Exceptionally high energy storage density for seasonal thermochemical energy storage by encapsulation of calcium chloride into hydrophobic nanosilica capsules

.


Introduction
Solar heat is available in large amounts during summer.However, to utilize it for heating in the winter when the demand is high, effective means of storing heat for months without exceedingly large thermal losses are needed [1].Thermochemical energy storage (TCES) via the sorption of water into hygroscopic salts offers both high energy storage density and very low losses in long-term heat storage [2,3].In sorption TCES, during charging phase, water bonded with the salt either in solid state as a salt hydrate or in liquid solution is evaporated by heating the salt hydrate or solution to a higher temperature [3].This dehydration is a highly endothermic reaction [3].During storage period the charged material is kept separated from water.When heating output is needed (discharging), water vapor, often in the form of humid air, is brought in contact with the salt or concentrated salt solution, leading to a highly exothermic hydration reaction and thus, an increase in temperature [3].
Many salt hydrates could in theory achieve high volumetric energy storage densities (VSD) of close to or even over 2 GJ/m 3 [4].This makes them suitable e.g., for seasonal thermal energy storage for residential building space heating, with the charging heat provided during summer by solar heat collectors.However, despite the impressive theoretical energy storage densities of many salt hydrates, their application in large scale systems has faced significant issues due to e.g., material degradation [5] and even VSD of 1 GJ/m 3 has shown to be difficult to achieve [6].Especially deliquescence can cause large problems in terms of reversibility of the hydration reaction.Deliquescence is a phenomenon where, if relative humidity (RH) exceeds the deliquescence relative humidity (DRH), the solid salt transitions into a liquid solution [3].Even the formation of a thin liquid layer on salt particle surfaces during sorption can lead to significant agglomeration upon subsequent solidification [7] while complete dissolution eliminates all surface area required for the sorption reactions.Thus, even partial deliquescence is often highly detrimental to the system performance.In addition, partial or complete melting of the higher salt hydrates can occur if, during heating, the higher hydrates melting point is achieved before dehydration to a lower level is completed [8].This can also lead to agglomeration.
Numerous attempts have been made to reduce agglomeration e.g., by infusing salts into different solid adsorbents such as silica gel and zeolite as well as non-reactive porous host matrices such as vermiculite and expanded graphite [5,9,10].Recently, microencapsulation of salt hydrates into porous shells has also gained interest with materials such as hydrophobic polymers [11,12] and hydrophilic hollow silica [6,12] tested.However, in many of the methods used, the amount of material needed for elimination of agglomeration is rather high.Thus, the gravimetric and especially the volumetric energy storage density of the system suffers with VSD usually dropping below 1 GJ/m 3 [6].Thus, the viability of the salt composite as an effective energy storage medium is reduced.New methods that reduce agglomeration with minimal use of non-reactive material are still needed for salt sorption to become a viable option for thermal energy storage.
So-called "dry water" consist of small, stable nanoparticleencapsulated water droplets that behave like a free-flowing powder [13].This concept can be used to encapsulate aqueous salt solutions as well [14].Dry water is formed when hydrophobically surface-treated fumed silica is mixed with water in a high-speed blender directly or with lower speed mixing after atomization [15].During the preparation of dry water, small water droplets form which are quickly surrounded by the hydrophobic nanoparticles that attach to the surface due to surface energies, trapping the water inside and resulting in a water-in-air "emulsion" [16].The result is a powder-like, free-flowing substance, that can be up to 98% water [15].Water cannot escape these capsules without mechanical stress, except via evaporation [13,15,17].
Recently, dry water has gained interest and the concept has been used e.g., to encapsulate K 2 CO 3 for CO 2 capture [14,[18][19][20].A patent from 2012 [21] also describes the encapsulation of different salts and CaO with hydrophobic silica nanoparticles and the use of such compositions for e.g., sorption TCES.However, no detailed data or scientific publications are available concerning the performance of such nanoencapsulated salts in sorption TCES.Additionally, prior reports of salt encapsulation with the dry water concept describe encapsulation only with high silica contents of over 10 wt% on dry salt basis.Due to the highly branched structure and resulting extremely low density of fumed silica, the overall density of dry water style-materials tends to drop drastically as silica content is increased [13].
In this study, we develop two novel modified dry water-style preparation methods for encapsulation of CaCl 2 with as little as 2 wt% SiO 2 nanoparticles in the resulting composition.We study the properties of such nanoparticle-encapsulated salts with emphasis on their performance as sorption TCES under repeated charge-discharge cycles, the achievable energy storage densities, reaction kinetics and the effect of preparation method on these properties.

Materials
Highly hydrophobic fumed silica particles (Aerosil R812S) were supplied by Evonik Industries AG.Its properties are listed in Table 1.

Abbreviations
released heat per unit mass (J/g) q encap,CaCl2 released heat per unit mass of capsules of CaCl 2 (J/g) q CaCl2 released heat per unit mass of CaCl Fumed silica consists of sintered branched networks, called aggregates, of primary amorphous silica (SiO 2 ) spheres of typical size 10-20 nm that are manufactured in a combustion process [15].This aggregate structure, briefly illustrated in Supplementary Fig. S1, causes the exceptionally low density of fumed silica [22].These aggregates, which can have sizes up to several hundred nm, are formed by the collision and fusion of primary silica spheres at high temperatures.Upon cooling, larger weakly bonded agglomerates of 20-30 μm size can form via mechanical entanglements or van der Waals forces, but these can easily be broken e.g., via sonication [15].However, the sintered aggregates usually remain intact as they have much higher energy O-Si-O bonds between them.In Aerosil R812S, the fumed silica aggregates have been further surface-treated with hexamethyldisilazane to make them hydrophobic.

