Introduction

Discovery of new battery materials is essential to further improve the gravimetric and volumetric energy density, as the dominant lithium-ion battery technology is reaching its theoretical limits1. All-solid-state batteries is a promising approach to increase the energy density utilizing metal anodes and divalent cationic conductors. Magnesium metal has a higher volumetric energy density (Mg: 3833 mAh cm−3 vs. Li: 2062 mAh cm−3) as compared to lithium allowing for more compact energy storage2,3. The main challenge is to facilitate fast ionic conductivity in the solid state, as the increased charge density of multivalent cations results in stronger interactions with the surrounding anion framework. In fact, there are only few reports of fast multivalent solid-state ionic conduction at ambient conditions4,5,6.

Recently, magnesium borohydride derivatives have received significant attention due to high ionic conductivities at close to ambient conditions and good compatibility with a magnesium metal anode7,8. Although the ionic conductivity of Mg(BH4)2 is negligible (σMg2+ ≈ 10−12 S cm−1 at 22 °C), the addition of neutral ligands have proven to increase the ionic conductivity to higher than 10−5 S cm−1 at room temperature (RT)4,8,9,10,11,12. Mg(BH4)2 readily reacts with a variety of neutral organic molecules forming new crystal structures with different magnesium coordination environments and lower dimensionality of the structural framework13,14. The ligand content of the fully coordinated Mg(BH4)2 ∙ xL (L = Ligand) can then be adjusted by ball milling with Mg(BH4)2 or by thermal treatment to prepare new crystalline compounds with lower x-values. The fully coordinated compound often consists of complex cationic units, e.g. [Mg(NH3)6]2+, and BH4− counter ions and displays a low ionic conductivity. Lowering the ligand content usually changes the local Mg2+ coordination to be coordinated by both the BH4- anions and the neutral ligands. This often results in one- or two-dimensional framework structures with increased ionic conductivity, and also influences the physical characteristics such as melting point and mechanical rigidity13.

Research on magnesium borohydride derivatives has so far focused on single ligand systems with great success, however, the effect of different ligands and formation of solid solutions has not yet been investigated. Ammonia and methylamine magnesium borohydride, Mg(BH4)2 ∙ NH3 and Mg(BH4)2 ∙ CH3NH2 are two of the best solid-state Mg2+ conductors and they are structurally similar4,15. They are both built from one-dimensional zig-zag chains, held together by either weak dihydrogen bonds or dispersion interactions. Solid solutions of metal borohydrides have previously been reported with either cation or anion mixing, but solid solution of the neutral ligands has yet to be reported16,17,18,19,20. Ligand mixing may alter the structure, introduce defects or provide additional interstitial sites, which can alter the Mg-ion mobility, and may allow fast ionic conductivity at lower temperatures.

Two fast magnesium electrolytes, Mg(BH4)2 ∙ NH3 and Mg(BH4)2 ∙ CH3NH2, are found to form solid solutions after mechanochemical treatment, which crystallizes in space group P212121. The solid solutions show a melting point depression with the lowest value observed for the equimolar composition at 60 °C. Furthermore, it is found that stacking faults are introduced during synthesis, however, these are eliminated upon heating above 40 °C. Stacking faults occur due to the difference in ligand geometry and increase the cationic conductivity. The equimolar composition of the solid solutions was found to have the highest cationic conductivity.

Results and discussion

Initial investigations

Mechanochemical treatment of Mg(BH4)2 ∙ NH3–Mg(BH4)2 ∙ CH3NH2 mixtures resulted in new solid solutions with varying composition, Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x (x = 0-1) as shown in Table 1. Initial sample characterization revealed that the high energy provided by the ball milling results in the formation of a solid solution. Using a high ball-to-sample mass ratio of 50:1 and a milling time of 120 min resulted in an almost complete reaction.

Table 1 Overview of investigated samples
Full size table

Bragg reflections from the solid solutions are positioned in between those of the two reactants (see Fig. 1a), suggesting structural similarities. This is further confirmed by infrared spectroscopy, where the spectra of sample s2-s4 can be described as a superposition of the parent compounds (see Fig. 1b). The Fourier-transform infrared spectroscopy (FTIR) spectra are assigned according to similar compounds reported in literature4,21,22. The B − H stretching modes (2030 to 2530 cm−1) and the deformation mode at 1240 cm−1 are constant throughout the samples, indicating that the local BH4 environment is unaffected by the formation of a solid solution, suggesting that the changes are only associated with an exchange of NH3 and CH3NH2. Modes above 3200 cm−1 are attributed to N − H stretching modes, while the weaker modes at 2925 to 3045 cm−1 are assigned to C − H stretching modes. Modes from 1400 to 1650 cm−1 are attributed to N − H bending. Modes from 950 to 1310 cm−1 are assigned to B − H and C − H bending as well as C − N stretching.

