Exploitation of thermochemical cycles based on solid oxide redox systems for thermochemical storage of solar heat. Part 1: Testing of cobalt oxide-based powders

Available online at www.sciencedirect.com

ScienceDirect Solar Energy 102 (2014) 189–211 www.elsevier.com/locate/solener

Exploitation of thermochemical cycles based on solid oxide redox systems for thermochemical storage of solar heat. Part 1: Testing of cobalt oxide-based powders Christos Agrafiotis ⇑, Martin Roeb, Martin Schmu¨cker, Christian Sattler Deutsches Zentrum fu¨r Luft- und Raumfahrt/German Aerospace Center (DLR), Linder Ho¨he, 51147 Ko¨ln, Germany Received 26 September 2013; received in revised form 26 December 2013; accepted 29 December 2013 Available online 11 February 2014 Communicated by: Associate Editor Michael Epstein

Abstract Thermochemical storage of solar heat exploits the enthalpy effects of reversible chemical reactions for the storage of solar energy. Among the possible reversible gas–solid chemical reactions, utilization of a pair of reduction–oxidation (redox) reactions of solid oxides of multivalent metals can be directly coupled to Concentrated Solar Power (CSP) plants employing air as the heat transfer fluid avoiding thus the need for separate heat exchangers. The redox pair of cobalt oxides Co3O4/CoO in particular, is characterized by high reaction enthalpies and thus potential heat storage capacity. Parametric testing of cobalt oxide-based powder compositions via Thermo-Gravimetric Analysis/Differential Scanning Calorimetry was performed to determine the temperature range for cyclic reduction–oxidation and optimize the process parameters for maximum reduction and re-oxidation extent. The heating/cooling rate is an important means to control the extent of the oxidation reaction which is slower than reduction. Complete re-oxidation was achieved within reasonable times by performing the two reactions at close temperatures and by controlling the heating/cooling rate. Under proper operating conditions Co3O4 powders exhibited longterm (30 cycles), complete and reproducible cyclic reduction/oxidation performance within the temperature range 800–1000 °C. No benefits occurred by using Ni, Mg and Cu cobaltates instead of “pure” Co3O4. The Co3O4 raw material’s specific surface area is an influential factor on redox performance to which observed differences among powders from various sources could be attributed. Presence of Na was also shown to affect significantly the evolution of the products’ microstructure, though not necessarily combined with improved redox performance. Ó 2014 Elsevier Ltd. All rights reserved. Keywords: Solar energy; Thermochemical cycles; Thermochemical heat storage; Redox reactions; Cobalt oxide; Cobaltates

1. Introduction The identification of suitable materials and processes for efficient and economically viable exploitation of solar energy for hydrogen and so-called “solar fuels” production has been a subject of intense research activity (Centi and Perathoner, 2010). This broad concept includes among

⇑ Corresponding author. Tel.: +49 22036014132; fax: +49 22036014072.

E-mail address: christos.agrafi[email protected] (C. Agrafiotis). 0038-092X/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.solener.2013.12.032

other routes, the utilization of Concentrated Solar Power (CSP) as the heat source required for performing the high-temperature endothermic reactions of the so-called thermochemical cycles – multi-step processes of two or more chemical reactions that form a closed cycle (Kodama, 2003). The redox-oxide-pair based thermochemical cycles in particular, operate on the principle of transition between the oxidized (higher-valence, MeOox) and reduced (lowervalence, MeOred) form of an oxide of a metal exhibiting multiple oxidation states (Steinfeld, 2005). In this concept,

190

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

Nomenclature y English letters C gas species concentration (moles of gas/m3) k temperature-dependent reaction rate constant (m3/moles/min) R reaction rate (moles of gas produced per minute per g of redox material) T absolute temperature (K) Greek letters W maximum oxygen storage capacity of the redox material (moles of O2 stored in the solid or “consumed from the gas phase” per g of redox material)

during the first, higher-temperature, endothermic thermal reduction (TR) step, the oxidized form of the oxide MeOox, releases a quantity of oxygen and transforms to its reduced state under the supply of external heat, according to the generalized reaction scheme: MeOox þ ðDH Þ ! MeOred þ O2 ðgÞ

ð1Þ

The second step involves (exothermic) oxidation of the reduced form of the oxide back to its oxidized state via an oxygen source (oxidant), establishing thus a cyclic process. If this oxygen source is water or CO2 this step (water- or carbon dioxide splitting, respectively) produces hydrogen or CO. If oxygen (e.g. in air) is used as an oxidant according to the following reaction scheme: MeOred þ O2 ðgÞ ! MeOox þ ðDH Þ

ð2Þ

the “net” result of reactions (1)+(2) is not the production of a particular chemical but the exploitation of the heat effects of the reactions for the “storage” of solar heat via the so-called ThermoChemical Storage (TCS). The operation principle of such a cycle uses the heat produced by e.g. a solar receiver during on-sun operation to power an endothermic chemical reaction like (1); should this reaction be completely reversible like (2) the thermal energy can be recovered completely by the reverse reaction taking place during off-sun operation (Ervin, 1977; Wentworth and Chen, 1976). Such a potential for Thermal Energy Storage (TES) integration within a Solar Thermal Power Plant (STPP) is one of the main differences between CSP and other renewable energy technologies. The concept of thermal storage is simple: throughout the day excess heat produced is diverted to a storage material. When electricity production is required after sunset, the stored heat is released into the steam cycle and the plant continues to produce electricity. Nowadays, most of the CSP plants have some ability to store heat energy for short periods of time and thus have

w

instantaneous actual oxygen storage capacity of the redox material normalized by the maximum one instantaneous concentration of adsorbed oxygen (moles of O2 per g of redox material).

Subscripts H 2O steam O2 oxygen oxidant gaseous oxidant, e.g. H2O, CO2, O2 ox oxidation/oxidized red reduction/reduced redox reduction–oxidation

a “buffering” capacity that allows them to smooth electricity production and eliminate the short-term variations other solar technologies exhibit during cloudy days. In fact, there are three types of implementing this TES approach based on the “nature” of heat to be stored: sensible, latent and thermochemical heat, i.e. heat produced through reversible chemical reactions like the ones above. The characteristics and the latest developments of these technologies are described in several recent reviews (Bauer et al., 2012; Fernandes et al., 2012; Gil et al., 2010; Kuravi et al., 2013; Petrasch and Klausner, 2012; Siegel, 2012). For TCS implementation is necessary that the chemical reactions involved are completely reversible. Several reversible reactions with significant heat effects have been proposed for exploitation: the most typical among gas–solid decomposition ones are those of metal hydroxides (Murthy et al., 1986), carbonates (Schaube et al., 2011) and oxides (Wentworth and Chen, 1976), under the reaction schemes (3)–(5) respectively: CaðOHÞ2 þ DH $ CaO þ H2 O

DH ¼ 100 kJ=molreact ð3Þ

CaCO3 þ DH $ CaO þ CO2

DH ¼ 167 kJ=molreact

Co3 O4 þ DH $ 3CoO þ 1=2O2

ð4Þ

DH ¼ 200 kJ=molreact ð5Þ

Thermochemical heat storage has several advantages over latent and sensible heat storage technologies: higher storage energy densities achievable, indefinitely long storage duration at near ambient temperature, heat-pumping capability and suitability for large scale. Solar thermal technologies via thermochemical conversion paths offer the prospect of inherent energy storage for continuous (24 h) generation of electricity, an increasingly significant issue as the world moves towards a truly renewable energy based economy.

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

Even though an analysis of possible thermochemical storage materials was given already in the 1970s (Wentworth and Chen, 1976) only recently thermochemical reactions regained an important role in the development of high temperature heat storage. Even so the development of those storage systems is still at an early stage compared to the status of development of sensible and latent heat storage ones (Gil et al., 2010). Several materials have been tested, e.g. sulphates (Tmar et al., 1981), calcium carbonate (Edwards and Materic´, 2012; Flamant et al., 1980; Kubota et al., 2000) or hydroxide (Murthy et al., 1986; Schaube et al., 2012). However decomposition reactions of carbonates or hydroxides produce CO2 or H2O respectively, which have to be separated/evaporated. In this respect, among the possible reversible gas–solid reactions with substantial thermal effects, the utilization of a pair of redox reactions involving solid oxides of multivalent metals like in (5) is most attractive for large-scale deployment especially in STPPs using air as the heat transfer fluid. In this case air can be used as both the heat transfer fluid and the reactant and therefore can come to direct contact with the storage material (oxide) with the two reduction/oxidation reactions producing simply oxygen-rich or oxygen-lean air (Gil et al., 2010). Such features render the particular technology ideally suited for large-scale deployment coupled to STPPs that use an open volumetric ceramic honeycomb receiver array technology like for example, the Solar Tower Ju¨lich (STJ), since it involves the same heat transfer fluid (air), has a simple set-up and can be matched to the plant’s operating temperature by proper selection of the storage materials. Currently, at rated operation conditions STJ’s heat storage system based on sensible heat, is cycled between 120 and 680 °C and supplies a storage capacity of about 9 MWh (Zunft et al., 2009). Therefore current engineering challenges in the field of redox-pair thermochemical cycles for solar-aided water and CO2 splitting that have to do with the synthesis of active redox materials (Agrafiotis et al., 2012; Furler et al., 2012; Stamatiou et al., 2010) as well as with the design and operation of novel solar receiver/reactor systems (Agrafiotis et al., 2005; Neises et al., 2012) are also common to TCS applications providing a large potential for the exploitation of “technology transfer” between the two fields. Some major differences between these two applications, however, are that in the case of TCS, the targeted material compositions are not characterized by high chemical affinity towards the production of a particular chemical end-product species (H2/CO) but rather from moderate to high heat effects of oxidation/reduction and high extent of reversibility. In addition, in H2O/CO2 splitting the thermal reduction step that involves the production of oxygen is favored only if the partial pressure of oxygen above the material is very low, i.e. under inert atmospheres. In contrast, thermochemical storage operation is targeted to materials that can be thermally reduced and oxidized under air atmosphere. The operating temperatures are also different: in the H2O/CO2 splitting cases the solar reactors are