Preparation of encapsulated CaCl 2
A high speed-blender Bosch SilentMixx model MMB66G7M with a ∼400 ml closed plastic container was used to mix the CaCl 2 solutions with the silica nanoparticles.The blender rotation speed was measured with a handheld Monarch Pocket-Tach Plus to be around 7000 rpm at full power.Two different preparation methods (Fig. 1) were used.Both methods include mixing a small amount of ethanol into the CaCl 2 solution before the high-speed mixing.We found that this was required for the creation of low SiO 2 mass fraction composites with complete encapsulation.The compositions prepared are shown in Table 2.
In method one, referred to as direct mixing method (dir), 0.9 ml of ethanol was first mixed with ∼50 g of 43 wt% CaCl 2 solution in the blender container at room temperature before pouring in varying amounts of fumed silica.The blender was then used at full power for s.The container was then turned upside down and lightly shaken by hand for a few seconds after which the sample was mixed with the blender for another 30 s at full power.
While the direct mixing method functions well for silica contents of ∼5 wt% or more (on dry salt basis), complete encapsulation was not achieved if very low amounts of silica were used.In method two, referred to as the phase inversion method (inv), a slightly higher amount of ethanol (see Table 2) was added to the CaCl 2 solution (43 wt% or wt%) before adding the fumed silica.If a 43 wt% solution was used, the preparation was done at room temperature but for 56 wt% solution, the solution, container, blender blade and the fumed silica were pre-heated together to ∼65 • C. In both cases, sample was then mixed at full power in the blender for 30 s which led to a formation of a continuous white substance (Fig. 1, middle).The precise structure of this phase is uncertain.The continuous structure would indicate that the structure is an airin-water foam with air bubbles encapsulated by the nanoparticles.However, small pieces could be cut from the structure.These cut pieces bear resemblance to the soufflé-like material described by Binks and Murakami [13] as they did not stick to surfaces and or easily disperse in CaCl 2 solution.The structure may also be of water-in-air-in-water type described by Binks et al. [23].The precise structure should be studied later but for simplicity, in this study we will refer to this phase as foam.
After forming the foam, the container was turned upside down and strongly shaken by hand until the sample had changed into a white powder-like substance, i.e., the sample had gone through phase inversion.The time required for phase inversion to occur varied from sample to sample but was generally less than 30 s.The sample was then further mixed for another 30 s in the blender which led to foam forming again and then the mixture was phase inverted again through handshaking.
Note that the sample mass used was slightly higher for the 56 wt% solution to keep the volume of sample approximately constant as the higher mass fraction solution has an increased density.Thus, the volume fraction of the water phase relative to air, a parameter that has been a Approximated based on relation between specific surface area and primary particle size presented in [22].b After 10 min of ultrasonic dispersion in ethanol.
c Measured with water intrusion method.found to affect dry water formation [13], was kept near constant between different compositions.
The precise amounts of materials used for each composition are listed in Table 2.The SiO 2 content of dried sample used for naming the samples was determined as y = mSiO2 mSiO2+m CaCl 2 , where m CaCl2 is the mass of anhydrous CaCl 2 in the encapsulated solution and m SiO2 is the mass of fumed silica used.
Note that in Table 2, ξ refers to mass fraction of CaCl 2 in the solution before addition of ethanol.Thus, the mass fraction of solution during encapsulation varies slightly between the compositions where 43 wt% CaCl 2 was used, as differing amounts of ethanol was added.For each composition, two batches were made to ensure the preparation methods were repeatable.
The preparation methods were quite sensitive to the amount of ethanol added.Small amounts of ethanol led to incomplete encapsulation when the direct mixing method was used whereas excessive amounts of ethanol led to irreversible foam formation, i.e., handshaking could no longer induce phase inversion.Irreversible phase inversion upon addition of a surfactant is known to occur for dry water (Binks et al., 2009), but the subsequent phase inversion back to water-in-air powder upon hand shaking seems to be a novel feature that has not, to our knowledge, been reported yet.While only CaCl 2 was encapsulated in this study, we expect the preparation methods to work on other salts as well.

Morphological characterization
Polarized optical microscope (Leica DM4500) was used to take images of the samples at room temperature.Scanning electron microscopy (SEM) was performed on a Zeiss Sigma VP microscope and a JEOL JSM-7500FA analytical field emission scanning electron microscope under vacuum and at 2-4 kV acceleration voltage.The powder samples were first attached on a metal stub with carbon tape and coated with a 4 nm layer of gold palladium alloy in a Leica EM ACE600 sputter coater prior to the SEM.Energy-dispersive X-ray spectroscopy (EDX) was also conducted at 15.0 kV acceleration voltage to detect the elemental composition of the capsules surface.No coating was used for this measurement.
Brunauer-Emmett-Teller (BET) surface area analysis was conducted on a Microtrac BELsorp Mini II.The pore size distribution was determined via the Dollimore-Heal (DH) method.Prior to the BET measurement, the powder samples were dried overnight and put in a Microtrac BELPrep Vac II Pre-Treatment Station for degasification under 150 • C for 2 h with flowing nitrogen gas.
X-ray diffraction (XRD) was carried out on the powder samples using a Rigaku SmartLab X-ray diffractometer with a rotating anode X-ray source (9 kW, Cu Kα1) and a HyPix-3000 2D detector.
Laser diffraction measurements were carried out with Malvern 2600 Mastersizer.Dried samples were dispersed in kerosene in a 12.5 ml cuvette with continuous mixing using a magnetic stirrer to prevent particles from instantly falling to the bottom.The volume concentrations determined by the Mastersizer software varied from 0.01 vol% to 0.05 vol% for different measurements.As larger particles still tended to sediment, measurements were started after just a few seconds of mixing.2000 sweeps were used for the measurement so that the total measuring time was around 15 s.Malvern's model independent droplet size distribution model was used to determine the particle size distributions, volume median diameters (D V,0.5 ) and 90th percentiles (D V,0.9 ), i.e., particle sizes below which 90% of the volume distribution is located.

Thermal conductivity
Thermal conductivity was measured with CTherm TCi analyzer.Before the tests, the analyzer was calibrated with distilled water, expanded polystyrene, LAF foam, PVC and Pyrex.
From each composition three wet and three dry ∼1.8 ml samples were analyzed.Here wet samples refer to samples taken in the condition they were after preparation while dry samples were dried overnight in an oven at 200 • C.

