Hydrogen Sorption in Erbium Borohydride Composite Mixtures with LiBH 4 and/or LiH

: Rare earth (RE) metal borohydrides have recently been receiving attention as possible hydrogen storage materials and solid-state Li-ion conductors. In this paper, the decomposition and reabsorption of Er(BH 4 ) 3 in composite mixtures with LiBH 4 and/or LiH were investigated. The composite of 3LiBH 4 + Er(BH 4 ) 3 + 3LiH has a theoretical hydrogen storage capacity of 9 wt %, nevertheless, only 6 wt % hydrogen are accessible due to the formation of thermally stable LiH. Hydrogen sorption measurements in a Sieverts-type apparatus revealed that during three desorption-absorption cycles of 3LiBH 4 + Er(BH 4 ) 3 + 3LiH, the composite desorbed 4.2, 3.7 and 3.5 wt % H for the ﬁrst, second and third cycle, respectively, and thus showed good rehydrogenation behavior. In situ synchrotron radiation powder X-ray diffraction (SR-PXD) after ball milling of Er(BH 4 ) 3 + 6LiH resulted in the formation of LiBH 4 , revealing that metathesis reactions occurred during milling in these systems. Impedance spectroscopy of absorbed Er(BH 4 ) 3 + 6LiH showed an exceptional high hysteresis of 40–60 K for the transition between the high and low temperature phases of LiBH 4 , indicating that the high temperature phase of LiBH 4 is stabilized in the composite. Debye-Scherrer using θ = 10 ◦ –70 ◦