191

heated directly by concentrated solar irradiation at temperatures around or exceeding 1400 °C for thermal reduction whereas TCS oxide materials should be capable of operation in the temperature range that can be achieved within the new generation of volumetric-receivers-based solar tower power plants i.e. approximately 100–1200 °C. Several oxide systems have been screened with respect to their thermochemical storage potential capability in the framework of a recent General Atomics-DLR bilateral collaboration (Wong, 2011). A number of such oxides has been eliminated either due to their redox temperatures being too low for efficient heat recovery (Cr5O12, Li2O2, Mg2O) or due to their raw materials and processing cost being too high (PtO2, Rh2O3, UO3). Based on the combination of demonstrated thermochemical redox activity and economic potential, Co3O4, BaO, Mn2O3, CuO, Fe2O3, Mn3O4 and V2O5 were selected for further developmental and design studies, whereas secondary oxide addition from abundant raw mineral sources was identified as a means to increase materials performance and potentially to decrease their cost to levels meeting eventually the USA.’s Department of Energy (DOE) storage cost and LCOE targets of $15/kWh and $0.09/kWh respectively. Co3O4 is considered among the most attractive of these systems since its reduction in air under atmospheric pressure (forward reaction of reaction scheme (5) above) takes place at about 900 °C, a temperature that can be achieved within the new generation of volumetric-receivers-based solar tower power plants. In addition, its energy density has been reported as 844 kJ/kg (Wong, 2011), among the highest of such oxide systems. Preliminary material tests of reaction scheme (5) showed good reaction kinetics and long-term material stability (Hutchings et al., 2006; Wong et al., 2010). Our research efforts in the field are two-fold: on the one hand, based on an extensive lab-scale experimentation screening plan, to improve the “chemical” characteristics of the process by investigating and “fine-tuning” the chemical composition and structural properties of redox pair oxide materials. The target is to quantify and maximize their thermochemical storage performance achieving in parallel fast reduction and re-oxidation kinetics, prolonged lifetime with constant activity and cyclability, combined with reasonable cost. On the other hand the major technical challenge for the industrial implementation of these applications lies in the proper design and operation of TCS reactors that will operate simultaneously and efficiently as heat exchangers. Our current goal is the development and manufacture of a pilot-scale TCS reactor/heat exchanger based on redox oxides, its coupling to an existing solar furnace facility and the on-site validation of the technology. In this respect this first part of the work presents results on the former issue, i.e. the identification of suitable oxide compositions with Co3O4 as the starting point and the optimization of their heat storage/release operating parameters, whereas implementation and testing of reactor concepts are the subject of the following parts.

192

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

2. Experimental Three raw Co3O4 oxides from different suppliers, namely VWR International bvba/sprl (Leuven, Belgium), Materion Advanced Chemicals (Milwaukee, USA) and Beijing Cerametek Materials Co., Ltd. (Beijing, China) were employed, hereafter to be referred as powders No1, No2 and No3, respectively. Their chemical analyses, as provided by the suppliers, are shown in Table 1. Their particle size distribution was measured in-house with a Malvern Mastersizer 2000Ò Low Angle Laser Light Scattering Analyzer (LALLS) and their specific surface area (BET) with a Thermo Scientifice SurferÒ Nitrogen porosimeter. The cobaltates were synthesized with Co3O4 powder No1 and the respective oxide of the second metal via co-firing in a box furnace (Nabertherm) under air. With respect to the other materials CuO and Na2CO3 were purchased from MERCK KGaA (Darmstad, Germany) and NiO and MgO from Materion Advanced Chemicals (Milwaukee, USA). Parametric ThermoGravimetric Analysis (TGA) and Differential Scanning Calorimetry (DSC) experiments were performed with a Netzsch STA 449 F3 JupiterÒ instrument. The TGA experimental protocol was as follows: a powder quantity of the order of 100 mg of the fully oxidized state of the redox oxide was placed on an approximately 2 cm2 flat alumina holder, covered with Pt foil in the TGA furnace. An Argon gas flow rate of 10 sccm was used as a protective gas at all times passing through the balance hosing before entering the furnace. The furnace started to be heated with a specific heat-up rate (°C/min) under air atmosphere with an air flow rate of 100 sccm.

As the temperature was raised, at some point a (reduction) temperature is reached (hereafter to be denoted as Tred throughout the manuscript) where the oxide should start dissociating according to the forward reaction of scheme (5) above. This reduction is characterized by weight loss (detectable by TGA), endothermicity (detectable by DSC) and oxygen evolution in the gas phase (detectable by mass-spectrometry). To ensure complete dissociation, the materials were heated to a targeted temperature above Tred (hereafter to be denoted as Thigh) and were left to dwell at this temperature for some time (the completion of reduction is induced from stabilization of weight observed). Then the material started to be cooled down with the same rate. At a certain temperature (hereafter to be denoted as Tox) the oxidation reaction starts (reverse reaction of scheme (5) above) accompanied by the reverse phenomena: weight increase, exothermicity and oxygen uptake from the gas phase. Theoretically the oxidation temperature Tox should be equal to the reduction (dissociation) temperature Tred observed during heating (Tred = Tox = Tredox) but in practice some deviations might occur due to hysteresis phenomena depending on the particular experimental set up as it will be shown below. If the reaction is completely reversible, the material should be oxidized back to its initial state and acquire its initial weight before reduction. To ensure complete oxidation, with the same rationale as above, the material was cooled down to a temperature slightly lower than that where oxidation began (hereafter to be denoted as Tlow) and left to dwell at this temperature for some time. Then, to ensure reproducibility, a second or more two-step (reduction/oxidation) cycles were performed under identical conditions. In all cases, after the completion of the last

Table 1 Chemical analysis of the Co3O4 powders employed as provided by the suppliers. Co3O4 Powder Supplier Code number

VWR International bvba/sprl No1

Materion advanced chemicals No2

Beijing Cerametek Materials Co., Ltd. No 3

Certificates of analysis by suppliers Size range Typical Purity Assay (Co)

N/A N/A >70%

325 mesh 99.5% 73.76%

3–10 lm 99.9% 73.7%

Impurities (%) Cd Ni Fe Zn Cu Mn Ca Mg Na Pb Cr Ga

<0.005 <0.1 <0.06 <0.01 – – – – – – – –

<0.0001 <0.0004 – – – – – <0.01 0.2980 <0.0002 0.002 <0.01

0.0042 0.0058 0.0012 0.007 0.008 0.0022 0.0048 0.011 0.0007 – –

Properties measured in-house Mean particle diameter (lm) Specific surface area (BET, m2/g)

3.97 1.8

6.78 4.0

10.0 19.0

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

programmed cycle, one more “half-step” – reduction - was performed under the same conditions. At the completion of this reduction step at Thigh, the TGA atmosphere was switched from air to 100 sccm Argon gas and the material was cooled down to room temperature. By this way the reduced state of the oxide could be “frozen” to be postanalysed comparatively to the initial, oxidized state. Structural characterization by X-Ray Diffraction analysis (XRD) was performed using a Siemens (Karlsruhe, Germany) D-500 Kristalloflex X-ray powder diffractometer (Cu Ka radiation) in the diffraction angle (2 theta) range of 10–80°. Microstructural observations by Scanning Electron Microscopy (SEM) were performed with a ZEISS ULTRA 55 FEG (Karl Zeiss, Oberkochen, Germany) instrument coupled with an INCA Pentafet x3 EDS Xray microanalysis system from Oxford Instruments. The first experiments involved TGA (weight change recording) only, aimed to identify the optimum process conditions, e.g. heat-up/cool-down rate, maximum and minimum operating temperature, dwell time at Thigh and Tlow etc. These parameters were progressively varied according to the experimental findings, focusing in particular on the establishment of conditions for achievement of complete re-oxidation. In parallel, before optimizing further the redox cycle parameters, several mixed Co3O4-based oxides were prepared and tested, in order to check any beneficial effect of a second metal on the performance of Co3O4. The materials targeted were cobaltates of the generic stoichiometry MeCo2O4 where Me = a bivalent metal; in this work Ni, Cu and Mg were tested (Angus et al., 2012; Klissurski and Uzunova, 1994). Such mixed oxides could in principle be prepared via the calcination under air of a mixture with the proper analogies of Co3O4 and the second metal oxide according to the following reaction scheme: 2Co3 O4 þ 3MeO þ 1=2O2 ! 3MeCo2 O4

