Transformation Kinetics of LiBH4–MgH2 for Hydrogen Storage

The reactive hydride composite (RHC) LiBH4–MgH2 is regarded as one of the most promising materials for hydrogen storage. Its extensive application is so far limited by its poor dehydrogenation kinetics, due to the hampered nucleation and growth process of MgB2. Nevertheless, the poor kinetics can be improved by additives. This work studied the growth process of MgB2 with varying contents of 3TiCl3·AlCl3 as an additive, and combined kinetic measurements, X-ray diffraction (XRD), and advanced transmission electron microscopy (TEM) to develop a structural understanding. It was found that the formation of MgB2 preferentially occurs on TiB2 nanoparticles. The major reason for this is that the elastic strain energy density can be reduced to ~4.7 × 107 J/m3 by creating an interface between MgB2 and TiB2, as opposed to ~2.9 × 108 J/m3 at the original interface between MgB2 and Mg. The kinetics of the MgB2 growth was modeled by the Johnson–Mehl–Avrami–Kolmogorov (JMAK) equation, describing the kinetics better than other kinetic models. It is suggested that the MgB2 growth rate-controlling step is changed from interface- to diffusion-controlled when the nucleation center changes from Mg to TiB2. This transition is also reflected in the change of the MgB2 morphology from bar- to platelet-like. Based on our observations, we suggest that an additive content between 2.5 and 5 mol% 3TiCl3·AlCl3 results in the best enhancement of the dehydrogenation kinetics.


Introduction
In the context of a low-carbon economy and the present global energy crisis, there is a rapidly growing demand for sustainable energy based on renewable energy sources, e.g., solar, wind power, and hydroelectric power [1]. Countless studies have been devoted to storing power from these intermittent sources. In addition to electrical energy storage, another important option is to store energy in the form of hydrogen in various media [2][3][4]. Hydrogen as a zero-carbon fuel can deliver an excellent energy density of approximately 120 MJ/kg. Currently, the most commercially utilized way to store hydrogen is by compressing hydrogen under ultrahigh pressure up to several hundred bars or by liquefying hydrogen below its boiling point down to −253 • C under atmospheric pressure and keeping it in specifically designed vessels [5,6]. However, harsh conditions are still necessary for both approaches, which will inevitably reduce the public acceptance of the extensive application of hydrogen. Therefore, storing hydrogen in solid-state materials is commonly regarded as the promising form of hydrogen storage [7]. LiBH 4 -MgH 2 is one of the promising reactive hydride composites for hydrogen storage [8][9][10]. By combining the complex metal hydride LiBH 4 and the metal hydride MgH 2 , the standard dehydrogenation enthalpy of the mixture is remarkably lowered by Figure 1a shows the XRD diffractograms of the samples with varying contents of 3TiCl 3 ·AlCl 3 additive in the as-milled state. Different peaks were detected, marking the reactants LiBH 4 and MgH 2 . In addition to this, LiCl can also be identified in the diffractogram when the additive content exceeds 2.5 mol%. This by-product is generated from the reaction between LiBH 4 and 3TiCl 3 ·AlCl 3 during the ball milling. With 10 mol% additives, the LiCl peaks are significantly enhanced, whereas the peaks for LiBH 4 are greatly weakened. This happens as LiBH 4 is consumed by the additive [20]. For the desorbed samples, the diffraction peaks of the LiBH 4 and MgH 2 were not observed, see Figure 1b, while the peaks of LiH and MgB 2 appeared. This indicates that the reaction between LiBH 4 and MgH 2 proceeded with the generation of MgB 2 .
