Increasing Hydrogen Density with the Cation-Anion Pair BH 4 −-NH 4 + in Perovskite-Type NH 4 Ca ( BH 4 ) 3

A novel metal borohydride ammonia-borane complex Ca(BH4)2·NH3BH3 is characterized as the decomposition product of the recently reported perovskite-type metal borohydride NH4Ca(BH4)3, suggesting that ammonium-based metal borohydrides release hydrogen gas via ammonia-borane-complexes. For the first time the concept of proton-hydride interactions to promote hydrogen release is applied to a cation-anion pair in a complex metal hydride. NH4Ca(BH4)3 is prepared mechanochemically from Ca(BH4)2 and NH4Cl as well as NH4BH4 following two different protocols, where the synthesis procedures are modified in the latter to solvent-based ball-milling using diethyl ether to maximize the phase yield in chlorine-free samples. During decomposition of NH4Ca(BH4)3 pure H2 is released, prior to the decomposition of the complex to its constituents. As opposed to a previously reported adduct between Ca(BH4)2 and NH3BH3, the present complex is described as NH3BH3-stuffed α-Ca(BH4)2.


Introduction
Complex hydrides material design has undergone major progress since the first reports suggesting LiBH4 as a solid state hydrogen storage material [1]. Semi-empirical chemical concepts combining the stabilities of the less stable and more stable borohydrides [2], for instance, have given way to more profound concepts that exploit different kinds of weak interaction [3]. Recently, we reported on a large family of mixed-metal borohydrides crystallizing in the perovskite structure type, the latter being a prime example of functional materials design [4]. As a very stable structure type, the perovskite form allows one to study the behavior of a large range of different metals in a borohydride environment, but also to apply an immense toolbox developed throughout decades of metal-oxide perovskite design. The most simple concept comprises cation-substitution on the cuboctahedral A-and octahedral B-sites of a material generalized as AB(BH4)3. A logical further step is to explore anion-mixing on the ligand site. It was shown that the borohydride perovskite host readily takes up halide anions on the borohydride site [4]. In this context, it comes to mind to test the nitrogen-based hydrogen molecules, which are currently being vigorously investigated as means to tailor hydrogen release temperatures [3]. Recently we described interesting lattice instabilities in the metal boroyhdride perovskites where supposedly close homopolar dihydrogen contacts between adjacent borohydride anions stabilize lattice distortions in the high-temperature phase of AB(BH4)3, while these are most commonly destabilized by temperature in metal oxide or halide perovskites [4].
Aside from ammonium borohydride NH4BH4 itself [5][6][7], the ammonium cation has been rather neglected by the hydrogen storage community, and so far considered only in conjunction with higher boranes such as NH4B3H8 [8], (NH4)2B10H10 or (NH4)2B12H12 [9] which impede the borazine evolution during thermolysis thanks to their stable borane cages, but never as a building block in a metal-based material. Attempts to tailor the thermal stability of NH4BH4, which decomposes slowly at room temperature (RT), have also been carried out to develop it as solid state hydrogen storage material [7,10].
The underlying concept of incorporating NH4 + was to increase the gravimetric hydrogen capacity ρm of perovskite-type KCa(BH4)3 by monovalent cation substitution on the K-site or divalent cation substitution on the Ca-site. Attempts to achieve this by the substitution K + -Na + or Ca 2+ -Mg 2+ fail, as the hypothetical compounds NaCa(BH4)3, KMg(BH4)3 or NaMg(BH4)3 all come to lie outside the experimentally established stability field of the structure type within borohydrides [4]. The ionic radius of monovalent NH4 + (1.43 Å), on the other hand, is close to that of K + (1.33 Å) and ammonium easily substitutes for potassium in many metal halides. The calculated tolerance and octahedral factors suggest NH4Ca(BH4)3 to be a stable compound. Its thermal stability and structure were recently reported in [4], accomplishing the long-term RT-stabilization of ammonium borohydride NH4BH4, with the extreme hydrogen density of 24.5 wt % H2 in a metal ammonium borohydride, NH4Ca(BH4)3, with a high density of ρm = 15.7 wt % H2. Herein, we modify the reported synthesis based on ammonium chloride NH4Cl to produce chlorine-free samples, taking up this study and focus on the actual decomposition mechanism of NH4Ca(BH4)3 via an unknown ammonia-borane complex, which is obtained as a single phase sample thanks to the modified procedure and whose structure is solved on the basis of synchrotron X-ray powder diffraction. The vividly applied concept to tailor hydrogen release properties, the hetero-polar contact between protic and hydridic hydrogen, is thus extended for the first time to a cation-anion pair, NH4 + -BH4 − , in a metal borohydride. The presented concept of cation-substitution by ammonium is easily extended to a large number of previously reported mixed-metal borohydrides [11].