Gennari et al. investigated the composite mixtures of 6LiBH 4 and RECl 3 with RE = Ce, Gd and found a decrease in decomposition temperature of LiBH 4 due to an in situ formed REH 2 phase [17].Additionally, the decomposed composite showed the ability to reabsorb 20% of the initial hydrogen content.Frommen et al. investigated the composite mixtures of LiBH 4 , LiH and RE(BH 4 ) 3 with RE = La, Er [27].It was reported that the desorption temperature of LiBH 4 in the mixture with RE = Er decreased by about 100 • C compared to pure LiBH 4 .Reabsorption at 340 • C under 100 bar H 2 yielded a hydrogen uptake of 66% after desorption against vacuum, and 80% after desorption against 5 bar H 2 , respectively, of the initially release of 3 wt % H 2 .
The established technique to synthesize RE borohydrides is the solid state metathesis approach, as it has the advantage of direct synthesis of a RE borohydride without complicated in vacuo manipulations or possible decomposition [28,29].The RE chloride and an alkali metal borohydride (i.e., Li, Na, K) are mixed in a one-step mechanochemical synthesis to form the RE borohydride and the alkali metal chloride via a metathesis reaction.LiBH 4 has proven to be an efficient precursor, but also other metal borohydrides e.g., NaBH 4 , have been successfully used [30][31][32].As an alternative method, the solvent-based synthesis of borohydrides has been used for over 50 years with the advantage of removing the alkali metal chloride (i.e., LiCl) [28], which is necessary to increase the hydrogen capacity.The halide-free synthesis in this paper is based on a mechanochemical synthesis [33] and wet-chemical extraction, as described by Hagemann et al. [34,35].
This work presents two new composite materials: Er(BH 4 ) 3 + 6LiH and 3LiBH 4 + Er(BH 4 ) 3 + 3LiH.The decomposition and reabsorption of hydrogen have been studied in detail by in situ synchrotron radiation powder X-ray diffraction (SR-PXD).Desorption-absorption cycling measurements were conducted in a Sieverts-type apparatus.The rehydrogenation properties of the 3LiBH 4 + Er(BH 4 ) 3 + 3LiH system were investigated during three desorption-absorption cycles.One cycle was studied for the Er(BH 4 ) 3 + 6LiH system.The latter was also investigated by impedance spectroscopy.Finally, thermal gravimetric (TG) and differential scanning calorimetry (DSC), as well as mass spectrometry (MS) were conducted.
The small amount of Er(BH 4 ) 3 present (Table 1) indicates that Equation (1) has almost gone to completion during milling.The in situ SR-PXD data show the phase transition of o-to hexagonal-LiBH 4 (h-LiBH 4 ) at 100 • C. The Bragg peaks of Er(BH 4 ) 3 start to decrease simultaneously and are gone at 240 • C. ErH 2 is present in significant amounts (44.1(4) wt %) directly after ball milling.Its lattice parameters increase more than expected from the thermal expansion from 186 • C, indicating that it is further hydrogenated, induced by the hydrogen pressure in the in situ SR-PXD measurements setup, with a likely intermediate state of ErH 2+δ with (0 ≤ δ ≤ 1) as shown in Figure S1.Bragg peaks from LiBH 4 disappear at 270 • C, caused by its melting.Above 280 • C, the sample oxidizes, as evident from strong Bragg peaks of Er 2 O 3 (not shown), and further analysis of the decomposition process was therefore not possible.
Figure 2a shows the differential scanning calorimetry (DSC) data of S1.The phase transition from o-to h-LiBH 4 occurs at 105 • C and its melting at 260 • C (peak temperatures), which is in good agreement with the SR-PXD data.A rather minor exothermic DSC event (dashed circle in Figure 2a) can be seen in analogy to the increase in the lattice parameter observed in SR-PXD starting at 186 • C, corresponding to the reaction from ErH 2 towards ErH 3 .The decomposition of the Li borohydride appears at 326 • C.There is a minor endothermic peak after the major endothermic event, which can be assigned to desorption of ErH 3 to ErH 2 (a similar decomposition temperature is observed for pure ErH 3 , as shown in Figure S2).Thermogravimetric (TG) data for S1 are also presented in Figure 2 and show mass loss of 3.4 wt % between RT and 400 • C, including a 30-min isotherm.
The powder X-ray diffraction (PXD) pattern of one complete desorption absorption cycle in the Sieverts apparatus is shown in Figure 2b.The main product after desorption at 400   ErH3, as shown in Figure S2).Thermogravimetric (TG) data for S1 are also presented in Figure 2 and show mass loss of 3.4 wt % between RT and 400 °C, including a 30-min isotherm.
The powder X-ray diffraction (PXD) pattern of one complete desorption absorption cycle in the Sieverts apparatus is shown in Figure 2b.The main product after desorption at 400 °C and 5-10 bar H2 is ErH3, formed by additional hydrogen absorption by ErH2 upon cooling.Minor Bragg peaks of ErB4 are also observed.Reabsorption at 340 °C under 100 bar H2 for 12 h resulted in the formation of LiBH4, as well as increased intensity of ErH3 peaks.The thermogravimetric and differential scanning calorimetry (TG-DSC) data for the decomposition of reabsorbed S1 is shown in Figure 2a (red curves).TG data show a mass loss of 3.0 wt %, corresponding to 88% of the initial released hydrogen.An increase in onset temperature of hydrogen release can be observed (∆T = 70 °C).In the DSC data, two peaks occur at rather low temperatures.The first peak is at 90 °C, but the process behind it is not completely understood (see Li-ion conductivity section).The second one at 105 °C corresponds to the o-to h-LiBH4 transition.The melting of LiBH4 occurs at 278 °C.The last endothermic event coincides with the mass loss and occurs at 390 °C, which is probably a superposition of two events.The first is proposed to be the decomposition of LiBH4, while the second is possibly the reduction of ErH3 to ErH2, see Figure S2.
In general, the desorption-absorption cycle shown in this section follows the reaction pathways outlined in Equations ( 2) and (3).4LiBH4(s) + ErH2(s) ErB4(s) + 4LiH(s) + 7.5H2(g) The thermogravimetric and differential scanning calorimetry (TG-DSC) data for the decomposition of reabsorbed S1 is shown in Figure 2a (red curves).TG data show a mass loss of 3.0 wt %, corresponding to 88% of the initial released hydrogen.An increase in onset temperature of hydrogen release can be observed (∆T = 70 • C).In the DSC data, two peaks occur at rather low temperatures.The first peak is at 90 • C, but the process behind it is not completely understood (see Li-ion conductivity section).The second one at 105 • C corresponds to the oto h-LiBH 4 transition.The melting of LiBH 4 occurs at 278 • C. The last endothermic event coincides with the mass loss and occurs at 390 • C, which is probably a superposition of two events.The first is proposed to be the decomposition of LiBH 4 , while the second is possibly the reduction of ErH 3 to ErH 2 , see Figure S2.
In general, the desorption-absorption cycle shown in this section follows the reaction pathways outlined in Equations ( 2) and (3).