Fig. 1: Initial structural characterization.
figure 1

a Normalized powder X-ray diffraction data of Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x (x = 0, 0.35, 0.51, 0.71 and 1) highlighting the hkl values of the individual Bragg reflections (data collected at T = 25 °C). All minor unassigned reflections are identified as either Mg(BH4)2 ∙ CH3NH2 or α-Mg(BH4)2. The unit cell axes of Mg(BH4)2 ∙ NH3 (Pnma) were transformed to match the standard setting of space group P212121. b Infrared spectroscopy of Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x (x = 0, 0.35, 0.51, 0.71 and 1). Vibrational modes are assigned according to refs. 37,38.

Full size image

Compared to Mg(BH4)2 ∙ NH3 (space group Pnma), the powder X-ray diffraction (PXD) pattern of Mg(BH4)2 ∙ CH3NH2 (space group P212121) contains additional low intensity Bragg peaks due to a lower crystal symmetry associated with ordered bridging borohydride groups in the latter and the lower symmetry around the CH3NH2 crystallographic sites. A minor amount of unreacted Mg(BH4)2 ∙ CH3NH2 is observed in the diffraction data of all samples, while all Mg(BH4)2 ∙ NH3 has reacted. This indicates that Mg(BH4)2 ∙ CH3NH2 dissolves in the structure of Mg(BH4)2 ∙ NH3. This observation is counterintuitive compared to cation and anion substitution in metal borohydrides, where the compound with the smaller ion usually dissolves in the structure with the larger ion14,20,23.

Structural investigations

The solid solutions were identified to crystallize in the lower symmetry space group P212121, similar to Mg(BH4)2 ∙ CH3NH2. While the structural framework is virtually identical with that of Mg(BH4)2 ∙ NH3, the Pnma space group is associated with disordering of the bridging [BH4]− anion and a mirror plane symmetry element on the ligand site, which is incompatible with the orientation of the CH3NH2 ligand. Furthermore, the higher symmetry space group (Pnma) would result in the extinction of some weak intensity reflections observed for the solid solution. The composition of the solid solutions, i.e. the occupancy of NH3 and CH3NH2 on the ligand position in the Mg(BH4)2 ∙ CH3NH2 structure, was extracted by Rietveld refinement of synchrotron powder X-ray diffraction data. The refined occupancies reveal a slightly higher CH3NH2 content as compared to the initial synthesis composition, consistent with minor amounts of α-Mg(BH4)2 in the pristine Mg(BH4)2 ∙ NH3 sample (s1), which was not accounted for in the mixing ratios.

The structure of the solid solutions Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x (x = 0.35, 0.51, 0.71) was solved using the structure of Mg(BH4)2 ∙ CH3NH2 as the starting configuration. Subsequently, NH3 was introduced on the atomic position derived from the structure of Mg(BH4)2 ∙ NH3. The occupancies of NH3 and CH3NH2 were initially set to match the synthesis composition and refined using the restraint that the sum of NH3 and CH3NH2 occupancies was kept constant at 1. Atomic positions and occupancy of the ligands were refined independently. The structure of Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x is built from one-dimensional zig-zag chains as shown in Fig. 2. The chains consist of tetrahedral units of [Mg(BH4)3L], with Mg in the center, and the neutral ligand L (NH3 or CH3NH2) coordinating through the lone pair on N. Two [BH4]− anions act as bridging ligands and coordinate through the edge (κ2) of the [BH4]− tetrahedron, while one [BH4]− anion acts as a terminal ligand coordinating through the face (κ3) of the tetrahedron. This results in a coordination number of 8 for Mg2+, in agreement with the parent compounds4,15. The average ligand positions are slightly contorted as compared to the parent compounds, which is likely due to changes in the local environment.

Fig. 2: Comparison of the average crystal structure of the solid solution to the parent compounds.
figure 2

The crystal structure of Mg(BH4)2 ∙ (NH3)0.49(CH3NH2)0.51 is similar to that of the parent compounds, but with contorted ligand positions.