Determination of sorption properties
The ability of the encapsulated salts to absorb water and release heat during a cycle of dehydration and subsequent rehydration was studied with simultaneous thermal analysis (STA) apparatus (Netzsch STA F3 Jupiter) connected to a modular humidity generator (ProUmid MHG32) and equipped with a vortex tube cooler.The humidity generator was calibrated with pure LiCl using the correlation between mass fraction and water vapor pressure presented in [24].Possible inaccuracy in equilibrium mass change measurements was estimated to be less than 1% and it mainly comes from inconsistency of the humidifier and baseline drift over long measurements.
For measuring sorption enthalpies and equilibrium water content, sample with mass ranging from 6 to 12 mg was placed in an open 25 μl aluminum crucible in the STA measurement chamber and heated to 150 • C at a rate of 10 K/min and kept there for 60 min.Then the sample was cooled to the desired temperature and once temperature had stabilized, the humidity generator was switched on to generate humid nitrogen flow at a rate of 150 ml/min for the isothermal hydration step which was continued until sample mass had stabilized.For comparison, pure CaCl 2 was also studied in STA with the same program.It was used as such without any grinding.During the experiments, purge and protective gas flows were set to 20 ml/min.The measured weight gain was used to determine the equilibrium water intake w in g-H 2 O per 1 g of dried sample w = meq−m dry m dry , where m eq is the sample mass in equilibrium with the surrounding water vapor and m dry is the sample mass after drying at 150 • C. Sorption ratio was determined as SR = , where w encap,CaCl2 and w CaCl2 are the equilibrium water intakes for the composite samples and pure CaCl respectively.Hydration level was determined as where M H2O and M CaCl2 are the molar masses of water and CaCl respectively.Based on measured reaction heats in STA, encapsulation ratio was determined as ER = , where q encap,CaCl2 and q CaCl2 are the measured releases of heat per unit mass of the encapsulated and pure CaCl 2 samples respectively.It has been previously observed by Gaeini et al. [11] that reaction heats of the salt hydration reactions can be underestimated by STA measurements by up to 15-30% due to e.g., higher thermal losses to surrounding gas phase in the relatively slow hydration reactions compared to much faster metal melting reactions used for calibration of STA.Furthermore, the STA device we used was only calibrated by melting samples at constant heating rates whereas the sorption reaction took place under isothermal conditions.Thus, to get more accurate estimates for gravimetric energy storage density (GSD) of different compositions, reaction enthalpy of pure CaCl 2 was first determined based on equilibrium hydration level as where Δ r H CaCl2⋅6 H2O is reaction enthalpy for hydration from anhydrous state to hexahydrate, Δ m H CaCl2⋅6 H2O is the melting enthalpy of the hexahydrate while Δ dil H(ξ 6 H2O →ξ eq ) and Δ con H⋅(n eq −6) are the dilution enthalpy and the condensation heat of water vapor from hexahydrate state, where n = 6, to equilibrium conditions at which hydration level is n eq , respectively.All reaction enthalpies are on molar basis, i.e. in units kJ molCaCl 2 except Δ con H which is in units kJ molH 2 O .A more detailed description of Eq. ( 2) is given in Supplementary information.The relevant thermodynamics properties are from [24][25][26][27].GSD of different compositions and pure CaCl 2 was then calculated as To evaluate an isotherm of water intake under different vapor pressures at 30 • C, the sample was dehydrated at 150 • C and then cooled to 30 • C. Once stable temperature was achieved, water vapor pressure was increased stepwise, and the sample was allowed to reach equilibrium conditions for each step while changes in mass were recorded.For these experiments small samples of inv-2.0 containing 1-2 mg of dry CaCl 2 were used.
The rehydration kinetics were evaluated with the same temperature program as the sorption equilibrium properties.Since sample masses are known to affect reaction kinetics quite significantly, for these measurements the samples were weighted so that there was approximately 5 mg of dry CaCl 2 in each sample.
For dehydration kinetics evaluation, another heating step was added to heat the sample to 200 • C at 10 K/min after equilibrium was achieved during absorption at 30 • C with water vapor pressure set to p H2O = 1.3 kPa.At the start of the heating step, water vapor supply was cut off and purge gas flow was increased to 100 ml/min to decrease water vapor build-up in the STA chamber.
Dehydration kinetics were also evaluated in isothermal conditions with the following program: First samples with ∼5 mg of CaCl 2 were dried by heating to 100 • C under pure nitrogen flow.Then, the samples were cooled down to 80 • C and subsequently hydrated to solid dihydrate state by setting the water vapor pressure to 4 kPa.Finally, humidity was set to zero and the samples were allowed to dry again.Based on the thermogravimetric data, hydration level n was then determined as a function of time with t = 0 set as the moment the humidity was set to zero.

Cycling stability
Cycling stability was tested in STA by running a temperature program that consisted of a heating step at 10 K/min to 120 • C, an isothermal segment at 120 • C for 30 min, cooling down at ∼3 K/min to 30 • C and an isothermal segment at 30 • C for around 2.5 h.This cycle was repeated 10 times for all compositions.In addition, for inv-2.0,an experiment with 30 cycles with the same temperature program was performed.Water vapor pressure was kept at ∼1.5 kPa with a flow rate of 150 ml/min to ensure that the samples hydrated to a completely liquid state during the rehydration steps.For the cycling experiments, relative water intake was defined as α i = mh i −mdry mh 1 −mdry , where m h1 and m hi are the masses of the sample at the end of the hydration steps of cycle one and i respectively.

Energy storage density
Based on known relationships between water vapor pressure and CaCl 2 hydration level at different temperatures taken from [24,28], ΔH was predicted at different hydration conditions.The method is covered in more detail in Supplementary information, but briefly, Eq. ( 2) was used for liquid states while for solid hydrate states, ΔH was determined simply as the reaction enthalpy of hydration from anhydrous CaCl 2 and n water molecules in gas state to the solid hydrate with n crystal waters, i.e., ΔH = Δ r H CaCl2⋅n H2O .For both liquid and solid states, GSD was then determined from Eq. ( 3).
The bulk density of different compositions ρ bulk was estimated by filling a 10 ml measuring cylinders with different compositions and recording the changes in mass.Void fraction, which is sometimes called (powder) porosity, was then determined as (4) where the average capsule solid density ρ capsule was determined for each composition as (5) where x is the mass fraction of SiO 2 on wet basis, determined as x = mSiO2 mSiO2+mCaCl 2 +mH 2 O +mEtOH with m EtOH = ρ EtOH V EtOH ≈ 789 kg m 3 ⋅V EtOH .The solid density of fumed silica particles is ρ SiO2 = 2200 kg/m 3 [22] and ρ aq is the density of the aqueous solution used to make the composition.Since the amount of ethanol added to the compositions was small, addition of ethanol was only assumed to affect the mass fraction of CaCl 2 in the solution and thus, ρ aq was calculated with the correlation presented in [24].
Average volume fraction of SiO 2 in the capsules was determined as Liquid volume fraction, which is the fraction of liquid solution volume in the bulk porous material, was then determined as Volumetric energy storage density (VSD) was determined by assuming that the encapsulation is performed with a liquid solution that has the same hydration level as the salt will achieve when it's in equilibrium with the surrounding water vapor at the hydration conditions, i. e., it was assumed that the composite material expands to its original volume after hydration.Note that this essentially means that the composition is optimized for the expected hydration conditions already during preparation and lower VSD values will likely be achieved if such optimization is not done.VSD was determined as VSD = φ liq GSD wet ρ aq,eq = φ liq ΔH M CaCl2 ξ eq ρ aq,eq (8) where GSD wet = GSD⋅ξ eq is the gravimetric energy storage density on wet basis while ξ eq is the CaCl 2 mass fraction and ρ aq,eq is the density of the CaCl 2 -water mixture inside the capsules, with both ξ eq and ρ aq,eq taken at the hydration conditions in equilibrium with the surrounding water vapor.