ð6Þ

The calcination temperature has to be high enough to provide for the mixed spinel formation but lower than 885 °C to avoid the thermal reduction of Co3O4 observed; thus the powder mixtures were dry-mixed and calcined at 800 °C for 4 h under air atmosphere. Subsequently, for specific experimental conditions, combined TGA/DSC experiments were performed to quantify the relevant heat effects. In several experiments the process gas effluent was diverted to a mass spectrometer (Pfeiffer, Omnistar) via which the Oxygen concentration was constantly monitored and recorded. However, given that the process is performed under air i.e. at oxygen concentration levels around 21 vol%, only minute changes in O2 concentration in the surrounding gas can occur due to the evolution/uptake of oxygen from the redox oxide. Thus the mass-spectrometer’s signal was used only qualitatively in conjunction with the TGA curves to corroborate such O2 evolution/uptake at a particular temperature; the exact amount of oxygen evolved/uptaken was inferred from the weight change curves of the TGA. Therefore, mass-

193

spectrometer’s signals, even though available for several experiments, are shown for only one. Finally, a further issue was to clarify whether the differences in redox behavior as well as in microstructure between the powders, observed during the course of the experiments (see below) could be attributed to differences in composition. The most dominant difference from Table 1 is the high Na content of Co3O4 powder No2, corresponding to a Na/Co atomic ratio of 0.01. Thus samples of the other two powders with the same atomic ratio, as well as samples of all three powders with a higher ratio of Na/ Co = 0.5, were prepared by mixing appropriately weighted amounts of Co3O4 and Na2CO3 powders and calcining under air at 800 °C for 4 h – with the same rationale as with that of the cobaltates synthesis – and tested via TGA. 3. Results 3.1. TGA cyclic reduction–oxidation experiments 3.1.1. Effect of upper and lower temperature limits of the cycle The first set of experiments was performed with ramp (heat-up/cool-down) rate of 10 °C/min, targeted to identify the upper and lower temperature limits and the necessary dwell time to achieve complete reduction and re-oxidation. The results (weight change and temperature curves vs. time) with three representative experimental protocols are shown comparatively in Fig. 1. In the first experiment (Fig. 1a), 1100 °C and 600 °C were selected as the upper and lower temperature limits Thigh and Tlow respectively, with dwell times of 1 h at each level and all Co3O4 powders were tested, exhibiting a qualitatively similar behaviour. Upon heating, reduction starts at T  885 °C, a value very close to that of 890 °C reported in the literature (Hutchings et al., 2006; Wong, 2011). Reduction is very fast and close to completion within a very short time during further heatup from 885 °C to 1100 °C. During dwell at 1100 °C, reduction is slowly driven to full completion and the samples’ weight is stabilized. The maximum weight loss recorded is very close to that calculated from the stoichiometry of reaction (5), 6.64%, i.e. the materials are practically fully reduced during the reduction step. Upon cooling, the weight remains constant until the same temperature of 885 °C (Tox) is “crossed”, where oxidation starts inferred from an abrupt weight loss. However, upon progressive cooling towards the lower level limit of 600 °C there exists a threshold temperature – in this particular experiment  720 °C – below which the rate of the oxidation reaction for powders No1 and No2 drops significantly and therefore, the weight curve becomes much less steep. As a result, during dwell at the selected temperature Tlow of 600 °C, the oxidation rate is very low and oxidation cannot be completed: the materials have not re-gained their original weight but less than 50% of it. After the end of dwell at 600 °C, during heating from 600 °C upwards, the rate starts to increase again, at first slowly, but as soon

194

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

100 98

T = 885 oC

t=3 hrs

96

100 98

T = 885 oC

t=1 hr

96 94 100 98

T = 885 oC

96

t=1 hr

o

T = 720 C

94 92

0

100

200

300

400

500

600

Temperature ( o C)

Weight change (%)

94

1200 1000 800 600 400 200 (c) 1150 (1 hr) - 800oC (3 hrs) 0 1000 800 600 400 200 o (b) 1150 (1 hr) - 800 C (1 hr) 0 Weight change (%) 1000 Temperature 800 Powder No1 Powder No2 600 Powder No3 400 Ramp rate: 10 oC/min 200 o (a) 1100 (1 hr) - 600 C (1 hr) 0 700 800 900

Time (min) Fig. 1. Effect of operational parameters (maximum and minimum temperature and dwell time) on the completion of oxidation step of Co3O4 powders tested in TGA with heatup/cooldown rates of 10 °C: (a) test of three powders between 1100 and 600 °C and dwell time of 1 h at the maximum and at the minimum temperature; (b) test of powder No1 between 1150 and 800 °C and dwell time of 1 h at the maximum and at the minimum temperature; and (c) test of powder No1 between 1150 and 800 °C and dwell time of 1 h at the maximum and 3 h at the minimum temperature.

as T  720 °C is “crossed”, abruptly, and the materials continue to oxidize gaining weight until the temperature of 885 °C is reached where they start reducing again. The performance of powder No3 has an identical qualitative trend; however, during cooling from  720 to 650 °C its oxidation reaction rate is higher and thus its oxidation proceeds to a much higher extent. The phenomena are repeated in an identical way during the second cycle. Thus the over-all time allocated for oxidation during cool-down from 870 to 600 °C, dwell of 1 h at 600 °C and heat-up from 600 to 885 °C together with the slow oxidation kinetics at 600 °C did not suffice to complete oxidation for all powders, even though the extent of oxidation among them differs. The other experiments depicted in Fig. 1 were performed with Co3O4 powder No1. In the experiment shown in Fig. 1b, a higher temperature – 800 °C instead of 600 °C – was selected as the lower temperature limit Tlow where the material should dwell; at the same time the upper temperature limit was increased from 1000 °C to 1150 °C. It can be seen that the reduction step is identical; however even though Tlow in this case is higher than 720 °C and thus the sample has re-gained a much higher percentage of its original weight still the oxidation step is not completed; no “flat” part of the weight curve (weight stabilization) can be observed. This means that still the time allocated to oxidation is lower than that needed for completion. Thus a third experiment between the same temperature limits (1150–800 °C) was performed but allocating 3 h dwell time at the lower temperature limit

for completion of oxidation. From the respective weight change curve shown in Fig. 1c, a significant “improvement” in the shape of the oxidation curves can be observed: they both have reached a plateau before reduction starts, indicating that the oxidation has been completed during the allocated time of 3 h. A closer inspection has revealed a slight reduction of the maximum weight “re-gained” after each oxidation step. From this set of experiments it can be concluded that the reduction step is relatively fast, complete and reproducible. In this respect, there is no particular need for either reaching upper temperatures much higher than the reduction one or for prolonged heating. On the contrary to ensure completion of the re-oxidation step within a reasonable time, the oxidation reaction has to be performed at a temperature high enough to provide for high oxidation rate i.e. only slightly lower than the reduction temperature. Thus to further optimize the cycle for Co3O4 and check for conditions of complete re-oxidation, further experiments were performed, “narrowing” the temperature operating range i.e. selecting the maximum and minimum operating temperatures Thigh and Tlow slightly above and below, respectively, of the reduction temperature Tredox. Fig. 2a shows an experiment performed with powder No1 with the same ramp rate of 10 °C/min but between Thigh = 930 °C, Tlow = 860 °C and respective dwell times of 1 and 3 h. The higher temperature plateau (Tlow) selected for re-oxidation in this case, resulted in a much faster

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

(a) 102

o

195

1200

6.0x10

1000

5.0x10

800

4.0x10

-7

o

Co3O4 No1: 930-860 C, 10 C/min

101 t=1 hr

100

T = 885 C

98 97

4.2x10

4.1x10

4.0x10

3.9x10

-7

t=3 hrs

-7

t=1 hr t=1 hr t=1 hr

-7

-7

600

-7

60

65

70

75

80

Time (min)

-7

Ion current [A]

3.4x10

95

-7

o

Weight change (%) Temperature

96

3.0x10

-7

Ion Current [A]

Ion current [A]

4.3x10

Temperature ( C)

99

400

2.0x10

200

1.0x10

0

0.0

-7

-7

Ion current [A]

Weight change (%)

-7

o

94

3.3x10

-7

3.2x10

-7

3.1x10

-7

3.0x10 130

135

140

145

-7

150

Time (min)

93 92 0

100

200

300

400

500

600

700

Time (min) 1200

(b)102

o

o

Co3O4 No1: 985-785 C, 10 C/min

101 o

T = 985 C

1000

100 o

T = 885 C

99 o

800

98 6.26 %

600

97

o

Temperature ( C)

Weight change (%)

T = 785 C

6.46 %

96 400 95 94

200 Weight change (%)

93

Temperature

0

92 0

25

50

75

100

125

150

175

200

Time (min)

Fig. 2. Further effect of operational parameters (maximum and minimum temperature and dwell time) on the completion of oxidation step for the Co3O4 powder No1 tested in TGA with heatup/cooldown rates of 10 °C: (a) test between 930 and 860 °C and dwell time of 1 h at the maximum and 3 h at the minimum temperature; and (b) test between 985 and 785 °C without dwell time neither at the maximum nor at the minimum temperature.

reaction rate and the achievement of a much higher weight gain during re-oxidation. Within these temperatures, it can be induced that 30 min at the reduction and 2 h at the oxidation plateau temperature suffice for full reduction and re-oxidation. In addition, a typical gas-phase oxygen concentration profile as recorded from the mass-spectrometer is shown. Small upward and downward peaks can be observed every time the temperature Tredox is crossed – the relevant points of the first cycle are “magnified” in the graphs embedded in Fig. 2a – verifying the relevant oxygen release/uptake. Finally, Fig. 2b shows the results of an experiment performed again with powder No1 with the same ramp

rate of 10 °C/min as before, between Thigh = 985 °C and Tlow = 785 °C, i.e. equidistant by 100 °C from the redox temperature of 885 °C, but without any dwell time at either one. Almost complete re-oxidation can be achieved in this case as well; the weight re-gained during re-oxidation is less by only 0.2% of that lost during reduction. This implies that an extended dwell at the oxidation temperature perhaps is not required, but completion of re-oxidation can be achieved by proper selection of the cool-down rate so that to allocate enough time for the completion of the oxidation reaction, as it will be shown below.