Results and Discussion
The substitution of K + for NH4 + results in a stable compound NH4Ca(BH4)3, which is presently investigated by means of two different synthetic mechanochemistry routes using different sources of the ammonium molecule. While the first employs a metathesis reaction based on ammonium chloride, in the second, chlorine-free, approach, NH4BH4 is loaded at low temperatures in the milling jar and reacted directly with Ca(BH4)2 in an addition reaction. The Cl-approach results in better crystallized compounds and a less complex phase composition (of novel phases) in as-milled samples, which is why the decomposition is discussed for these samples, despite the chlorine-content. On the other hand, the main decomposition product occurring during thermolysis of both samples is obtained as a well crystallized single phase composition in the NH4BH4-based one, when milled in the presence of a solvent (ether). The crystal structure of the main decomposition product is hence solved and discussed on the basis of these data.

Formation of NH4Ca(BH4)3
In samples synthesized from NH4Cl, perovskite-type NH4Ca(BH4)3 is formed according to a general metathesis reaction used for borohydride-perovskites, developed for cases when one of the precursors cannot be provided as a borohydride: A number of different nominal NH4Cl:Ca(BH4)2:LiBH4 compositions were investigated, NH4Ca(BH4)3 forming in the less NH4Cl-rich mixtures 1:1:1, 1:2:1 and 1:2:2, but neither in 2:2:1 nor 3:1:3. At least three other phases with very low decomposition temperatures were observed as impurities or the main phase in 2:2:1 and 3:1:3 mixtures in as-milled samples prepared from the chloride, presumably containing NH4 + as indicated by their low decomposition temperatures, but these could not be identified and will not be further discussed presently.
The formation of NH4Ca(BH4)3 from the borohydrides, i.e., Ca(BH4)2 and NH4BH4, was attempted in various trials. To this end milling jars were preliminarily cooled to 243 K, in order to stabilize NH4BH4 (unstable at RT) in a ratio 1:1 next to solid Ca(BH4)2. Soft milling (30 min at 500 rpm) of this mixture resulted in a multiphase powder with an unsatisfactory phase yield of NH4Ca(BH4) (black curve in Figure 1). The low yield may be owed to partial decomposition of NH4BH4, since the jars were not cooled during milling and the inside-temperature reached approximately 310 K.
The procedure was subsequently modified to a solvent-assisted protocol where diethyl ether was added to the mixture and the inside-temperature was kept below 270 K, resulting in a significantly improved phase yield of NH4Ca(BH4)3 (blue curve in Figure 1). At temperature below RT, and neglecting impurities, the formation from the borohydrides can hence ideally be approximated as: Despite differences in the phase composition with respect to impurities the perovskite-type borohydride NH4Ca(BH4)3 is the main phase in both the presently discussed chloride (1:1:1) and ether-based approaches. Furthermore, the fact that the lattice parameter a is identical within error (5.655(3) and 5.651(5) Å at 313 K for NH4Ca(BH4)3 produced by chloride-and ether-based ball milling) as indicated by identical Bragg peak position in Figure 1 demonstrates the efficiency of Reaction (1) in terms of its capability to produce a borohydride-pure product without chloride on the anion-site, showing that the metathesis is completed in such an approach.