Another way to verify that the dehydrogenation process of LiBH 4 -MgH 2 is completed is to monitor the amount of hydrogen released from the material over time. As shown in Figure 1c, the more additive that was added to the system, the less hydrogen was released, which is consistent with the intensified LiCl peaks shown in Figure 1a,b. This means that less LiBH 4 is available for the reaction with MgH 2 . One anomaly is that the sample without additives only released less than 10 mol% H 2 , which is below the expectation of a theoretical value of 11.5 mol% H 2 . This can be attributed to the partial oxidation of LiBH 4 or/and MgH 2 , the low purity of the raw materials from the suppliers, or the inhomogeneous dispersion of reactants over ball milling, and so on. It is also notable that the additive improves the dehydrogenation kinetics very effectively. In contrast to the curve with respect to hydrogen release from the sample without additives, the incubation stage almost entirely disappeared with the addition of only 1 mol% 3TiCl 3 ·AlCl 3 . In general, the duration of the dehydrogenation process is shortened from about 12 h without additives to about 4 h with additives. Furthermore, these curves seem to have different sigmoidal patterns, which might relate to the different MgB 2 growths during the reaction between LiBH 4 and Mg.
that the additive improves the dehydrogenation kinetics very effectively. In contrast to the curve with respect to hydrogen release from the sample without additives, the incubation stage almost entirely disappeared with the addition of only 1 mol% 3TiCl3·AlCl3. In general, the duration of the dehydrogenation process is shortened from about 12 h without additives to about 4 h with additives. Furthermore, these curves seem to have different sigmoidal patterns, which might relate to the different MgB2 growths during the reaction between LiBH4 and Mg.  Figure 2a exhibits the morphology of the reactants LiBH4-MgH2 prior to dehydrogenation. The corresponding diffraction pattern shows the diffraction spots representing MgH2. The missing crystallographic information for LiBH4 in the diffraction pattern can presumably be attributed either to oxidation or to electron beam damage, as LiBH4 is both air-and electron-beam-sensitive [21,22]. This speculation can be to some extent confirmed by Figure 2b, which shows the same region as Figure 2a, and delivers the elemental distribution of magnesium and oxygen, as obtained by local EDX. The distribution of Mg directly represents the distribution of MgH2, since Mg only exists as MgH2 in this region. What's interesting is that the MgH2 grains are embedded in some of the oxygen-containing material. It is thus reasonable to assume that this material corresponds to oxidized LiBH4 present in the surrounding area.  Figure 2a exhibits the morphology of the reactants LiBH 4 -MgH 2 prior to dehydrogenation. The corresponding diffraction pattern shows the diffraction spots representing MgH 2 . The missing crystallographic information for LiBH 4 in the diffraction pattern can presumably be attributed either to oxidation or to electron beam damage, as LiBH 4 is both air-and electron-beam-sensitive [21,22]. This speculation can be to some extent confirmed by Figure 2b, which shows the same region as Figure 2a, and delivers the elemental distribution of magnesium and oxygen, as obtained by local EDX. The distribution of Mg directly represents the distribution of MgH 2 , since Mg only exists as MgH 2 in this region. What's interesting is that the MgH 2 grains are embedded in some of the oxygen-containing material. It is thus reasonable to assume that this material corresponds to oxidized LiBH 4 present in the surrounding area. Figure 2c exhibits the STEM-HAADF image of the material after dehydrogenation of LiBH 4 -MgH 2 . The diffraction pattern was taken from the corresponding 4D-STEM data stack. Accordingly, the crystals growing in the same direction are MgB 2 , and their growth direction is 1210 . This is in agreement with previous studies [19]. The uniform distribution of magnesium and boron recorded via STEM-EELS (Figure 2d) further proves the formation of MgB 2 with a bar-shaped morphology. Given the crystal structure and the rectangular bar-like morphology of MgB 2 , the other two directions that are vertical to Figure 2c exhibits the STEM-HAADF image of the material after dehydrogenation of LiBH4-MgH2. The diffraction pattern was taken from the corresponding 4D-STEM data stack. Accordingly, the crystals growing in the same direction are MgB2, and their growth direction is [12 ̅ 10]. This is in agreement with previous studies [19]. The uniform distribution of magnesium and boron recorded via STEM-EELS (Figure 2d) further proves the formation of MgB2 with a bar-shaped morphology. Given the crystal structure and the rectangular bar-like morphology of MgB2, the other two directions that are vertical to [12 ̅ 10] were determined to be [101 ̅ 0] and [0002], see Figure 2e. The disappearance of LiH may be due to electron beam damage [23].  