3LiBH 4 + Er(BH 4 ) 3 + 3LiH (S2)
Figure 3a shows the ex situ lab PXD patterns of 3LiBH 4 + Er(BH 4 ) 3 + 3LiH (S2) in the as-milled, desorbed, and reabsorbed states.The Bragg peaks of the reactants LiBH 4 , Er(BH 4 ) 3 and LiH, are observed after ball milling.In addition, ErH 2 is formed in minor amounts, showing that a reaction has already started during the milling process.The products after desorption at 400 • C and 5-10 bar H 2 are mainly ErB 4 and several Er-hydrides.Cubic (c-) ErH 3 (Fm-3m) is the main phase, but small amounts of ErH 2 , ErH and trigonal (t-) ErH 3 (P-3c1) are also present.As suggested above, ErH 3 appeared during cooling under hydrogen pressure.Minor Bragg peaks corresponding to ErB 12 are detected as well.Reabsorption at 340 • C under 100 bar H 2 for 12 h led to the complete consumption of the Er borides.Trivalent Er-hydride was found after reabsorption in both the cubic and trigonal modifications.
Interestingly, the Bragg peaks of ErB 4 show less intensity in the ex situ PXD pattern of desorbed S1 in Figure 2b than in desorbed S2, although the conditions were the same.This observation let us to a hypothesis that the excess of LiBH 4 somehow increases the crystallinity of the material.This is supported by the observations in another recently published study, where the decomposition of pure Er(BH 4 ) 3 resulted in a completely amorphous material [36].The crystallinity was improved by addition of LiH and by reabsorption.The products were ErH 3 and LiBH 4 , and the final material was crystalline.Whether this observation is caused by amorphous Er-B-H species, which simply decrease when LiBH 4 is formed, or if the process is triggered directly by LiBH 4 /ErH 3 formation, cannot be concluded.
TG-DSC data of S2, stored for three months in a glove box, is presented in Figure 3b.TG data show a mass loss of 3.9 wt % up to 400 • C, and DSC data show three endothermic events, two sharp and one broad.The two sharp events are the phase transition and melting of LiBH 4 occurring at 111 and 271 • C, respectively.These events are not related to any mass loss in the TG data.The third endothermic peak is broad, and therefore probably a superposition of the decomposition of Er(BH 4 ) 3 and LiBH 4 with peak temperatures at 336 and 391 • C, respectively.Figure 3bii shows H 2 (m/z = 2) desorption between 275 and 400 • C. Minor amounts of B 2 H 6 (m/z = 26) are also seen in Figure 3bi, as well as S(CH 3 ) 2 (m/z = 62) between 75 and 125 • C.
Figure 4a shows in situ SR-PXD desorption data of S2, which was stored for nine months in a glove box.The RT data show significant Bragg peaks from ErH 2 .Compared to the rather minor peaks of ErH 2 directly after ball milling in Figure 3a, it becomes clear that ErH 2 is formed during the storage in Ar.This observation suggests that the reaction in Equation ( 1) even occurs at ambient temperature for S2.Equation ( 1) is an intermediate reaction, showing the formation of LiBH 4 , initiated during the milling of Er(BH 4 ) 3 and LiH, as well as the formation of ErH 2 .Bragg peaks of the other phases appear for the as-milled S2 (Figure 3a) at RT.The atomic positions for Er in Er(BH 4 ) 3 were refined (Figure S3 and Table S1) and are within the standard deviations of those reported for Y in Y(BH 4 ) 3 [5].The in situ SR-PXD desorption starts with a heating ramp from RT to 400 • C with a heating rate of 5 • C•min −1 and a 30-min isotherm at 400 • C. The gap in the data is due to a lost connection to the detector for 15 min in the temperature interval of 30-80 • C. The phase transition of LiBH 4 occurs at 100 • C. The Bragg peaks of Er(BH 4 ) 3 decrease and are gone at 195 • C, which is 55 • C lower compared to the in situ SR-PXD desorption data of S1.In any case, these observations are in agreement with earlier reports, where the halide free yttrium borohydride turns into "an X-ray amorphous solid" without melting at ~200 • C [35].Therefore, we assume an amorphisation, as all Bragg peaks of Er(BH 4 ) 3 disappear and the DSC data indicate no melting.The amorphisation temperatures in S1 and S2 are, however, 50-100 • C lower than the reported amorphisation temperature of pure Er(BH 4 ) 3 [36].The melting of LiBH 4 can be observed by the in situ SR-PXD data at 295 • C. The hydrogenation of the bivalent to trivalent erbium hydride occurs simultaneously, which is evident from a shift to lower 2θ angles for the Er hydride peaks (Figure 4a and Figure S1) and in good agreement with the in situ SR-PXD desorption data of S1.The Bragg peaks of ErH 3 increase in intensity until the heating is stopped after 30 min at 400 • C, concluding that the timeframe of the in situ SR-PXD desorption measurement for S2 does not allow for the formation of ErB 4 , which is the suggested desorption product, see Equation (2).Three desorption-absorption cycles were measured for S2 in a Sieverts-type apparatus.The volumetric desorption data after 5 h at 400 • C and 5-10 bar H 2 is illustrated in Figure 5.The first, second and third desorption show a hydrogen release of 4.2 wt %, 3.7 wt % and 3.5 wt %, respectively.During the first desorption of S2, Er(BH 4 ) 3 is decomposed to Er-hydride according to the ex situ PXD data, which resulted in a slower hydrogen release at the beginning of the first desorption (black curve in Figure 5).As Er(BH 4 ) 3 is not formed on rehydrogenation, this step is not repeated after the first desorption, thus resulting in a faster desorption for the second/third reabsorption.Three desorption-absorption cycles were measured for S2 in a Sieverts-type apparatus.The volumetric desorption data after 5 h at 400 °C and 5-10 bar H2 is illustrated in Figure 5.The first, second and third desorption show a hydrogen release of 4.2 wt %, 3.7 wt % and 3.5 wt %, respectively.During the first desorption of S2, Er(BH4)3 is decomposed to Er-hydride according to the ex situ PXD data, which resulted in a slower hydrogen release at the beginning of the first desorption (black curve in Figure 5).As Er(BH4)3 is not formed on rehydrogenation, this step is not repeated after the first desorption, thus resulting in a faster desorption for the second/third reabsorption.The TG-DSC data of S2 after the first reabsorption is shown in Figure 6.The DSC signal is similar to the ball milled sample in Figure 3b.The phase transition of LiBH4 shifts to a slightly lower temperature, 103 °C (111 °C for ball milled S2).The melting of LiBH4 still occurs at 270 °C.The third minor endothermic peak at 322 °C in Figure 6 cannot be explained.Although the set temperature for the isotherm is 400 °C, the fourth peak was observed at a peak temperature of 403 °C (due to a slight temperature overshoot) in Figure 6.It is suggested to be the decomposition of LiBH4, and hence slightly higher than for S2 after ball milling.The mass loss starts simultaneously with the fourth endothermic DSC event, also suggesting the decomposition of LiBH4.The fifth peak in the isothermal regime after 7 min at 400 °C is suggested to be the reduction of ErH3 to ErH2 as the mass  6 cannot be explained.Although the set temperature for the isotherm is 400 • C, the fourth peak was observed at a peak temperature of 403 • C (due to a slight temperature overshoot) in Figure 6.It is suggested to be the decomposition of LiBH 4 , and hence slightly higher than for S2 after ball milling.The mass loss starts simultaneously with the fourth endothermic DSC event, also suggesting the decomposition of LiBH 4 .The fifth peak in the isothermal regime after 7 min at 400 • C is suggested to be the reduction of ErH 3 to ErH 2 as the mass loss continues through this event (Figure S2).The TG data in Figure 6 show a mass loss of 4.5 wt % for reabsorbed S2.
Although it was reported that hydrogen pressure has a beneficial effect on desorption behavior, the TG and Sieverts data for the reabsorbed S2 show a weight loss of 4.5 and 3.7 wt %, respectively.The total hydrogen content of S2 is 9.0 wt %, but only 6.0 wt % can be obtained when forming LiH, meaning that our TG data shows a hydrogen uptake of 75%, compared to the nominal maximum value.loss continues through this event (Figure S2).The TG data in Figure 6 show a mass loss of 4.5 wt % for reabsorbed S2.Although it was reported that hydrogen pressure has a beneficial effect on desorption behavior, the TG and Sieverts data for the reabsorbed S2 show a weight loss of 4.5 and 3.7 wt %, respectively.The total hydrogen content of S2 is 9.0 wt %, but only 6.0 wt % can be obtained when forming LiH, meaning that our TG data shows a hydrogen uptake of 75%, compared to the nominal maximum value.To summarize, we observe from the PXD and SR-PXD data an increasing amount of ErB4 during thermal decomposition and an increasing amount of ErH2+δ during reabsorption for S2.This is in good agreement with the following reaction pathway, with an assumed start of decomposition similar to reaction products in Equation (1) The theoretical storage capacity of Equation (4) of 5.1 wt % H was not achieved, which is possibly caused by the uncompleted decomposition reactions, as residuals of ErH3 have been found in decomposed samples.
The comparison of S2 to the recently investigated 6LiBH4 + ErCl3 + 3LiH composite by Frommen et al. [27] shows that the two systems behave very similar with respect to rehydrogenation.During the first reabsorption, 80% to 85% of the original hydrogen content could be reabsorbed for both systems at similar conditions.The weight loss for our system shows 4.5 wt % after one desorption-absorption cycle while 2.4 wt % was reported for the other composite [27], meaning that the absorption in our composite has almost double the hydrogen capacity after the first cycle.
There are some differences in the hydrogenation products.In our halide-free S2 composite, c-ErH3 and t-ErH3 are both formed during rehydrogenation with 27.4(2) wt % and 14.4(1) wt %, respectively.LiBH4 is the major reabsorption product with refined 58.1(2) wt %, suggesting, as reported recently, that LiH plays a crucial role in the formation of new LiBH4 at the employed conditions [36].In Reference [27] only c-ErH3 was observed after rehydrogenation, but no t-ErH3, similar to our absorbed S1 sample.

