Synthesis and characterization of Magnesium-Iron-Cobalt complex hydrides

The formation, structure and deuterium desorption properties of Mg2FexCo(1−x)Dy (0 ≤ x ≤ 1 and 5 ≤ y ≤ 6) complex hydrides were investigated. The synthesis was carried out by reactive ball milling, using a mixture of powders of the parent elements in D2 atmosphere. The formation of quaternary deuterides was identified from Rietveld refinements of powder X-Ray diffraction and powder neutron diffraction patterns, and from infrared attenuated total reflectance analysis. It was observed that the crystal structure of deuterides depends on the transition metal fraction. For Co-rich compositions, i.e. up to x = 0.1, hydrides have the tetragonal distorted CaF2-type structure (space group P4/nmm) of Mg2CoD5 at room temperature. For Fe-rich compositions, i.e. x ≥ 0.5, a cubic hydride is observed, with the same K2PtCl6-type structure (space group Fm3¯\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{{\bf{3}}}$$\end{document}m) as Mg2FeD6 and as Mg2CoD5 at high temperatures. For x = 0.3, both the cubic and the tetragonal deuterides are detected. Differential scanning calorimetry coupled with thermogravimetric and temperature programmed desorption analyses show rather similar deuterium desorption properties for all samples, without significant changes as a function of composition. Finally, hydrogen sorption experiments performed for Mg2Fe0.5Co0.5H5.5 at 30 bar of H2 and 673 K showed reversible reactions, with good kinetic for both absorption and desorption of hydrogen.

Fe is fully immiscible in Mg 8 , while Co forms a stable intermetallic compounds (MgCo 2 ) 9 . Thus, neither of the TMs form stable intermetallic compounds Mg 2 TM to be used as precursor for the hydride 10 . This means that the synthesis of ternary complex hydride is challenging, since the direct hydrogenation of the intermetallic phase is not possible. Thus, it is necessary to proceed with a reaction between the TMs and MgH 2 or Mg under hydrogen atmosphere, as summarized by the two reactions below: It should be noted that, Eq. 2 can occur via the intermediate formation of MgH 2 , after which the reaction proceeds according to Eq. 1 11 . In general, the route governing the formation of Mg 2 TMH x complexes strongly depends on the synthesis method and processing conditions 10,[12][13][14][15][16][17] . Regardless the route, it has been proposed that hydrogen is attracted to Mg(H 2 )/TM interfaces in order to reduce the interfacial energy, which is positive due to both topological disorder and the positive heat of mixing of Mg and Fe or Co 4 .
A conventional method of synthesis, such as annealing, requires a reaction at high temperature (e.g. >673 K) and high hydrogen pressure (e.g. 50-100 bar) for relatively long reaction times (i.e. more than one day), and only provides yields around 50% 18,19 . On the other hand, mechanochemical synthesis methods in a reactive atmosphere allow for the formation of those hydrides at lower hydrogen pressures (i.e. ≤50 bar) and close to RT, reaching yields of more than 80% 20,21 . The reactive mechanochemical synthesis has the advantage of enhancing hydrogen sorption kinetics, due to the formation of fresh surfaces and reduced particle sizes 22 , allowing hydride formation already after few hours of milling 21,22 .
In the last few years, some interest has been devoted to Mg-based complex hydrides with more than one TM and the formation of quaternary hydrides with Mg, Fe and Co has been already reported 4,11,23 . Baum et al. 11 presented the formation of Mg 2 (FeH 6 ) 0.5 (CoH 5 ) 0.5 , hinting at a complex formation process, due to the immiscibility of Fe and Co with Mg. Deledda and Hauback 4 reported in more details the structure and thermal stability of Mg 2 (FeH 6 ) 0.5 (CoH 5 ) 0.5 , showing that the quaternary hydride is isostructural to Mg 2 FeH 6 and to the high-temperature phase of Mg 2 CoH 5 , containing both [FeH 6 ] 4− and [CoH 5 ] 4− complex anions 4 . Moreover, they showed a hydrogen desorption temperature of 570 K, intermediate to 560 K for Mg 2 FeH 6 and 585 K for Mg 2 CoH 5 . Finally, Zélis et al. 23 investigated the synthesis of mixed Mg 2 (FeH 6 ) (1−x) (CoH 5 ) x systems, with different Fe-Co contents (x = 0.25, 0.5, 0.75). However, no detailed structural and thermal characterizations were reported, suggesting that further studies on this system are necessary.