For a more detailed comparison, we first focused on the extreme case where 10 mol% 3TiCl 3 ·AlCl 3 was added, and then continued with the results on samples with a lower additive content. Figure 3a exhibits the desorbed LiBH 4 -MgH 2 with 10 mol% 3TiCl 3 ·AlCl 3 , with the major crystal also being MgB 2 according to the diffraction pattern. Figure 3b displays the EDX map of Mg and Ti, showing the distribution of MgB 2 and Ti-containing materials. Similar to the previous observations without additives, there are parallel MgB 2 crystals oriented in the same direction, yet with a much smaller size. To determine the MgB 2 morphology in this case, STEM tomography analyses were carried out on the region shown in Figure 3a. Figure 3c displays the tomography images reconstructed from the EDX map of Mg at different angles. One exemplarily selected piece of MgB 2 in the yellow box marked in both Figure 3a,c was further studied; see Figure 3d. In contrast to the bar-like MgB 2 crystals displayed in Figure 2c,d, a regular hexagonal platelet appeared in the case of the 10 mol% additive. As illustrated in Figure 3e, given the hexagonal close-packed (hcp) crystal structure of MgB 2 , it can be immediately determined that this corresponds to the basal plane {0002} [24]. The six-fold symmetrical surface planes can also be indexed. The two possible candidates for these planes are the primary prism plane 1010 and the second prism plane 1210 , as shown in Figure 3e. According to recent studies, the bar-like MgB 2 crystals nucleate on Mg grains, whereas the nucleation of platelet-like MgB 2 occurs on TiB 2 nanoparticles [19]. Based on the edge-to-edge matching model [25][26][27], their respective orientation relationships and the corresponding misfits have been previously determined, which are summarized in Table 1.

Observation of MgB 2 Using TEM
For a more detailed comparison, we first focused on the extreme case where 10 mol% 3TiCl3·AlCl3 was added, and then continued with the results on samples with a lower additive content. Figure 3a exhibits the desorbed LiBH4-MgH2 with 10 mol% 3TiCl3·AlCl3, with the major crystal also being MgB2 according to the diffraction pattern. Figure 3b displays the EDX map of Mg and Ti, showing the distribution of MgB2 and Ti-containing materials. Similar to the previous observations without additives, there are parallel MgB2 crystals oriented in the same direction, yet with a much smaller size. To determine the MgB2 morphology in this case, STEM tomography analyses were carried out on the region shown in Figure 3a. Figure 3c displays the tomography images reconstructed from the EDX map of Mg at different angles. One exemplarily selected piece of MgB2 in the yellow box marked in both Figure 3a,c was further studied; see Figure 3d. In contrast to the barlike MgB2 crystals displayed in Figure 2c,d, a regular hexagonal platelet appeared in the case of the 10 mol% additive. As illustrated in Figure 3e, given the hexagonal close-packed (hcp) crystal structure of MgB2, it can be immediately determined that this corresponds to the basal plane {0002} [24]. The six-fold symmetrical surface planes can also be indexed. The two possible candidates for these planes are the primary prism plane {101 ̅ 0} and the second prism plane {12 ̅ 10}, as shown in Figure 3e. According to recent studies, the barlike MgB2 crystals nucleate on Mg grains, whereas the nucleation of platelet-like MgB2 occurs on TiB2 nanoparticles [19]. Based on the edge-to-edge matching model [25][26][27], their respective orientation relationships and the corresponding misfits have been previously determined, which are summarized in Table 1.  With this information, we can calculate the elastic strain energy densities hkil for any given lattice plane hkil with the equation hkil = 1 2 Y hkil δ 2 hkil , where Y hkil and δ hkil represent the Young's modulus and the atomic misfit in a certain direction, respectively [28]. The corresponding Young's modulus Y hkil for MgB 2 can be then calculated by substituting the Miller indices of the lattice plane hkil, the elastic constants C 11 = 365 GPa, C 12 = 98 GPa, C 13 = 65 GPa, C 33 = 203 GPa and C 44 = 58 GPa, and the lattice constants a = 3.0851 Å and c = 3.5201 Å into Equation (1) [29,30]. The compliances S ij in the equation can be transferred from the elastic constants C lk based on the given crystal structure [31].