Li-Ion Conductivity
The Li-ion conductivity of absorbed S1 was measured with alternating current (AC) impedance spectroscopy from RT to 140 °C (Figure 7).Further details to the experimental setup are given in Reference [37] and the references therein.The measurement was motivated by the hypothesis that the high temperature phase of LiBH4 was stabilized in the composite, thus shifting the phase To summarize, we observe from the PXD and SR-PXD data an increasing amount of ErB 4 during thermal decomposition and an increasing amount of ErH 2+δ during reabsorption for S2.This is in good agreement with the following reaction pathway, with an assumed start of decomposition similar to reaction products in Equation ( 1) The theoretical storage capacity of Equation (4) of 5.1 wt % H was not achieved, which is possibly caused by the uncompleted decomposition reactions, as residuals of ErH 3 have been found in decomposed samples.
The comparison of S2 to the recently investigated 6LiBH 4 + ErCl 3 + 3LiH composite by Frommen et al. [27] shows that the two systems behave very similar with respect to rehydrogenation.During the first reabsorption, 80% to 85% of the original hydrogen content could be reabsorbed for both systems at similar conditions.The weight loss for our system shows 4.5 wt % after one desorption-absorption cycle while 2.4 wt % was reported for the other composite [27], meaning that the absorption in our composite has almost double the hydrogen capacity after the first cycle.
There are some differences in the hydrogenation products.In our halide-free S2 composite, c-ErH 3 and t-ErH 3 are both formed during rehydrogenation with 27.4(2) wt % and 14.4(1) wt %, respectively.LiBH 4 is the major reabsorption product with refined 58.1(2) wt %, suggesting, as reported recently, that LiH plays a crucial role in the formation of new LiBH 4 at the employed conditions [36].In Reference [27] only c-ErH 3 was observed after rehydrogenation, but no t-ErH 3 , similar to our absorbed S1 sample.