Morphological properties
All preparation methods used produced a white powder-like substance which could easily be poured from one container to another.Small liquid stains could be observed on glass surfaces after pouring, suggesting that some capsules had broken.However, the amount of leaked solution was extremely small.Upon holding the samples in the glass containers for over a year, no further leakage was observed by visual inspection.
Optical microscope images (Fig. 2 and Fig. S2) show that most capsules are white and somewhat translucent.The non-spherical shape of most capsules is likely due to the silica particles jamming together at the air-liquid interface preventing the droplet from relaxation into spherical geometry [23,29].However, there were some near perfectly spherical capsules that were often almost completely transparent.While at least some transparent capsules with diameters usually much less than 50 μm were visible on all compositions, the transparent spherical capsules were by far most prevalent on inv-2.0 (Fig. 2a) with diameters reaching 200 μm.The spherical shape in combination with transparency suggests that the nanoparticle aggregates in these capsules are not strongly entangled, allowing the liquid to relax to a spherical shape.However, we expect that nanoparticles still exist on the interface as the droplets should not be stable otherwise.It has also been found elsewhere that fumed silica nanoparticles can create nearly transparent systems [22].
The higher prevalence of the spherical, near transparent capsules on inv-2.0seen in Fig. 2a might be due to the higher mass fraction CaCl 2 solution used for their preparation.Higher salt mass fraction leads to higher viscosity and higher surface tension, even though the latter effect is uncertain as the effect of the surfactant (ethanol) might eliminate the difference in surface tension.Still, these differences could make larger capsules with incomplete surface coverage more durable during the preparation, leading to a higher chance of them surviving under the shear forces endured during the mixing.
To study the morphology in detail, SEM images of different compositions were taken.The lower magnification SEM images presented in Fig. 3 confirm the irregular shape of most of the capsules.These images also suggest that higher silica content leads to lower particle size while the phase inversion method (Fig. 3a and b) results in a narrower distribution in particle size compared to the direct mixing method (Fig. 3c).Similar differences between inv-4.9 and dir-4.9 are also visible in the optical images (Fig. 2c and d).However, the optical images suggest that inv-2.0 has a smaller particle size than inv-2.5 (Fig. 2a and b), breaking the trend of higher silica content leading to smaller particle size.This is possibly due to incomplete surface coverage on capsules of inv-2.0 which would allow more surface area to be encapsulated with the same amount of silica.
The higher magnification SEM images shown in Fig. 4 show that the coating consists of SiO 2 nanoparticles that are at least partly agglomerated.Some of the SiO 2 agglomerates appear around 20 μm in size (Fig. 4a and d) which is consistent with the agglomerate sizes found in pure fumed silica powders [15,22].However, the agglomerates may also be small capsules attached to larger ones.Some pores of over 1 μm are visible on the surface of the larger capsule's shells (Fig. 4d) indicating that mass transfer through the shell is not completely limited to Knudsen diffusion, unless the materials are used at low total pressures.The structures of the shells seem similar for different compositions except for inv-2.0 for which the surfaces appear smoother with some larger cracks (Fig. 4b).This is likely due to the thinner and more incomplete surface coverage of nanoparticles on these capsules which allows the salt surface to be visible.
The particle size distributions from laser diffraction measurements shown in Fig. 5 further confirm the narrower spread in particle size with the phase inversion method.Apart from inv-2.0,only one strong peak is visible for compositions made with the phase inversion method while the compositions made with the direct mixing method show two strong peaks resulting in a wider spread.For inv-2.0 there appears to be two overlapping peaks.The cause for this is unknown but it is perhaps related to the incomplete encapsulation of some capsules.
The difference in particle size distribution spread is especially high for samples with 4.9 wt% silica with dir-4.9showing a much higher share of large capsules with size over 300 μm than inv-4.9.It is known that for liquid marbles, which are millimeter sized liquid droplets encapsulated with nonwetting particles, durability under compression pressure increases linearly as radius decreases [30].Thus, the phase inversion method, which produces less of the larger particles, might be more desirable for systems where the capsules are under relatively large pressures, for example under the weight of the capsules above them or while enduring collisions with container walls and other capsules in a fluidized bed type reactor.
The volume median diameters (D V,0.5 ) and 90th percentiles (D V,0.9 ), i. e. particle size below which 90% of the volume distribution is located, are shown in Table 3.These were determined via laser diffraction measurements.Overall, D V,0.5 of dried capsules of different compositions ranged from around 90 to 210 μm.D V,0.5 generally decreases with increasing silica content, especially when going from inv-2.5 to inv-4.9.However, as an exception to this rule, inv-2.0displays a much smaller D V,0.5 than inv-2.5.This further suggests that the capsule shells in inv-2.0 are thinner and less densely covered with silica than in inv-2.5.
After dispersing the sample in kerosene, D V,0.5 decreased with time when repeated measurements were taken, suggesting that larger particles may have been slowly sedimenting beneath the laser light despite the mixing with magnetic stirrer.Thus, it is possible that some larger capsules were not detected as accurately as the smaller ones.

Table 3
Volume median diameter (D V,0.5 ) and 90th percentiles (D V,0.9 ) of dried compositions based on laser diffraction measurements and BET surface area of dried compositions and fumed silica (Aerosil R812S).The results of BET surface area analysis are included in Table 3.The measured BET surface area of Aerosil R812S (223.8 m 2 /g) corresponds well with the manufacturer's reported range of 195-245 m 2 /g [22].For the composites, an increasing silica content seems to lead to an increase in surface area.This is likely due to two factors.First, as shown earlier, an increase in silica content leads to smaller capsule size and therefore an increase in overall surface area of a sample.Thus, the BET results further confirm this correlation between particle size and silica content.Second, fumed silica has a very high specific surface area so any increase in the amount of silica will be visible in the results.However, composition inv-2.0deviates from the general trend as its BET surface area is higher than that of inv-2.5.This further confirms the smaller particle size of inv-2.0 compared to inv-2.5 found in laser diffraction measurements.
Fig. 6 shows EDX mapping indicating the presence of calcium and silicon on composition dir-7.2.Calcium can be seen spread rather evenly on the entire surface.The varying intensity of silicon indicates that the thickness of the silica layer varies on the surface, even on dir-7.2 which has a relatively high silica content.When comparing Fig. 6 and Fig. S3 it can be seen that the intensity of chloride matches well with that of calcium which is expected as they are chemically bonded and thus cannot exist separately.Similarly, silicon and oxygen intensities match well due to chemical bonding of the elements.A layered image showing a combination of the different elements is shown in Fig. S4 and EDX elemental diagram is shown in Fig. S5.X-ray patterns of the dried fumed silica encapsulated CaCl 2 (composition dir-4.9)and pure fumed silica are also shown in Fig. S6.