Combining the calculated Young's modulus Y hkil for the selected orientations and the related misfit δ hkil from Table 1 gives the elastic strain energy densities hkil for MgB 2 ; see Table 2. According to the experimental observations and the determined elastic strain energy presumably induced during the formation of MgB 2 , the large energy density of more than 7 × 10 8 J/m 3 along the 0002 direction explains why both MgB 2 bars (on Mg) and MgB 2 platelets (on TiB 2 ) appeared to be rather thin in this direction. For Mg, the nucleation and growth of MgB 2 along the 1010 direction on Mg also led to a significant amount of strain energy density of up to 3·10 8 J/m 3 . The related high strain energy at the interface may account for the primary growth along the 1210 direction, which is perpendicular to the interface between MgB 2 and Mg, leading to the morphology of a rectangular bar. In contrast for TiB 2 , the nucleation of MgB 2 along either 1010 or 1210 on TiB 2 is equivalent from an energetic point of view, leading to a significantly smaller strain energy density of only 4.7·10 7 J/m 3 . This value is about an order of magnitude smaller than that of the corresponding growth on Mg, which explains the more isotropic morphology of the hexagonal MgB 2 platelets. MgB 2 bars are also distinct from MgB 2 platelets in terms of their aspect ratio, which is much larger than one, indicating that they grow predominantly in one direction.
Another interesting point is the observation of a parallel alignment of the MgB 2 crystals observed in both cases (with or without additives). We can understand where the parallel alignment of the MgB 2 bars has come from, as Mg grains are large enough to provide sufficient surface area for the nucleation of several MgB 2 bars on the same plane; see Figure 2b. Since the nucleation of MgB 2 follows a specific crystallographic orientation with respect to Mg, it is natural for the MgB 2 crystals nucleating on the same Mg plane to grow in the same direction. However, given the distance between two parallel MgB 2 platelets up to several hundred nanometers (Figure 3a), and the size of TiB 2 nanoparticles (Figure 3b), it is not likely for two MgB 2 platelets to nucleate and grow on the same TiB 2 nanoparticle. From this perspective, one assumption is that some TiB 2 nanoparticles may be attached to Mg grains in certain orientations with respect to Mg to minimize their interfacial energy during the dehydrogenation process. The nucleation of MgB 2 is then more likely to first occur on these attached nanoparticles. This is not only because the nucleation of MgB 2 based on TiB 2 requires less strain energy per unit volume, but also because the diffusion distance for Mg atoms is much shorter, as these TiB 2 nanoparticles are directly attached on the surface of Mg grains.
To understand the dependence of the MgB 2 morphology on the additive content, samples added with lower contents of 3TiCl 3 ·AlCl 3 were also studied. Figure 4a shows a STEM-HAADF image of the sample with 1 mol% additives after dehydrogenation. Again, the corresponding diffraction pattern confirms the existence of MgB 2 . By comparing this image with the corresponding elemental distribution of Mg, the parallel-oriented crystals of MgB 2 can be recognized in the agglomerate. As indicated in Figure 4a, two parallel MgB 2 crystals were selected for electron tomography measurements. It turns out that both pieces of MgB 2 have a bar-like morphology (Figure 4b). In some other regions, a MgB 2 morphology similar to Figure 3 can also be observed, which indicates the generation of MgB 2 platelets (Figure 4c). These observations indicate that the nucleation of MgB 2 on Mg has occurred. Besides, the nucleation of MgB 2 on TiB 2 was not as dominant as in the case of 10 mol% additives, where no more MgB 2 bars were observed. This also implies that there is a competition between the nucleation on Mg grains and TiB 2 nanoparticles for Mg and B atoms to generate MgB 2 bars or MgB 2 platelets.
Another interesting point is the observation of a parallel alignment of the MgB2 crystals observed in both cases (with or without additives). We can understand where the parallel alignment of the MgB2 bars has come from, as Mg grains are large enough to provide sufficient surface area for the nucleation of several MgB2 bars on the same plane; see Figure 2b. Since the nucleation of MgB2 follows a specific crystallographic orientation with respect to Mg, it is natural for the MgB2 crystals nucleating on the same Mg plane to grow in the same direction. However, given the distance between two parallel MgB2 platelets up to several hundred nanometers (Figure 3a), and the size of TiB2 nanoparticles ( Figure  3b), it is not likely for two MgB2 platelets to nucleate and grow on the same TiB2 nanoparticle. From this perspective, one assumption is that some TiB2 nanoparticles may be attached to Mg grains in certain orientations with respect to Mg to minimize their interfacial energy during the dehydrogenation process. The nucleation of MgB2 is then more likely to first occur on these attached nanoparticles. This is not only because the nucleation of MgB2 based on TiB2 requires less strain energy per unit volume, but also because the diffusion distance for Mg atoms is much shorter, as these TiB2 nanoparticles are directly attached on the surface of Mg grains.