196

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

3.1.2. Effect of a second metal cation introduced in the Co3O4’s spinel structure The XRD spectra of the products of reaction (6) above are compared to that of the reactant Co3O4 powder No1 in Fig. 3a (diffraction angle range 10–80°) and 3b (diffraction angle range 35–44° where the major peaks of the primary phases are located). Among the products only the Co–Cu one is single-phase spinel; the other products contain, in addition to the major mixed spinel phase formed, small quantities of the respective bivalent oxide phase, NiO, MgO and ZnO respectively. A small amount of bivalent

(a)

CoO exists also in the reactant Co3O4 No1 and the major peak of Co3O4 is slightly shifted to higher diffraction angles with respect to the “standard” one. All these powders - except the one with Zn which could face Zn volatilization issues at these temperatures - were comparatively tested for cyclic reduction/oxidation with heat-up/cool-down rates of 10 °C, between 1150 and 800 °C and dwell times of 1 and 3 h at the maximum and minimum temperature respectively. These conditions - wide temperature range spanned and long reaction times - were selected to ensure complete re-oxidation of the mixed

10000

o o

o: ZnO

o o

o

o

o

Intensity (arbitrary units)

8000

No CuO

6000

+ +

+: MgO

4000

x

x

x

X: NiO

2000

*

* : CoO

*

0 10

15

20

25

30

35

40

45

50

55

60

* 65

70

75

80

o

Diffraction angle (2 theta) ( )

(b)

o

Co3O4 No1 reactant and products of 3MeO (Me=Ni,Mg,Cu,Zn)+2Co3O4(No1) fired at 800 C for 4 hrs

12000

Co3O4 No1,

Ni-Co-O,

Mg-Co-O,

Cu-Co-O,

Zn-Co-O

--- Ref Co3O4

10000

Intensity (arbitrary units)

o o: ZnO

8000

No CuO

6000

+ +: MgO

4000

x

x X: NiO

2000

* : CoO

*

0 35

36

37

38

39

40

41

42

43

44

o

Diffraction angle (2 theta) ( ) Fig. 3. Phase composition (XRD) comparison of the various cobaltate powders after calcination at 800 °C for 4 h under air: (a) diffraction angle range 10– 80o; and (b) diffraction angle range 35–44o.

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

197

 During further heat up from Tredox to Thigh the relevant weight change curves of Co3O4 and MgCo2O4 are “smooth” implying that these materials suffer no further phase changes. On the contrary, the “steps” and “angles” occurring in the respective curves of NiCo2O4 and CuCo2O4 advocate for further phase transformations occurring in these compositions. This is particularly true for the Co–Cu system where the two consecutive “steps” in the weight loss could be attributed to a first dissociation of the spinel CuCo2O4 to

compositions. The respective weight change curves are shown in Fig. 4a, whereas a “magnification” of the first cycle is shown in Fig. 4b. The following remarks can be made with respect to Co3O4 and the cobaltates – NiCo2O4, MgCo2O4 and CuCo2O4:  All four samples are reduced around the same Tredox temperature of 885 °C. However their weight loss is different: Co3O4 exhibits the highest and the other three lower and similar among them.

1200

(a) 101 T

100

1000 99 800

97 600 96

o

Weight change (%)

o

T = 885 C

Temperature ( C)

98

400

95

94 200 Co3O4 No1

Wt. (%)

93

NiCo2O4 MgCo2O4

o

MeCo2O4, (Me=Ni, Mg, Cu): 1150-800 C

CuCo2O4

0

92 0

100

200

300

400

500

600

700

800

900

Time (min) 1200

(b) 101 100

1000 99

Wt. (%)

800

97 600 96 400

95

Temperature ( oC)

Weight change (%)

T

T = 885 oC

98

Co3O4 No1

94

NiCo2O4 MgCo2O4

200

CuCo2O4

93

MeCo2O4, (Me=Ni, Mg, Cu): 1150-800oC 0

92 0

100

200

300

400

Time (min) Fig. 4. Weight change comparison among the various cobaltate powders tested with heatup/cooldown rates of 10 °C in reduction–oxidation between 1150 and 800 °C and dwell times of 1 and 3 h at the maximum and minimum temperature respectively: (a) full duration of experiment; and (b) first step only.

198

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

Co3O4 and CuO and to a subsequent reduction of CuO to Cu2O which is known to take place at a temperature around 1050–1100 °C. XRD has revealed that the reduced form consists of the two individual reduced oxides Cu2O and CoO indicating that the spinel is eventually probably dissociated irreversibly to the individual oxides. This situation is similar during cool-down from Thigh to Tredox.  During isothermal dwell at 1150 °C, the profiles of NiCo2O4 and MgCo2O4 are similar in shape to that of Co3O4 indicating minor weight loss with a slow rate and eventual weight stabilization. In contrast, the Co– Cu–O system seems to loose further weight in a linear fashion, attributed to possible volatilization since the operating temperature is already very close to the melting point of Cu2O (1232 °C).  During further cool-down from Tredox to Tlow two different material “groups” can be distinguished: Co3O4 and NiCo2O4 are characterized by higher oxidation rates and approach almost complete weight recovery. On the contrary the oxidation rates of MgCo2O4 and CuCo2O4 are very slow and during dwell at Tlow these two materials cannot recover their initial weight. The first part of the weight curve of MgCo2O4 in particular, is identical to that of NiCo2O4 both with respect to the shape of the curve as well as to the total weight loss; however during oxidation its behaviour is similar to that of CuCo2O4. This behaviour is identical during the second cycle. A comparison among the four reduction products from Co–O, Ni–Co–O, Mg–Co–O and Cu–Co–O is shown in Fig. 5a (diffraction angle range 10–80°) and 5b (diffraction angle range 35–60°). In all cases the major reduction product is CoO. In the case of Cu–Co–O system it is impossible to distinguish the peaks of CoO from those of Cu2O, thus the product consists probably of both reduced oxides. In the cases of Ni–Co–O and Mg–Co–O the other bivalent metal (Ni, Mg) does not form a separate oxide but is incorporated within this structure (Co, Me)O causing a shift on the relevant peak with respect to that of “pure” CoO. With respect to the two materials that seem to work reproducibly from cycle-to cycle i.e. “pure” Co3O4 and NiCo2O4 only the product of the reduction of Co3O4 is single-phase reduced phase CoO; the reduction product of NiCo2O4 still contains some minor amount of the non-reduced phase Co3O4. However, since there are no traces of NiO present it seems that NiCo2O4 does not dissociate to NiO and Co3O4 upon heating; probably retains the mixed spinel form, the extent of which, though, during reduction to CoO, is not complete with a percentage of the “original” spinel (containing the Ni) not being reduced and thus exhibiting less weight change during reduction than that of Co3O4. In summary, partial substitution of the Co+2 cation by another bivalent ion in the spinel’s structure – at least with the cations tested in the present study Ni, Cu and Mg – did

not improve the redox performance of “pure” Co3O4. Therefore, further parametric experiments were performed with “pure” Co3O4 powders. 3.1.3. Effect of ramp rate and reactant powder A conclusion from the experiments above is that pure Co3O4 can operate in a satisfactory way cycled between 985 and 785 °C without dwell at either the upper or the lower temperature limit. The time allocated for oxidation as the powder is cooled down from Tredox to Tlow and heated back to Tredox again, depends on the cool-down/ heat-up (ramp) rate. Thus, experiments with three different ramp rates, 5, 10 and 15 °C/min were performed with all Co3O4 powders to investigate further any difference in performance due to the raw material’s source. The results are shown comparatively in Fig. 6 (powder No1 was cycled 5 times at the highest ramp rate of 15 °C/min). A first qualitative conclusion valid for all powders is that under the conditions tested, reduction is always full during the time allocated for heating the powder from Tredox to Thigh and cooling it back to Tredox again. The lower part of the weight change curves is always flat under all ramp rates tested. However, focus should be placed on the effect of process parameters on the shape of the weight change curve during re-oxidation (upper part of the curve). The ramp rate affects the re-oxidation extent of powder No1; the lower the ramp rate the higher the extent of re-oxidation and the more the respective part of the weight curve tends to “flatten” and become a “mirror” image of that during reduction. On the contrary the ramp rate does not have an effect on powders No2 and 3: in all cases both materials seem capable of re-gaining their initial weight within the time allocated from crossing the redox temperature until reaching it again. The corresponding weight change curves are flat at both the upper and lower parts under all circumstances tested. Even though the difference between the maximum and the minimum of the weight curves is almost the same for all powders, some qualitative differences could be distinguished and quantified from the analysis of the respective weight evolution curves. Reduction and especially, re-oxidation are faster for powders No2 and 3. Approximate completion times for the case of ramp rate of 5 °C/min are 9 min for reduction and 30 min for oxidation compared to 20 and 32 min respectively for powder No1. A synopsis of the reduction and oxidation temperatures for the three powders as a function of ramp rate is given in Table 2. For all powders and all experiments the values do not differ from cycle to cycle in the same experiment. For powder No1 the reduction and oxidation “ignition” temperatures are practically equal to each other, whereas for powders No2 and 3, oxidation starts at about 20–30 °C lower than where reduction started – irrespective of the ramp rate for both powders. In summary, in this set of experiments powders No2 and 3 exhibited better and reproducible performance, being capable of full reduction and, mostly, oxidation within