Li-Ion Conductivity
The Li-ion conductivity of absorbed S1 was measured with alternating current (AC) impedance spectroscopy from RT to 140 • C (Figure 7).Further details to the experimental setup are given in Reference [37] and the references therein.The measurement was motivated by the hypothesis that the high temperature phase of LiBH 4 was stabilized in the composite, thus shifting the phase transition to lower temperatures.Evidently, the absorbed S1 showed an endothermic DSC event as low as 90 • C, which we first considered to be the phase transition in LiBH 4 (see DSC data, red curve in Figure 2a).
Figure 7 shows the Li + ionic conductivity for absorbed S1, and the values are systematically lower than for pure LiBH 4 as expected, since LiBH 4 only makes up 21 wt % in the present sample (The low conductivity values for the samples can be explained by minor amounts of LiBH 4 after rehydrogenation.For absorbed S1 the content is approx.2.5 molar equivalent of LiBH 4 corresponding to 21 wt %.This value was calculated from the mass loss shown by TG data of absorbed S1).However, upon heating above 110 • C, the conductivities reach about four orders of magnitude higher than the RT modification, which is in good agreement with the conductivity enhancement in pure LiBH 4 [38].The conductivity in absorbed S1 (Figure 7) shows a slightly steeper slope than pure LiBH 4 [38], which manifested in the absence of the "conductivity jump" in our data.We first suggested that this observation was caused by the initial ball milling of S1.In a recent publication, it was shown that ball milled LiBH 4 -LiI samples have a higher conductivity before annealing caused by the formation of a "defect rich microstructure," which influences the conductivity positively [39].However, the sample chosen for impedance spectroscopy was heated to 400 • C during the first decomposition and then reabsorbed at 340 • C.This heat treatment should heal all defects and disregard their influence [39].In summary, the high temperature phase transition is observed at 110 • C, which is in agreement with pure LiBH 4 [38,40].From the Li-ion conductivity measurements, the DSC event at 90 • C for absorbed S1 cannot be explained by the high temperature phase transition of LiBH 4 .
transition to lower temperatures.Evidently, the absorbed S1 showed an endothermic DSC event as low as 90 °C, which we first considered to be the phase transition in LiBH4 (see DSC data, red curve in Figure 2a).
Figure 7 shows the Li + ionic conductivity for absorbed S1, and the values are systematically lower than for pure LiBH4 as expected, since LiBH4 only makes up 21 wt % in the present sample (The low conductivity values for the samples can be explained by minor amounts of LiBH4 after rehydrogenation.For absorbed S1 the content is approx.2.5 molar equivalent of LiBH4 corresponding to 21 wt %.This value was calculated from the mass loss shown by TG data of absorbed S1).However, upon heating above 110 °C, the conductivities reach about four orders of magnitude higher than the RT modification, which is in good agreement with the conductivity enhancement in pure LiBH4 [38].The conductivity in absorbed S1 (Figure 7) shows a slightly steeper slope than pure LiBH4 [38], which manifested in the absence of the "conductivity jump" in our data.We first suggested that this observation was caused by the initial ball milling of S1.In a recent publication, it was shown that ball milled LiBH4-LiI samples have a higher conductivity before annealing caused by the formation of a "defect rich microstructure," which influences the conductivity positively [39].However, the sample chosen for impedance spectroscopy was heated to 400 °C during the first decomposition and then reabsorbed at 340 °C.This heat treatment should heal all defects and disregard their influence [39].In summary, the high temperature phase transition is observed at 110 °C, which is in agreement with pure LiBH4 [38,40].From the Li-ion conductivity measurements, the DSC event at 90 °C for absorbed S1 cannot be explained by the high temperature phase transition of LiBH4.However, a rather wide hysteresis is observed in Figure 7, which is in the range of 40-60 K.That is in strong contrast to the hysteresis of pure and ball milled LiBH4, which is only 4 K [41], but rather close to the LiBH4-LiCl system in which the hysteresis is about 20-40 K [40].The presence of lithium halides can be ruled out, as it would significantly lower the phase transition temperature [40].It has been observed [42] that nanoconfined LiBH4 in mesoporous silica scaffolds shows a remarkable Li-conductivity in the temperature range of RT-140 °C.According to Blanchard et al. [42], the high conductivity is a consequence of two different fractions of LiBH4, a bulk LiBH4 fraction and a thin (1.0 nm) interfacial layer of LiBH4.Assuming a morphological change in the grain size and in the inter-grains arrangement, it could be assumed that the formation of the abovementioned layer between LiBH4 and ErH3 after several cycles of hydrogenation has happened.This can explain the However, a rather wide hysteresis is observed in Figure 7, which is in the range of 40-60 K.That is in strong contrast to the hysteresis of pure and ball milled LiBH 4 , which is only 4 K [41], but rather close to the LiBH 4 -LiCl system in which the hysteresis is about 20-40 K [40].The presence of lithium halides can be ruled out, as it would significantly lower the phase transition temperature [40].It has been observed [42] that nanoconfined LiBH 4 in mesoporous silica scaffolds shows a remarkable Li-conductivity in the temperature range of RT-140 • C. According to Blanchard et al. [42], the high conductivity is a consequence of two different fractions of LiBH 4 , a bulk LiBH 4 fraction and a thin (1.0 nm) interfacial layer of LiBH 4 .Assuming a morphological change in the grain size and in the inter-grains arrangement, it could be assumed that the formation of the abovementioned layer between LiBH 4 and ErH 3 after several cycles of hydrogenation has happened.This can explain the inertia of the system to undergo the transition to the RT phase.The ionic conductivity contribution of this layer is dominant with respect to the bulk LiBH 4 , however, it is not visible by diffraction due to its nanoscopic nature.This effect of nanoconfined LiBH 4 could explain the rather low endothermic DSC event, which were discussed above, as similar events were also reported in Reference [42].