To understand the dependence of the MgB2 morphology on the additive content, samples added with lower contents of 3TiCl3·AlCl3 were also studied. Figure 4a shows a STEM-HAADF image of the sample with 1 mol% additives after dehydrogenation. Again, the corresponding diffraction pattern confirms the existence of MgB2. By comparing this image with the corresponding elemental distribution of Mg, the parallel-oriented crystals of MgB2 can be recognized in the agglomerate. As indicated in Figure 4a, two parallel MgB2 crystals were selected for electron tomography measurements. It turns out that both pieces of MgB2 have a bar-like morphology (Figure 4b). In some other regions, a MgB2 morphology similar to Figure 3 can also be observed, which indicates the generation of MgB2 platelets (Figure 4c). These observations indicate that the nucleation of MgB2 on Mg has occurred. Besides, the nucleation of MgB2 on TiB2 was not as dominant as in the case of 10 mol% additives, where no more MgB2 bars were observed. This also implies that there is a competition between the nucleation on Mg grains and TiB2 nanoparticles for Mg and B atoms to generate MgB2 bars or MgB2 platelets. When the additive content was raised up to 2.5 mol%, we still observed MgB 2 bars in some areas (Figure 5a,b). After further increasing the additive content to 5 mol%, the majority of the observed MgB 2 crystals were platelet-like, as shown in Figure 5c,d. This can be further verified by the tomography analysis on the piece of MgB 2 highlighted in the yellow box in Figure 5c. These tomography images display a MgB 2 platelet at different angles. Based on these observations, it seems to imply a turning point in the competition between the nucleation on Mg and TiB 2 . When the additive content was raised up to 2.5 mol%, we still observed MgB2 bars in some areas (Figure 5a,b). After further increasing the additive content to 5 mol%, the majority of the observed MgB2 crystals were platelet-like, as shown in Figure 5c,d. This can be further verified by the tomography analysis on the piece of MgB2 highlighted in the yellow box in Figure 5c. These tomography images display a MgB2 platelet at different angles. Based on these observations, it seems to imply a turning point in the competition between the nucleation on Mg and TiB2.

Kinetic Modeling
According to the TEM results, when varying contents of 3TiCl 3 ·AlCl 3 are added to LiBH 4 -MgH 2 , the dehydrogenation processes of LiBH 4 -MgH 2 differ. This difference is not only reflected in the change in MgB 2 morphology, but also in the change in the hydrogen release rate. Regarding the additive-promoted dehydrogenation process, the shortened incubation stage that relates to the accelerated nucleation of MgB 2 plays an essential role, as shown in Figure 1c. In addition, the subsequent step of MgB 2 growth that accompanies a massive amount of hydrogen release is also crucial to the improvement of the dehydrogenation kinetics of LiBH 4 -MgH 2 . This aspect is discussed in the following. Figure 6a is extracted from Figure 1c and shows the hydrogen release over time for the respective samples, during the second step of the reaction, ranging from the start of the massive hydrogen release to the quasi-end of the dehydrogenation, when each curve is nearly flat. The process of counting starts with −3.2 wt% of hydrogen release to ensure the self-decomposition of MgH 2 is completed and all the released hydrogen in this case is coming from the reaction between LiBH 4 and Mg. To keep the analysis consistent, the data with hydrogen re lease α ranging from 15% to 85% were selected for the plot of ln[− ln(1 − α)] vs. ln(t); see Figure 6c. The slope is determined by linear regression with the corresponding R-square value evaluating the quality of the fit for the respective samples, as shown in Figure 6d All the fitting parameters can be found in Table S1. As can be seen, with the increase in additive content, the Avrami exponent decreases. Based on TEM observations of bar-like and platelet-like MgB2 crystals, the MgB growth was determined to be mainly two-dimensional, as the morphology extension in the third dimension along the c axis is almost negligible. The value of dimensionality d i therefore located between two and three. From this perspective, m = 1 and m = 2 can be assigned to the two most extreme cases: without additives (n = 2.32) and with 10 mol% additives (n = 1.25), since only one MgB2 morphology exists for either case. Based on these values, the growth rate-controlling steps were determined to be mainly interface controlled or mainly diffusion-controlled, respectively. The change in the growth rate controlling step is also in agreement with the discussed decrease in the elastic strain en ergy density at the interface between MgB2 and TiB2 compared with