**Cycle Stability and Hydration Behavior of Magnesium Oxide and Its Dependence on the Precursor-Related Particle Morphology**

**Georg Gravogl 1,2, Christian Knoll 2,3, Jan M. Welch 4, Werner Artner 5, Norbert Freiberger 6, Roland Nilica 6, Elisabeth Eitenberger 7, Gernot Friedbacher 7, Michael Harasek 3, Andreas Werner 8, Klaudia Hradil 5, Herwig Peterlik 9, Peter Weinberger 2, Danny Müller 2,\* and Ronald Miletich <sup>1</sup>**


Received: 31 August 2018; Accepted: 2 October 2018; Published: 7 October 2018

**Abstract:** Thermochemical energy storage is considered as an auspicious method for the recycling of medium-temperature waste heat. The reaction couple Mg(OH)2–MgO is intensely investigated for this purpose, suffering so far from limited cycle stability. To overcome this issue, Mg(OH)2, MgCO3, and MgC2O4·2H2O were compared as precursor materials for MgO production. Depending on the precursor, the particle morphology of the resulting MgO changes, resulting in different hydration behavior and cycle stability. Agglomeration of the material during cyclization was identified as main reason for the decreased reactivity. Immersion of the spent material in liquid H2O decomposes the agglomerates restoring the initial reactivity of the material, thus serving as a regeneration step.

**Keywords:** particle morphology; magnesium hydroxide; magnesium carbonate; magnesium oxalate; magnesium oxide; cycle stability; in-situ powder X-ray diffraction (PXRD); hydration reactivity; thermochemical energy storage; thermochemistry

#### **1. Introduction**

Energy management is a major challenge for our society, requiring equal measures of political and scientific involvement [1]. Energy supply, sustainable, environmentally benign energy production, and efficient utilization are key issues in managing global energy use [2]. Energy management may, in many cases, be better expressed as 'heat management', as heat is the most ubiquitous form of energy.

In nearly all types of electrical power plants, as well as in most industrial processes, heat is used as the driving force and operating medium. Within this context, the utilization of waste heat, accounting for two-thirds of overall global energy production, is an extensively investigated field [3]. The use of waste heat flows includes several aspects, one of them being temporal decoupling of waste heat availability and demand, as the two are not necessarily correlated. The necessary storage may be realized using materials for sensible, latent, or thermochemical storage of energy (heat) [4–9]. All three energy storage concepts offer advantages in specific areas of application [6,9,10].

Thermochemical energy storage (TCES) features long-term storage, a wide range of compatible temperatures, applicability as a heat pump system, and finally, high energy storage densities [10–13]. Based on these aspects, medium-temperature waste heat (up to 450 ◦C and extensively available from industrial processes) is perfectly suitable for TCES systems. An attractive TCES material for medium-temperature applications is the system Mg(OH)2–MgO with a storage temperature around 350 ◦C [14]. Both Mg(OH)2 and MgO are industrial base materials and are, therefore, available in large quantities at low prices.

Mg(OH)2–MgO as a TCES material is well known for this purpose, with many aspects related to its application in energy storage already investigated in literature. Kinetic investigations of dehydration and rehydration [15,16], mechanistic aspects of the conversion [17,18], modification of the material by additions of lithium salts [16,19,20], by coating or use of composite material [21,22], by dotation [23], and finally also applicability in form of a chemical heat pump [24] were reported. Nonetheless, two key issues preventing industrial application remain unaddressed: First, rehydration reactivity (completeness), and second, the cycle stability. Whereas for the limited cyclability observed thus far, no satisfying solution has been found, the rehydration reactivity is addressed by the addition of lithium salts [16,19,20], which are quite expensive. On a molecular level, reactivity could also be tuned by dotation of Mg(OH)2 with Ca2+-ions [23].

On an industrial scale MgO is produced via calcination of Mg(OH)2 or MgCO3 [25]. Both precursors are found in natural deposits, but whereas MgCO3 is an industrially mined raw material, Mg(OH)2 is produced from serpentinite or processing seawater [26]. However, aerobic calcination of any other Mg compound may result in formation of MgO by stepwise decomposition. In Scheme 1, this is shown at the example of the mentioned industrial precursor, as well as for magnesium oxalate dihydrate.

**Scheme 1.** Thermal decomposition of various MgO precursors: (1) Mg(OH)2, (2) MgCO3, (3) MgC2O4·2H2O.

All so far performed investigations on the rehydration of MgO for thermochemical energy storage purposes have largely neglected the origin of the MgO. As Mg(OH)2, MgCO3, MgC2O4·2H2O, and MgO crystallize in crystallographically and stereochemically different systems (Table 1), and feature notably different particle morphologies, MgO samples originating from different precursors can not necessarily be expected to have the same properties with respect to rehydration and cycle stability. This assumption is supported by previous kinetic studies on the H2O-dissociation on MgO. Compared to the isotypic CaO [27], the lower hydration reactivity of MgO [28] is mainly caused by the kinetic barrier of the water dissociation on the surface [29]. The disfavored H2O-dissociation as first step in formation of Mg(OH)2 occurs mainly at surface defects, edges, step edges, or corner sites, exhibiting

a lower dissociation energy barrier [30]. This suggests that by variation of the particle morphology and origin of the MgO, the rehydration behavior should be affected. While all precursors result in compositionally indistinguishable MgO sample stoichiometries, the particle size and morphology, crystallographic orientation, and thus the orientation of the reactive surfaces of the material are not necessarily the same. To verify this hypothesis, MgO obtained by calcination of Mg(OH)2, MgCO3, and MgC2O4·2H2O was investigated regarding hydration reactivity and cycle stability.


**Table 1.** Comparison of the crystallographic parameters of selected MgO precursors and MgO.

#### **2. Materials and Methods**

#### *2.1. Material*

Mg(OH)2 powder (particle size ≤5 μm) and MgCO3 (particle size ≤200 μm) were supplied by RHI-AG (X-ray fluorescence analysis (Bruker AXS GmbH, 76187 Karlsruhe, Germany)) of the materials revealed no significant impurities). MgC2O4·2H2O (98.5% purity) was purchased from abcr (GmBH, 76187 Karlsruhe, Germany) and the particle fraction ≤200 μm was used as supplied. The materials were calcined in an electric furnace under air and a static atmosphere for 4 h at variable temperatures (Mg(OH)2: at 375 ◦C; MgCO3: at 550 ◦C, 600 ◦C, 650 ◦C; MgC2O4·2H2O: at 650 ◦C). For subsequent rehydration, the in-situ calcined material from the (powder X-ray diffraction) P-XRD measurement was kept for 24 h in liquid water under ambient pressure-temperature conditions.

#### *2.2. BET Surface*

The specific surface of the samples was determined by nitrogen sorption measurements, which were performed on an ASAP 2020 (Micromeritics) instrument. The samples (amounting between 100–200 mg) were degassed under vacuum at 80 ◦C overnight prior to measurement. The surface area was calculated according to Brunauer, Emmett, and Teller (BET, Micromeritics Instrument Corp., Norcross, GA, USA) and t-plot methods [35].

#### *2.3. Powder X-ray Diffraction with In-Situ Hydration (P-XRD)*

Hydration of calcined samples was performed in an Anton Paar XRK 900 (Bruker AXS GmbH, 76187 Karlsruhe, Germany) sample chamber, connected to an evaporation coil kept at 300 ◦C (see Figure S1a). Using an HPLC-pump, water was evaporated at rates from 1 g H2O min−<sup>1</sup> up to 3 g min−<sup>1</sup> and the resulting steam was passed through the sample (1 mm thickness) with 0.2 L min−<sup>1</sup> nitrogen as carrier gas. The sample is mounted on a hollow ceramic powder sample holder, allowing for complete perfusion of the sample with the water vapour (see Figure S1b). As the sample is completely penetrated by the X-rays, the obtained diffractograms represent an average across the total sample with respect to the quantitative phase proportions. The diffractograms were evaluated using the PANalytical program suite HighScorePlus v3.0d. A background correction and a Kα<sup>2</sup> strip were performed. Phase assignment is based on the ICDD-PDF4+ database (International Diffraction Data-Powder Diffraction File), the exact phase composition, shown in the conversion plots, was obtained via Rietveld-refinement [36] in the program suite HighScorePlus v3.0d. All quantifications based on P-XRD are accurate within of ±2%. The rehydration rates were calculated based on the phase

composition derived from the diffractograms, normalizing the percentages of Mg(OH)2 and MgO to a total of 100%.

#### *2.4. Scanning Electron Microscopy (SEM)*

SEM (Thermo Fisher Scientific, 168 Third Avenue, Waltham, MA 02451, USA)) images were recorded on gold coated samples with a Quanta 200 SEM instrument from FEI under low-vacuum at a water vapour pressure of 80 Pa to prevent electrostatic charging.

#### *2.5. Small-Angle X-ray Scattering (SAXS)*

The samples were prepared either as powder between two pieces of tape or in a sealed capillary. Patterns were recorded using a microsource with X-rays from a copper target (Incoatec High Brilliance, wavelength 0.1542 nm, CuKα), a point focus (Nanostar from Bruker AXS) and a 2D detector (VÅNTEC 2000). The X-ray patterns were radially averaged and background corrected to obtain scattering intensities in dependence on the scattering vector *q* = (4π/*λ*) sinθ, with 2θ being the scattering angle.

The fit function from *Beaucage* [37] to describe scattering intensities of complex systems with a broad size distribution consists of a power law and Guinier's exponential form,

$$I(q) \propto \text{Gexp}\left(\frac{-q^2 R\_{\overline{\chi}}^2}{3}\right) + B \left[\frac{\left(\text{erf}\left(q R\_{\overline{\chi}} / \sqrt{6}\right)\right)^3}{q}\right]^{d\_f} \tag{1}$$

where *G* and *B* are the numerical prefactors, *df* is the fractal dimension, *Rg* is the radius of gyration and *erf*(*x*) is the error function. To describe the particle interference and thus the tendency of particles to agglomerate, additionally a structure factor from a hard sphere model was used [38,39],

$$I(q) \propto \left( G \exp\left(\frac{-q^2 R\_{\mathcal{S}}^2}{3}\right) + B \left[\frac{\left(\varepsilon r f(q R\_{\mathcal{S}} / \sqrt{6})\right)^3}{q}\right]^{d\_f} \right) S(q) \tag{2}$$

with

$$S(q) = 1/\left(1 + 24\eta \, G\_{int}(2q\mathbb{R}\_{HS})/\left(2q\mathbb{R}\_{HS}\right)\right) \tag{3}$$

and *RHS* being the hard sphere radius describing a typical distance of objects, *η* the hard sphere volume factor for characterizing the amount of agglomeration, and *Gint* a function derived in Kinning et al. [38].