SEM images (see Figures S4-S6) were collected, but their resolution is not sufficient to observe a 1.0-nm thin interfacial layer.Further measurements with TEM are necessary to conclude with certainty that an interfacial layer of LiBH 4 has formed.Due to the paramagnetic properties of erbium, it was not possible to conduct NMR.
The synthesis of Er(BH 4 ) 3 has been described in Reference [36].The composite mixtures were ball milled using a Fritsch Pulverisette 6 planetary mill (Fritsch, Idar-Oberstein, Germany) employing an 80-mL tungsten carbide-coated steel vial and balls.A ball to powder ration of 40:1 was used.All sample descriptions of the composite mixtures are given in Table 1, including their composition and synthesis method as well as refined lattice parameters and phase fractions obtained by Rietveld refinements.
The samples were stored and handled in an MBraun glove box (MBraun Inertgas-Systeme GmBH, Garching, Germany) fitted with a recirculation system and oxygen/humidity sensors with H 2 O/O 2 levels below 1 ppm.All procedures outside of the glove box were performed using in vacuo or Schlenk line techniques under a purified Ar atmosphere.An in-house manufactured Sieverts-type apparatus [43] was used for hydrogen desorption-absorption cycling experiments.Desorption was performed using a temperature ramp of 5 • C•min −1 from room temperature to 400 • C under 3 bar H 2 , followed by a 12 h isothermal step.Absorption was conducted for 12 h at 340 • C and 100 bar H 2 .
Powder X-ray diffraction (PXD) data were collected using a Rigaku SmartLab diffractometer (Rigaku, Tokyo, Japan).The samples were measured within rotating glass capillaries, with an inner diameter of 0.5 mm and sealed with silicone grease.All measurements were completed in Debye-Scherrer geometry using CuKα radiation with λ = 1.54056Å, over the scattering angles 2θ = 10 • -70 • .
In situ synchrotron radiation powder X-ray diffraction (SR-PXD) was executed at Swiss Norwegian Beam Lines, BM01 [44], at the European Synchrotron Radiation Facility (ESRF), Grenoble, France.The cycling experiments were conducted using a heating/cooling rate of 5 • C•min −1 with a typical H 2 pressure of ~100 bar for absorption and ~3 bar for desorption.Hydrogen pressure was employed for all volumetric and in situ SR-PXD experiments, as similar conditions were reported to increase gas desorption compared to desorption under vacuum [17,27,45].Data were collected using a Pilatus 2 M detector (DECTRIS Ltd., Baden-Daettwil, Switzerland) and a sample-to-detector distance of 146 mm at wavelengths of 0.77787 Å. Wavelength and sample-to-detector distance were calibrated from an NIST LaB 6 standard.Exposure time was set to 30 s, giving a temperature resolution of 2.5 K/pattern.The sample was contained in a single-crystal sapphire tube (inner diameter 0.8 mm) connected to the cell with Vespel ferrules and Swagelok fittings.Hydrogen was introduced and removed from the cell with an in-house built computer-controlled gas rig.The cell was rotated by 10 • to improve powder averaging during each exposure.The single-crystal reflections from the sapphire tubes were masked out manually in Fit2D [46].Rietveld refinements were performed with GSAS and Expgui software [47,48].Thompson-Cox-Hastings pseudo-Voigt functions with three Gaussian and one Lorentzian parameter were used to model the Bragg peak profiles [49].The background was fitted with a shifted Chebyshev polynomial with up to 36 terms.The atomic positions for the Er(BH 4 ) 3 were taken from Y(BH 4 ) 3 [5] and refined with the data shown in Figure S3 at room temperature.BH 4 units were treated as rigid bodies with B-H distances of 1.13 Å.
Simultaneous thermogravimetric and differential scanning calorimetry (TG-DSC) experiments were carried out with a Netzsch STA 449 F3 Jupiter instrument (Netzsch, Bavaria, Germany).In some measurements, mass spectrometry (MS) was performed simultaneously by connecting a Hiden Analytical HPR-20 QMS (Hiden, Warrington, UK) to the TG-DSC.The samples were loaded in aluminum crucibles (~5-10 mg) and were heated up to 400-500 • C with a heating rate of 5 • C•min −1 .An argon flow of 70 mL•min −1 was used as a protection gas and purge gas.All given temperatures are peak temperatures, unless stated otherwise.
Li-ion conductivity was measured by impedance spectroscopy, employing an HP 4192A FL impedance analyzer (Keysight Technologies, Santa Rosa, CA, USA).The frequency range was 5 Hz to 10 MHz with a signal amplitude of 60 mV.The temperature was varied by 10 • C for each measurement from RT to 140 • C, and back to RT.Three sets were performed for each measurement in order to improve reproducibility.A 3-ton mechanical axial press was used for pressing pellets with a typical diameter of 6.35 mm and a thickness of 0.5-1.0mm.The samples were placed in a BDS 1200 Novocontrol sample cell (Novocontrol Technologies GmbH Co. KG, Montabaur, Germany) under an argon atmosphere between two gold ion-blocking electrodes.A single parallel resistor(R)-constant phase element (CPE) was used as an equivalent circuit at low temperatures, where the noise is higher.At higher temperatures, additional effects were taken into account, such as the polarization of the electrode interfaces (modeled with a single CPE) and, eventually, a second R-CPE parallel circuit due to grain boundary contributions.A Hitachi S-4800 Scanning electron microscope (SEM) (Hitachi, Tokyo, Japan) equipped with a Noran System Six energy dispersive spectrometer (EDS) was employed for investigating the absorbed S1 after the conductivity measurements.