Full size image

Thermal investigations

While the structures of Mg(BH4)2 ∙ NH3 and Mg(BH4)2 ∙ CH3NH2 are similar, the weak dihydrogen and dispersion interactions within the crystal structure differ. For the structure of Mg(BH4)2 ∙ NH3, weak dihydrogen bonds between Hδ+ on NH3 and Hδ- on BH4− are present between the chains, and are proposed to be important for the interstitial Mg2+ ion conduction mechanism15. These interactions are weaker in Mg(BH4)2 ∙ CH3NH2, as the hydrophobic -CH3 moieties are located between the chains, and also changes the preferred conduction pathway4. The differences are also reflected by the difference in melting point, i.e. 90 °C for Mg(BH4)2 ∙ NH3 and 75 °C for Mg(BH4)2 ∙ CH3NH2, respectively. A melting point depression is observed for the solid solutions Mg(BH4)2 ∙ (NH3)1−x (CH3NH2)x (see Fig. 3), with the lowest, Tmp = 60 °C, for x = 0.51 (s3). Melting point was determined at the point where no diffraction was observed (see Fig. 4b and Supplementary Fig. 1). Eutectic melting compositions have also been reported for several other borohydride derivatives8,10,11,24,25.

Fig. 3: Melting points.
figure 3

Melting points of Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x (x = 0, 0.35, 0.51, 0.71, 1) extracted from the temperature where no diffraction is observed from synchrotron radiation powder X-ray diffraction.

Full size image
Fig. 4: Structural investigations.
figure 4

a Unit cell volume (filled symbols) and unit cell axis length (open symbols) as a function of composition (x) of Mg(BH4)2 ∙ (NH3)1−x (CH3NH2)x determined by Rietveld refinements at 20 °C. b Temperature resolved synchrotron powder X-ray diffraction data of Mg(BH4)2 ∙ (NH3)0.65(CH3NH2)0.35 (s2) (λ = 0.826366(3) Å and ΔT/Δt = 2 °C min−1). The melting point of the solid solution is marked with a dashed line. c The relative change of the unit cell parameters as a function of temperature for s2-s4. d The strain parameter for the [110] direction as a function of temperature for s2-s4. The dashed line marks release of strain at T > 40 °C. Error bars were extracted from the refinement software Fullprof.

Full size image

The unit cell parameters and the occupancy of NH3 and CH3NH2 were extracted from Rietveld refinements of the PXD data for s2-s4 at room temperature (see Supplementary Fig. 2). The unit cell axis length and volume per formula unit (V/Z) as a function of x in Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x is displayed in Fig. 4a, where the lines indicate the linear interpolation between x = 0 and x = 126. In general, the extracted parameters of V/Z agree well with Vegard’s law, supporting the existence of a solid solution with full solubility of the components. The largest discrepancy is observed for the x = 0.51 (s3), with a larger than predicted unit cell volume. The largest change in unit cell parameters is observed along the b-axis, consistent with the directional orientation of the neutral ligands and the larger size of CH3NH2 compared to NH3. The b-axes for the solid solutions are longer than predicted by Vegard’s law, suggesting a mismatch in size between the two ligands along this axis, possibly introducing some strain in the structures. This is also evident by significant peak broadening as observed for the (020) reflection as compared to e.g. (102) (see Figs. 1a and 4d). This effect is most pronounced in sample s2, likely due to the high NH3 content. The peak broadening likely arises from imperfect stacking of the chains, consistent with a predominant effect along the ligand direction (b-axis). This was accounted for in the structural model by refining the anisotropic strain parameter along the [010], [110], and [011] crystallographic directions, which resulted in a good congruence between the model and the data.