#### **3. Discussion and Results**

To combine the apparent particle morphology with the crystallographic features of the lattice as given in Table 1, SEM-images of the original and calcined materials are compared in Figure 1. The first row corresponds to SEM-images of the various MgO precursors; in the second row the resulting MgO samples, obtained after thermal decomposition, are shown. Whereas Mg(OH)2 particles feature euhedral idiomorphic shapes with characteristic faces following hexagonal symmetry (Figure 1a), both MgCO3 (Figure 1b) and MgC2O4·2H2O (Figure 1c) reveal subidiomorphic irregular particle shapes occasionally showing typical rhombohedral (Figure 1b) or foliated (Figure 1c) cleavage faces, which in the case of MgC2O4·2H2O correspond to its layer structure.

The particle morphology of the materials changes during calcination (Figure 1, second row), leading to three differently textured MgO samples. Whereas for using Mg(OH)2 as the precursor material (Figure 1a), calcination results in an apparently unchanged particle morphologies, the MgO crystallites obtained from both MgCO3 (Figure 1b) and MgC2O4·2H2O (Figure 1c) precursors are characterized by a clear surface fragmentation, which can be attributed to larger degree of structural reconstruction on the release of volatile components. In contrast, the H2O release from Mg(OH)2 to

MgO follows a simple change from hcp to ccp arrangement of the octahedral subunits and hence preserves the particles in its shape to a large extent.

**Figure 1.** SEM pictures of (**a**) Mg(OH)2; (**b**) MgCO3; (**c**) MgC2O4·2H2O before calcination (first row) and after calcination (second row).

On the nanoscale, small-angle X-ray scattering (SAXS) reveals a transformation of the material from a dense solid to a highly porous material on calcination (see Figure S2). The nanostructure of MgO was modelled by a unified Guinier/power law [37], resulting in a radius of gyration for the size of the particles and an agglomeration with a structure factor from a hard sphere model, describing the agglomeration of particles with a typical distance 2*RHS* and the packing density with a hard sphere volume ratio *η* [38,39]. The detailed fit parameters are found in the supporting information (Table S1). In general, the gyration radius of MgO particles calcined from MgCO3 and MgC2O4·2H2O is about 6.6 and 5.1 nm, respectively, in comparison to about 2 nm if calcined from Mg(OH)2. In contrast, the values of *η* = 0.18 and a fractal dimension of *df* = 2.8 indicate, that MgO from Mg(OH)2 consists of small, agglomerated particles with a wide size distribution, whereas MgO from other precursors is built up of larger, denser nanoparticles (*η* close to zero, *df* = 4).

In order to allow a better comparability between the different precursor materials investigated within this study, MgO obtained by calcination from Mg(OH)2 was used as reference material [28]. In Scheme 2, schematic representation of the calcination and rehydration conditions applied for its preparation is shown.

$$\begin{array}{ccccc} \mathsf{Mg(O\mathsf{H})\_2} & \xrightarrow{\Lambda,\ 3\mathsf{T}\mathsf{5}\ \mathsf{T}\mathsf{6}} & \mathsf{Mg\mathsf{O}} & \xrightarrow{\mathsf{v\mathsf{A}\mathsf{O}\mathsf{O}\mathsf{O}\mathsf{N}}} & \mathsf{Mg(O\mathsf{H})\_2} & & \mathsf{6}\mathsf{T}\ \mathsf{%} \\\\ \mathsf{O}\mathsf{N}\mathsf{-}\mathsf{o} & & \mathsf{O}\mathsf{N}\mathsf{-}\mathsf{l} & & \mathsf{O}\mathsf{N}\mathsf{-}\mathsf{2} \\ \end{array}$$

**Scheme 2.** Conditions for calcination and rehydration of MgO obtained from Mg(OH)2.

To correlate the particle morphologies with rehydration reactivity and cycle stability, rehydration experiments using the different MgO samples were monitored by in situ powder X-ray diffraction (P-XRD). This allows for a direct observation and quantification of the reaction progress. As in previous experiments, the rehydration reactivity of the MgO produced from Mg(OH)2 was found quite limited [28]. To eventually increase the reaction rate, an even larger excess of water vapour was introduced into the reaction chamber. Increasing the vapour flow from 1 g min−<sup>1</sup> to 3 g min−<sup>1</sup> enhanced the rehydration conversion of MgO from 44% to 67% (Figure 2a). To assess the cycle stability for the increased vapour flow, five consecutive rehydration–calcination cycles were performed (Figure 2b). Similar to previous experiments with lower vapour flows the rehydration yield decreased over 5 cycles to a final Mg(OH)2 conversion of only 14%. Even after the first cycle the rehydration conversion was depleted to 43%.

**Figure 2.** (**a**) Rehydration of Mg(OH)2-originating MgO with various water vapour flow rates; (**b**) Cycle stability of Mg(OH)2-originating MgO.

For MgO produced by calcination of Mg(OH)2 a strong correlation between reactivity, accessible surface area and calcination temperature has been established [28]. Higher calcination temperatures promote sintering of the particles, leading to a decreased porosity, increased MgO crystal size, and decreased rehydration yield.

To assess the possibility of a similar effect for MgO originating from MgCO3, initial studies of the correlation between calcination temperature, BET surface and rehydration reactivity were made. For this purpose, samples of MgCO3 were calcined at 550 ◦C, 600 ◦C, and 650 ◦C for 3, 6, 9, and 12 h. The MgO formed from MgCO3 calcined for 6 h at 600 ◦C had the highest surface area (Figure S3). Nevertheless, attempted rehydration of all samples by water vapour in the P-XRD failed, showing no Mg(OH)2 formation within 120 minutes. In Scheme 3, representation of the conditions applied for calcination and rehydration of MgCO3-derived MgO is given.

**Scheme 3.** Conditions for calcination and rehydration of MgO obtained from MgCO3.

Based on the assumption, that a different chemical history of the MgO would have an impact on the reactivity during rehydration, a varied rehydration rate would have been expected. Observing no conversion to Mg(OH)2 under the applied conditions was, however, quite unexpected. On prolonged exposure to water vapour over 24 h for MgO **CO3-1** (see Scheme 3) a very sluggish formation of Mg(OH)2 below 10% was observed. To ascertain whether rehydration of this material could be driven by longer exposure to a vast excess of reactant, the samples were stored in liquid water. After 24 h reaction time in water at room-temperature, according to P-XRD measurements the material had been completely transformed to Mg(OH)2 (**CO3-3**) To repeat the in situ rehydration study for this material, the calcination step was repeated at 375 ◦C using the conditions developed for Mg(OH)2 [28]. After a new calcination step the BET surface of the various samples **CO3-4** was found to be slightly higher

than for **CO3-1**, the MgO originating directly from MgCO3 (Figure S4). In contrast to the first attempt, now for those materials the rehydration experiments in the P-XRD were repeated successfully for all materials (for detailed rehydration rates see Figure S5). Ranked according to their final conversion to Mg(OH)2, the most reactive material within this series was obtained by calcination of MgCO3 at 600 ◦C for 6 h (Figure 3) and subsequent rehydration in liquid water. A final conversion to 84% Mg(OH)2 was not only by far the highest yield for the MgCO3-originating series, but also notably more than for MgO originating from Mg(OH)2 (67% final conversion).

**Figure 3.** Final conversion for rehydration of the various MgCO3-originating MgO samples **CO3-4** in the P-XRD.

SEM images demonstrate, that the particle morphology of MgCO3 (Figure 4a) is retained after calcination, although the formerly distinct edges and surfaces are now covered by smaller scales (Figure 4b). During the hydration of the calcined material in liquid water the large particles disintegrate into smaller platelets (Figure 4c), although lacking the characteristic hexagonal morphology as characteristic for euhedrally grown Mg(OH)2 (see Figure 1). A subsequent calcination of material rehydrated in liquid water retains the afore mentioned platelet morphology (Figure 4d).

**Figure 4.** SEM images of (**a**) MgCO3, (**b**) calcined MgCO3 (**CO3-1**), (**c**) calcined MgCO3, rehydrated for 24 h in liquid water (**CO3-3**), (**d**) material from image c after calcination (**CO3-4**).

The changing particle shape observed in the SEM images is attributed to the volume work going along with the Mg(OH)2 formation. To enable this rehydration-related rearrangement of the material, water in its liquid form seems crucial as an agent triggering the rehydration process. SAXS intensities (Figure S6) show that on (repeated) rehydration of the MgCO3-derived MgO-samples the particle morphology is widely unchanged. From the larger scattering intensity a highly porous nanostructure, retained during rehydration, may be extrapolated. At the same time, a general decrease in particle size was also observed, being in good agreement with the SEM images (Figure 4).

The carbonate-derived MgO **CO3-4** was also investigated in terms of cycle stability (Figure 5). Similar to the material **OH-1** originating from Mg(OH)2, also in the case of **CO3-4** a decrease in rehydration reactivity was detected, although to a lesser extent than observed for **OH-1**. Over five cycles the rehydration conversion drops to 57% (84% in the 1st, 75% in the 2nd cycle).

**Figure 5.** Cycle stability of MgCO3-originating MgO **CO3-4**.

As a third precursor for preparation of reactive MgO, MgC2O4·2H2O was investigated (see Scheme 4).

**Scheme 4.** Conditions for calcination and rehydration of MgO obtained from MgC2O4·2H2O.

Since MgC2O4·2H2O decomposes stepwise via MgCO3 (see Scheme 1), only samples calcined in the furnace at 600 ◦C for 6 h were investigated. A comparison of the SEM images in Figure 6, compares the morphology of the different samples: initial oxalate material **C2O4-0** (Figure 6a), the calcined material **C2O4-1** (Figure 6b), MgO after rehydration in liquid water **C2O4-3** (Figure 6c), and a new calcined material **C2O4-4** (Figure 6d). Similar to the MgCO3-case, calcination of the initial material resulted in partial fragmentation, whereas subsequent treatment with liquid water and re-calcination forced the material to adopt a lamellar-structured particle morphology. In contrast to the MgO originating from MgCO3, thinner platelets were formed, those structure is preserved after calcination. Moreover, even the rehydration of the oxalate-based MgO did not yield the typical hexagonally shaped morphologies of euhedral brucite crystallites.

**Figure 6.** SEM images of (**a**) MgC2O4·2H2O, (**b**) calcined MgC2O4·2H2O (**C2O4-1**), (**c**) calcined MgC2O4·2H2O, rehydrated for 24 h in liquid water (**C2O4-3**), (**d**) material from image **c** after calcination (**C2O4-4**).

Both SEM and SAXS data show a comparable picture as observed for MgCO3. Rehydration of MgO **C2O4-1** in liquid water results in conservation of the original particle morphology to a wide extent (Figure S7). The porous nanostructure formed is preserved during rehydration. Accordingly, the increased SAXS intensities observed for the rehydrated material seem to arise from the formation of smaller particles (SEM images, Figure 6).

To determine the reactivity of the MgC2O4·2H2O-based MgO, both the material obtained by direct calcination of MgC2O4·2H2O (**C2O4-1**) and that resulting from the rehydration–calcination sequence (**C2O4-4**) were subject to rehydration in the P-XRD. Unlike the case of MgCO3, the directly calcined material **C2O4-1** was found to be reactive to rehydration, resulting in a final conversion of 68% (Figure 7a).