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

(a) 10000

199

Phase composition of reduction products after 2 cycles between 1150-800oC Co3O4, Ni-Co-O, Mg-Co-O, Cu-Co-O --- Ref CoO

9000

Intensity (arbitrary units)

8000 7000 6000 5000 4000 3000 2000 1000 0 10

30

20

40

50

60

70

80

o

Diffraction angle (2 theta) ( )

(b) 10000

o

Phase composition of reduction products after 2 cycles between 1150-800 C Co3O4, Ni-Co-O, Mg-Co-O, Cu-Co-O --- Ref CoO, --- Ref Co3O4

9000

Intensity (arbitrary units)

8000 7000 6000 5000 4000 3000 2000 1000 0 35

40

45

50

55

60

o

Diffraction angle (2 theta) ( ) Fig. 5. Phase composition (XRD) comparison of the reduction products of the various cobaltate powders after cyclic reduction–oxidation with heatup/ cooldown rates of 10 °C between 1150 and 800 °C and dwell times of 1 and 3 h at the maximum and minimum temperature respectively in diffraction angle range: (a) 10–80o; and (b) 35–60o.

the temperature range 800–1000 °C, practically nonaffected from the process parameters. 3.2. TGA/DSC cyclic reduction–oxidation experiments The next step involved combined TGA/DSC experiments under selected operating conditions in order to quantify the heat effects of the redox reactions. An initial comparative experiment between all Co3O4 powders under the ramp rate of 5 °C/min demonstrated that powder No2 absorbed/released more heat per step under the same

experimental conditions; therefore parametric studies on the effect of ramp rate on the heat effects of the redox reactions were performed with that powder. From the respective experiments with 5, 10 and 15 °C/min comparatively shown in Fig. 7, it is clear that the ramp rate affects the heat absorbed/released. As the ramp rate increases the DSC peaks become sharper – since the reaction takes place with greater speed in the higher temperature region and thus is completed within a narrower temperature interval (Hatakeyama and Liu, 1998) – and consequently the heat effects recorded are lower, as the integration of the areas

200

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211 102

Co3O4: 985-785oC

100

1200 900

98

600

96 300

94

15 oC/min

98 96

900 600 300

94

o

Weight change (%)

Wt. (%), Powder 1 Wt. (%), Powder 2 Wt. (%), Powder 3 T (oC), Powder 1 T (oC), Powder 2 T (oC), Powder 3

10 oC/min

92

1200

100

900

98

Temperature ( C)

1200

92 100

600

96 300

94

5 oC/min

0

92 0

50

100

150

200

250

300

350

Time (min) Fig. 6. Weight change comparison of the three Co3O4 powders, during reduction–oxidation between 985 and 785 °C without dwell time, as a function of heating/cooling rate.

Table 2 Onset of reduction and oxidation for the Co3O4 powders as a function of ramp rate. Co3O4 Powder No 1/No 2/No 3 Ramp rate (°C/min)

Tred (°C)

Tox (°C)

DT (°C)

5 10 15

885/912/923 885/893/– 885/890/924

880/887/890 885/873/– 885/870/900

5/25/33 0/20/ 0/20/24

under the relevant curves demonstrated. The calculated areas for all experimental conditions converted from mW/mg to kJ/kg are provided in Table 3 and compared to the theoretical value taken as 844 kJ/kg (Wong, 2011) in the last column, ranging between 43% and 66% of it. It should be noted that the heat of the second/fourth steps (oxidations) is always lower than the heat of the respective first/third/fifth step (reductions). Finally the maxima of the heat absorption/release curves (given on the same Table) occur at higher/lower temperature values, respectively, than the redox temperature. The maximum of the heat absorption takes place always around 915–920 °C irrespective of the ramp rate, whereas the maximum of heat release takes place between 835 and 850 °C with the value of the particular temperature increasing slightly with a decrease of the ramp rate. These observations are in full accordance with those reported in previous studies (Hutchings et al., 2006). 3.3. Long-term TGA cyclic reduction–oxidation experiments Powders No1 and No2 were tested in the TGA between 785 and 985 °C with a ramp rate of 5 °C/min and no dwell for 30 cycles, so that a preliminary comparison could be

made with respect to their longer-term performance, stability and phase composition/ microstructure evolution. This number of 30 cycles was selected as a first, representative and practical laboratory-scale, “long-term” experiment based also on the fact that in studies of up to 500 cycles that are considered as a reasonable industrial validation test period, the major microstructural changes (grain growth) have been reported to take place within the first 50 cycles (Wong, 2011). The results of the relevant experiment for powder No1 are shown in Fig. 8a, whereas in Fig. 8b the first two cycles of the 30-cycles experiment are compared to the 2-cycles experiment under the same conditions. The respective results for powder No2 are shown in a similar way in Fig. 9a and b. The comparison of the first two cycles of the 30-cycles experiment to the “original” 2-cycle experiments (Figs. 8b and 9b) demonstrates the full reproducibility of the process for both powders. Powder No2 performed repeatable cyclic reduction/oxidation operation for 30 times without a significant activity loss. Even though there is a slight “shift” of the weight curve towards lower values at an increased number of cycles (to the right) the maximum distance between the top and the bottom of the weight curve remained practically constant (6.75% for the 1st cycle vs. 6.50% for the 30th cycle). For powder No1 the “shift” of the weight curve towards lower values at an increased number of cycles is more intense and the maximum distance between the top and bottom of the weight curve at the 30th cycle is smaller than that at the 1st (5.69% vs. 6.57% respectively). In conclusion, powder No1 exhibited lower performance than No2 not only with respect to extent of re-oxidation, but also with respect to “robustness” vs. operating conditions as well as long-term performance stability.

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

(a) 102

201 1200

0

96

400

94 0

25

50

75

100

DSC

100

800

98 96

400

94 0

20

40

60

80

100

120

140

160

180 DSC

o

Rate: 10 oC/min

92

-4

0 200 1200

100

2 0 -2 -4 4 2

800

98

4

DSC (mW/mg)

0 150 1200

125

Temperature ( C)

Weight change (%)

-2

Rate: 15 oC/min

92

102

2

800

DSC

98

102

4

TGA T

100

0

96

400

94

Co3O4 Powder No 2: 985-785oC

-2

Rate: 5 oC/min

0

92 0

50

100

150

200

250

300

-4

350

Time (min)

(b) 14

Co3O4 Powder No 2: 985-785oC

12

Rate: 15oC/min Rate: 10oC/min Rate: 5oC/min

DSC (mW/mg)

10

8

6

4

2

0

-2 0

50

100

150

200

250

300

350

Time (min) Fig. 7. Effect of heating/cooling rate on: (a) weight change and reduction/oxidation heat effect (TGA/DSC); (b) reduction/oxidation heat effect only (DSC), for Co3O4 powder No 2 tested between 985 and 785 °C (endotherms: pointing upwards, exotherms: pointing downwards).

3.4. Pre- and post-characterization To identify possible causes for the differences in behavior among the Co3O4 powders used, extensive characterization techniques were employed both on the as-received powders as well as on their respective (reduction) products after cyclic operation. 3.4.1. As-received Co3O4 powders The XRD spectra of the Co3O4 powders employed as provided by the suppliers, are compared in Fig. 10a. The only visible difference is that powder No1 contains a small

quantity of the bivalent oxide CoO whereas the other two are single-phase Co3O4 spinels. This difference could perhaps explain the initial weight increase observed on powder No1 just before the first reduction (Figs. 1, 2, 6 and 8) attributed to oxidation of the existing small quantity of CoO under heating-up in air. This initial weight increase is absent in the respective curves of the other two powders. The mean particle diameter and specific surface areas of the powders as measured in-house are also provided in Table 1. They are similar to those reported in the literature for Co3O4 powders tested in similar studies (Hutchings et al., 2006). The major part of all powder particles lies

202

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

Table 3 Heats of reduction/oxidation reactions calculated per step from the relevant areas under the curves of the DSC experiments, as a function of powder employed and ramp rate and comparison to the theoretical value (taken as 844 kJ/kg); lower rows: respective temperatures at peak maxima. Co3O4 powder No2 No2 No2 No1 No3

No2 No2 No2 No1 No3

Ramp rate (°C/min)