The temperature dependent structural evolution of the solid solutions of Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x was investigated using in situ synchrotron radiation powder X-ray diffraction (see Fig. 4b and Supplementary Fig. 3). Sequential Rietveld refinements were performed in the temperature range 20 to 55 °C. Beyond this temperature, the refinements are unreliable due to melting of the samples. During heating, the Bragg reflections of Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x move towards lower scattering angles, consistent with thermal expansion of the unit cell. In the case of Mg(BH4)2 ∙ (NH3)0.65 (CH3NH2)0.35 (s2), the sample melts at 63 °C, leaving α-Mg(BH4)2 as the only diffracting compound (see Fig. 4b). Interestingly, the Bragg reflections of Mg(BH4)2 ∙ (NH3)0.65(CH3NH2)0.35 (s2) become sharper prior to melting and with a negative thermal expansion along the b-axis. While the peaks in general become sharper, the anisotropic strain associated with the [110] direction is also released, explaining the decreasing a- and b-axis lengths (see Fig. 4b, c). Strain along the [010] and [011] directions were also investigated (see Supplementary Fig. 3), where a significant decrease in strain is observed for s2 in the [010] direction. This effect of stacking fault elimination is less pronounced, but still observed for Mg(BH4)2 ∙ (NH3)0.49(CH3NH2)0.51 (s3) and Mg(BH4)2 ∙ (NH3)0.29 (CH3NH2)0.71 (s4), however an increased thermal expansion is observed for these compositions above 45 °C (see Fig. 4c and S3). Thus, the increasing temperature results in a strain release, effectively reducing the unit cell size and relaxing the structure to a more thermodynamically stable configuration.

Ionic conductivity

The Mg2+ conductivity was assessed using electrochemical impedance spectroscopy (EIS) measurements of a symmetric cell with Mo-blocking electrodes, Mo|Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x | Mo (x = 0, 0.35, 0.51, 0.71, 1). Previous studies of both Mg(BH4)2 ∙ NH3 and Mg(BH4)2 ∙ CH3NH2 have demonstrated magnesium plating and stripping, and the Mg(BH4)2 ∙ xNH3 system has been used in all-solid-state magnesium battery cells4,7,8,15. Mg2+ is therefore assumed to be the only mobile ion. The data were fitted with an equivalent circuit, as described in the experimental section to extract the ionic conductivity as a function of temperature (see Fig. 5). In some cases the ionic conductivity of the solid solutions can increase by several orders of magnitude compared to their pure counterparts, as reported for e.g. Na2(B10H10)x(B12H12)1−x, Li1+2xZn1−xPS4 and Li6−xPS5−xCl1+x27,28,29, however in this case the ionic conductivity of samples s2-s4 are in between that of the parent compounds. Here, the highest conductivity of the solid solutions is achieved for x = 0.51 (s3) with σ(Mg2+) = 7.3 ∙ 10−6 S cm−1 at 40 °C, which is around one order of magnitude lower than Mg(BH4)2 ∙ CH3NH2 (s5). Interestingly, x = 0.51 (s3) has one order of magnitude higher ionic conductivity than x = 0.35 (s2) and x = 0.71 (s4). This suggests that the conduction paths of the solid solutions are altered, as compared to the parent compounds, as ionic conductivity depends on the specific composition and structure, rather than the relative content of the more conductive component (Mg(BH4)2 ∙ CH3NH2).

Fig. 5: Ionic conductivity.
figure 5

a Ionic conductivity of Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x (x = 0, 0.35, 0.51, 0.71 and 1) during 1st heating (open symbols) and 3rd cooling (closed) after thermal cycling between 30 and 40 °C three times. b Comparison of the best performing solid solution (Mg(BH4)2 ∙ (NH3)0.49(CH3NH2)0.51 (s3)) to other reported magnesium-ion conductors4,8,9,10,15,30,39.

Full size image

The effect of stacking fault elimination was also noticed during the initial heating in the EIS experiment, where the highly strained sample s2 initially showed a higher Mg2+ conductivity, which decreased above 30 °C. As a result of the structural strain release, the ionic conductivity was significantly lower in the subsequent cooling and heating cycles. A similar, but slightly less pronounced, effect is observed for s4. Interestingly, s2 and s4 have similar Mg2+ ionic conductivities and activation energies for the cationic migration (Ea) after heat treatment, Ea ≈ 1.4 eV (see supplementary Table 1). The activation energy of s3 is significantly decreased after the heat treatment (from 1.27 to 0.91 eV) and this low activation energy state is maintained all the way to −20 °C. The results presented here suggest that introducing differently sized ligands may induce stacking faults, and thereby increase the amount of void space and improve conductivity.

Note that previous reports on Mg(BH4)2 ∙ CH3NH2 show varying ionic conductivities for different synthesis batches, and a fading ionic conductivity after several months of storage. A slight excess of CH3NH2 results in the ionic liquid Mg(BH4)2 ∙ 2CH3NH2 and the release of this excess may therefore result in a fading of the ionic conductivity over time. As trace amounts of Mg(BH4)2 were found in the samples presented in this work, it is unlikely that excess CH3NH2 is present. Moreover, the ionic conductivity of the samples were measured several months after synthesis to ensure a stable chemical composition.