**Figure 7.** (**a**) Direct rehydration and cycle stability of MgC2O4-originating MgO **C2O4-1**; (**b**) Rehydration and cycle stability of MgC2O4-originating MgO **C2O4-4**.

The most remarkable finding is, that several batches of **C2O4-4**, the sample previously rehydrated in liquid H2O and subsequently calcined, could be fully rehydrated in the 1st cycle, but in the 2nd cycle the rehydration conversion decreased to 64% (Figure 7b). After five cycles both samples gave a comparable final conversion of slightly less than 50% Mg(OH)2.

Assessing the technological feasibility of MgC2O4·2H2O as precursor material, on the one hand the material was completely rehydrated to Mg(OH)2 within the first cycle after an initial treatment with liquid water. On the other hand, due to a large decrease in conversion rate during the successive cycles, a modest overall performance, and a relatively higher price compared to Mg(OH)2 and MgCO3, MgC2O4·2H2O is most likely not suitable as a competitive MgO precursor. Therefore, MgC2O4·2H2O was not subjected further studies.

Based on the conversion-enhancing effect of rehydrating calcined material in liquid water, both a sample of spent MgO originating from Mg(OH)2 and one from MgCO3 after the 5th rehydration cycle was calcined a further time and then rehydrated for 24 h in liquid water. At that point, P-XRD analysis showed complete transformation to Mg(OH)2 for both samples. Both Mg(OH)2-samples were now subjected further five rehydration–calcination cycles in the P-XRD, followed by a further regeneration in liquid water and another five rehydration–calcination cycles in the P-XRD. The conversion rates for 15 consecutive cycles, including two regeneration steps after five cycles are shown in Figure 8.

In the case of MgO originating from Mg(OH)2, shown in Figure 8a, the conversion rate after the first regeneration (Cycle 6) was slightly enhanced compared to the very 1st cycle. This effect was even more pronounced after the second regeneration (Cycle 10), revealing an even further increased reactivity. Nevertheless, the depletion evidenced in the second cycle was retained even after the second cycles after regeneration, as observed for Cycles 6 and 12. These results could be reproduced on various batches.

**Figure 8.** Selected conversion rates from a series of 15 consecutive calcination–hydration cycles, including two regeneration steps in liquid water after the 5th and the 10th cycle. (**a**) Mg(OH)2-originating MgO after regeneration; (**b**) MgCO3-originating MgO after regeneration.

In the case of MgO derived from MgCO3, a different but even more promising effect was observed: The spent material could be completely regenerated to reproduce the reactivity observed for the first cycle. Even the second cycle after each regeneration process (Cycle 7 and 12) was comparable to the "first" second cycle. Even in this case the effect was reproducible on various batches.

To better understand the physical processes during regeneration, SEM images of material during several stages of regeneration and cycling were compared. Despite differences in initial particle morphology (Figure 9a,d), MgO originating from Mg(OH)2 or MgCO3 shows similar evolution of reactivity during repeated calcination–rehydration cycles. After five consecutive cycles, resulting in aged material of depleted rehydration reactivity (see Figures 2 and 5), the particle morphology of Mg(OH)2-derived MgO (Figure 9b) seems nearly unaffected. In contrast, for material originating from MgCO3, the larger spherical aggregates are retained (Figure 9e). After regeneration for 24 h in liquid water both materials reveal a lamellar, platelet morphology devoid of the characteristic hexagonal brucite particle shape (Figure 9c,f).

**Figure 9.** SEM images of various intermediates during calcination/rehydration/regeneration for Mg(OH)2- (left) and MgCO3-originating MgO (right); (**a**) Mg(OH)2-originating MgO; (**b**) Mg(OH)2-originating MgO after 5 rehydration–calcination cycles; (**c**) material of image b after regeneration for 24 h in liquid H2O; (**d**) MgCO3-originating MgO; (**e**) MgCO3-originating MgO after five rehydration–calcination cycles; (**f**) material of image **e** after regeneration for 24 h in liquid H2O.

To directly monitor the rehydration/regeneration process, both a sample of MgO originating from Mg(OH)2 and from MgCO3 were observed during rehydration in liquid water by in situ SAXS (Figures S8 and S9). Initial SAXS curves (black) and final SAXS curves (red) are highlighted to better visualize the data and clearly show the change in the inner structure of the particles. The main difference is the time required for structural recovery, which is about three-fold shorter for Mg(OH)2-derived material compared to that from MgCO3 (Figure S8), most likely due to the considerably larger particle sizes favored by the latter.

Within various repeated experiments the regeneration process described for spent MgO was found to be reproducible—on one hand for the material of the same origin, on the other for materials of different origins. We suggest that during the regeneration process due to the comparably long reaction time and the vast excess of water, a complete conversion to Mg(OH)2 as well as a regeneration of the particle morphology occurs. Both effects complement each other, restoring the original reactivity of the material.

The possibility of a regeneration of spent material is of utmost importance for assessing the economic feasibility of a TCES material and energy storage process, as by prolonging the life-time the materials investment costs are minimized. Additionally, by implementation of a continuous regeneration step into the process, regenerating after each discharging–charging cycle a defined amount of material, permanent high activity of the TCES material circumventing efficiency losses by ageing would be ensured.

#### **4. Conclusions**

MgO obtained by calcination of Mg(OH)2, MgCO3, and MgC2O4·2H2O was compared regarding its rehydration reactivity and cycle stability to assess its applicability in thermochemical energy storage. The three different MgO-precursors led to three MgO samples featuring different particle morphologies with identical chemical compositions. Whereas Mg(OH)2 and MgC2O4·2H2O resulted in reactive MgO that could be rehydrated by water vapour to Mg(OH)2 directly following calcination, material originating from MgCO3 resulted in no conversion on contact with water vapour. Only after rehydration in liquid water and subsequent calcination of the thus formed Mg(OH)2, 84% of the resulting material could be rehydrated by water vapour. All materials investigated showed decreased rehydration reactivity during consecutive calcination–rehydration cycles, with MgCO3-derived MgO showing the smallest decline in reactivity. A regeneration step, consisting of rehydration of the spent material in liquid water over 24 h, restored the initial reactivity allowing for recycling of the material. In the case of Mg(OH)2 derived material, the initial reactivity could even be improved by repeated regeneration of the material in liquid water.

The results reported herein confirm, that the reactivity of MgO towards rehydration is strongly correlated to origin and physicochemical history of the material–an aspect so far neglected in the research on TCES materials. The correlation between chemical history and performance of storage materials may stimulate additional to coating, chemical dotation, etc., the consideration of a further, easily tunable parameter for the research on novel TCES systems.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/8/10/795/s1, Figure S1: Rehydration setup, reaction chamber and sample holder used for the in situ studies. Figure S2: SAXS intensities of starting materials and materials after calcination. Figure S3: BET surfaces of the MgCO3-originating MgO samples. Figure S4: BET surfaces of the MgCO3-originating MgO samples after rehydration. Figure S5: Rehydration rates of MgCO3-originating MgO samples in the P-XRD. Figure S6: SAXS intensities of materials from MgCO3 precursor. Figure S7: SAXS intensities of materials from Mg2C2O4·2H2O precursor. Figure S8: In situ SAXS intensities during regeneration in liquid water for 24 h. Figure S9: Kinetics of conversion to hydroxide during regeneration in liquid water. Table S1: Fit data for calcined materials

**Author Contributions:** Experimental investigation: G.G., C.K.; Proof-reading and language: J.M.W.; Evaluation of P-XRD data: W.A., K.H.; Provision of samples and scientific contribution: N.F., R.N.; SEM images and interpretation: E.E., G.F.; SAXS measurements and interpretation: H.P.; Project administration: A.W.; Conception of the study, writing, review, and editing: D.M.; Supervision and funding acquisition: M.H., P.W., R.M.

**Funding:** This research was funded by the Austrian Research Promotion Agency (FFG Forschungsförderungsgesellschaft), project 845020, 841150 and project 848876.

**Acknowledgments:** The X-ray center (XRC) of TU Wien is kindly acknowledged for the access to the powder X-ray diffractometer.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Synthesis of Me Doped Mg(OH)2 Materials for Thermochemical Heat Storage**

#### **Elpida Piperopoulos 1,2,\*, Marianna Fazio <sup>1</sup> and Emanuela Mastronardo 3,4**


Received: 13 June 2018; Accepted: 19 July 2018; Published: 26 July 2018

**Abstract:** In order to investigate the influence of metal (Me) doping in Mg(OH)2 synthesis on its thermochemical behavior, Ca2+, Co2+ and Ni2+ ions were inserted in Mg(OH)2 matrix and the resulting materials were investigated for structural, morphological and thermochemical characterization. The densification of the material accompanied by the loss in porosity significantly influenced the hydration process, diminishing the conversion percentage and the kinetics. On the other hand, it increased the volumetric stored/released heat capacity (between 400 and 725 MJ/m3), reaching almost three times the un-doped Mg(OH)2 value.

**Keywords:** magnesium hydroxide; thermochemical heat storage; metal doping

#### **1. Introduction**

The Renewable Energy Directive establishes an overall policy for the production and promotion of energy from renewable sources in the European Union (EU). The EU target for 2020 is to achieve at least 20% of its total energy requests with renewables. EU countries have already agreed on a new renewable energy target of at least 27% as climate goals for 2030. On 30 November 2016, the European Commission published a proposal for a revised Renewable Energy Directive to make the EU a global leader in renewable energy. Renewable energy can be produced from a wide variety of sources including solar, wind, hydro, geothermal, tidal and biomass. By using more renewables to meet its energy needs, the EU lowers its dependence on imported fossil fuels and makes its energy production more sustainable. Due to climatic variability, the means of storing these types of renewable energy have become an urgent consideration [1]. This has led to the search for efficient and sustainable methods of storing energy and a considerable effort to understand how energy storage works, how existing methods can be improved and how new ones can be developed. Thermal energy storage (TES) transfers heat to storage media during the charging period and releases it at a later stage during the discharging step. It can be usefully applied in solar plants, or in industrial processes. Through TES systems, heat can be stored in the form of sensible [2] or latent heat [3] or in the form of chemical energy (thermochemical storage) [4]. Sensible heat storage is achieved by varying the temperature of a storage material. Latent heat storage is realized changing a material phase at a constant temperature, while the thermochemical storage promotes a reversible chemical reaction. Sensible heat storage is well-documented. Latent heat storage, using phase change materials (PCMs), has been heavily researched and is widely used domestically and industrially. Thermochemical heat storage (TCS) is still at an early stage of laboratory and pilot research in spite of its attractive application for long-term energy storage and higher stored/released heat values [5,6]. Storage density, in terms of the amount of energy per unit of volume, is important for optimizing the use of these kind of materials [7] as it is relevant to their transportation and

application in concentrated systems [6]. In 1978, Bowery et al. [8] investigated the practical feasibility of a BaO2/BaO system for high-temperature heat storage. Theoretical calculations discovered that the endothermic reaction occurred when the temperature exceeds 754 ◦C, and the calculated energy storage density was about 2.9 GJ/m3. Subsequently, the reaction was found difficult to achieve complete conversion; even if the temperature rose to 1027 ◦C, the theoretical conversion rate of BaO2 had a maximum of 85%. Since then, several TCS materials have been studied and many strategies have been adopted to improve these storage materials [9]. Carrillo et al. studied the effect that co-doping of Mn oxides with Fe and Cu has on the redox temperatures of both forward and reverse reactions [10,11]. Block et al. tested several compositions of eight binary metal oxide systems as well as the pure metal oxides (cobalt oxide, iron oxide, copper oxide and manganese oxide) in terms of their ability to store energy thermochemically [12]. The calcium oxide hydration/dehydration reaction is proposed as a suitable reaction couple for thermochemical energy storage systems for its high energy density (0.4 kWh/kg) and low material cost (50 €/t) [13–15]. Sakellariou et al. prepared mixed calcium oxide–alumina compositions, assessed in terms of their cyclic hydration–dehydration performance in the temperature range of 200–550 ◦C. One of the main purposes of using Al as additive was related to materials structural enhancement [16]. A suitable TCS system storing in lower temperature range between 200 ◦C and 400 ◦C, which has been examined in this study, is the dehydration/hydration reaction of magnesium hydroxide/oxide:

$$\text{Mg(OH)}\_{2}\text{(s)} \leftrightarrow \text{MgO(s)} + \text{H}\_{2}\text{O(v)}\ \Delta\text{H}\_{0} = \pm 81 \text{ kJ/mol} \tag{1}$$