Sorption properties
The sorption properties depicting the material's ability to absorb water and release heat are collected into Table 4 for different compositions and pure CaCl 2 .Measurements were made at a hydration temperature of 30 • C with water vapor pressure set to ∼1.3 kPa.At least one sample from both prepared batches of each composition was measured to ensure repeatability.Equilibrium values of the mass fraction of CaCl 2 in the solution ξ eq ≈ 0.443 and hydration level n eq ≈ 7.76, corresponding to the measured water intake w ≈ 1.26 of pure CaCl 2 , were used in calculating ΔH and GSD with Eqs. ( 2) and (3).Note, that the measured heat release per mass of dry sample q is technically an estimate of GSD as well but due to known underestimation of hydration reaction heats explained earlier in Chapter 2.5, GSD estimate is given based on measured water intake w and Eqs. ( 2) and (3) while q is used to calculate ER.
As can be seen in Table 4, values of both encapsulation ratio ER and sorption ratio SR determined based on STA measurements correlate with the amount of SiO 2 used for each composition.Thus, the values suggest that the encapsulation was complete in all compositions and that the presence of SiO 2 nanoparticles does not significantly affect the sorption properties of CaCl 2 .Based on the difference between values of q and GSD of pure CaCl 2 , STA seems to underestimate reaction heats by about 4%.
The equilibrium hydration level of inv-2.0calculated with Eq. ( 1) based on STA measurements at different water vapor pressures at 30 • C is shown in Fig. 7.The formation of a solid dihydrate took place at low vapor pressure below 0.3 kPa.Once vapor pressure of 0.92 kPa, which is the vapor pressure of a saturated solution of CaCl 2 at 30 • C, was reached, the salt inside the capsules behaved as can be expected for an aqueous CaCl 2 solution, i.e., hydration level followed the known relationship between water vapor pressure and CaCl 2 mass fraction.
However, conversion to solid tetrahydrate was not found at any vapor pressure.Instead, as vapor pressure was increased above a threshold level of ∼0.68 kPa, CaCl 2 inside the capsules reached a hydration level of 5.16 which could then be continuously increased and decreased back by altering the vapor pressure slightly.Thus, it seems very likely that the CaCl 2 inside the capsules was in liquid state.
The CaCl 2 mass fraction in this solution was clearly over the

Table 4
Measured heat released in STA per mass of dry sample (q), encapsulation ratio (ER), equilibrium water intake (w), sorption ratio (SR) and the estimated GSD based on w and thermodynamic properties of CaCl 2 .All values are based on measurements at a hydration temperature of 30 solubility limit of ∼50% at 30 • C [24].Thus, it seems that the formed solution inside the capsules is supersaturated and forms at a relative humidity value that is much smaller than the DRH that would be expected based on the thermodynamic equilibrium.In recent years, it has been found that small samples of CaCl 2 can deliquesce into supersaturated solutions at far lower RH values than the vapor pressure of a saturated CaCl 2 solution would suggest [32,33].As relative humidity is increased, due to deliquescence being a much faster process than the formation of a higher solid hydrate (tetrahydrate or hexahydrate), the dihydrate can deliquesce before the sample has time to hydrate to a higher solid hydrate in any measurable amount [32].This can occur since the dihydrate has a lower DRH than the higher hydrates and due to nucleation barrier preventing crystallization, the formed metastable liquids may not crystallize into the thermodynamically stable hydrate for very long periods of time.The phenomenon appears to be more common the smaller the salt particle is [32].Since the hydrophobic silica capsules create rather small particles of CaCl 2 (Table 3), it is likely that this phenomenon is occurring during their hydration.
To support this hypothesis, we evaluated the relative humidity at which 2-3 mg samples of inv-2.0start to transition from dihydrate state to a higher hydration level at different temperatures ranging from 25 • C to 40 • C and measured the equilibrium hydration level just above the RH threshold level.Starting with a sample in STA dried at 150 • C and cooled to the desired temperature with only pure N 2 flow present, RH was increased at isothermal conditions to around 8% at which point the salt transitioned into the dihydrate state.Once the sample mass had stabilized to the dihydrate state, RH was increased stepwise by 1% if no notable change in mass was detected in 3 min.In these measurements, once mass gain started the samples appeared to always deliquesce and no evidence of transition onto a tetrahydrate or hexahydrate was observed at any of the temperatures.At 25 • C, deliquescence occurred at ∼16% RH and at other temperatures measured (30 • C, 35 • C, and 40 • C), the measured DRH was between 15% and 17%.The results match well with the dihydrate DRH value of 18.5% measured by Guo et al. [33] at 25 • C and the value of 16% predicted by Kelly and Wexler [34] at the same temperature.Considering that Guo et al. [33] reported an absolute uncertainty of ± 1% and we estimate the absolute uncertainty to be ± 2% for RH in our measurements, both of these fall within our range of accuracy.Thus, it seems very likely that instead of going through the thermodynamic equilibrium states, during the sorption process the CaCl 2 particles inside the SiO 2 shells first hydrate into the dihydrate and then deliquesce into a supersaturated solution.
It is possible that the hydration to tetrahydrate was just too slow to be observed before RH was increased.However, such slow process would likely have little practical significance for sorption TCES, as heating power extractable from the reaction would be extremely low.

Thermal conductivity
The average thermal conductivities and corresponding standard deviations (SD) of different compositions are shown in Table 5.The standard deviation for samples within each batch was low; typically around for 0.001 W/(mK), maximum value 0.004 W/(mK) was found for wet inv-2.0sample.Generally, the thermal conductivities are quite low as expected for a powder-type material.In wet state, i.e., when the salt inside the capsules is in an aqueous solution state, the thermal conductivities are higher.This is probably due to the higher proportional share of fumed silica in the dried state as we measured Aerosil R812S thermal conductivity in powder form to be very low (0.040 W/(mK)).Low thermal conductivity of fumed silica also explains the decreasing trend of thermal conductivity with higher silica content.However, the thermal conductivity of inv-2.0 in wet state is comparatively low.This is probably due to the higher concentration of the salt solution which decreases the thermal conductivity of the aqueous solution [24].