that between MgB and Mg during the formation of MgB2. For the samples with a lower additive content  Figure 6b exhibits the min-max normalized hydrogen release between zero and one for each sample. Notably, it takes more than twice the time to release about 80% of hydrogen from the sample without additives as it does with additives. This means that both the nucleation and the growth of MgB 2 platelets are faster than those of MgB 2 bars. It is thus intuitive to consider a positive correlation between the hydrogen release rate and the amount of 3TiCl 3 ·AlCl 3 . However, after approximately 50% of hydrogen was released, the hydrogen release rate of the sample with 10 mol% additives was smaller than that of the samples with smaller amounts of additives. One reasonable interpretation is that the mutual impingement between the generated phases becomes even more furious, as more nucleation centers of TiB 2 exist in the surrounding environment, which inversely decelerates the dehydrogenation process. This might also be an indication of an oversaturation of the sample with additives. Nonetheless, it is still important to study extreme cases so that the physical mechanisms behind the growth of MgB 2 can be distinguished and clarified.
The Johnson-Mehl-Avrami-Kolmogorov (JMAK) equation generally describes the kinetics of phase transformation for solids. To obtain an in-depth understanding of the dehydrogenation kinetics and determine the rate-controlling step for the MgB 2 growth for the respective samples, this equation was applied to model the process of the MgB 2 growth; see Equation (2) [32,33]: where α denotes the fraction of released hydrogen over time (representing the volume fraction of the directly correlated MgB 2 phase), k denotes the reaction rate constant that depends on temperature and n refers to the Avrami exponent. The numerical value of the Avrami exponent n can be regarded as an indicator of the growth dimensionality for MgB 2 crystals and the related rate-controlling steps [32,34]. In general, n is equal to d/m, where d represents the dimensionality of crystal growth with the conditions that 1 ≤ d ≤ 2 refers to one-dimensional growth (e.g., needle), 2 ≤ d ≤ 3 refers to twodimensional growth (e.g., platelet and sheet), and 3 ≤ d ≤ 4 refers to three-dimensional growth (e.g., sphere). The value of m indicates the rate-controlling step for the phase transformation, with m = 1 referring to the interface-controlled growth, and m = 2 representing the diffusion-controlled growth. To determine n for each sample, we can rewrite Equation (2) as: From the equation, n can be immediately determined from the slope by plotting ln[− ln(1 − α)] against ln(t). To keep the analysis consistent, the data with hydrogen release α ranging from 15% to 85% were selected for the plot of ln[− ln(1 − α)] vs. ln(t); see Figure 6c. The slope is determined by linear regression with the corresponding R-square value evaluating the quality of the fit for the respective samples, as shown in Figure 6d. All the fitting parameters can be found in Table S1. As can be seen, with the increase in additive content, the Avrami exponent decreases.
Based on TEM observations of bar-like and platelet-like MgB 2 crystals, the MgB 2 growth was determined to be mainly two-dimensional, as the morphology extension in the third dimension along the c axis is almost negligible. The value of dimensionality d is therefore located between two and three. From this perspective, m = 1 and m = 2 can be assigned to the two most extreme cases: without additives (n = 2.32) and with 10 mol% additives (n = 1.25), since only one MgB 2 morphology exists for either case. Based on these values, the growth rate-controlling steps were determined to be mainly interface-controlled or mainly diffusion-controlled, respectively. The change in the growth rate-controlling step is also in agreement with the discussed decrease in the elastic strain energy density at the interface between MgB 2 and TiB 2 compared with that between MgB 2 and Mg during the formation of MgB 2 . For the samples with a lower additive content, where both MgB 2 bars and platelets were observed, it can therefore be expected that the interface-controlled and the diffusion-controlled growth may affect the dehydrogenation process simultaneously. Their Avrami exponents are thus more likely to reflect the simultaneous contribution from both growth mechanisms with varying weights for each case. This can be additionally confirmed by the decrease in their R-square values, indicating a worsened fit. When increasing the additive content up to 10 mol%, the R-square value improved again. This can be explained by the dominance of the diffusion-controlled growth, which is in agreement with the fact that only MgB 2 platelets have been observed in this case.