The above system offers several advantages, high storage capacity, medium operating temperature range (as reported above), long-term storage of reactants and products, low heat loss and non-toxicity of the materials [17]. Through the endothermic dehydration reaction, heat can be stored and released when required by the reverse exothermic hydration reaction. This system has been widely studied to improve storage material performances, as mass and volume energy density, kinetics and ciclability. Shkatulov et al. studied LiNO3-doped Mg(OH)2 storage material that exhibits a decrease in the dehydration start temperature by 76 ◦C [18]. Junichi et al. developed a 6.8 wt.% LiCl/Mg(OH)2 system that drops the dehydration temperature of magnesium hydroxide, from 277 ◦C to 233 ◦C, being able to store 816 MJ/m3 volumetric heat storage capacity [19]. Muller et al. found that calcium doping of magnesium oxide results in significantly increased water dissociation rates, thus enhancing both hydration rate and reaction completeness of hydration compared to pure MgO [20]. Zamengo et al. prepared a Mg(OH)2/MgO system supported on expanded graphite. The pelletized storage material, decreasing the tablets volume required to store the same amount of thermal energy of Mg(OH)2 pellets of almost 13.6%, increases volume energy density [21]. In previous studies it was found that, synthesizing Mg(OH)2 in presence of a cationic surfactant (cetyl trimethyl ammonium bromide—CTAB), an optimum CTAB concentration exists and it exhibits the highest volumetric stored/released heat capacity, ∼560 MJ/m3 two times higher than that measured over Mg(OH)2 prepared in absence of CTAB [22]. The purpose of this work is to investigate the influence of metal (Ca2+, Co2+ and Ni2+) doping in Mg(OH)2 synthesis on its structural and morphological properties and consequently on its thermochemical behavior.

#### **2. Materials and Methods**

#### *2.1. Samples Preparation*

The Metal (Me) doped Mg(OH)2 samples were synthesized by precipitation method. The following raw materials were used: Mg(NO3)2·6H2O, 99%, supplied by Sigma-Aldrich (St. Louis, MO, USA), as magnesium source, ammonia solution (NH4OH, 30 wt.% Carlo Erba) as precipitating agent and Ca(NO3)2, Co(NO3)2 and Ni(NO3)2 respectively for Ca2+, Co2+ and Ni2+ doping metals. The precipitation was carried out as follows: 50 mL of a solution containing Mg2+ and Me ion (Ca2+ or Co2+ or Ni2+) were gradually added (2.5 mL/min) through a peristaltic pump to 150 mL of NH4OH

solution (*ph* = 11.8), under magnetic stirring. The final solution was aged at ambient temperature for 24 h, then it was vacuum filtered (0.22 μm); the collected solid was washed with deionized water and dried in a vacuum oven (Binder, Tuttlingen, Germany) at 50 ◦C overnight. Table 1 reports the code of samples and the chemical composition of solutions for all the preparations.


**Table 1.** Sample code, chemical compositions of the solutions. Mg2+ and OH<sup>−</sup> molar concentration were 0.01 M and 0.063 M in each preparation.

#### *2.2. Samples Characterization*

Quantitative analysis of calcium, nickel and cobalt present into the solid was performed by means of ICP-MS spectrometer (PERKIN-ELMER, model NexION 300×, Waltham, MA, US). Approximately 3 mg (*wtmeasured*) of each synthesized sample (*wtsynthesized*) was dissolved in the minimum volume of concentrated HNO3, and then deionized water was added until the final volume of 10 mL (*Vf*) was reached. Exactly 100 μL (*V*1) of this solution, mixed of 100 μL of concentrated HNO3, were diluted up to 10 mL (*V*2) and then analyzed. The grams of dopant (*Me*) present in the samples are calculated as follows:

$$\text{Me } (\text{g}) = \{ (\frac{[\text{Me}]\_{ICPMS}}{V\_1 \text{ (l)}}) \times V\_2 (\text{l}) \} \times \frac{V\_f (\text{l})}{1000} \text{\} \times \frac{wt\_{\text{syntheizad}} \text{ (g)}}{wt\_{\text{measured}} \text{ (g)}} \tag{2}$$

Pore volume was calculated by Barrett-Joyner-Halenda (BJH) method using the nitrogen desorption isotherm measured at −196 ◦C with a Quantachrome Autosorb-iQ MP (NOVA 1200, Boynton Beach, FL, USA) instrument. Samples were degassed prior to analysis under vacuum at 120 ◦C for 3 h. Each sample's mean particle size was determined by Dynamic Light Scattering (DLS) technique. DLS was measured at 25 ◦C using a Zetasizer Nano ZS instrument (Malvern Instruments, Malvern, UK) equipped with a helium-neon 4 mW laser (wavelength *λ*<sup>0</sup> = 632.8 nm). The scattering angle was equal to 173◦. Prior to measurements, samples were sonicated for 30 minutes in ethylene glycol. The bulk density of samples was measured by weighing a known volume of solids (*V* (mL)) and calculated by the formula:

$$
\rho = m \, (\text{kg}) / \text{V} (\text{m}^3) \tag{3}
$$

The as-prepared samples were analyzed by means of scanning electron microscopy (SEM, Quanta 450, FEI, Hillsboro, OR, USA) and X-Ray Diffraction (XRD, Bruker D8 Advance, Bruker, Billerica, MA, USA) to determine their morphology and crystal structure.

SEM analysis were performed on Cr-metallized samples and operating with an accelerating voltage of 10 kV under high vacuum conditions (6.92 × <sup>10</sup>−<sup>5</sup> Pa).

#### *2.3. Thermochemical Performance*

The evaluation of the thermochemical behavior of the prepared samples under cyclic heat storage/release experiments was performed using a customized thermogravimetric unit (STA 449 F3 Jupiter Netzsch, Selb, Bavaria, Germany) that allowed us to carry out a succession of dehydration and hydration reactions. The thermogravimetric apparatus was equipped with a water vapor generator for the vapor supply during the hydration reaction. A cyclic heat storage/release experiment was carried out on a mass of ~15 mg as reported elsewhere [17,23,24]: the sample was first dried at 125 ◦C in inert atmosphere (under N2 flow: 100 mL/min) for 60 min to remove the physically adsorbed water. Then, the temperature was increased at 10 ◦C/min up to the dehydration temperature (*T*<sup>d</sup> = 350 ◦C) and dehydration reaction proceeded over 120 min under isothermal conditions. After the complete dehydration reaction, the temperature was decreased (cooling rate = −10 ◦C/min) to the hydration temperature (*T*<sup>h</sup> = 125 ◦C). The hydration reaction proceeded over 120 min, during which the water vapor necessary for the re-hydration reaction was supplied by the water vapor generator at 2.2 g/h and mixed with 35 mL/min N2 as carrier gas (*p*H2O = 57.8 kPa). After the fixed hydration time, the water vapor supply was stopped and the sample was kept at 125 ◦C for 30 min under a constant N2 flow (100 mL/min) to remove physically adsorbed water from the sample. This procedure was repeated for each heat storage/release cycle. In this study, for a preliminary comparison, the samples were subjected to 3 cycles experiments. To be consistent with previous studies [17,22,23,25] the materials performances were expressed in terms of reacted fraction (*β*(%)) defined by Equation (4):

$$
\beta(\%) = (1 - \frac{\Delta m\_{\text{real}}}{\Delta m\_{\text{th}}}) \times 100,\tag{4}
$$

where Δ*mreal*(%) was the instantaneous real mass change and Δ*mth*(%) was the theoretical mass change due to the dehydration of 1 mol Mg(OH)2, respectively expressed by Equations (5) and (6):

$$
\Delta m\_{\text{real}}(\%) = \frac{m\_{\text{in}} - m\_{\text{inst}}}{m\_{\text{in}}} \times 100,\tag{5}
$$

$$
\Delta m\_{th} \text{(\%)} = (\frac{M\_{\text{Mg(OH)2}} - M\_{\text{MgO}}}{M\_{\text{Mg(OH)2}}} \times 100) = 30.89\%,\tag{6}
$$

where *min*(*g*) and *minst*(*g*) were respectively the initial sample mass and the instantaneous mass during TG analysis. While, *MMg*(*OH*)2(g/mol) and *MMgO*(g/mol) were respectively the molecular weight of Mg(OH)2 and MgO.

The dehydration and hydration conversions (Δ*βd*/*h*(%)) were calculated respectively by Equations (7) and (8):

$$
\Delta \beta\_d(\%) = \beta\_d^i - \beta\_{d'}^f \tag{7}
$$

$$
\Delta \beta\_{\mathbb{H}}(\%) = \beta\_{\mathbb{H}} - \beta\_{d'}^{f} \tag{8}
$$

where *β<sup>i</sup> <sup>d</sup>* and *<sup>β</sup><sup>f</sup> <sup>d</sup>* were respectively the reacted fraction at the beginning and at the end of the dehydration treatment. While, *β<sup>h</sup>* was the final reacted fraction of MgO at the point of water supply termination.

The stored/released heat capacity per volume unit (*Q<sup>V</sup> <sup>s</sup>*/*<sup>r</sup>* (MJ/m3)) was calculated using Equation (9):

$$\mathcal{Q}\_{s/r}^{V} \left(\mathrm{M}\right) / m^{3} \right) = -\frac{\Delta H^{0}}{\mathcal{M}\_{\mathrm{Mg}(OH)2}} \times \Delta \mathcal{J}\_{d/h} \times \rho \tag{9}$$

where Δ*H*<sup>0</sup> (kJ/mol) is the enthalpy of reaction and *ρ* (kg/m3) the bulk density of the sample.