Thermolysis of NH4Ca(BH4)3
The ammonium metal-borohydride perovskite NH4Ca(BH4)3 starts decomposing at approximately 380 K in both, chloride and chloride-free samples. The respective temperature-dependent in-situ synchrotron diffraction data are shown for the chloride sample in Figure 2. The suggested decomposition mechanism for the chloride synthesis involves the formation of β-Ca(BH4)2 and LiBH4 via an intermediate decomposition product, which occurs as an impurity already in some as milled mixtures in both approaches (Supplementary Figures S1 and S2). This decomposition intermediate is an unreported AB-complex of Ca(BH4)2 (AB = NH3BH3) and is discussed in detail below. As temperature is increased, calcium borohydride then proceeds to take up chlorine according to Ca(BH4)2−xClx in Cl-mixtures, as discussed below. Interestingly, chloride seems to stabilize the β-phase of Ca(BH4)2, which has previously been observed [12], whereas in ether-milled samples the phase composition post-decomposition is a mixture of β-and γ-phases. Ideally, the decomposition mechanism for the chloride-free protocol consists in inverting Reaction (2), i.e., the stored NH4BH4 would be released. In reality, however, the decomposition involves the intermediate 1, i.e., the novel AB-complex Ca(BH4)2·AB. Hence, while NH4Ca(BH4)3 may be considered a means of storing (irreversibly) ammonium borohydride in the solid state, it cannot be released. The idealized decomposition mechanism of NH4Ca(BH4)3 can therefore be written as: While this reaction is stoichiometric and observed in the chloride free samples, the situation is more complex when the presence of chlorine is taken into account, leading to multiple processes releasing and absorbing chlorine which are outlined below. Figure 3 shows the previously published [4] analysis of volatile decomposition products with the corresponding thermal and gravimetric signals, measured on the chloride-sample. Notably, there is no diborane released during the whole thermal decomposition of NH4Ca(BH4)3, which in turn does release one molecule of hydrogen around 380 K (observed on Cl sample) and transforms to Ca(BH4)2·AB, in both chloride and ether milled samples, according to Reaction (3).
In chloride-based samples the decomposition proceeds directly to the AB-complex. This gas release is readily identified as the first mass loss in Figure 3, where mass spectrometry (top panel) shows that the released product's composition is nearly pure hydrogen (very small signal of NH3 at 380 K). Ca(BH4)2·AB then decomposes to release additional hydrogen above 400 K, accompanied by a minor amount of ammonia NH3 (Figure 3), similar to previously reported reaction mechanisms for AB complexes involving recovery of β-Ca(BH4)2 [13,14]. This corresponds to the second mass loss and is visible as a continuous evolution in the spectrometric detection of H2. The decomposition temperature of the AB complex compares quite well with those of the first two reported members of the metal borohydride AB family [15]. In the following we briefly discuss the release and uptake of chlorine by different phases during thermal decomposition of the mixture NH4Cl:Ca(BH4)2:LiBH4 1:1:1 and referring to the in-situ diffraction data shown in Figure 2. Chloride substitution has recently been investigated by ball-milling CaCl2 + Ca(BH4)2 [16]. Therein, anion-substitution is achieved by heating as-milled mixtures of unreacted precursors, which results in halide-substituted β-Ca(BH4)2-xClx. The authors report different compositions with discrete lattice parameters depending on the nominal composition of precursors in the respective sample. The chlorine uptake by β-Ca(BH4)2 during the decomposition of NH4Ca(BH4)3 is fundamentally different, since it is not CaCl2 acting as a chlorine-source. In this case, the chlorine is provided by LiCl, which is formed during the metathesis Reaction (1). Concurring with the decomposition of the ammonium perovskite LiCl continuously transforms into hexagonal HT-phase Li(BH4)xCl1−x, which crystallizes together with halide-substituted β-Ca(BH4)2 ( Figure 2) according to Reaction (4): While the dissolution of chloride is known to stabilize the hexagonal HT-phase of LiBH4, we observe a hitherto unreported behavior of the HT-phase (β) of Ca(BH4)2. A detailed examination of the in-situ diffraction data provided in Figure 2 is essential to the following discussion. At approximately 350 K the remnant Ca(BH4)2 not reacted but amorphized during ball-milling starts recrystallizing, accompanying the formation and crystallization of the AB-complex Ca(BH4)2·AB. Despite the temperature increase the lattice parameters of β-Ca(BH4)2 decrease as crystallinity increases (see Bragg signal at 15° shifting to higher angles with temperature, in Figure 2), due to chlorine-uptake originating from LiCl, while the Li is used to form Li(BH4)xCl1−x. At temperatures of approximately 400 K, coinciding with the decomposition of the AB-complex in the chloride-based sample, halide-substituted β-Ca(BH4)2 abruptly forms two phases of well-defined chemical composition and discrete lattice parameters (labeled Phase1 and Phase2 in Figure 2), accompanied by a third phase of intermediate composition and much broader diffraction peaks. Rietveld refinement was performed on data collected at 413 K and including 3 different β-Ca(BH4)xCl2-x compositions. The result of the refinement is shown in Figure 2. Two compositions Ca(BH4)0.9Cl1.1 and Ca(BH4)1.2Cl0.8 could be refined reliably corresponding to unit cell volumes of 194.8(9) and 185.3(7) Å 3 , respectively. the evolution of the cell volumes at higher temperature is shown in Supplementary Figure S3. These values are far below that of the reported unit cell volume of pure homoleptic β-Ca(BH4)2 at 305 K, 208.1 Å 3 [17], in agreement with the smaller anion radius of Cl − . The anion-composition of the third component corresponding to the broad signals in Figure 2 could not be determined reproducibly but is thought to lie between the two determined stoichiometries. The line width is approximately a factor 2.5 larger than in both other compositions. This may be due to a continuous distribution of chlorine dissolution on the BH4-site. But also a shorter coherence length could be at the origin of the signal-width, suggesting intermediate chemical compositions on very short length scales, for instance at the grain-boundaries between compositions Ca(BH4)0.9Cl1.1 and Ca(BH4)1.2Cl0.8. To our knowledge the coexistence of various discrete anion-compositions has not been reported for β-Ca(BH4)2, despite anion-substitution having been quite well studied in this compound [12,16].