Sorption kinetics
The isothermal dehydration and hydration curves from STA measurements are shown in Fig. 8 for different compositions.The dehydration of samples when heated at a rate of 10 K/min is shown in Fig. S7.
The isothermal dehydration at 80 • C from the dihydrate state (Fig. 8a) seems to occur in two steps with on average around 1.3-1.4water molecules per each salt molecule leaving rather quickly in less than 15 min.The complete dehydration takes up to 100 min due to the decrease in dehydration kinetics that occurs after the first 1.3-1.4water molecules have left.The cause of the sudden decrease in dehydration kinetics around this mark is yet unknown.
Apart from inv-2.0, the dehydration kinetics improve with increasing BET surface area.In addition, apart from inv-2.0, the dehydration kinetics improve with decreasing D V,0.5 except the order of dir-7.2 and inv-7.2 is reversed.This is likely due to the much higher share of relatively large capsules in dir-7.2compared to inv-7.2 that can be seen in Fig. 5 and is also apparent in its much higher D V,0.9 value.In larger capsules, there is a longer distance for water molecules to travel before they reach the crystal surface.In addition, larger capsules also have a smaller ratio of surface area to particle volume.Thus, the findings agree well with expectations from theory.
The slower kinetics of inv-2.0might be caused by tighter packing of the particles during the dehydration.All the samples are expected to become more porous as they lose water.However, the samples were placed in the STA chamber in the condition they were after preparation and inv-2.0 is made with 56 wt% CaCl 2 solution while the other samples are made with a 43 wt% solution.Thus, inv-2.0 is likely less porous Fig. 7. Measured equilibrium hydration level of inv-2.0 at 30 • C at different water vapor pressures with pure CaCl 2 equilibrium shown for comparison.Equilibrium properties of pure CaCl 2 equilibrium are based on [24,31].Although a hydrate of form CaCl 2 ⋅(1/3) H 2 O is also known to exist [26], it was left out of for simplicity.

Table 5
Thermal conductivities in units W/(mK) of different compositions at 23 during dehydration, reducing its reaction kinetics compared to the other samples.
When heating the sample at a rate of 10 K/min, the differences in dehydration kinetics were overall very small, especially for the early part of the dehydration down to around hydration level of three.To get more detailed curves, only the removal of the last three water molecules is shown in Fig. S7.It is possible that early dehydration is mainly limited by water vapor transfer from the top of the sample to STA chamber surroundings, which hides the differences between compositions.However, for the removal of the last two water molecules, inv-2.0 and dir-4.9show slightly slower dehydration kinetics as does inv-2.5 albeit to a lesser extent.The order in kinetics of dir-4.9 and inv-2.5 is reversed from the isothermal dehydration measurements.This might be explained by larger thermal gradients forming inside the samples.With the high heating rate of 10 K/min, the samples completely dehydrate from a state with hydration level of 7.76 to anhydrous state in less than 15 min.During the dehydration, the temperature in the middle of the sample is likely smaller than on the edges due to limited heat transfer from the sample edges.The temperature gradient is amplified due to the high evaporation rate cooling down the sample.As dir-4.9 has a lower thermal conductivity than inv-2.5 (Table 5) and a lower bulk density (Table 6), i.e., the sample with same mass of CaCl 2 had a larger volume, larger temperature gradients are expected to form inside dir-4.9than inv-2.5.This might have slowed down the dehydration of dir-4.9enough to reverse the order.
Note that if the sorption material is dried by hot air flowing through it, as is often the case for sorption systems, differences in powder thermal conductivity will not affect the dehydration kinetics as they have here.Thus, results from the isothermal dehydration experiments are more relevant for most practical applications.
The isothermal hydration kinetics are depicted in Fig. 8b.The hydration kinetics overall show quite small differences between different samples.However, even considering for possible inaccuracy in measurements, inv-2.0does show somewhat slower kinetics than the others.As with dehydration kinetics, this might be due to the tighter packing of this composition.Despite the small differences observed here, the compositions might show significantly larger differences in hydration kinetics in practical scale storage systems.Since there is no gas flow through the powder sample inside the STA chamber, reaction kinetics might be mainly limited by water vapor transfer external to the capsules rather than internal transfer or reaction kinetics.In practical scale storage systems humid air would be flown through the powder or the powder would be placed in near-vacuum conditions.In both cases it is expected that particle size would affect reaction kinetics to a much higher degree, although this will have to be confirmed with further studies.

Cycling stability
All compositions were tested for cycle stability in repeated dehydrationhydration cycles which is crucial for long-term performance of the material.For each composition, at least one sample from both prepared batches of each composition was tested to ensure repeatability of the preparation method.Fig. 9a shows the thermogravimetric analysis and temperature curves for cycling experiments with inv-2.5 and Fig. 9b shows the relative water intake α i for each cycle i for different compositions.Similar graph for compositions with 7.2 wt% SiO 2 is included in Supplementary information (Fig. S8).All tested samples showed excellent stability under 10 repeated cycles.Furthermore, inv-2.0showed no signs of degradation under 30 cycles as shown in Fig. S9.As inv-2.0 has the lowest amount of silica in the composition, it was expected to be the most susceptible to degradation due to agglomeration.Thus, it is very

Table 6
Densities and volume fractions of different compositions.The bulk density ρ bulk is based on measurements with volumetric flask.Eqs. ( 4)-( 7) are used for determining φ V , ρ capsule , φ SiO2,caps and φ liq .Density of aqueous solution ρ aq is based on the correlation presented in [24].likely that all the other compositions would perform equally well.The test conditions were such that CaCl 2 went through complete dissolution during the hydration step.Thus, any significant breakage of the capsules should have led to liquid solution escaping the capsules and contacting other leaked liquid, leading to significant agglomeration upon recrystallization during the subsequent heating step.If this had occurred, a notable declining trend in the hydration level at the end of each sorption step should have been visible, as hydration kinetics would have been hindered.However, no such trend was found within measurement accuracy; even for the measurement with the most decrease in relative water intake (sample of inv-2.5, shown in Fig. 9b), the last hydration showed over 98% water intake compared to the first cycle, i.e. α 10 > 0.98 while another sample of the same composition (shown in Fig. 9a) resulted in α 10 ≈ 1.01.
The results in Fig. 9, Fig. S8 and Fig. S9 show that the capsules are very stable and very little if any solution is leaking out of them at any stage.Thus, agglomeration seems to be completely prevented.The capsules were durable even under conditions where they expanded beyond their original volume, as inv-2.0 was prepared with a 56 wt% CaCl 2 solution but hydrated to a less than 43.5 wt% solution during the hydration steps, corresponding to 41% increase from original capsule volume.Previously, Liu et al. [30] have shown that liquid marbles made from Aerosil R812 hydrophobic fumed silica can withstand almost 45% increase in surface area upon deformation under compression before rupturing due to eventual poor surface coverage of particles on its surface.It seems likely that due to the core of dry water-style capsules being made of a liquid, and the nanoparticle aggregates not strongly attached to one another, the aggregates can rearrange themselves on the surface.This ensures that enough of the surface remains covered to prevent liquid leakage and agglomeration to other capsules even if the volume and thus also surface area of the capsules expands to some extent.If we assume encapsulated salt solutions have similar durability as encapsulated water droplets, the encapsulated salts might last expanding by up to 75% from original volume, although the limits were not tested in this study.