While the ionic conductivity of s3 is lower than the precursor Mg(BH4)2 ∙ CH3NH2 (s5), the ionic conductivity is comparable or even better than some of the best reported Mg2+ conductors, e.g. Mg(BH4)2 ∙ NH2(C2H4)NH2, Mg(BH4)2 ∙ 1.6NH3 and Mg(BH4)2 ∙ 2NH3BH3, in particular below room temperature due to the lower activation energy for the cationic migration8,9,30. A common approach to lower the activation energy and increase the ionic conductivity at low temperatures is the addition of inert nanoparticles, which has the additional effect of providing mechanical stability8,11,31. To test this strategy, a nanoconfined sample consisting of s3 and Al2O3 (13 nm, 50 wt%) was investigated (s6). However, this sample demonstrated an increased activation energy, Ea ≈ 1.4 eV, which resulted in a lower ionic conductivity below 38 °C (s6) (see Supplementary Fig. 4). Furthermore, it was not possible to mechanically stabilize a molten state as seen for the Mg(BH4)2 ∙ xNH3 system8.

Conclusion

In conclusion, three solid solution samples were synthesized with the composition Mg(BH4)2 ∙ (NH3)1−x(CH3NH2)x (x = 0.35, 0.51, and 0.71), resulting in the first reported mixed ligand metal borohydride systems. The refined unit cell volumes in the compositional range 0 < x < 1, follow Vegard’s law revealing full solubility of the system, Mg(BH4)2 ∙ NH3 − Mg(BH4)2 ∙ CH3NH2. The crystal structures are analogous to the parent compounds built from zig-zag chains of tetrahedral [Mg(BH4)3L] units connected via two bridging [BH4]− groups, while the neutral ligand (L = NH3 or CH3NH2) partially occupies the same crystallographic site. Interestingly, it appears that Mg(BH4)2 ∙ CH3NH2 dissolves in the structure of Mg(BH4)2 ∙ NH3, which is unusual as the latter contains the smaller ligand. Temperature resolved powder X-ray diffraction revealed microstructural strain along the [110] direction, likely caused by stacking faults, which was eliminated upon heating above 40 °C. This was also reflected in the ionic conductivity, which decreased upon strain release. The ionic conductivity of the samples was found to be in between the parent compounds, but did not correlate directly with the relative composition (x). The highest conductivity was found for the solid solution Mg(BH4)2 ∙ (NH3)0.49(CH3NH2)0.51 with σ(Mg2+) = 7.3 ∙ 10−6 S cm−1 at 40 °C, which is also the composition with the lowest melting point (60 °C). In contrast to other related compounds, the addition of inert and insulating nanoparticles did not have a positive impact on the Mg2+ conductivity. Furthermore, it was not possible to mechanically stabilize the eutectic melt at room temperature. However, this work opens new avenues for ligand-assisted solid state ionic conductors, where solid solutions between compounds with different ligands can alter the physical properties including the ionic conductivity.

Methods

Synthesis

Storage of chemicals and synthesis are performed under inert conditions using either an argon filled gloveboxes or Schlenk line techniques. An overview of investigated samples can be found in Table 1. Synthesis of α-Mg(BH4)2 was performed as described in refs. 32,33. Anhydrous dimethylsulphide borane, (CH3)2S ∙ BH3 (100%, 10 mL, 0.105 mol), was mixed with anhydrous toluene (99.8%, 40 mL) in a 250 mL round-bottomed flask with a filter. While stirring, di-n-butylmagnesium (Mg(C4H9)2, 1.0 M in heptane with up to 1 wt% triethylaluminum, 28 mL, 0.028 mol) was gradually added over two minutes, ensuring an excess of (CH3)2S ∙ BH3 to prevent the formation of partly substituted borohydrides33. A white precipitate formed upon addition of Mg(C4H9)2 and the suspension was left to react for 20 h while stirring at room temperature. The product precipitate, Mg(BH4)2·½S(CH3)2, was washed with toluene (3 ×10 mL). The dry powder was subsequently heated in Schlenk tubes to 140 °C for 1.5 h under argon flow and 3 h under a dynamic vacuum to form pure α-Mg(BH4)2.