The aim of this work is to synthetize and investigate the properties of Mg 2 Fe x Co (1−x) D y complex deuterides with different Fe-Co and D contents (0 ≤ x ≤ 1 and 5 ≤ y ≤ 6), comparing results with the ternary Mg 2 CoD 5 and Mg 2 FeD 6 compounds. Deuterium was used instead of hydrogen to allow the structural study with Powder Neutron Diffraction (PND). The use of neutron diffraction is crucial for characterizing the crystalline structure, since it allows to distinguish Fe and Co (which have very similar X-ray scattering cross section) and to determine the occupancy and the position of deuterium (i.e. H). The quaternary deuterides/hydrides synthesized in this study where found to be isostructural either with Mg 2 CoD 5 (tetragonal P4/nmm), for x = 0.1, or with Mg 2 FeD 6 (cubic Fm3m) for x ≥ 0.46. For x = 0.3, two hydrides are formed: Mg 2 (FeD 6 ) 0.3 (CoD 5 ) 0.7 (tetragonal P4/nmm) and Mg 2 (FeD 6 ) 0.4 (CoD 5 ) 0.6 (cubic Fm3m). All hydrides have a similar hydrogen desorption process, with a maximum desorption temperature T max ≅ 550 K and activation energy of desorption Ea des ≅ 95 kJmol −1 . The enthalpy of desorption has been used to determine the thermodynamics of tetragonal and cubic solid solutions. Rehydrogenation of Mg 2 (FeH 6 ) 0.5 (CoH 5 ) 0.5 occurs at 673 K in 30 bar of hydrogen with relatively good kinetics. experimental Synthesis. The synthesis of quaternary deuterides, with formula Mg 2 Fe x Co (1−x) D y was achieved by Reactive Ball Milling (RBM) using a deuterium atmosphere inside the milling vial. The nominal amount of iron and cobalt is given by x (x = 0.1, 0.3, 0.5, 0.7 and 0.9) and refers to the nominal content of Fe and Co in the starting elemental powder mixtures. For simplicity, we refer to those samples as Fe0.1, Fe0.3, Fe0.5, Fe0.7, Fe0.9, respectively. The ternary compounds Mg 2 CoD 5 and Mg 2 FeD 6 , referred as sample Fe0.0 and Fe1.0, respectively, were also prepared for comparison.
The synthesis was carried out using elemental powder of Fe (200 mesh), Co (350 mesh) and Mg (350 mesh) with a purity level over 99% (purchased from Alfa Aesar) and deuterium gas (purchased from Nippon Gases) with purity >99.5%. No further gas purification process was applied. The milling was carried out in a Fritsch Pulverisette 6 (P6) planetary ball milling, using a specially designed hardened steel vial, commercialized by Evico Magnetics. The vial is rated to 150 bar and equipped with a temperature and pressure monitoring system. In this work, milling was carried out in 50 bar of D 2 , at RT, for 20 hours at 400 rpm, using 10 mm diameter hardened steel balls, with a ball-to-powder weight ratio of approximately 40:1. The amount of deuterium absorbed during milling was calculated applying the perfect gas law from the changes in pressure and temperature recorded by the monitoring system and by taking into account the free volume within the vial. After the synthesis, about 200 mg of the as-milled powders were annealed to reduce the internal stresses created by milling, allowing an accurate structural characterization. The thermal treatment was performed at 473 K in 50 bar of D 2 for a period of approximately 48 hours. All samples were handled in a glovebox in a purified Ar atmosphere.
Structural characterization. Powder X-ray Diffraction. Powder X-Ray Diffraction (PXD) analysis was performed using a Bruker D8 A25 diffractometer. It was equipped with Mo K-α radiation and a Lynxeye detector. The powder samples were packed in glass capillaries with a diameter of 0.5 mm. The PXD scan speed was 2 s per step, with steps of 0.04° from 5° to 45° in 2θ. Rietveld refinements were carried out with the software Maud 24 and Thermal characterization. Differential Scanning Calorimetric coupled with Thermal Gravimetric Analysis. The thermal stability was studied by Differential Scanning Calorimetry coupled with Thermal Gravimetric Analysis (DSC-TGA). Measurements were performed on a STA 449 F3 Jupiter instrument produced by Netzsch, and carried out with an Argon flow of 50 ml/min. The as-milled powders were heated at different heating rates (1, 5, 10, 20 and 40 K/min), from RT up to a maximum of 873 K. The maximum peak temperatures T max of the endothermic events obtained at different heating rates were used to calculate the desorption activation energy (Ea des ) using the Kissinger method 28 , after having verified the isokinetic conditions 28 . In addition, the enthalpy of hydrogen desorption (ΔH des ) was estimated by integrating the area of the endothermic peaks in the DSC trace.