#### **3. Results and Discussion**

#### *3.1. Me Doped Mg(OH)2 Preparation*

In the first instance, it was evaluated whether, under the preparation condition of the present work, each ion could precipitate as hydroxide. Precipitation of hydroxide from the solution through the reaction (10)

$$\rm{Me^{\eta+} + nOH^{-} \to Me(OH)\_{n} \text{ (s)}}\tag{10}$$

occurs when the supersaturation conditions are reached. Supersaturation conditions are defined as:

$$\text{[Me}^{\text{ll}+}\text{]}\cdot\text{[OH}^{-}\text{]}^{\text{n}} > \text{K}\_{\text{sp}}.\tag{11}$$

where [Me*<sup>n</sup>*+] and [OH<sup>−</sup>] represent the concentration expressed as molarity (M) of cation and hydroxyl ions, *n* represent the hydroxyl's stoichiometric coefficient and Ksp is the thermodynamic equilibrium constant of solubility product. As shown in Table 2 supersaturation conditions are satisfied in case of Mg(OH)2, Co(OH)2 and Ni(OH)2 formation but not for Ca(OH)2, whatever the calcium concentration used being the ionic product [Me*<sup>n</sup>*+]·[OH−] *<sup>n</sup>* < Ksp.

**Table 2.** Evaluation of supersaturation conditions for Mg(OH)2, Co(OH)2, Ni(OH)2 and Ca(OH)2 formation under conditions used in the present work. Y: Yes, N: No.


As will be further explained, in reality neither Ni(OH)2 nor Co(OH)2 solids form (Figure 1). This is due to the fact that with a large excess of ammonia, cobalt and nickel hydroxides redissolve forming hexaminocobalt(II) (Co(NH3)6 2+) and hexaminonickel(II) (Ni(NH3)6 2+) ions as ammonia substitutes as a ligand [25]. As shown in Figure 1 no solid formation occurs even after 24 h. In case of Co2+ solution, pink colored due to presence of Co(H2O)6 2+, upon addition of NH4OH color rapidly changes to yellow then to a deep red-brown. This is due to oxidation of hexaminocobalt(II) to hexaminocobalt(III) ions by air [25]. In case of Ni2+, light green colored by the complex Ni(H2O)6 2+, addition of ammonia causes a color change to light blue typical of Ni(NH3)6 2+ complex [25].

**Figure 1.** Formation of cobalt and nickel hexamine complexes. Starting aqueous Co2+ solution 0.002 M (**a**); Upon addition of NH4(OH) (**b**) and after mixing for 24 h (**c**). Starting aqueous Ni2+ solution 0.002 M (**d**); Upon addition of NH4(OH) (**e**) and after mixing for 24 h (**f**).

After these preliminary evaluations MH, MH-Ca, MH-Co, and MH-Ni were prepared according to the procedure reported in the experimental section. The Me content on the final sample, (*gMe*/*gsample*)%, is reported in Figure 2. As general feature, regardless the type of Me, the load increases with the initial amount present into the solution. At given Me initial concentration the *gMe*/*gsample* content varies among the different type of Me; in case of Ni the lowest amount of loaded Me is obtained while the highest amount is achieved for Co containing samples.

**Figure 2.** Me content per g of sample obtained by precipitation vs. Me concentration in the starting solution.

Considering that no calcium, cobalt and nickel hydroxide form, it can be assumed that these metal ions are included into Mg(OH)2 host matrix.

#### *3.2. Structure and Morphology of Samples*

In Figure 3a–d XRD analysis of MH and MH-Me samples are shown. The diffractograms are acquired in a 2θ range between 10◦ and 80◦. Mg(OH)2 spectrum (Figure 3a) presents the reflection peaks typical of hexagonal brucite (2θ: 18.5◦, 32.5◦, 38◦, 51◦, 58.5◦, 62◦, 68◦, 72◦), in agreement with standard data, (JCPDS 7-0239 and JCPDS 25-0284). The three most intense peaks (2θ = 18.5◦, 38.0◦, 58.5◦) appear sharp and narrow as a result of high degree of crystallization of hexagonal brucite. Reflection peaks of MH-Ca and MH-Ni samples, regardless the amount of calcium or nickel, match with those of pure Mg(OH)2 in terms of peaks position (Figure 3b,c). They are intense and narrow thus suggesting that the high crystallization degree of brucite is maintained. For MH-Ca2 and MH-Ca3 samples (Figure 3b) is also present a peak at 29.4◦, related to CaCO3 (JCPDS 47-1743) likely due to the slight carbonation of calcium ions by CO2 present in the atmosphere. The main difference with respect to pure MH concerns the change in the relative intensity among the two main peaks relative to (001) and (101) plane. The intensity of Mg(OH)2 (001) plane's peak, which corresponds to the basal plane of brucite, becomes stronger than the diffraction peak for the (101) plane. As reported in Table 3 Entries 1–7, the intensity ratio of reflections *I*001/*I*<sup>101</sup> increases from 0.78 (MH sample) up to values ranging between 0.95–1.34 for MH-Ca e MH-Ni. No clear correlation is observed between the increase of *I*001/*I*<sup>101</sup> and the metal content. From these results it is possible to conclude that in presence of calcium and nickel, ions preferential growth along the (001) hexagonal basal plane of brucite occurs leading to a layered structure, e.g., flakes or platelets, with high aspect ratio along the *c*-axis. [26–28]. Wu et al. have already reported that the strength of (001) plane became stronger than that of (101) plane upon hydrothermal modification of Mg(OH)2 in presence of CaCl2 [29]. MH-Co samples, instead, shows a peculiar feature. The spectra shown in Figure 3d present reflection peaks, centered at the same position of those of brucite (Figure 3a), with a progressive intensity decrease and peak broadening, as Co content increases, that indicate the lowering of crystallization degree. Rietveld refinement reported in Table 3 confirms that metal ions are included into Mg(OH)2 host matrix because

a volume cell (*V*(Å3)) increase is observed, in relation to the metal load. At lower metal load for all Me-doped samples the volume cell remains almost similar to the MH sample's one, but increasing Me load it increases till 41.5 Å3 for MH-Co3 sample (Entry 9 in Table 3), which presents the higher amount of Co in the matrix. Only for MH-Ni2 and MH-Ni3 (Entries 6 and 7 in Table 3) it decreases. Substituting Mg ion (*rMg*2+ = 0.72 Å) with Ca ion (*rCa*2+ = 100 Å) it is simple to understand, according to Vegard's law [30], the cell volume change, while it is more difficult in the case of Co (*rCo*2+ = 0.70 Å) and Ni (*rNi*2+ = 0.70 Å) ions, which ion radius are similar to Mg's one. In these cases, probably, atoms are substituted interstitially leading to a lattice expansion [31].

**Figure 3.** XRD patterns of MH (**a**), MH-Ca (**b**), MH-Ni (**c**) and MH-Co (**d**) samples.

The morphology of MH and MH-Me samples is evaluated by means of SEM analysis, shown in Figure 4a–k. MH sample (Figure 4a) presents as large aggregates prevalently formed by magnesium hydroxide hexagonal platelets, in agreement with XRD findings; in addition, a few rounded shaped particles (red arrows) are also visible. The evolution of the sample morphology as the result of the doping by calcium and nickel appears very similar. In particular, increasing the amount of calcium and nickel in the solid, large agglomerates of highly stacked hexagonal brucite particles form (Figure 4b–g). The use of cobalt as doping ion, instead, gives rise to a dramatic change in the morphology with respect to MH, especially at higher cobalt load. Indeed, while for MH-Co1 brucite platelets having the peculiar stacked configuration are still visible (Figure 4h), in case of MH-Co2 and MH-Co3 it is clearly observed the progressive formation of amorphous hydroxide. In particular, for MH-Co2 sample it can be seen magnesium hydroxide platelets (Figure 4i white arrow) embedded into large portions of badly crystallized material (Figure 4i black arrow). Increasing the cobalt content, MH-Co3 sample, the crystalline hexagonal brucite is practically not visible anymore or it is very difficult to distinguish, and large sheets of poorly crystallized hydroxide represent the material's main component (Figure 4j). SEM analysis is in agreement with the XRD results that evidence the progressive amorphization of Mg(OH)2 increasing the cobalt content (Figure 4h–j).

It is noteworthy from the low magnification images of MH-Co3 sample (Figure 4k) that the large sheets of unshaped badly crystallized hydroxide are very densely packed, forming a continuous and extended rough surface.

Referring to the mechanism of hexagonal Mg(OH)2 growth, based on the model of anion coordination polyhedron (ACP) [32] where the nucleation seeds Mg(OH)6 <sup>4</sup><sup>−</sup> first form the growth units (Figure 5a) that pile up with each other forming large dimension growth units in the same face (*x*, *y*) (Figure 5b) which then connect one to another along the *z* axis forming (001) planes (Figure 5c) and finally the hexagonal structure. It can be argued that calcium and nickel ions promote the growth along the *z* axis (Figure 5c) then the hexagonal structure, as inferred by the increase of the intensity ratio *I*001/*I*<sup>101</sup> while cobalt ions, instead, strongly hinder the piling of growth units in the *x*, *y* plane and then the crystal formation.

The mean particle size of investigated samples, as inferred by DLS analysis, are reported in Table 3.


**Table 3.** Intensity ratios and morphological properties of investigated samples.

\* Measured by means of Dynamic Light Scattering analysis.

MH shows the highest value centered at 181 nm. A decrease of the mean particle size is obtained for all MH-Me1 (Entries 2, 5, 8). At higher Me content, two different behaviors have been observed depending on the type of metal. For calcium and nickel doped samples mean particle size returns progressively to increase, although it is lower than that of MH, with the metal content (Entries 3, 4, 6, 7). For MH-Co2 instead, mean particle size continues to decrease in MH-Co2 (Entry 9) while abruptly increases for MH-Co3 sample, containing the highest metal content (Entry 10).

**Figure 4.** SEM images of investigated samples. MH (**a**); MH-Ca1 (**b**); MH-Ca2 (**c**); MH-Ca3 (**d**); MH-Ni1 (**e**); MH-Ni2 (**f**); MH-Ni3 (**g**); MH-Co1 (**h**); MH-Co2 (**i**); MH-Co3 (**j**,**k**).

**Figure 5.** Mg(OH)2 growth, based on the model of anion coordination polyhedron (ACP). Growth unit (**a**); Large dimension growth units in the same face (**b**); Hexagonal structure (**c**).

It is noteworthy that for MH-Ca and MH-Ni samples DLS data really reflects the size of the crystalline hexagonal platelets which seems to be influenced by the cations content. Wu et al. have already reported an increase of Mg(OH)2 particle size increasing the calcium content during hydrothermal treatment of hydroxide [29]. The authors suggest that calcium promotes the formation of Mg(OH)<sup>+</sup> which may be favorable for the formation of nucleation seeds Mg(OH)6 <sup>4</sup><sup>−</sup> which represents the growth unit for Mg(OH)2 growth [32]. A similar effect can be depicted for nickel ion.