Conclusions
This work characterized two new composite materials consisting of Er(BH 4 ) 3 , LiBH 4 and/or LiH.The composites react during ball milling, storage as well as decomposition in a two-step reaction where Er-hydrides and LiBH 4 are formed in the first step.ErB 4 is formed in a second step during thermal decomposition.The composites can be cycled between the first and second decomposition steps by applying hydrogen pressure.Volumetric measurements show a hydrogen capacity of 88% after the first cycle of the initially released hydrogen content and seem stable after the third cycle, with 95% rehydrogenation compared to the second cycle.With a weight loss of 4.5 wt % after the first desorption-absorption cycle, the hydrogen capacity had almost doubled compared to earlier investigated systems, which included LiCl.
Li-ion conductivity measurements of absorbed Er(BH 4 ) 3 + 6LiH showed an exceptional high hysteresis of 40-60 K for the transition between the high and low temperature phases of LiBH 4 , which may be a good starting point to investigate further, for all solid state Li-ion batteries.
synchrotron radiation powder X-ray diffraction experiments.Matteo Brighi and Radovan Černý conducted the impedance spectroscopy experiments, analysed the data and added valuable sections in the discussion of the results.Torben R. Jensen and Bjørn C. Hauback acted as main supervisor for this work.They contributed to initiate the work, followed up the experiments in Aarhus and Kjeller, respectively, and helped with the preparation of the manuscript.