Sample density
Measured bulk densities of different compositions (ρ bulk ) are shown in Table 6.The values listed are measured without any tapping or shaking apart from very lightly shaking the cylinder to even the top layer of the powder for measurements.It was found that density could be increased by slightly tapping or shaking the measuring cylinder.For example, for the inv-2.5 the density could easily be increased up to 1.12 g/ml, i.e., by almost 8% with this method.This is likely due to the capsules reorienting themselves to a denser packing under the shaking.It is also possible that the capsules, containing a deformable liquid solution and inside a shell consisting of separate nanoparticles, might be reshaped slightly upon tapping.
Results for densities and volume fractions can be found in Table 6.
ρ bulk and φ liq are plotted as a function of φ SiO2,caps in Fig. 10, which shows that both ρ bulk and φ liq correlate negatively with increasing volumetric fraction of SiO 2 in the capsules.However, no significant difference between the two preparation methods can be seen.Strong negative correlation between volumetric fraction of SiO 2 and bulk density was also observed by Binks and Murakami [13], who found that the bulk density of dry water dropped by a factor of 2.5 when silica content was increased from 2 wt% to 9 wt%.The strong decrease in bulk density with increasing silica content is likely caused by the highly porous structure of the branched fumed silica aggregates.This porous structure should lead to very low bulk density for the shell of the capsules as well as for any fumed silica agglomerates that remain between the capsules.This is supported by the extremely low bulk density of fumed silica.For Aerosil R812S the bulk density is only 60 g/l [22] i.e., over 97 vol% of Aerosil  R812S powder is air and similar void fraction may be present in the capsule shells.As can be seen from Eq. ( 8), VSD is directly proportional to φ liq and therefore the best composition inv-2.5 should have over 53% higher VSD than inv-7.2which has the lowest φ liq .

Energy storage density
Values of gravimetric energy storage density (GSD) and volumetric energy storage density (VSD) are some of the most important criteria for feasibility of long-term TCES systems as they highly affect the cost of the system and storage space required.For calculation of GSD from Eq. ( 3) and VSD from Eq. ( 8), tetrahydrate and hexahydrate are neglected as they are not expected to form based on the STA measurements.Thus, the only solid hydrates that are expected to form during the hydration are the monohydrate and the dihydrate.Once sufficient RH for solid dihydrate formation is achieved, further increase in water vapor pressure leads to formation of a supersaturated solution once the dihydrate DRH is reached.Based on Fig. 7 and values of w recorded at the DRH limit at different temperatures, the correlation presented in [24] should provide accurate predictions even for the supersaturated solutions with absolute uncertainties of less than 2% in RH.Thus, the predictions of GSD and VSD should have good accuracy.
As data on the dihydrate melt density could not be found, a rough density estimate 0.8⋅ρ CaCl2⋅2 H2O = 1.48 g/cm 3 , where ρ CaCl2⋅2 H2O is the density of solid dihydrate, was used for determination of VSD of the dihydrate.This estimate may be optimistic as the encapsulation would have to be done at very high temperature above the melting point of the dihydrate T m = 176 • C, which has not been tested and as such, a solution with lower mass fraction of CaCl 2 may be needed so that the composition can be prepared at a lower temperature.For the monohydrate, a diluted solution with a weight fraction of the dihydrate and density of 1.48 g/ cm 3 was assumed for the encapsulation.
As can be seen in Table 6, φ liq,inv−2.0= 0.66< φ liq,inv−2.5= 0.72, even though the two composites have similar φ SiO2,caps .Thus, the use of higher mass fraction of CaCl 2 in preparation may affect the achievable liquid volume fraction.To avoid overly optimistic estimates in some conditions, φ liq = 0.66 was used to calculate VSD.For GSD, SR = 0.975, i.e., the SR value of inv-2.5 was used.Both GSD and VSD are plotted as a function of dew point in Fig. 11.The horizontal lines in both graphs are due to the dihydrate and monohydrate reaction enthalpies and densities which vary very little with hydration temperatures.Therefore, lines for different hydration temperatures overlap.The horizontal lines extend beyond the plotted dew point range as the lower hydrates can form even with dew point below T dew = − 7 • C.
It can be seen from Fig. 11 that increasing the water evaporation temperature (dew point T d ) with constant hydration temperature (T h ), i. e., decreasing the temperature lift (T h − T d ) will increase GSD exponentially once deliquescence threshold is achieved but VSD does not increase much and eventually peaks and starts to decrease.This is due to the increase in volume that occurs with each absorbed water molecule and the decrease in density as the mass fraction of salt in the solution gets lower.Additionally, the heat released per absorbed water molecule gets lower and eventually approaches water condensation heat as the salt mass fraction approaches zero, i.e., when the dew point gets close to the hydration temperature.
Based on Fig. 11a, VSD values of 0.82 GJ/m 3 are achievable even with high temperature lifts exceeding 60 • C as CaCl 2 hydrates to the dihydrate stage with very low water vapor pressures.With temperature lifts of around 30 • C it should still be possible to achieve high VSD values of around 1.4 GJ/m 3 .Such high VSD values easily exceeding 1 GJ/m have not been reported elsewhere in sorption energy storage literature to our knowledge.To put the results into context, Donkers et al. [4] suggest that 7-12 GJ of thermal energy storage capacity is needed to overcome seasonal fluctuations on a typical Western European 120 m residential building, assuming the passive house standard of 15 kW h/m 3 or 28 kW h/m 3 for built or renovated buildings respectively, is applied.Based on an evaporation temperature of 10 • C, which according to Donkers et al. [4] is a reasonable estimate for temperature at which water can be evaporated if a 7 m deep borehole in the Netherlands is used as the heat source, and a hydration temperature of 30 • C, VSD of just over 1.5 GJ/m 3 could be achieved with inv-2.5.This suggests that in some cases less than 5 m 3 of material storage space would be needed for storing solar heat for the whole winter in new, well insulated buildings.
Based on Fig. 11b, GSD values of around 1100 kJ/kg are achieved if only dihydrate forms but with temperature lifts of around 30 • C, GSD is around 2400 kJ/kg.The latter value is higher than what is usually achievable even with pure CaCl 2 in similar conditions as the highest hydrate that can achieve a temperature lift of 30 • C is the tetrahydrate, for which GSD ≈ 2200 kJ/kg.The increase in GSD for the composite is due to the early deliquescence and resulting higher hydration level compared to the stable tetrahydrate (Fig. 7).The difference between the composite material and pure CaCl 2 gets even larger when vapor pressure is increased but is kept below the vapor pressure of a saturated CaCl solution.For example, the maximum difference at hydration temperature of 30 • C occurs when the vapor pressure is 0.91 kPa (dew point 5.6 • C), i.e., just below the vapor pressure of saturated CaCl 2 solution.At this point, GSD ≈ 2850 kJ/kg which is 30% higher than the GSD of the stable form of CaCl 2 in these conditions, i.e., the tetrahydrate.For the temperature lift from 10 • C to 30 • C, GSD is over 3300 kJ/kg.
In Table 7, the GSD and VSD of our composition inv-2.5 is compared to the best performing salt hydrate sorption composites reported in recent reviews [5,6].The comparison is done by determining GSD inv−2.5 and VSD inv−2.5 under the same hydration conditions, i.e., with the same T d and T h as used in the referenced studies.Only those composites studied under conditions of T h > 20 were considered as lower temperatures were not considered reasonable for domestic heating applications.GSD inv−2.5 and VSD inv−2.5 are determined based on Eqs. ( 3) and ( 8) similarly as was done for Fig. 11 with SR = 0.975 and φ sol,liq = 0.66.Clearly, inv-2.5 is expected to perform notably better than any other state-of-the-art composite in all studied hydration conditions.Furthermore, the low content of additive material should keep costs low.
While Fig. 11 details a conservative estimate for VSD, values higher by 9% are estimated if the measured solution volume fraction of inv-2.5, φ liq,2.5 p.i. = 0.72, is used.If we also assume that the density increase achieved by tapping and shaking the powder is achievable in energy storage applications, 18% increase in VSD can be achieved as solution volume fraction would increase to φ liq = 0.78.Thus, with water evaporation temperature at 10 • C and a hydration temperature of 30 • C, values as high as VSD ≈ 1.8 GJ/m 3 might be achievable.However, the effects of such density increase to kinetic properties and capsule stability will need to be confirmed in another study.