Based on these results, the best additive content that accelerates the dehydrogenation process the most can be determined. It may lay between 2.5 mol% and 5 mol% additive content. According to the interpretation, this is because the growth rate-controlling step changes from interface-controlled for 2.5 mol% additives to diffusion-controlled for 5 mol% additives (Avrami exponent n < 1.5). In this range, the formation of MgB 2 platelets becomes gradually dominant. The transition in the growth rate-controlling step is also consistent with the TEM observations, where much fewer MgB 2 bars were found when the additive content was increased to 5 mol%.
To further validate the results of the JMAK equation, several other frequently utilized models to study the kinetics of hydrogen storage materials were utilized for comparison; see Table 3. The fit quality of these models was evaluated by the reduced time method, by plotting the theoretical reduced time (t/t 0.5 ) theoretical against the experimental reduced time (t/t 0.5 ) experimental , where t 0.5 denotes the time when 50% of hydrogen has been released [35,36]. Figure 7a-e shows the plots of (t/t 0.5 ) theoretical vs. (t/t 0.5 ) experimental of all these kinetic models for each sample using the data with the hydrogen release α ranging from 15% to 85%. The fitting parameters can be found in Tables S2-S6. The optimal fitting has a slope equal to one and an intercept equal to zero, so a straight line of y = x is drawn on each figure for reference. In Figure 7f, the respective intercept and slope values for the JMAK model are summarized. In general, the JMAK model performs best in fitting the data to describe the second reaction step of dehydrogenation process of LiBH 4 -MgH 2 over time. Table 3. The introduction to the applied kinetic models [32,34,37,38].

Model k*t=
Two-dimensional growth of contracting volume (2D CV) Three-dimensional growth of contracting volume (3D CV)

Material Preparations
The reactants were purchased in the form of powder: MgH2 (95% purity, Rockwood

Kinetics Measurements
The volumetric measurements of hydrogen release were performed using a custombuilt Sievert's-type apparatus. The milled sample (~170 mg) was charged into the stainless steel sample holder of the measuring apparatus. The samples were annealed from room temperature to 400 • C at a heating rate of 10 • C min −1 under a hydrogen atmosphere of 4 bar. After reaching the target temperature of 400 • C, the materials were kept under isothermal conditions for several hours.

XRD Experiments
The ex situ XRD experiments were performed using a Bruker D8 Discover diffractometer equipped with a Cu X-ray source (λ = 1.54184 Å) and a 2D VANTEC detector. The operating voltage and current were 50 kV and 1000 mA, respectively. The diffraction patterns were acquired in the 2θ range from 10 • to 90 • with a step size of 0.005 • , ∆2θ = 10 • and the exposure time for each step of 400 s. To prevent oxidation of the materials, the specimens were packed onto a sample holder and sealed with an airtight dome made of poly (methyl methacrylate) (PMMA), which was carried out in an argon-filled glovebox (O 2 , H 2 O < 0.5 ppm).

TEM Experiments
TEM experiments were performed using a Themis-Z 60-300 (Thermo Fisher Scientific Inc., Waltham, MA, USA) equipped with a monochromator and double aberration correctors (probe and image Cs correctors) operated at 300 kV. TEM sample preparation was carried out under an argon atmosphere in a glovebox (O 2 , H 2 O < 0.5 ppm). Sample powders were dispersed in toluene and ultra-sonicated for 1 min before being distributed on lacey carbon-coated gold TEM grids with the item number S166-A3-V (Plano GmbH). Subsequently, they were transferred under argon from the glovebox into the microscope with a vacuum transfer holder (model number 648, AMETEK Gatan Inc.).