Thermal programmed desorption. Thermal Programmed Desorption (TPD) was performed using an instrument built in-house, which is equipped with a diaphragm pump and a turbomolecular pump to reach 10 −5 mbar. As-milled powders were heated at 5 K/min from RT up to 823 K in vacuum.
Volumetric measurements by Sievert's Method. Hydrogen sorption measurements were performed by Sievert's method with a volumetric apparatus from AMC (Pittsburgh). Analysis were performed with hydrogen gas (purchased from Nippon Gases) with purity >99.9999%. Measurements were carried out in isothermal conditions at 673 K. Desorption was obtained in vacuum for a maximum of 10 h, using a rotary pump (10 −2 bar), while absorption was performed at 10 bar and 30 bar of H 2 for a maximum of 35 h.

Results and discussion
Synthesis with reactive ball milling. Figure 1a shows D 2 absorbed by the powders as a function of milling time. After an incubation time ranging from 30 to 50 min, absorption starts and continues for about 8-10 h. A high rate of hydrogenation is observed during the first 2 hours, followed by a slower absorption up to the end of the hydrogenation. The latter can be linked to the progressive consumption of fresh Mg/Fe-Co interfaces, which results in slower hydrogenation rates towards the end of the milling process 4 . This trend can be better observed by plotting the absorption rate as a function of milling time (Fig. 1b). The higher the amount of iron, the higher the rate of hydrogenation. Even a small amount of Fe drastically increases the hydrogen absorption kinetics, as seen by comparing Fe0.0 (no Fe) and Fe0.1 (nominal amount of Fe 0.1) in Fig. 1b. www.nature.com/scientificreports www.nature.com/scientificreports/ The deuterium content obtained after RBM is reported as a function of the nominal amount of iron in Fig. 2. A rather linear trend is observed, reflecting the ability of Fe to coordinate more deuterium atoms (six) with respect to Co (five). On the other hand, there are no significant differences between the two ternary hydrides and Fe or Co rich ones (i.e. Fe0.0-Fe0.1 and Fe1.0-Fe0.9), respectively. Samples Fe0.3, Fe0.5 and Fe0.7 also absorb a comparable amount of deuterium.  Table 1 summarizes the structural information obtained from the Rietveld refinements, including the formula obtained from the refinements of the occupancy. From Table 1, we can see that adding Fe in the tetragonal structure, is causing a decrease in the unit cell parameter c, while a is increasing. The addition of Co in the cubic structure of Mg 2 FeD 6 causes a decrease in unit cell dimension. Figures 3 and S3 show the experimental and calculated PND patterns for sample Fe0.3 and Fe0.7, respectively.

Structural characterization of as
The refinements confirm the formation of a cubic phase in Fe09, Fe0.7 and Fe0.5. For the latter, results agree with ref. 4 . The hydrogen volumetric capacity calculated from the unit cell volume is 138 g/l for Mg 2 Fe 0.5 Co 0.5 D 5.5 and 143 g/l for Mg 2 Fe 0.7 Co 0.3 D 5.7 , suggesting similar hydrogen storage capacities.
For sample Fe0.3 the structural analyses are more challenging. The peaks at 60° and 110° in 2θ (Fig. 3), hints at the presence of a structure which is not purely cubic. The peak at 60° shows a shoulder, which might suggest the influence from a tetragonal phase. Indeed, refinements with a single cubic phase resulted in a poor fit. On the other hand, refinements with a single tetragonal phase with Mg 2 CoH 5 -type structure did not yield a satisfactory fit either. Therefore, a refinement was performed considering the presence of the two pure ternary hydrides, Mg 2 CoD 5 and Mg 2 FeD 6 , but it was impossible to reach convergence and the unit cell parameters were far from  www.nature.com/scientificreports www.nature.com/scientificreports/ the expected values. However, refinements with both a cubic phase with Co substituting Fe and a tetragonal phase with Fe substituting Co, together with a ~9 wt% of unreacted metallic Fe, yield excellent fit to both the PND and PXD data. To decrease the number of refinable parameters, the composition of the cubic phase was fixed to Mg 2 (Co 0.6 Fe 0.4 )D 5.4 as estimated from the unit cell parameter (a = 6.419(1) Å). The Fe/Co ratio in the tetragonal phase was refined with a soft constraint on keeping the overall elemental composition of the model phases similar to the nominal composition. The deuterium coordination around the transition metal site was set to a fully occupied square pyramid, similar to that in Mg 2 CoD 5 , with an additional, partly occupied apical site below the basal plane. The occupancy of the addition apical site was set to be equal to the occupancy of Fe on the transition metal site, thus accounting for the expected octahedral D coordination around the Fe atoms. The difference between refined and nominal elemental compositions was less than 7 wt% for all elements. A model where both the apical sites in the tetrahedral phase were partly occupied was also tested but yielded slightly poorer fits.