Conclusions
In this work, we showed that CaCl 2 can be encapsulated with hydrophobic fumed silica nanoparticles to create capsules that prevent liquid CaCl 2 solution from leaking out while still allowing water vapor to permeate.As a result, the new material composite provides exceptionally promising properties for long-term thermal energy storage.
Two different methods, direct mixing method and phase inversion method, both with varying amounts of silica, were compared.The structural properties of formed capsules were investigated with various methods including optical microscopy, SEM and laser diffraction confirming the encapsulation of the salt into capsules with volume median diameters ranging from around 90 μm-210 μm.Sorption and energy storage properties were studied with STA.
The novel phase inversion method required as little as 2% weight silica for encapsulation.Sorption properties were evaluated under various temperatures and vapor pressures.Gravimetric energy storage densities under hydration conditions of p H2O = 1.3 kPa and T = 30 ᵒC were in some cases over 97% of that of pure CaCl 2 , suggesting that the silica nanoparticles in the shell have little effect on reaction enthalpies of encapsulated CaCl 2 .However, in hydration conditions under which a solid tetrahydrate or hexahydrate was expected to appear, a liquid supersaturated solution seemed to form instead.This is likely due to the dihydrate undergoing deliquescence before the formation of a higher hydrate had time to develop.
Compositions with larger average capsule size generally showed decreased dehydration kinetics while increased silica content, which was found to lead to smaller capsules, ensured faster dehydration kinetics.However, in some cases a wide particle size distribution and resulting relatively high share of large capsules was more important factor in limiting reaction kinetics than average capsule size.
The direct mixing method, which produced capsules with higher spread in capsule size, showed decreased dehydration kinetics compared to the phase inversion method.However, as the solution concentration during preparation was also found to affect dehydration and hydration kinetics, the relationship between kinetics and preparation conditions seems to be rather complex and should be investigated further.
None of the tested samples, which contained approximately 5 mg of dry CaCl 2 , showed signs of degradation in sorption properties under ten dehydration-hydration cycles performed under conditions in which samples went through complete deliquescence during hydration.Furthermore, the composition with the lowest amount of silica (2 wt%) showed no signs of degradation under 30 cycles.This suggests that the compositions are stable when used as a sorption TCES material.
Volumetric energy storage densities were predicted for a wide range of different hydration conditions based on the STA results.The results showed that volumetric energy storage density of up to 1.4 GJ/m 3 is achievable on bulk material level, without considering other system components, with temperature lifts as high as 30 ᵒC.Such high values easily exceeding 1 GJ/m 3 are unprecedented in sorption TCES.Further optimization of the material could in best-case scenario yield volumetric energy storage densities of over 1.6 GJ/m 3 under similar conditions and with lower temperature lift of 20 ᵒC, values close to 1.8 GJ/m 3 might be achievable.Thus, hydrophobic fumed silica encapsulation of CaCl 2 , especially with the phase inversion method, shows excellent potential for seasonal thermal energy storage.However, larger scale experiments are still needed to prove the materials durability for practical applications.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Table 7
Comparison of GSD and VSD of other state-of-the-art sorption materials to that of inv-2.5 (GSD inv−2.5 and VSD inv−2.5 ).Salt content is for the referenced materials, inv-2.5 has a salt content of 97.5 wt%.T d and T h are the dew point and hydration temperature respectively and they are the same for the referenced material and our material.a Value from 10th cycle as there was some degradation in the material.
b Hydration conditions are just on the DRH limit of our material.In the cited study [6], the conditions were similarly on a threshold where hydration took a step up.
Thus, the higher estimate where deliquescence takes place is given.

Fig. 4 .
Fig. 4. SEM images of samples a) inv-7.2b) inv-2.0c) dir-4.9d) section of dir-4.9magnified further.Red circles highlight possible larger SiO 2 agglomerates.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article).

Fig. 5 .
Fig. 5. Particle size distributions of different compositions measured with laser diffraction.

Fig. 6 .
Fig. 6.EDX mapping of dir-7.2.The two images are of the same area in the sample.Purple color on the left-hand image shows presence of calcium while the green color on the right-hand image indicates presence of silicon.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article).

Fig. 8 .
Fig. 8. (a) Isothermal dehydration of samples at 80 • C (b) Hydration curves for isothermal hydration of samples at hydration temperature T h = 30 • C with p h = 1.3 kPa.

Fig. 9 .
Fig. 9. (a) Ten dehydration/hydration cycles for inv-2.5.Water vapor pressure was set to 1.5 kPa.(b) Relative water intake for each hydration cycle for different samples.

Fig. 10 .
Fig. 10.Bulk density of composition and liquid volume fraction in different compositions as a function of average volume fraction of silica nanoparticles in the capsules.Linear fit is made for φ liq .

Fig. 11 .
Fig. 11.(a) Volumetric energy storage density and (b) gravimetric energy storage density achievable at different output temperatures as a function of dew point.Deliquescence thresholds are present on all curves even though they are marked only on isotherms of 25 • C and 30 • C.

Table 2
Prepared compositions.Number in the name gives the SiO 2 wt% of dried sample.inv refers to samples made with the phase inversion method while dir refers to samples made with the direct mixing method.ξ is the mass fraction of CaCl 2 in the solution before ethanol addition.V EtOH is the amount of ethanol added during preparation.T prep refers to the approximate temperature of the solution and fumed silica during preparation.
• C in both wet and dry state.