Figure 2 .
Figure 2. (a) TG-DSC data of ball milled S1 (black curves) and reabsorbed S1 (S1 abs) (red curves), showing TG data (top) and DSC signal (bottom).X-axis in time (min) as the heating from RT to 400 °C proceeded with a 30-min isotherm.Temperature (°C) is given on the secondary y-axis to the right; (b) Ex situ PXD data of S1 showing one absorption-desorption cycle with ball milled (bottom), desorbed (middle), and reabsorbed (top) composite.λ = 1.54056Å.

Figure 2 .
Figure 2. (a) TG-DSC data of ball milled S1 (black curves) and reabsorbed S1 (S1 abs) (red curves), showing TG data (top) and DSC signal (bottom).X-axis in time (min) as the heating from RT to 400 • C proceeded with a 30-min isotherm.Temperature ( • C) is given on the secondary y-axis to the right; (b) Ex situ PXD data of S1 showing one absorption-desorption cycle with ball milled (bottom), desorbed (middle), and reabsorbed (top) composite.λ = 1.54056Å.

Figure 5 .
Figure 5. Hydrogen release in the first three desorption during cycling of S2 in a Sieverts-type apparatus.The hydrogen pressure is 5-10 bar.

Figure 5 .
Figure 5. Hydrogen release in the first three desorption during cycling of S2 in a Sieverts-type apparatus.The hydrogen pressure is 5-10 bar.

Figure 6 .
Figure 6.TG (top) and DSC (bottom) data of S2 after the first reabsorption.The x-axis is in time (min) as the measurement proceeds isothermally for 30 min after the heating ramp.Temperature (°C) is shown on the y-axis to the right.

Figure 6 .
Figure 6.TG (top) and DSC (bottom) data of S2 after the first reabsorption.The x-axis is in time (min) as the measurement proceeds isothermally for 30 min after the heating ramp.Temperature ( • C) is shown on the y-axis to the right.

Figure 7 .
Figure 7. Li-ion conductivity measured by AC impedance spectroscopy for S1 abs.Conductivity was measured every 10 °C from RT to 140 °C and back down to RT. See text for further discussion.

7 .
Li-ion conductivity measured by AC impedance spectroscopy for S1 abs.Conductivity was measured every 10 • C from RT to 140 • C and back down to RT. See text for further discussion.
Figure S7: SR-PXD Rietveld refinement and difference plot of S1, Figure S8: PXD Rietveld refinement and difference plot of S2, Table • C and 5-10 bar H 2 is ErH 3 , formed by additional hydrogen absorption by ErH 2 upon cooling.Minor Bragg peaks of ErB 4 are also observed.Reabsorption at 340 • C under 100 bar H 2 for 12 h resulted in the formation of LiBH 4 , as well as increased intensity of ErH 3 peaks.

Table 1 .
Compositions and synthesis method of the investigated samples.The products after ball milling are given with lattice parameters and fractions from Rietveld refinements of diffraction data at room temperature (RT).Refinements are shown in TableS1.Sample S2 was stored in a glove box for nine months before measurements were taken.Estimated standard deviations are given in parentheses.

Table 1 .
Compositions and synthesis method of the investigated samples.The products after ball milling are given with lattice parameters and fractions from Rietveld refinements of diffraction data at room temperature (RT).Refinements are shown in Tables S1.Sample S2 was stored in a glove box for nine months before measurements were taken.Estimated standard deviations are given in parentheses.