Zélis et al. 23 previously reported on the quaternary hydride Mg-Fe-Co-H with a Fe:Co ratio of 0.25:0.75. Based on PXD analysis, they assigned a tetragonal symmetry P4/nmm, isostructural with Mg 2 CoH 5 . Moreover, using Mössbauer spectroscopy, a non-cubic symmetry was revealed for the Fe sites 23 . Thus, considering the results reported by Zelis et al. 23 and the one obtained in this work, we can conclude that for 0 ≤ x ≤ 0.25 there exists a tetragonal hydride Mg 2 Fe x Co (1−x) D y with 5 ≤ y ≤ 5.25. Around x = 0.3, a two-phase region is found where the tetragonal hydride Mg 2 Fe x Co (1−x) D y coexists with a cubic hydride, as that formed for 0.5 ≤ x ≤ 0.9. It is worth noting that these results agree with Verbovytskyy et al. 29 who found a tetragonal P4/nmm symmetry, isostructural with Mg 2 CoH 5 , for the quaternary hydride Mg 2 Ni 0.5 Co 0.5 H 4.4 and suggested that quaternary Mg 2 M x M' y H x hydrides display the same the structure of the parent ternary phases.
For all samples, ATR-IR spectra are shown in Fig. 4. As reported in refs. 3,6 , [FeD 6 ] 4− shows one stretching band, while [CoD 5 ] 4− shows two bands, which reflect the octahedral and square-base pyramidal symmetry, respectively. They appear in the same frequency range. The observed frequencies are in good agreement with literature 3,6 : for sample Fe1.0, the [FeD 6 ] 4− stretching is observed at 1261 cm −1 , while, for Fe0.0, the two bands for   Table 1 shows also the amount of hydride phase formed, obtained from the refinements with PXD patterns, using crystal structure information established by PND. The amount of hydride formed after milling is in all cases higher than 80%. Such high yields have been obtained thanks to the use of high-energy milling techniques, confirming that it is more suitable for the synthesis of Mg-Fe-Co hydrides with respect to conventional sintering methods 18,19 . Using the quantitative results, a mass balance was performed to calculate the amount of D 2 absorbed in all samples (Fig. 2). In most cases, the results confirm the amount obtained by RBM reported in Fig. 2, except for the Fe0.0 and Fe0.1, which have an error of 9 and 8%, respectively.
It can be concluded that the quaternary hydride structure is influenced by the amount of complex anions present. In all five mixtures investigated in this study, Fe and Co randomly occupy the same site, creating solid solutions with a tetragonal structure of Mg 2 CoD 5 for Co-rich samples and with a cubic structure of Mg 2 FeD 6 for Fe-rich ones. The hydride formed are Mg 2 (FeD 6 Fig. S4), the desorption clearly presents two overlapping peaks, while in the analysis at lower heating rates (1, 5, 10 K/min in Figs. S5-S7) only a broad desorption peak is detected. This agrees with previous reports 5,7,17 which claim that hydrogen desorption of Mg 2 CoH 5 can involve the formation and successive decomposition of other Mg-Co hydrides (i.e. Mg 6 Co 2 H 11 30 or Mg 3 CoH 5 5 ). However, in Fe0.1 the double peak is less pronounced, suggesting that small amounts of Fe can affect the desorption mechanism of the tetragonal Mg 2 CoD 5 phase. This is not observed for Fe0.3. In this sample, two hydrides need to desorb (section 3.2), but two separate signals are not detected. This indicates similar desorption temperatures of the two hydrides. In the samples Fe0.0 and Fe0.1, a small exothermic peak is also observed at about 650-700 K and is associated to the formation of the intermetallic compound MgCo 5,17 . This is a metastable phase, as presented in the introduction, only MgCo 2 is stable in the Mg-Co system 9 .