Calculated area under the DSC curve (mW/mg) 1st peak

2nd peak

3rd peak

4th peak

5th peak

Average of 5

15 10 5 5 5

8.77 9.96 11.49 9.16 6.28

7.88 8.41 9.00 6.77 5.73

7.64 9.03 9.64 9.43 6.42

7.61 8.02 8.17 6.72 5.57

8.32 6.17 8.48 9.60 6.50

8.04 8.32 9.36 8.34 6.10

15 10 5 5 5

Temperature at peak maximum (°C) 917 920 918 918 922

835 835 847 847 844

915 919 918 918 921

837 845 851 851 838

917 920 917 917 922

below 10 microns. Powder No1 is slightly finer than No2 and 3, however powder No3 has the highest specific surface area: almost ten times that of No1 and five times that of No2 – 1.75, 4.04 and 19 m2/g respectively. 3.4.2. Products’ phase composition evolution (XRD) The XRD spectra of the products of powders No1 and No2 after 2 and 30 cycles in the TGA under the same conditions, are compared in Fig. 10b and c respectively. On the one hand the reduction is complete: there are no traces of Co3O4 present in all reduction products. On the other hand, the two products after reduction look identical: only a slight shift of the peaks of the sample after 30 cycles towards higher diffraction angles can be observed. 3.4.3. Products’ microstructure evolution (SEM) To observe the microstructural evolution of the powders used, “comparative” SEM analysis was performed at the same magnifications when possible, on such powder samples in three “states”: as received from the suppliers, after 2 and after 30 redox cycles in the TGA. The respective SEM photographs are shown in Fig. 11a–c respectively, in consecutive magnifications of 20,000, 50,000 and 100,000 times to facilitate direct comparison. In their as-received state (Fig. 11a) the powders exhibit similar microstructural features. They all consist of relatively large agglomerates with dimensions of the order of 10 microns. These agglomerates in turn consist of loosely packed (20 k), smaller dense grains with dimensions of few microns (50 k). Even under the highest magnification (100 k) the morphology of the powders looks very similar. However strong microstructural differences are revealed in the products after 2 cycles (Fig. 11b). The product from powder No1 consists of dense globular grains with hexagonal-like facets which are clearly seen in even higher magnifications (50 k). In contrast, the grains of the product of powder No2 consist of terraced crystal planes resulting in a step-shaped morphology. The product of powder No 3 exhibits also multi-layered crystal planes microstructure.

DH (KJ/kg)

DH (% of th.)

482.6 499.1 561.4 500.2 366.0

57.2 59.1 66.5 59.3 43.3

This microstructural difference is much more intense in the respective products after 30 cycles (Fig. 11c) becoming evident even from the lowest magnification of 20 k. The grains of the product of powder No1 maintain more or less their globular, equiaxed shape. On the contrary, those of powder No2 form sintering contacts leading to small-scale closed ring structures. This difference in morphology becomes much more striking at the higher magnifications of 50 and 100 k. The “multi-layered” structure is also visible in high magnifications on the globules of powder No1 (Fig. 11c, 100 k), but not to the extent observed on the product of powder No2. In fact in the respective micrographs these layers look like regular (or slightly distorted) hexagons with step growth morphology. This microstructure has even resulted in pores the walls of which are formed by regular polygonal geometrical shapes (Fig. 11d). It is interesting to note that the microstructure of Co3O4 after 500 cycles reported in other studies (Wong, 2011) even though at low magnifications resembles very much that of powder No2 shown in the top photograph of Fig. 11c – i.e. elongated, curved, ring-like structures – is not characterized from the same step-shape morphology in high magnifications. 3.5. Effect of Na content of the starting powder As already mentioned, since the most striking difference in composition among the three Co3O4 powders is the high Na content of Co3O4 powder No2, samples of all Co3O4 powders with varying Na content prepared in-house as described above, were comparatively tested under the same conditions, i.e. between 785 and 985 °C with a ramp rate of 5 °C/min and no dwell at either temperature level for 2 cycles. The results are shown in Fig. 12. Addition of a small amount of Na (Na/Co = 0.01) seemed beneficial for Powder No1 improving its weight loss and re-gain during the two redox steps and practically immaterial for powder No3. On the contrary, addition of a high Na percentage (Na/Co = 0.5) essentially “destroyed” the performance of

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

(a) 102

203 1200

Co3O4 powder No 1, 30 cycles: 985-785oC, 5 oC/min

101 1000 100

800

98 600

97

o

Temperature ( C)

Weight change (%)

99

96 400 95

94

200

93 Weight change (%) Temperature

0

92 0

500

1000

1500

2000

2500

Time (min) 1200

(b) 102

Co3O4 Powder No 1: 985-785oC, 5 oC/min; 2 vs. 30 cycles 2 cycles 30 cycles

101

1000

100 800

98 600 97

o

Temperature ( C)

Weight change (%)

99

400

96

95 200 94 Weight change (%) Temperature

0

93 0

50

100

150

200

250

300

350

Time (min) Fig. 8. Long-term (30 cycles) cyclic reduction–oxidation behaviour of Co3O4 powder No 1 in the TGA between 985 and 785 °C with a heating/cooling rate of 5 °C/min: (a) 30 cycles experiment; (b) comparison of the first two cycles of the long-term reduction/oxidation experiment vs. a two-cycle experiment with the same powder under the same conditions.

all three powders. Not only the weight loss during reduction is much lower (3%) but only in the case of powder No3 the reduction proceeded to completion i.e. the weight curve has acquired and maintained a “flat” shape during the process. The “angles” and steps in the cases of the other

two powders indicate, like in the case of mixed Cu/Co systems, formation of mixed Na–Co oxide phases and relevant phase transformations like possible melting preventing complete reduction. This is supported by the relevant SEM photographs of the Co3O4–Na2Co3 mixture with

204

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

(a) 102

1200

Co3O4 powder No 2, 30 cycles: 985-785oC, 5 oC/min

101 1000 100

800

98 600

97

96

Temperature ( oC)

Weight change (%)

99

400 95

94

200

93 Weight change (%) Temperature

0

92 0

500

1000

1500

2000

2500

Time (min)

(b) 102

1200

Co3O4 Powder No 2, 2 vs. 30 cycles: 985-785oC, 5 oC/min

101

2 cycles 30 cycles

1000

99

800

98 600

97 96

400

Temperature (oC)

Weight change (%)

100

95 94

200

93 Weight change (%) Temperature

92 0

50

100

150

200

250

300

0 350

Time (min) Fig. 9. Long-term (30 cycles) cyclic reduction–oxidation behaviour of Co3O4 powder No 2 in the TGA between 985 and 785 °C with a heating/cooling rate of 5 °C/min: (a) 30 cycles experiment; and (b) comparison of the first two cycles of the long-term reduction/oxidation experiment vs. a two-cycle experiment with the same powder under the same conditions.

Co3O4 powder No1 and Na/Co ratio = 0.01 before and after cyclic TGA shown in Fig. 13. In the photographs on the left column (calcination product before TGA) that can be compared directly to those of the first column of Fig. 11a, Co3O4 particles are clearly distinguished distrib-

uted among Na-oxide-based larger hexagonally-shaped layers. After cyclic TGA (right column, to be compared with first column of Fig. 11b) the features of both these components can be still distinguished even though local melting and sintering has occurred. This terraced hexa-

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

(a)

205

Co3O4 reactant No1

7000

Co3O4 reactant No2 Co3O4 reactant No3

Intensity (arbitrary units)

6000

---

Ref Co3O4

---

Ref CoO

5000 4000 3000 2000 1000 0 10

20

30

40

50

60

70

80

o

Diffraction angle (2 theta) ( )

(b)

5000

o

Co3O4 No1 after TGA cycling between 985-785 C

4500

Intensity (arbitrary units)

4000

---

After 2 cycles After 30 cycles Ref Co3O4

---

Ref CoO

3500 3000 2500 2000 1500 1000 500 0 10

20

30

40

50

60

70

80

o

Diffraction angle (2 theta) ( )

(c) 5000

o

Co3O4 No2 after TGA cycling between 985-785 C

4500

Intensity (arbitrary units)

4000

---

After 2 cycles After 30 cycles Ref Co3O4

---

Ref CoO

3500 3000 2500 2000 1500 1000 500 0 10

20

30

40

50

60

70

80

o

Diffraction angle (2 theta) ( ) Fig. 10. X-ray diffraction phase composition comparison of: (a) the three Co3O4 reactant powders employed; (b), (c) the reduction products of powders No 1 and No 2, respectively, after 2 and 30 redox cycles between 985 and 785 °C with a heating/cooling rate of 5 °C/min.

206

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

Fig. 11. Comparison of microstructure evolution for the three Co3O4 powders employed. SEM photographs of the powders under consecutive magnifications of 20,000, 50,000 and 100,000 times: (a) in their as-received state; (b) reduction products after 2 redox cycles; (c) reduction products of powders No1 and 2 after 30 redox cycles; (d) SEM photographs of the reduction product of Co3O4 powder No2 after 30 redox cycles, under magnifications of 40,000 (top) and 100,000 (bottom) times.