The utilized beam current ranged from 50 to 150 pA for TEM experiments. Selected area electron diffraction (SAED) patterns were collected using an OneView camera (AME-TEK Gatan Inc.). Scanning TEM (STEM) images were recorded via a high-angle annular dark-field (HAADF) detector with a convergence angle of 21.5 mrad and a camera length of 93 mm. Energy-dispersive X-ray spectroscopy (EDX) spectrum-imaging (SI) was executed with a Super-X windowless EDX detector (Thermo Fisher Scientific Inc.) using the same parameters for STEM imaging.
Electron tomography was carried out using a Fischione 2020 tomography holder in STEM mode with a convergence angle of 8.58 mrad. Other parameters stayed the same as mentioned above. HAADF-STEM tilt series with image dimensions of 1024 × 1024 pixels were collected using Xplore3D (Thermo Fisher Scientific Inc.) over a tilt range with an increment of 2 • from −70 • to 70 • for the desorbed 2LiBH 4 -MgH 2 with 1 mol% 3TiCl 3 ·AlCl 3 , from −74 • to 76 • for the desorbed 2LiBH 4 -MgH 2 with 5 mol% 3TiCl 3 ·AlCl 3 and from −66 • to 72 • for the desorbed 2LiBH 4 -MgH 2 with 10 mol% 3TiCl 3 ·AlCl 3 . Electron tomography combined with EDX mapping was also carried out on the desorbed 2LiBH 4 -MgH 2 with 10 mol% 3TiCl 3 ·AlCl 3 with a tilt range from −66 • to 69 • with 3 • increments. Each EDX map of Mg was constructed from 900 frames with image dimensions of 256 × 256 pixels and a dwell time of 12 µs. The alignment of the tilt series was performed in IMOD using the cross-correlation method. The aligned tilt series were then reconstructed using the algorithm of simultaneous iterative reconstruction technique (SIRT) with 100 iterations in Inspect3D (Thermo Fisher Scientific Inc.) [39]. The 3D visualization was realized using Avizo 2020.2 (Thermo Fisher Scientific Inc.).
4D-STEM measurements were carried out in µ-probe mode using the OneView camera, with a convergence angle of 0.47 mrad, a camera length of 720 mm, an acquisition time of 13.2 ms for each pixel, and a dose of~4.7 × 10 7 e nm −2 . STEM electron energy-loss spectroscopy (EELS) SI was acquired using a Continuum 970 HighRes imaging filter (GIF) (AMETEK Gatan Inc.) in dual-EELS mode with 13.64 ms acquisition time for each low-loss spectrum, 136.4 ms for each high-loss spectrum, 21.5 mrad convergence angle, 40 mrad collection angle, and 0.15 eV per channel energy dispersion. The EELS SI data were denoised by principal component analysis (PCA), which effectively reduces the random noise generated during signal recording [40,41].

Summary
The second dehydrogenation step of the reactive hydride composite LiBH 4 -MgH 2 is controlled by the nucleation and growth of MgB 2 in respective orientations. The observed different MgB 2 morphologies can be directly correlated to the required elastic strain energy per unit volume. The nucleation of MgB 2 on Mg requires an energy up to 2.9 × 10 8 J/m 3 in the 1010 direction, whereas it needs only 4.7·10 7 J/m 3 for the nucleation of MgB 2 on TiB 2 in the 1010 or 1210 directions. The formation of MgB 2 may occur primarily on those TiB 2 nanoparticles that adhere to Mg grains, which leads to the generation of parallel MgB 2 platelets. According to the JMAK equation parameter interpretation, the growth ratecontrolling steps for MgB 2 bars or platelets are interface-controlled or diffusion-controlled, respectively. The change in the growth mechanism is consistent with the decreased elastic strain energy density determined for the nucleation of MgB 2 on TiB 2 and the change in the morphology of MgB 2 when additives were included. Based on the second dehydrogenation growth kinetics, the best additive content that accelerates the dehydrogenation process of LiBH 4 -MgH 2 the most is between 2.5 mol% and 5 mol%. However, given the consumption of LiBH 4 by additives, the trade-off between a reduced hydrogen storage capacity and improved kinetics also needs to be carefully considered in practice.