From the comparison of the derivative of TG curves with DSC and TPD traces at 5 K/min (Figs. S8, S6 and S9, respectively), it can be clearly observed that the weight loss is associated with a deuterium desorption event. As seen from the DSC signals (Figs. 5 and S4-S7) and Table 2, summarizing the values of the maximum temperature of the desorption peaks registered from TPD analysis at 5 K/min, the ternary and quaternary hydrides have similar temperatures of desorption. This is in contrast with TPD analysis at 2 K/min in ref. 4 , in which it was found that Mg 2 FeH 6 has a maximum of desorption temperature at 560 K, Mg 2 CoH 5 at 585 K (a single desorption peak is reported) and Mg 2 (FeH 6 ) 0.5 (CoH 5 ) 0.5 at 570 K. This discrepancy can be related to the mechanism governing hydrogen desorption in Mg 2 CoH 5 , which is not yet fully explained. The occurrence of a double peak observed here has been also reported earlier, even for low heating rates 5,30 and is related to the formation and decomposition of Mg 6 Co 2 H 11 30 or Mg 3 CoH 5 5 . Zépon et al. 31 proposed two different mechanisms, depending on temperature, for hydrogen desorption. At low temperatures, hydrogen is released without any structural changes of the hydride. However, at temperatures above 573 K, another mechanism involving the formation of Mg 2 CoH x<5 , which then decomposes up to the end, takes place 31 . The discrepancy between this work and ref. 4 . can therefore be linked to the complexity of the Mg 2 CoH 5 decomposition. Thus, we cannot confirm that the thermal behaviour www.nature.com/scientificreports www.nature.com/scientificreports/ of the quaternary hydrides is intermediate to that of ternary hydrides 4 , but it can be concluded that it is rather similar. Figure 2 reports the D 2 weight loss registered from TG analysis at 5 K/min. The results are slightly lower than the amount absorbed during the synthesis. This underestimation is likely caused by a slight sample oxidation during the TG measurements, since the instrument is not placed in a glovebox and samples are air sensitive. Table 2 presents the experimental values of ΔH des obtained from the DSC analysis and the Ea des calculated applying the Kissinger method (see section 2.3.1.) on DSC traces and on the derivative of the TG curves (Fig. S10). As can be seen from Table 2, the Ea des data obtained from the two techniques are similar, except for Fe0.0. For the ternary hydrides, ΔH des and Ea des are in good agreement with literature values 14,32,33 (Table 2), except for Ea des of Mg 2 CoH 5, which is significantly lower than 114.8 kJ/mol reported by Norek et al. 5 . Results for the hydrides for samples from Fe0.1 to Fe 0.9 are comparable, with values of Ea des 89-99 kJmol −1 and ΔH des 69-74 kJ/mol D2 , respectively.
In summary, all hydrides investigated in this work have a comparable thermal stability, which is not significantly influenced by the relative amount of the transition metals. Indeed, the thermal behaviour is linked to the strength of the TM-D bond and correlates to the amount of energy necessary to break the bond to release D 2 (H 2 ). In complex hydrides, TM-D is a covalent bond, which means that the hydrides are stable, and relatively high temperatures are necessary for hydrogen desorption. In this case, the Fe-D and Co-D bonds have similar strengths and changing the amount of complex anions in the hydrides should not have a big influence on thermal stability and desorption temperatures.
Structural and microstructural analysis after hydrogen desorption. The structural and microstructural analysis after hydrogen desorption was performed on samples after TPD analysis. The PXD patterns (Fig. S11) confirm previously reported results 4 . A bcc-(FeCo) solid solution and elemental Mg are observed, in samples from Fe0.3 to Fe0.9, while for the ternary Mg 2 FeD 6 hydride, Fe and Mg form. An intermetallic Mg-Co compound is also observed in Co-rich samples, i.e. from Fe0.0 to Fe0.3. As mentioned previously, Mg and Co only form one stable intermetallic compound, MgCo 2 9 . Other metastable phases, such as MgCo or Mg 2 Co, have been reported after the decomposition of Mg 2 CoH 5 5,7,16,17,34 . The exact stoichiometry of the Mg-Co intermetallic compound formed after desorption of Mg 2 CoH 5 is widely discussed in the literature 5,7,16,17,34 , since the temperature influences the nature of this product 17 , due to the formation of metastable intermetallic phases. Here it was not possible to define the exact stoichiometry or crystal structure of the observed Mg-Co compound, but we exclude that it is MgCo 2 or MgCo, as no diffraction peaks match with the MgCo or MgCo 2 structures.