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

207

Fig 11. (continued)

gon-like step growth morphology can be distinguished to some extent in the magnifications of 50 and 100 k, even though not in the extent observed on powder No 2 (Fig. 11b and c). 4. Discussion Co3O4 powders from different sources exhibited differences both with respect to redox performance as well as

microstructure evolution. During the oxidation step there is a striking difference between the experiment shown in Fig. 1a and the other two experiments (Fig. 1b and c) in the shape of the weight change curve during heat-up after the end of the dwell at the lower temperature limit until reaching the redox temperature. During this period no sharp weight increase is observed in the experiments of Fig. 1b and c; on the contrary the weight gain seems to follow the “parabolic” behavior observed during the

208

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211 1200

102 o

o

Co3O4 powder comparison 985-785 C, 5 C/min

101 1000 100

800 T

98 600

97

o

Temperature ( C)

Weight change (%)

99

96 400 95

Co3O4 No1 Co3O4 No2 Co3O4 No3

94

200

0.01 Na-doped Co3O4 No1 0.01 Na-doped Co3O4 No3 0.5 Na-doped Co3O4 No1

93

0.5 Na-doped Co3O4 No2 0.5 Na-doped Co3O4 No3

0

92 0

50

100

150

200

250

300

350

Time (min) Fig. 12. Effect of Na addition to Co3O4 powders during cyclic reduction–oxidation in the TGA between 985 and 785 °C with a heating/cooling rate of 5 °C/min.

isothermal step. This difference in behavior can be explained from the modeling of the oxide redox thermochemical cycles proposed in previous works (Agrafiotis et al., 2013; Kostoglou et al., 2011). In these works a phenomenological model was proposed to determine the simplest overall reaction rate expression that can describe the experimental data. With respect to the oxidation step – water splitting in that particular case – a reaction rate depending not only on the partial pressure of oxidant (steam in that case) but also on the instantaneous concentration of the empty oxygen storage sites of the redox solid material was proposed. Among candidate kinetic expressions from the extensive list referred to gas–solid reactions (Francis et al., 2010; Botas et al., 2012), an oxidation rate (or equivalently, a hydrogen production rate) of nth order with respect to steam concentration CH2 O and first order with respect to the empty oxygen storage sites on the surface of the redox material (unimolecular decay law in gas–solid reactions terminology) was assumed as a first starting point. In addition to the water splitting case, a similar reaction rate expression can be used for any other oxidation reaction i.e. carbon dioxide splitting or oxidation from oxygen in air. In all cases the oxidation rate can be written as the factor of a temperature-dependent rate constant and an oxidant-concentration-depended function proportional to the number of empty sites available for oxygen storage on the redox material at a particular temperature: Rox ¼ k ox Wð1  yÞCnoxidant

ð7Þ

where Rox is the oxidation reaction rate in terms of moles of O2 from the gas phase consumed per gram of redox

material per minute, k ox is the temperature-dependent rate constant (units: m3/moles/min) of the oxidation reaction, W is the maximum oxygen storage capacity of the redox material (expressed in moles O2 “stored in the solid” or “consumed from the gas phase” per g of redox material) and Coxidant is the gas phase concentration (in moles/m3) of the oxidant e.g. H2O, CO2 or O2 as in the particular case. It is noted that the exponent n can take in general values from 0 to 1 depending on the dominant step in the oxidation reaction. Finally, y is the instantaneous actual oxygen storage of the solid redox material normalized by the maximum one and takes values between 0 and 1. The quantity (1  y) is proportional to the number of empty sites for oxygen storage on the redox material. The term “instantaneous” means the actual percentage of the storage sites being filled with oxygen from the oxidant at the “termination” of the previous step of the cycle – depending on the “history” of that previous step. This distinction is necessary since W may have been inaccessible in practice. The quantity Wð1  yÞ which is the instantaneous oxygen storage capacity of the redox material (in moles of “stored” oxygen per g of redox material) can be considered as the instantaneous “driving force” for the reaction. Such a simplification is possible only in the case of a linear relation between Rox and y as the one described in Eq. (7). The dependence of the rate constant k ox on temperature can be an Arrhenius-type one if the rate-controlling step during oxidation is a chemical reaction per se (depending exponentially on temperature) or a less-strong one if another elementary step (e.g. adsorption-like) is the rate-controlling one. In a similar way, for the reduction step, a simple first-order expression for the reaction rate is:

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

209

Fig. 13. Comparison of microstructure evolution for Co3O4 powder No1 doped with Na in an atomic ratio Na/Co = 0.01. SEM photographs of the powders under consecutive magnifications of 5000, 20,000, 50,000 and 100,000 times. Left column: after calcination of the Co3O4–Na2CO3 mixture at 800 °C for 4 h; right column after cyclic TGA of that powder between 985 and 785 °C with a heating/cooling rate of 5 °C/min for two cycles.

Rred ¼ k red Wy=2

ð8Þ

where Rred is the reduction reaction rate in terms of moles of O2 produced per g of redox material per minute, k red is the temperature-dependent rate constant (units: m3/moles/min) of the reduction reaction and W and y are defined as above. The instantaneous concentration of adsorbed oxygen (moles of O2 per g of redox material) can be defined as a new variable w, where the variables w, W and y are related by w ¼ Wy=2

ð9Þ

Eq. (7) can explain the difference in the weight gain curve during oxidation among the three experiments shown in Fig. 1. The first lower temperature limit of 600 °C selected is too low, the temperature dependent rate constant is too small and as a result, during the one-hour dwell not too many of the initially empty oxygen storage sites are covered, thus the quantity (1  y) is still relatively high. As

a consequence, after the end of the dwell time at 600 °C upon heating to 885 °C, not only the temperaturedepended rate constant k ox increases but also W (the theoretical oxygen storage capacity of the material increases with increasing temperature) whereas the quantity (1  y) is still relatively high. Therefore their product results in a steep increase of the reaction rate between 600 and 885 °C. On the contrary in the other two experiments, most of the oxidation takes place during dwell at 800 °C which is high enough for the rate to be significant, thus most of the surface sites are filled with oxygen with the material almost recovering its initial weight. As a consequence, during subsequent heating from 800 °C to 885 °C the almost negligible driving force for further reaction (1  y) counterbalances the increase of the temperature-depended rate constant and W; thus only a marginal further increase in the reaction rate is observed during this stage. Similarly, the very low reduction rate during dwell at Thigh, is explained by Eq. (8): even though at this high temperature

210

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

the temperature-dependent rate constant is high, most of the oxygen storage sites are already “empty” when reaching this temperature, y is very low and therefore the driving force for further reaction minimal. Therefore, prolonged dwell of the material at this temperature induces an insignificant further effect on the completion of reduction. As already mentioned, the model proposed is just a fundamental mechanism describing the reaction of the surface layers of the redox material. Phenomena such as reaction in the bulk material and associated oxygen transport kinetics are not considered. It was also found that this linear model follows the experimental hydrogen/oxygen production reaction rates from their beginning up to a reduction of about 30% of their initial values i.e. the time during which the steep changes in the oxygen concentration take place in this work. Therefore it can adequately explain not only the evolution of oxygen concentration in the first stages of oxidation, but the differences observed among powders of different surface area as well. With respect to the microstructure difference, the faceted-layered structure observed can be rather safely attributed to the presence of Na. “Artificial” addition of Na in the starting Co3O4 powders resulted in the creation of such microstructure even though not the extent observed in the powder already containing Na from the supplier. It is known that Co3O4 crystals can be obtained in various structures such as e.g. nanocubes and rhombododecahedrons (Sun et al., 2013; Yang and Sasaki, 2010) depending on the synthesis conditions and parameters. The resemblance of the latter polyhedra structures to the morphology depicted in Fig. 11b, c and d is striking. Moreover, it is a well-known phenomenon that crystal surface energies can be affected significantly by small amounts of impurities (Berkovitch-Yellin et al., 1985). Thus, the observation of step-shaped morphology in Na-containing specimen No2 may be interpreted in terms of certain energetically unfavourable crystal planes which are replaced by a sequential and step-wise arrangement of energetically favourable planes. It seems that a closely controlled Na percentage and calcination cycle is required to maintain that structure. It has been reported (El-Shobaky et al., 1983) that doping of Co3O4 with Na2O or Li2O increases the reactivity of cobaltous oxide CoO for re-oxidation, an observation in accordance with the phenomena observed in this study. However in our case, even though morphologically interesting, this occurring structure did not seem to have an effect on the redox performance of the powder. Further studies are under way to elucidate such phenomena and propose suitable mechanisms correlating redox performance to chemical/microstructural features. 5. Conclusions Since the oxidation/reduction reactions of several multivalent oxides are accompanied by significant heat effects, such redox oxides can be used to thermochemically store and release solar heat in CSP plants using air as heat

transfer fluid. Among such oxides Co3O4 has several advantages like capability of reduction/oxidation within a temperature range reachable in future air-operated solar thermal power plants, accompanied by high heats of the respective reactions. TGA experiments with three different Co3O4 powders have demonstrated that re-oxidation is slower than reduction and therefore takes longer; this difference can be alleviated by “narrowing” the upper-lower temperature range operation and performing the two reactions at as close temperatures as possible. The heating/cooling rate is also an important means to control the extent of re-oxidation reaction; lower rates allow for higher residence times at higher temperatures and thus higher degrees of re-oxidation. In conclusion, it was shown that Co3O4 can operate in a satisfactory and cyclic mode within the temperature range 800–1000 °C. The heat effects of its reduction and oxidation were quantified. No benefits occurred by using Ni, Mg and Cu cobaltates instead of “pure” Co3O4. However, further experiments are under way with other additives as well as other oxide systems. These studies have a two-fold purpose: on the one hand to improve the stability, longterm performance, heat generation characteristics and/or raw materials cost and on the other hand to determine the optimal operation conditions (temperature range, heating/cooling rate) of other redox oxides. The performance of Co3O4 during cyclic redox operation depends on the kind of powder used. First results indicate that, within the ranges tested, particle size is not critical and surface area is an influential factor on redox performance on which differences among powders from various sources should be primarily attributed. Presence of Na was also shown to affect significantly the evolution of the product’s microstructure, though not necessarily combined with improved redox performance. In summary, powders from two out of the three sources exhibited full reoxidation under a wide range of operation conditions. Thermochemical solar heat storage with redox oxides is a promising route for increasing the storage density of Solar Thermal Power Plants. A necessary condition though for its large-scale implementation is the development of efficient, integrated thermochemical reactors/heat exchangers, suitably incorporated within the plants’ infrastructure. Ongoing research is focused on the design and testing of such redox-oxide-based integrated TCS reactors/heat exchangers with enhanced transport, thermal and heat recovery properties. Design issues concern the maximum amount of redox powder that can be incorporated and exploited efficiently in various reactor concepts and effective control of the charging/discharging phases to maximize off-sun operation potential. These results will be reported in future publications. Acknowledgements Dr. C. Agrafiotis would like to thank the European Commission for funding of this work within the Project