Synthesis of s1 and s5 was performed by first synthesizing Mg(BH4)2 ∙ 6 L (L = NH3, CH3NH2) via the solid-gas reaction between dry NH3/CH3NH2 and α-Mg(BH4)2 at 0 °C and −10 °C, respectively, as described in refs. 4,15,34. Mg(BH4)2 ∙ L was synthesized according to refs. 4,15, where Mg(BH4)2 ∙ 6 L were ball milled with Mg(BH4)2 in molar ratio 1:5 for two minutes with 60 repetitions at 350 rpm with a two minute break between repetitions (total milling time 120 min). A ball-to-sample ratio of 50:1 was used. WC balls and vials were used with a ball size of Ø 10 mm. Subsequent PXD analysis showed that that sample s1 contained minor amounts of Mg(BH4)2 ∙ 2NH3 and 505 mg of sample (s1) was ball milled with 80 mg Mg(BH4)2 using the same procedure. Synthesis of the solid solutions (s2-s4) was performed by ball milling stoichiometric amounts of s1 and s5 in the molar ratios of 2:1 (s2), 1:1 (s3) and 1:2 (s4) for two minutes with 60 repetitions with a two-minute break between repetitions at 350 rpm. Synthesis of s6 was performed by ball milling s3 with 50 wt% Al2O3 nanoparticles (13 nm) using the same milling program as for s2-s4. Before measuring electrochemical impedance spectroscopy, a pellet was cast by melting sample s6 at 65 °C for 30 min into a 5 mm pellet die, before leaving it under the weight of the piston for two days to solidify.

Characterization

Liquid-state nuclear magnetic resonance (NMR) spectroscopy was performed on a Bruker Ascend 400 MHz spectrometer equipped with a 1H-13C-15N 5 mm TXI liquid state probes. Samples were dissolved in D6 DMSO in NMR tubes.

Synchrotron radiation powder X-ray diffraction data were acquired at the I11 beamline at the diamond light source, during heating with a constant rate of 2 °C min−1 in the temperature range 20 to 80 °C and using a wavelength of λ = 0.826366(3) Å.

The software FOX was used for indexing of the unit cells of the three solid solutions35,36. The structure of β-Mg(BH4)2 ∙ CH3NH2 was used for subsequent Rietveld refinements using the software Fullprof4. The background were described by linear interpolation between selected points, while Pseudo-Voigt profile functions were used to fit the diffraction peaks. As hydrogen is a weak scatterer of X-rays; BH4-, methylamine and ammonia were refined as rigid bodies. To this end, accurate hydrogen positions could not be obtained, however soft constrains were used to obtain a more “chemically” accurate structure. NH3 was added to the same site as CH3NH2 and occupancies were set as the milling ratios to be refined afterwards. An anisotropic strain model was added as significant diffraction peak broadening was observed. The α-Mg(BH4)2 structure was added to the 2:1 (s2) refinement to account for the impurity. Two additional structures (Mg(BH4)2 ∙ CH3NH2 and α-Mg(BH4)2) were added to account for the excess Mg(BH4)2 ∙ CH3NH2 and α-Mg(BH4)2 in 1:1 (s3). Only the Mg(BH4)2 ∙ CH3NH2 structure was added to the 1:2 (s4) refinement. The results can be seen in Fig. S2. Initially, the unit cell parameters and profile functions were refined, followed by atomic position and occupancy refinements of NH3 and CH3NH2 around the same crystallographic site.

EIS measurements were performed from 1 × 107 Hz to 1 Hz using a Biologic MTZ-35 impedance analyser. To eliminate capacitance, resistance, and inductance in the setup a custom-made symmetrical molybdenum sample holder equipped with a 4-probe setup was used. Samples were cold pressed in a hydraulic press at 1 GPa at room temperature for one minute. Samples were heated using a custom-made furnace at 0.2 °C min−1 in the temperature range 24 to 40 °C and continually measured during heating. Additional measurements were performed at 20, 13 and 6 °C to assess low temperature conductivity. Data were fitted using a Q1/(Q2 + R2) equivalent circuit where R2 is a charge transfer resistance, Q1 will act as a capacitor (α1 ≈ 1) essentially creating a non-ideal RC circuit with Q2 acting like a mass transfer element (α2 ≈ 0.7). We assume the resistance of the setup to be negligible.

FTIR was conducted with a Shimadzu QATR-S spectrometer placed in an argon filled glovebox. FTIR spectra were measured in the range 500 cm−1 to 4000 cm−1 with 4 cm−1 and averaged over 32 scans.