In the case of MH-Co2 and MH-Co3 samples instead, DLS analysis likely reflects the size of crystalline hexagonal platelets (very few, especially in the case of MH-Co3) in addition to the size of the amorphous portions of the sample. Considering that the preparation of the samples for DLS analysis provides that they are sonicated, it is likely that, due to the lowest mechanical resistance, amorphous phase is fragmented in an uncontrolled way. Therefore, the resulting particle size observed is not the direct evidence of a such influence of cobalt ion during the Mg(OH)2 growth.

Table 3 also lists the material's properties such as apparent density *ρ* (kg/m3). From the reported data it is evident that the apparent density of all MH-Me samples is significantly higher than that of pure MH. In general, the apparent density enhancement ranges between 68% (sample MH-Co2) up to 200% (sample MH-Co3). The strong enhancement of density is visually demonstrated in Figure 6.

**Figure 6.** Volume occupied by ~69 mg of MH-Co3 (**on the left**) and MH (**on the right**) samples.

Apparent density is defined as the average density of the material and includes the volume of pores within the particle boundary [33]. Generally, the higher the density, the smaller the pore volume in the sample. The almost general behavior of doped samples (MH-Ca and MH-Ni), in fact, reflects a higher density of the material and a lower value of the porosity, except for the MH-Ni1 sample, which morphology (Figure 4e) appears to be less stacked than the samples with the highest metal load and more similar to MH-Co1 and MH-Co2 (Figure 4h,i), which show a comparable pore volume (Entries 8 and 9 in Table 3). The same peculiar morphology was found for Mg(OH)2 prepared in the presence of CTAB, which promotes the formation of well separated Mg(OH)2 particles, lowering the hydroxide mean particle diameter and increasing the bulk density likely due to the peculiar stacked configuration of hydroxide particles, reported elsewhere [22].

The increase of apparent density could be due to two concomitant effects, which are the lowering of particle size and the strong agglomeration of magnesium hydroxide particles (MH-Ca and MH-Ni sample) or, as in case of MH-Co, to the densely packed amorphous material formation, as evidenced by SEM and by the lowest value of pore volume (0.245 cm3/g) detected for MH-Co3 sample.

#### *3.3. Thermochemical Behavior*

The thermogravimetric data are calculated assuming the metal doping negligible and Mg(OH)2 at 100 wt.%. The curves in Figure 7 are relative to the third cycle, when the thermochemical behavior of the samples was observed to be stable [24].

**Figure 7.** TG analysis, influence of metal loading. Reacted fraction in dehydration and hydration reactions of MH-Ca (**a**,**b**), MH-Ni (**c**,**d**), MH-Co (**e**,**f**).

For all the doped materials, the percentage of MH reacted fraction during dehydration and hydration decreases (Figure 7). MH-Ca and MH-Ni follow the opposite trend observable for morphological properties in Table 3. In fact, Mg(OH)2 conversion progressively decreases, increasing metal load from MH-Ca1 to MH-Ca2 (Entries 2 and 3 in Table 4), and then it remains almost stable for MH-Ca3 (Entry 4 in Table 4). The same behavior is observed for MH-Ni. The mean particle size, as described before, progressively increases following the same criteria. For MH-Co1, instead, conversion (%) continues to increase to MH-Co2 (Entry 9 in Table 4) while abruptly decreases for MH-Co3 sample, which presents the highest mean particle size (Entry 10 in Table 3). Additionally, for hydration, a similar trend is observed. If the 1st cycle dehydration reaction is analyzed, it can be observed that the conversion percentage of all the samples is equal to MH conversion and in some cases also higher (Entries 2, 3, 5, 6 and 7 in Table 4) or quite low (Entries 4 and 10 in Table 4). Therefore, the limiting process that influences the material behavior is the hydration. The *βh*%, during first cycle, as shown in Table 4, does not reach MH hydration with the exception of MH-Ca1 (Entry 2 in Table 4), which also maintains the higher conversion percentage in the following cycles. This behavior seems to be related to the main particle size reported in Table 3. As discussed above, the smaller particle size is strictly correlated with the higher density of the doped samples. This morphology strongly influences the magnesia hydration. As reported by Tang et al. [34], MgO hydration process follows common MgO dissolution/Mg(OH)2 precipitation mechanism, well accepted in the literature [35–37]. Initially water vapor is chemisorbed on the MgO and then physically adsorbed to form a liquid layer on the surface of the solid (chemical control of the reaction). This layer of water reacts with the MgO to form a surface layer of Mg(OH)2, that covers surfaces and pores of MgO particles. As a result, the diffusion of water vapor is hindered inside the particles, which reduces the overall reaction rate and the rehydration conversion *βh*% (diffusion controlled). When density is high, because of the small particle size and the packed morphology described in Figure 4, the porosity of the material is very poor and the water permeability is difficult. Figure 8 shows the SEM analysis of the investigated samples after cycling. For brevity, only MH and doped samples with highest metal load are reported (MH-Ca3, MH-Ni3, MH-Co3). It can be observed that coalescence is more favored for doped samples. For MH-Ca3 and MH-Ni3, the particle size increase is clearly observable (compare Figure 4 (white arrows) with Figure 8 (white circles)), MH-Co3 keeps its packed structure, formed by large sheets of poorly crystallized hydroxide (red arrows in Figure 8). Probably, also in this case, the high density plays a very important role influencing the change of morphology during the dehydration/hydration cycles. The presence of a lower porosity of the material and a smaller particle size favors the coalescence of the latter in larger particles; this decreases heat transfer property and leads to further loss of bulk porosity diminishing the MgO rehydration kinetics [24].

**Figure 8.** *Cont.*

**Figure 8.** SEM images of investigated samples after cycling: MH (**a**); MH-Ca3 (**b**); MH-Ni3 (**c**); MH-Co3 (**d**).

Also noteworthy is the fact that, from the slope of the dehydration and hydration curves, doped samples exhibit similar dehydration kinetics with respect to un-doped MH (see Figure 7a,c,e). On the contrary, hydration kinetics is highly affected by the doping, which, in general, decreases the kinetics (see Figure 7b,d,f). This is evident especially in Ni-doped samples (see Figure 7d). Hence, depending on the final application of the storage technology, heat can be released at a required rate by tuning MH with a proper dopant cation and amount.


**Table 4.** Comparison between dehydration/hydration conversions (Δ*βd*/*h*) at first and third cycles.

Looking at stored/released heat capacity by unit volume (*Qs*/*<sup>r</sup> <sup>V</sup>*) (Figure 9a–f) it can be seen that MH shows the lowest stored/released heat capacity that is 285 MJ/m<sup>3</sup> with respect to MH-M samples for which a higher *Qs*/*<sup>r</sup> <sup>V</sup>* is generally observed, as a consequence of the higher apparent density (Table 3).

**Figure 9.** Stored and released heat per volume unit of MH-Ca (**a**,**b**), MH-Ni (**c**,**d**), MH-Co (**e**,**f**).

The highest value 725 MJ/m<sup>3</sup> is achieved on MH-Co3 (Figure 9e), which is almost three times higher than MH's value. This value is, so far, the highest reported in the literature for pure Mg(OH)2 heat storage material. The stored heat increases with an increase in the metal load doping. Released heat per volume unit in doped samples almost never reaches the 100% of stored heat.

#### **4. Conclusions**

The present study clearly suggests that morphological characteristics (porosity, mean particle size) and apparent density are significantly influenced by the Me (Ca2+, Ni2+, Co2+) doping during the preparation of Mg(OH)2 through precipitation. It was found that, considering that no calcium, cobalt and nickel hydroxides precipitate during the synthesis, these metal ions are included into Mg(OH)2 host matrix, as confirmed by Rietveld refinement of XRD analysis. All the investigated samples show an apparent density increase. MH-Co3, which presents badly crystallized and highly packed hydroxide, reaches a higher density than the MH sample of 200%. Apparent density describes two concomitant effects that are the lowering of particle size and the strong agglomeration of magnesium hydroxide particles (MH-Ca and MH-Ni samples) with a consequent decrease in sample porosity. In MH-Co case, the high density is due to a densely packed amorphous material formation. A correlation between morphological properties and the thermochemical behavior of Mg(OH)2 is found. In particular, for all the investigated doped samples a lower reacted fraction is obtained in comparison with the not-doped material. However, because of the higher apparent density, the doped samples exhibit higher volumetric stored/released heat capacity. The highest value is reported for MH-Co3 sample (725 MJ/m3), and it is almost three times higher than MH's value. In future development, the doped samples will be further investigated to enhance their performance, while maintaining the high density, and they will be tested for several cycles, to investigate their stability in real applications.

**Author Contributions:** Conceptualization, E.P.; Funding acquisition, E.P.; Investigation, M.F. and E.M.; Project administration, E.P.; Supervision, E.P.; Validation, M.F.; Visualization, E.P., M.F. and E.M.; Writing–original draft, E.P.; Writing–review and editing, E.P., M.F. and E.M.

**Funding:** This research was funded by INSTM (Consorzio Interuniversitario Nazionale per la Scienza e Tecnologia dei Materiali) grant number INSTMME002 "Thermochemical materials for heat storage: development and characterization".

**Acknowledgments:** This study was conducted as part of the project "Thermochemical materials for heat storage: development and characterization" sponsored by INSTM (Consorzio Interuniversitario Nazionale per la Scienza e Tecnologia dei Materiali), within the IEA SHC Task 58 "Material and Component Development for Thermal Energy Storage".

**Conflicts of Interest:** The authors declare no conflicts of interest. The funders had no role in: the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.

#### **References**


cycles of Mn and Co oxides: Pure oxides versus mixed ones. *Sol. Energy Mater. Sol. Cells* **2014**, *123*, 47–57. [CrossRef]


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Solid-State Reactions for the Storage of Thermal Energy**

#### **Stefania Doppiu 1,\*, Jean-Luc Dauvergne <sup>1</sup> and Elena Palomo del Barrio 1,2**


Received: 20 December 2018; Accepted: 4 February 2019; Published: 7 February 2019

**Abstract:** In this paper, the use of solid-state reactions for the storing of thermal energy at high temperature is proposed. The candidate reactions are eutectoid- and peritectoid-type transitions where all the components (reactants and reaction products) are in the solid state. To the best of our knowledge, these classes of reactions have not been considered so far for application in thermal energy storage. This study includes the theoretical investigation, based on the Calphad method, of binary metals and salts systems that allowed to determine the thermodynamic properties of interest such as the enthalpy, the free energy, the temperature of transition, the volume expansion and the heat capacity, giving guidelines for the selection of the most promising materials in view of their use for thermal energy storage applications. The theoretical investigation carried out allowed the selection of several promising candidates, in a wide range of temperatures (300–800 ◦C). Moreover, the preliminary experimental study and results of the binary Mn-Ni metallic system are reported. This system showed a complex reacting behavior with several discrepancies between the theoretical phase diagram and the experimental results regarding the type of reaction, the transition temperatures and enthalpies and the final products. The discrepancies observed could be due both to the synthesis method applied and to the high sensitivity of the material leading to partial or total oxidation upon heating even if in presence of small amount of oxygen (at the ppm level).