C. Agrafiotis et al. / Solar Energy 102 (2014) 189–211

“FP7-PEOPLE-2011-IEF, 300194” under the Intra-European Fellowship Marie Curie Actions of the 7th Framework Programme. The help and input of Dipl.-Phys. Friedemann Call from DLR with the TGA/DSC experiments is gratefully acknowledged. References Agrafiotis, C., Roeb, M., Konstandopoulos, A.G., Nalbandian, L., Zaspalis, V.T., Sattler, C., Stobbe, P., Steele, A.M., 2005. Solar water splitting for hydrogen production with monolithic reactors. Sol. Energy 79, 409–421. Agrafiotis, C.C., Pagkoura, C., Zygogianni, A., Karagiannakis, G., Kostoglou, M., Konstandopoulos, A.G., 2012. Hydrogen production via solar-aided water splitting thermochemical cycles: combustion synthesis and preliminary evaluation of spinel redox-pair materials. Int. J. Hydrogen Energy 37, 8964–8980. Agrafiotis, C., Zygogianni, A., Pagkoura, C., Kostoglou, M., Konstandopoulos, A.G., 2013. Hydrogen production via solar-aided water splitting thermochemical cycles with nickel ferrite: experiments and modeling. AIChE J. 59, 1213–1225. Angus, K., Thomas, P., Guerbois, J.-P., 2012. Synthesis and characterisation of cobaltite and ferrite spinels using thermogravimetric analysis and X-ray crystallography. J. Therm. Anal. Calorim. 108, 449–452. Bauer, T., Steinmann, W.-D., Laing, D., Tamme, R., 2012. Thermal energy storage materials and systems. Ann. Rev. Heat Transf. 15. Berkovitch-Yellin, Z., Van Mil, J., Addadi, L., Idelson, M., Lahav, M., Leiserowitz, L., 1985. Crystal morphology engineering by” tailormade” inhibitors; a new probe to fine intermolecular interactions. J. Am. Chem. Soc. 107, 3111–3122. Botas, J.A., Maruga´n, J., Molina, R., Herrado´n, C., 2012. Kinetic modelling of the first step of Mn2O3/MnO thermochemical cycle for solar hydrogen production. Int. J. Hydrogen Energy 37, 18661–18671. Centi, G., Perathoner, S., 2010. Towards solar fuels from water and CO2. ChemSusChem 3, 195–208. Edwards, S.E.B., Materic´, V., 2012. Calcium looping in solar power generation plants. Sol. Energy 86, 2494–2503. El-Shobaky, G.A., Ghoneim, N.M., Morsi, I.M., 1983. Thermal behaviour of cobaltic and cobaltous oxides as influenced by doping with some alkali metal oxides. Thermochim. Acta 70, 325–337. Ervin, G., 1977. Solar heat storage using chemical reactions. J. Solid State Chem. 22, 51–61. Fernandes, D., Pitie´, F., Ca´ceres, G., Baeyens, J., 2012. Thermal energy storage: “how previous findings determine current research priorities”. Energy 39, 246–257. Flamant, G., Hernandez, D., Bonet, C., Traverse, J.-P., 1980. Experimental aspects of the thermochemical conversion of solar energy; decarbonation of CaCO3. Sol. Energy 24, 385–395. Francis, T.M., Lichty, P.R., Weimer, A.W., 2010. Manganese oxide dissociation kinetics for the Mn2O3 thermochemical water-splitting cycle. Part I: experimental. Chem. Eng. Sci. 65, 3709–3717. Furler, P., Scheffe, J., Gorbar, M., Moes, L., Vogt, U., Steinfeld, A., 2012. Solar thermochemical CO2 splitting utilizing a reticulated porous ceria redox system. Energy Fuels 26, 7051–7059. Gil, A., Medrano, M., Martorell, I., La´zaro, A., Dolado, P., Zalba, B., Cabeza, L.F., 2010. State of the art on high temperature thermal energy storage for power generation. Part 1—Concepts, materials and modellization. Renew. Sustain. Energy Rev. 14, 31–55. Hatakeyama, T., Liu, Z. (Eds.), 1998. Handbook of Thermal Analysis. John Wiley & Sons Ltd., Baffins Lane, Chichester, West Sussex PO19 lUD, England.

211

Hutchings, K.N., Wilson, M., Larsen, P.A., Cutler, R.A., 2006. Kinetic and thermodynamic considerations for oxygen absorption/desorption using cobalt oxide. Solid State Ionics 177, 45–51. Klissurski, D., Uzunova, E., 1994. Synthesis and features of binary cobaltite spinels. J. Mater. Sci. 29, 285–293. Kodama, T., 2003. High-temperature solar chemistry for converting solar heat to chemical fuels. Prog. Energy Combust. Sci. 29, 567–597. Kostoglou, M., Lekkos, C.P., Konstandopoulos, A.G., 2011. On mathematical modeling of solar hydrogen production in monolithic reactors. Comput. Chem. Eng. 35, 1915–1922. Kubota, M., Kyaw, K., Watanabe, F., Matsuda, H., Hasatani, M., 2000. Study of decarbonation of CaCO3 for high temperature thermal energy storage. J. Chem. Eng. Jpn. 33, 797–800. Kuravi, S., Trahan, J., Goswami, D.Y., Rahman, M.M., Stefanakos, E.K., 2013. Thermal energy storage technologies and systems for concentrating solar power plants. Prog. Energy Combust. Sci. 39, 285–319. Murthy, M.S., Raghavendrachar, P., Sriram, S.V., 1986. Thermal decomposition of doped calcium hydroxide for chemical energy storage. Sol. Energy 36, 53–62. Neises, M., Tescari, S., de Oliveira, L., Roeb, M., Sattler, C., Wong, B., 2012. Solar-heated rotary kiln for thermochemical energy storage. Sol. Energy 86, 3040–3048. Petrasch, J., Klausner, J., 2012. Integrated solar thermochemical cycles for energy storage and fuel production. Wiley Interdiscipl. Rev.: Energy Environ. 1, 347–361. Schaube, F., Wo¨rner, A., Tamme, R., 2011. High temperature thermochemical heat storage for concentrated solar power using gas–solid reactions. J. Sol.Energy Eng. 133, 031006–031006. Schaube, F., Koch, L., Wo¨rner, A., Mu¨ller-Steinhagen, H., 2012. A thermodynamic and kinetic study of the de- and rehydration of Ca(OH)2 at high H2O partial pressures for thermo-chemical heat storage. Thermochim. Acta 538, 9–20. Siegel, N.P., 2012. Thermal energy storage for solar power production. Wiley Interdiscipl. Rev.: Energy Environ. 1, 119–131. Stamatiou, A., Loutzenhiser, P., Steinfeld, A., 2010. Solar syngas production from H2O and CO2 via two-step thermochemical cycles based on Zn/ZnO and FeO/Fe3O4 redox reactions: kinetic analysis. Energy Fuels 24, 2716–2722. Steinfeld, A., 2005. Solar thermochemical production of hydrogen – a review. Sol. Energy 78, 603–615. Sun, H., Ang, H.-M., Tade, M.O., Wang, S., 2013. Co3O4 Nanocrystals with predominantly exposed facets: synthesis, environmental and energy applications. J. Mater. Chem. A. Tmar, M., Bernard, C., Ducarroir, M., 1981. Local storage of solar energy by reversible reactions with sulfates. Sol. Energy 26, 529–536. Wentworth, W.E., Chen, E., 1976. Simple thermal decomposition reactions for storage of solar thermal energy. Sol. Energy 18, 205–214. Wong, B., 2011. Thermochemical Heat Storage for Concentrated Solar Power, Final Report for the US Department of Energy, San Diego, CA, USA. Wong, B., Brown, L., Schaube, F., Tamme, R., Sattler, C., 2010. Oxide Based Thermochemical Heat Storage. In: 16th Solar PACES International Symposium, Perpignan, France. Yang, J., Sasaki, T., 2010. Morphological control of single crystalline Co3O4 polyhedrons: selective and nonselective growth of crystal planes directed by differently charged surfactants and solvents. Cryst. Growth Des. 10, 1233–1236. Zunft, S., Ha¨nel, M., Kru¨ger, M., Dreißigacker, V., 2009. High-Temperature Heat Storage for Air-Cooled Solar Central Receiver Plants: A Design Study. Solar PACES International Symposium, Berlin, Germany.