Mg is hard to be detected in all samples (Fig. S11). This can be better visualized in Fig. S12, which shows the patterns after desorption for sample Fe0.3, Fe0.5 and Fe0.7. Quantitative analysis on the latter two results in Mg amounts less than 10 wt.% which are not representative of the nominal composition (≈ 46 wt.%).
To better understand the elemental distribution after desorption, a SEM analysis with EDS mapping was performed on sample Fe0.7, comparing powder morphologies and the distribution of Mg, Fe and Co after milling, after annealing and after decomposition. Reactive milling produces a fine powder morphology (Fig. 6a), and this is observed also in powders after annealing (Fig. 6c). The average particles size is of the order of 10 μm, with some powder agglomerations. In the desorbed powder (Fig. 6e), a similar powder morphology is present. The EDS maps for the as-milled powder (Fig. 6b), show a uniform distribution of Mg, Co and Fe elements, suggesting a homogeneous distribution of the hydride. Moreover, the measured weight percentage of elements, i.e. Mg 41 wt.%, Fe 40 wt.% and Co 19 wt.%, agrees well with the nominal composition, i.e. Mg 46 wt.%, Fe 37 wt.% and Co 17 wt.%, thus confirming the effectiveness of the milling process. The annealed sample (Fig. 6d), still presents a homogeneous distribution of the elements, representative of the quaternary hydride, but some regions with only Fe and Co are also observed. This indicates that annealing promotes the growth and phase segregation of the (FeCo) phase. The overall elemental composition (Mg 37 wt.%, Fe 43 wt.% and Co 20 wt.%) is very similar to that measured before annealing and agrees with the nominal composition. On the contrary, in the desorbed sample (Fig. 6f), Mg is confined to fewer regions and is detected in low quantities, whereas iron and cobalt are predominant. The result of the elemental analysis (Mg 5 wt.%, Fe 66 wt.% and Co 29 wt.%) agrees with what was found from PXD, suggesting that desorption causes a strong phases separation. Moreover, we cannot exclude a  www.nature.com/scientificreports www.nature.com/scientificreports/ extending up to positive values for higher Fe-contents, since no tetragonal phase has ever been reported for those compositions.
Based on the schematic representation in Fig. 7 and by applying the common tangent construction, we can identify a two-phase region for 0.25 < x < 0.4 where the tetragonal and cubic phases coexist. The occurrence of a phase mixture in such a region is supported by our results for Fe0.3, which show the presence of both tetragonal Mg 2 (FeH 6 ) 0.3 (CoH 5 ) 0.7 and cubic Mg 2 (FeH 6 ) 0.4 (CoH 5 ) 0.6 . It should be stressed that the estimated ΔH f for Fe0.3, which is listed in Table S1 and included in Fig. 7, is the value obtained from the desorption of both phases, since two separate endothermic peaks were not observed (see Fig. 5). Thus, an estimate of ΔH f for each compound was not possible.
The reversibility of hydrogen release and uptake reactions was tested at isothermal conditions at 673 K for sample Fe0.5 in the as-milled state. Desorption in vacuum was very fast and in less than 30 min all hydrogen was desorbed. On the contrary, a rather slow kinetics of absorption was observed at 10 bar of H 2 , with only 0.22 wt.% H 2 absorbed in 35 hours. PXD measurements performed after absorption at 10 bar revealed that only a small amount of MgH 2 were formed. Moreover, SEM-EDS analysis performed after absorption still show significant segregation of the unreacted FeCo and Mg phases, as observed for the samples after decomposition (section 3.5).
A much faster absorption kinetics is observed when the H 2 pressure is increased up to 30 bar. Indeed, most of the hydrogen is absorbed within 2 hours and the final hydrogen content is 4.4 wt.% H 2 . This is comparable to the amount registered during desorption (i.e. 4.3 wt.% H 2 ). PXD measurements on the sample after the rehydrogenation (Fig. 8) show that Mg 2 (FeH 6 ) 0.5 (CoH 5 ) 0.5 is the main reaction product, while MgH 2 is also formed as a secondary phase.
EDS elemental mapping show a quite uniform elemental distribution, as observed for the as-milled sample (section 3.5). The formation of the quaternary hydride Mg 2 (FeH 6 ) 0.5 (CoH 5 ) 0.5 confirms that hydrogen sorption reactions are reversible also for the mixed transition metal quaternary systems. Rehydrogenation is possible