**Keywords:** solid state reactions; thermal energy storage; nanocrystalline materials; ball milling

#### **1. Introduction**

The scope of this investigation is the development of performance materials with high energy density, reversibility, long cycle life, compact, low cost and with the potential to build "simple" thermal energy storage systems. This field of research is one of the priorities for i) helping the penetration and dispatchability of renewable energies; ii) contributing to create a low-carbon society for environmental protection; and iii) increasing energy efficiency and decreasing energy demand as targeted in the main road maps [1–3] related to energy policy and environment developed in recent years. In this context, the development of TES materials will play a major role, for example, in helping to the re-utilization of wasted heat (e.g., in industrial processes) and in guaranteeing non-stop energy production, e.g., in solar energy power plants [4,5].

As is already well known, thermal energy can be stored using different processes: sensible, latent and thermochemical storage [5–9]. The energy capacity in these processes increases progressively from sensible to thermochemical processes. Unfortunately, this is accompanied by an increase of the complexity of the TES system that has to be developed, implying a substantial increase of the costs. For example, in the case of gas–solid reactions, the TES system should be composed of two reactors to keep the reactants separated (gas and solid) up to when the discharging process is needed (putting in contact the reactants to promote the exothermic reaction to recover the energy). This

solution is technologically much more complex than the case of sensible storage, where the storage material, solid or liquid, is placed in a unique reactor. As a consequence, thermochemical storage is still at the prototyping/demonstration level and implies, so far, high investment costs. Latent heat storage (using solid–liquid or solid–solid phase transitions) is a more accessible technology nowadays at the demonstration level. Sensible storage is a mature technology already commercialized in many applications.

The main idea of this study is the use of solid-state chemical reactions as materials for thermal energy storage at high temperature. In particular, the focus is given to euteuctoid and peritectoid reactions [10] occurring in binary metals and salt-based systems. The goal was the identification of reactions fulfilling the requirements needed to be used as TES materials (high storage capacity, good thermal conductivity, mechanical and chemical stability, complete reversibility in charging/discharging cycles, affordable cost, etc.) and to obtain the experimental proof of their feasibility and reversibility.

These types of reactions have not been considered so far for the application addressed in this paper. The great advantages and novelties that they are potentially expected to bring in the TES field are as follows:


These reactions offer many advantages, but can also present drawbacks, such as problems connected to the atomic diffusion in the solid-state (slow reaction kinetics) and the poor heat transfer rate in the solid-state especially when salt mixtures are taken into account.

The work included the selection of the most promising materials by a deep analysis of the existing databases of binary metals and salt systems [11], determining all the theoretical thermodynamic properties needed for the evaluation of the TES performances. As a result of the selection process, the Mn-Ni system was chosen for experimental investigation and feasibility study.

#### **2. Materials and Methods**

#### *2.1. Materials Selection*

The search for eutectoid and peritectoid reactions with suitable reaction temperatures was based on available phase diagrams for multi-component systems (ASM International, Scientific Group Thermodata Europe, ThemoTech Inc, etc. (GU2 7YG, Guildford, United Kingdom) and focused on metallic and salt binary systems. The theoretical performances were evaluated by using the CALPHAD (CALculation of PHAse Diagram) method. Regular and sub-regular solution models were used to obtain the Gibbs energy functions of various solution phases. The excess Gibbs energy of each phase was represented by the Redlich-Kister formalism, with binary interaction parameters following the form of power series [12–14]. These parameters were optimized by using the optimization module of FactSage7.0 software (7.0, GTT-Technologies, Herzogenrath, Germany) [15]. In particular, the selection was performed by using the set of evaluated and optimized thermodynamic databases for inorganic systems, such as light metal, alloy, molten salt, oxide. All the key thermodynamic properties (e.g., enthalpy of reaction, specific heats, densities, volume change during the charge/discharge process) of identified eutectoids and peritectoids were obtained assuming equilibrium conditions.

#### *2.2. Nanocrystalline Materials Production*

Two compositions in the Mn-Ni phase diagram were selected for experimental study: the peritectoid Mn75-Ni25 and the eutectoid Mn52-Ni48 (molar ratio). Mn and Ni powder were supplied by Alfa Aesar with purities of 99.3% and 99.8%, respectively. To avoid air contamination, the handling

and sampling were carried out under controlled atmosphere in an Argon glove box (Brown) with levels of oxygen and humidity lower than 0.1 ppm.

To maximize the reactivity in the solid state, the materials were subjected to mechanochemical treatment (Ball milling) to achieve powders with a controlled degree of nanocrystallization. The mechanical milling was used only for the synthesis of nanostructured powder. The goal here was not to promote the reaction by ball milling but the preparation of highly reactive materials (high amount of defects, high specific surface area, and high contact area) and activate the reaction subsequently by thermal treatment.

For this purpose, a Spex mixer mill (875 RPM), using stainless steel vials and balls, was used. Two different milling procedures for the preparation of nanocrystalline materials were applied:


#### *2.3. Structural Analysis*

The structural analysis of the materials was performed by X-Ray diffraction analysis using a Bruker D8 Discover equipped with a LYNXEYE XE detector with monochromatic Cu Kα1 radiation of λ = 1.54056 Å. Patterns were recorded in a 2θ angular range 10–120◦ with a step size of 0.02◦ and a step time of 1.5 s. The measurements were performed at room temperature. The structural evolution upon heating was studied by in situ XRD measurements by using a Bruker Advance D8 instrument with cobalt radiation (λCoα1 = 1.78886 Å/λCoα2 = 1.79277 Å). The equipment operated in Brag-Brentano theta-theta geometry, with an operating power of 30 kV and 50 mA. The samples were placed in a nickel-coated high-vacuum chamber designed for the use in the range from room temperature up to 1200 ◦C (HTK 1200N) under a high vacuum, inert and reactive atmosphere. The sample was mounted on an alumina sample holder avoiding any contact with the wall of the chamber and in contact with the temperature sensor.

Information about the phases formed and their relative percentages, the crystallite sizes and the microstrain level were obtained from the X-ray patterns by using a full profile fitting procedure [16] based on the Rietveld method [17].

The morphology of the material was studied by Scanning Electron Microscopy (SEM, ) using a Quanta 200 FEG scanning electron microscope (FEI Company, Hillsboro, OR, USA) operated in high-vacuum mode at 30 kV and with a back-scattered electron detector (BSED). In addition, energy-dispersive X-ray spectroscopy (EDX) analyses were carried out in order to obtain chemical composition maps.

#### *2.4. Reactivity and Thermodynamic Characterization*

The reactivity of the materials was tested by Differential Scanning Calorimetry (DSC) technique using a Thermal Analysis Q2000 model. These techniques allowed the determination of the reaction temperatures and the reaction enthalpies. For all the measurements the heating rate was 5 K/min with three or twenty cycles between 450 and 660 ◦C including isothermal steps of 30 min between subsequent heating and cooling processes. The structural changes of the materials after DSC experiments were determined by XRD analysis.

#### **3. Results and Discussion**

#### *3.1. Materials Selection Results*

More than 200 binary phase diagrams (metals and salts) were analyzed using available databases and the FactSage7.0 software (7.0, GTT-Technologies, Herzogenrath, Germany) in the temperature range of 300–1000 ◦C. The criteria of selection were based on availability of the materials, no toxicity and relatively low cost. The modelling made it possible to identify all the transitions of interest in the systems studied (eutectoids, peritectoids, peritectics, eutectics, etc.), together with the associated energy densities and main thermodynamic and thermophysical parameters. In this study, all the systems with eutectoid and peritectoid transitions with theoretical volumetric energy densities lower than 100 kWh/m3 were discarded. The results of the theoretical modelling allowed the selection of several potential candidates, in a wide range of temperatures, shown in Table 1, that could be used for further experimental investigation.


**Table 1.** Results of the selection process.

Another aspect considered in this study was how to compare these reactions with other types of thermal energy storage processes (sensible, latent or thermochemical), thinking about their possible integration into a real application. To that end, several aspects have to be considered: (i) we deal with chemical reactions where all the components, reagents and products, are in the solid state; (ii) the reaction mechanism is governed by the atomic diffusion in the solid state; and (iii) the reaction occurs at a well-defined constant temperature. Considering the solid-state nature of these materials, they can be assimilated to a sensible storage material with the difference that at a certain temperature an extra contribution to the sensible heat is given by the enthalpy of the reaction. As a result, depending on the reaction and its energy and on the thermophysical properties, such as the Cp, for the reactions proposed in this paper a higher overall energy density is expected. In Figure 1, the energy density obtained by the sum of the sensible heat contribution and the reaction enthalpy, considering a range of temperature of 100 K around the reaction temperature, is compared to the best sensible storage materials.

**Figure 1.** Theoretical volumetric energy densities of some selected solid-state reactions (black squares). The energy corresponding to a ΔT of 100 K (red squares), together with the values relevant to the best sensible storage materials (ΔT = 100 K), is also pictured.

This allows achieving, for almost all the systems considered, theoretical energy densities between 250–350 kWh/m3. These values are higher than the sensible storage materials considered nowadays. For example, the magnetite, the best material identified so far, presents an energy density of 120 kWh/m3 for a ΔT = 100 K. These results are very promising and confirm the great theoretical potential of these reactions. It is noteworthy that the Mn75-Ni25 shows theoretical energy densities considerably higher compared to the other systems (583 kWh/m3 for the reaction and 707 kWh/m3 when a ΔT of 100 K is considered). Due to these results, the Mn-Ni system was the first choice for the experimental investigation.

To determine univocally the composition with the highest energy density, for each system studied and each transition of interest, four compositions around the theoretical one were analyzed (two before and two after). In particular, the Mn-Ni system presents three solid-state reactions (both eutectoid and peritectoid) below 600 ◦C at the compositions Mn75-Ni25, Mn52-Ni48 and Mn25-Ni75 with promising theoretical energy densities as shown in Figure 2.

**Figure 2.** (**a**) Theoretical energy densities of the solid-state reactions corresponding to the compositions Mn75-Ni25, Mn52-Ni48 and Mn25-Ni75. The relative volumetric energy densities are also reported in the Figure. In (**b**), the Mn-Ni phase diagram is shown, with the corresponding transition highlighted (colored dots).

#### *3.2. Synthesis of Mn75-Ni25 and Mn52-Ni48 Nanocrystalline Materials*

The reactions studied in this paper are governed by the diffusion in the solid state. As a consequence, it is imperative to find synthesis routes in order to maximize the atomic diffusion by decreasing the atomic diffusion path length (small grain sizes as well as small particle sizes), introducing structural defects (dislocation, grain boundaries step, kink and corner atoms, etc.) and promoting high intermixing degree to guarantee the maximum contact between the reagents (high specific surface area). It is well known that a powerful tool for achieving these results is given by mechanochemical techniques [18]. The two procedures applied for the synthesis of nanocrystalline Mn75-Ni25 and Mn52-Ni48 led to the formation of samples with different microstructures and similar degree of nanocrystallization (grain sizes around 15 nm) and, as wanted, with no evidence of the formation of the intermetallic Mn3Ni and MnNi predicted in the phase diagram. The two compositions studied showed different behavior depending on the milling treatment applied. In Figure 3, the XRD patterns for the four samples are reported together with the results of the fitting procedure using the Rietveld method. Only the Mn and Ni reflections are detected in the XRD patterns.

