specially marked in the first cycles [6]. 4. Conclusions245 Nanosilica particles have been successfully coated with CaO by Atomic layer deposition (ALD) in a non–asisted fluidized bed reactor operated at atmospheric pressure and using glass beads as solid precursor carriers directly fed into the fluidized bed. The CO2capture capacity of the processed material has been tested in multi–cyclic carbonation/calcination tests. A stable capture capacity250 is seen during carbonation indicating that nanosilica acts as a thermally stable support to mitigate sintering of CaO in the calcination stages. Acknowledgements This work was supported by the Andalusian Regional Government (Junta de Andalucia, contracts FQM-5735 and TEP-7858) and Spanish Government255 Agency Ministerio de Ciencia e Innovacion (contracts FIS2011-25161 and CTQ2011- 27626). References [1] J. Valverde, P. Sanchez-Jimenez, L. Perez-Maqueda, Role of precalcination and regeneration conditions on postcombustion CO2capture in260 the ca-looping technology, Applied Energy 136 (0) (2014) 347 – 356. doi:10.1016/j.apenergy.2014.09.052. URL http://www.sciencedirect.com/science/article/pii/ S0306261914010009 [2] S. E. Edwards, V. Materic, Calcium looping in solar power265 generation plants, Solar Energy 86 (9) (2012) 2494 – 2503. doi:10.1016/j.solener.2012.05.019. URL http://www.sciencedirect.com/science/article/pii/ S0038092X12001971 [3] G. Flamant, D. Hernandez, C. Bonet, J. Traverse, Experimental as-270 pects of the thermochemical conversion of solar-energy -decarbonation of CaCO3, Solar Energy 24 (4) (1980) 385–395. doi:10.1016/0038-092X(80) 90301-1. 8 [4] J. M. Valverde, Ca-based synthetic materials with enhanced CO2capture efficiency, J. Mater. Chem. A 1 (2013) 447–468. doi:10.1039/C2TA00096B.275 URL http://dx.doi.org/10.1039/C2TA00096B [5] G. Grasa, R. Murillo, M. Alonso, J. C. Abanades, Application of the random pore model to the carbonation cyclic reaction, AIChE Journal 55 (5) (2009) 1246–1255. doi:10.1002/aic.11746. URL http://dx.doi.org/10.1002/aic.11746280 [6] P. S´anchez-Jimenez, L. Perez-Maqueda, J. Valverde, Nanosilica supported cao: A regenerable and mechanically hard CO2sorbent at ca-looping conditions, Applied Energy 118 (0) (2014) 92 – 99. doi:10.1016/j.apenergy.2013.12.024. URL http://www.sciencedirect.com/science/article/pii/285 S0306261913010234 [7] R. L. Puurunen, Surface chemistry of atomic layer deposition: A case study for the trimethylaluminum/water process, Journal of Applied Physics 97 (12) (2005) –. doi:10.1063/1.1940727. URL http://scitation.aip.org/content/aip/journal/jap/97/12/290 10.1063/1.1940727 [8] S. M. George, Atomic Layer Deposition: An Overview, Chemical Reviews 110 (1) (2010) 111–131. doi:10.1021/cr900056b. [9] C. Detavernier, J. Dendooven, S. Pulinthanathu Sree, K. F. Ludwig, J. A. Martens, Tailoring nanoporous materials by atomic layer deposition, Chem.295 Soc. Rev. 40 (2011) 5242–5253. doi:10.1039/C1CS15091J. URL http://dx.doi.org/10.1039/C1CS15091J [10] D. M. King, X. Liang, A. W. Weimer, Functionalization of fine particles using atomic and molecular layer deposition, Powder Technology 221 (SI) (2012) 13–25. doi:10.1016/j.powtec.2011.12.020.300 [11] C. Marichy, M. Bechelany, N. Pinna, Atomic layer deposition of nanostructured materials for energy and environmental applications, Advanced Materials 24 (8) (2012) 1017–1032. doi:10.1002/adma.201104129. URL http://dx.doi.org/10.1002/adma.201104129 [12] M. Knez, A. Kadri, C. Wege, U. Gosele, H. Jeske, K. Nielsch, Atomic layer305 deposition on biological macromolecules: Metal oxide coating of tobacco mosaic virus and ferritin, Nano Letters 6 (6) (2006) 1172–1177. doi:10. 1021/nl060413j. [13] S. George, J. Ferguson, A. Weimer, Atomic layer controlled deposition on particle surfaces, US Patent 6,613,383 (2003).310 [14] S. George, J. Ferguson, A. Weimer, J. Wank, Insulating and functionalizing fine metal-containing particles with conformal ultra-thin films, US Patent 6,713,177 (2004). 9