**Figure 3.** X-ray diffraction results of the samples (**a**) Mn75-Ni25 and (**b**) Mn52-Ni48 prepared by one-step or two-step synthesis. The results of the fitting procedure using the Rietveld method are also reported. Black line: experimental pattern. Red line: theoretical pattern. Blue line: difference between experimental and theoretical patterns.

The two-step synthesis caused, for all Mn75-Ni25 samples tested in this investigation, the gradual disappearance of the diffraction peaks corresponding to Nickel (PDF number: 01-071-3740 4-850) (see Figure 3a pattern below). Surprisingly, the gradual disappearance of Ni reflections is not accompanied by a substantial variation of the cell parameters of Mn (PDF number: 01-089-2105 32-637), as a consequence of the substitution of Mn atoms by Ni in the Mn net with the formation of a Mn1−xNix solid solution (see Table 1). The solubilization of Ni into Mn should lead to the decrease of the volume of the primitive cell due to the smaller size of Ni compared to Mn (reference value for Mn being the sample milled 4 h, see Table 2).

**Table 2.** Results of the Rietveld analysis for Mn75-Ni25 and Mn52-Ni48.


The sample prepared by one-step synthesis (where the Ni is clearly visible after 4 h of milling, Figure 3a pattern above) shows the expected behavior (decreasing of the Mn cell parameter), while the sample prepared by two-step synthesis (where the Ni is hardly visible after 4 h of milling, Figure 3a pattern below) shows an opposite trend, with slightly higher cell parameters compared to the Mn milled for 4 h. Further studies were then carried out, by applying mechanical milling progressively higher (30 min, 1 h, 2 h, 4 h and 8 h), in order to clarify this behavior and correlate the disappearing of Ni reflection with the structural modification of Mn. For higher milling times (8 h and 16 h) no Ni was detected in the mixture after mixing (XRD). In Figure 4, the SEM micrograph with the corresponding EDX analysis of the sample milled for 8 h are reported, confirming the results obtained by XRD investigation.

**Figure 4.** SEM micrograph (**a**) and EDX analysis (**b**) of the samples Mn75-Ni25 (two-step synthesis).

For this sample, the particles size distribution was determined by following two approaches a) using a particles size analyzer (Master sizer 3000, Malvern), and b) using the software ImageJ (version 2.0, an open source Java-based software) [19,20]. Following the two approaches, similar results were obtained within media of around 4 μm (SEM/ImageJ) and around 6 μm (particle size analyzer). The difference between the crystallite sizes and the particle sizes is not surprising, due to the phenomena occurring during milling that, in the case of metals, can promote the cold welding of the particles with a consequent increase of their sizes without decreasing the overall reactivity of the material if the crystallite sizes remain small. The behavior under milling and the final structure obtained, depending on the conditions applied, are now subject to deep study due to the very different reacting behavior observed for the different samples, as will be explained later in the text.

#### *3.3. Reactivity upon Heating*

To study the reactivity upon heating and cooling, all the samples synthesized were studied by differential scanning calorimetry and simultaneous thermal analysis techniques. The goal was to determine the reactivity, to quantify the energy relative to the solid-state reaction (peritectoid, eutectoid), and to perform preliminary study on the reversibility by cycling test. The results of the DSC tests are shown in Figure 5.

**Figure 5.** DSC results of the samples: (**a**) Mn52-Ni48 prepared by two-step synthesis, (**b**) Mn52-Ni48 prepared by one-step synthesis, (**c**) Mn75-Ni25 prepared by two-step synthesis, and (**d**) Mn75-Ni25 prepared by one-step synthesis.

Regarding the DSC results several aspects can be highlighted:

There was a discrepancy between the theoretical phase diagram (see inset Figure 1) and the experimental results for the composition Mn52-Ni48, where no reactivity was detected in that range of composition between room temperature and 350 ◦C (eutectoid reaction expected at 251 ◦C). This result was confirmed by further experiments carried out under the same experimental conditions.

The reacting behavior of the composition Mn75-Ni25 is strongly influenced by the experimental conditions applied for the synthesis of the material (no reactivity in case of one-step synthesis).

Following the theoretical phase diagram at the composition Mn75-Ni25, the reaction between Mn and Ni should lead to the formation of the intermetallic Mn3Ni. This reaction was expected to be exothermic, while an endothermal event was observed in the experimental results at higher temperature than the predicted one (630 ◦C instead of 566 ◦C). Moreover, the XRD patterns performed after DSC measurements reveal, for most of the samples, only the presence of Mn, Mn(1−x)NixO and traces of Ni. The detection of the formation of a small amount of one new "unknown phase" was possible only after many experiments, where the extensive oxidation of the sample during the DSC experiment was limited.

The reaction is reversible with a progressive increase of the enthalpy upon cycling. It is noteworthy that the value of the enthalpy for this system cannot be given precisely due to the very high reactivity of the material upon heating, leading to oxidation even when controlled atmosphere (level of oxygen below 0.1 ppm) or vacuum is applied. The energy obtained experimentally is considerably lower than the predicted one (around 10 J/g instead of 300 J/g). This behavior is amplified and clearly visible by performing cycling experiments in the DSC apparatus (up to 20 cycles), as shown in Figure 6.

**Figure 6.** Cycling results of the sample Mn75-Ni25 synthesized by the two-step synthesis. (**a**) Heating steps, (**b**) cooling steps.

The cycling results show the progressive increase of the enthalpy of the reaction up to a certain limit, when the enthalpy starts to decrease to very low values (the DSC peak almost disappears). During the cooling process, a partial displacement of the peak during the first cycles can also be observed, reaching a stationary regime after eight cycles. In these experiments, two competitive effects are occurring at the same time, the progressive oxidation of the sample, which causes the increase of the inactive phase in the mixture (no contribution to the reaction heat), and the increase of the reaction enthalpy due to the intrinsic behavior of the mixture (still under investigation).

The X-ray diffraction analysis after the cycling experiment (20 cycles) did not allow correlation of the final structure with the reactivity observed. Only in the case of one of the samples cycled three times was it possible to obtain an XRD diffractogram (reported in Figure 7) in which the formation of a new unknown phase was detected (diffraction peaks at 2θ angular range of 36.2◦, 47.2◦ and 68.9◦). This phase could correspond to the intermetallic Mn3Ni predicted in the phase diagram; unfortunately, no crystallographic information is available for this phase, hindering its univocal determination.

**Figure 7.** X-ray diffraction results of Mn75-Ni25 sample (two-step synthesis) after DSC experiment.

To shed light on the reacting behavior observed and to determine the phenomena occurring upon heating several "in situ" XRD experiments were planned and carried out applying the same heating protocol used in the DSC experiments. The goal was to correlate the transitions observed in the DSC with the corresponding structural modification in the mixture. Unfortunately, none of the attempts made in order to obtain these measurements were successful, due to the extremely high reactivity of the mixture in the presence of traces of oxygen. Different trials were carried out under vacuum, under overpressure of N2 and in dynamic atmosphere (continuous vacuum and N2 flux). In all cases, the complete oxidation of the Mn75-Ni25 mixture was observed with the formation of the mixed oxide Mn1−xNixO (see Figure 8). It is noteworthy that, for the composition studied, the formation of the two solid solutions Mn1−xNixO and Ni1−xMnxO should be detected, while only one phase was detected [21].

**Figure 8.** "In situ" X-ray diffraction measurements of Mn75-Ni25 sample (two-step synthesis).

The same problems of oxidation were encountered when trying to perform measurements to test the thermophysical properties of Mn75-Ni25 upon heating (for examples in the case of LFA measurements), not making it possible to obtain further experimental results.

#### **4. Conclusions and Perspectives**

In this paper, solid-state reactions are proposed as possible candidates for thermal energy storage applications at high temperature. This study allowed the selection of several reactions with theoretical energy density above 100 kWh/m3. Two Mn-Ni compositions (Mn75-Ni25 and Mn52-Ni48) were chosen for the experimental study. The reaction was activated by thermal treatment after the preliminary preparation of nanostructured powders by ball milling techniques. The behavior of this system revealed considerably more complex than expected. More and more questions arose with the proceeding of the experimental investigation. For example, i) it is well known that mechanical milling is a powerful technique to increase the solubility limit in binary metallic systems; however, the solubilization of considerably high amount of nickel in the manganese net was accompanied by only a slight variation of the cell parameters. ii) A high degree of solubilization of Ni into Mn in the case of one-step synthesis where the two elements were milled at high milling intensity for up to 4 h was expected, while the results show an opposite behavior, with a higher solubilization degree being reached in the mixture prepared by two-step synthesis (the two elements milled together only for 15 min in mild conditions). iii) The direct milling of the two elements (one-step synthesis) led to a total absence of reactivity, even if similar degrees of nanocrystallization and homogenization were achieved, compared to the two-step synthesis. In addition, finally, what type of reacting event is connected to the reversible and progressively increasing peak observed in the DSC analysis? We are now pursuing different strategies to understand all of the behavior observed. The work is more focused on the material science point of view to explain the peculiar reactivity observed. Regarding the performance of this material for thermal energy storage applications, two main aspects can be considered. On one hand, the discrepancies between the theoretical reaction enthalpy and the experimental one are probably due to the progressive oxidation of the sample during thermal cycling. In this respect, it is hard to be definitive with regard to the performance as thermal energy storage material up to when it will be possible to test the material avoiding its oxidation. On the other hand, a material so reactive is not suitable for thermal energy storage applications because of the extremely controlled experimental conditions needed to avoid its degradation. This could be an important technological constraint leading, most probably, to investment costs that are too high for large-scale applications.

**Author Contributions:** Conceptualization, S.D. E.P.; Methodology, S.D.; Software, J-L.D.; Validation, S.D., E.P.; Investigation, S.D.; Resources, E.P.; Data curation, S.D., J-L.D., E.P.; Writing—original draft preparation, S.D.; Writing—review and editing, S. D., J-L.D., E.P.; Visualization, S.D., J-L.D.; Project administration, S.D.; Funding acquisition, S.D., E.P.

**Funding:** This research was funded by the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 752520". SOLSTORE project (Solid-state reactions for thermal energy storage).

**Acknowledgments:** The authors acknowledge Cristina Luengo Vilumbrales and Maria Jáuregui for the help and commitment in the experimental measurements.

**Conflicts of Interest:** The authors declare and confirm that there is no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Nanomaterials* Editorial Office E-mail: nanomaterials@mdpi.com www.mdpi.com/journal/nanomaterials

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18