Ether-Samples
In the samples which were prepared by ball milling solidified NH4BH4 in ether at low temperatures the decomposition mechanism is not influenced by the kinetics of chlorine-exchange between the involved pseudobinary compounds β-Ca(BH4)2 and LiBH4.
While the transformation of NH4Ca(BH4)3 to the novel AB-complex is observed as in the chloridesample, the decomposition now involves a further intermediate 2 (Supplementary Figure S2) that has not been characterized (T-interval 370-385 K, given the vicinity to the solved structure of Ca(BH4)2·AB we assume 2 to be a further AB-complex, possibly a phase transformation then takes place at 380 K), but indexed to an orthorhombic cell of a = 11.602, b = 14.339, c = 14. 296 Å, V = 2378 Å 3 . This phase does not occur in chloride-syntheses. It is unlikely that there is a hydrogen loss occurring between intermediate 2 and Ca(BH4)2·AB, since the reported product of the first two decomposition steps of NH4BH4 is the diammoniate of diborane [(NH3)2BH2](BH4), which itself is an isomer in equilibrium with AB [18], and no intermediate deprotonated molecule exists. Given the small stability field of approximately 5 K and the vicinity to the stability field of Ca(BH4)2·AB, intermediate 2 may be a further AB-derivative of different stoichiometry, or merely a superstructure to Ca(BH4)2·AB, related by a polymorphic transformation.
In ether milled mixtures Ca(BH4)2·AB begins to decompose to β-and γ-Ca(BH4)2 at 400 K, proceeding until 430 K (Supplementary Figure S2), which is in surprising agreement with the decomposition-onset of the previously reported AB-complex of Ca(BH4)2, determined at 398 K, and ending at 428 K [14]. Hence, while the second decomposition step follows the established mechanism of AB-complexes, the perovskite adds an additional hydrogen release prior to this.
We have not identified further decomposition products. The previously reported AB-complexed Ca(BH4)2 decomposes to its constituents. AB melts at 377 K therefore its possible existence cannot be determined from the diffraction patterns. A difference in structure between the two (discussed below) could suggest a slightly modified decomposition pathway, the reported AB-complex [14] possibly showing a greater predisposition to dissociate to Ca(BH4)2 and AB due to its structural topology, which is built of defined layers of both precursors, while Ca(BH4)2·AB reported herein contains AB molecules whose nearest neighbor interactions are considered to be weaker with respect to those taking place in AB itself.

Crystal Structure of NH4Ca(BH4)3
In the ammonium calcium borohydride perovskite the NH4 + cation occupies the A-site of the perovskite structure, where Ca is the B-cation and the BH4 − anion is the ligand. We have chosen to model the structure in an ordered manner in a subgroup symmetry which allows for an identical unit cell metric to the primitive perovskite unit. A fully disordered structure can be modelled in Pm-3m, however the fit is not better. The ordered description is supported, albeit not fully justified, by X-ray diffraction data collected down to temperatures of 100 K which did not show a change in diffraction pattern that would suggest a phase transition due to ordering. All reported phase transitions in borohydride perovskites occur above RT without any further events down to 100 K [4,19].
Furthermore, the dominant weak interaction in NH4Ca(BH4)3 is of attractive nature and structurally points towards an ordered model. Future neutron studies, as planned also for other metal borohydride perovskites may however result in a correction of this space group symmetry. Our structural description currently is in P-43m, which is a subgroup to Pm-3m, and allows maximizing homopolar dihydrogen contacts BH···HB between hydride anions, which are constrained to lower values by symmetry in other subgroups to Pm-3m, as for instance P432. One boron position and one nitrogen position provide orientation between the molecular species that are rotated by 90° around the principal cubic direction with respect to each other (Figure 4).

Crystal Structure of Ca(BH4)2·NH3BH3
The dehydrogenation product of the ammonium borohydride perovskite is the calcium borohydride ammonia-borane complex Ca(BH4)2·AB, which contains one AB molecule per Ca(BH4)2 unit and whose structure was solved ab-initio on a single phase sample from diffraction data collected at 414 K ( Figure 5) during the decomposition of ether-milled NH4Ca(BH4)3. Interestingly, an AB complex of Ca(BH4)2 has previously been reported, which contains two AB molecules per unit cell [14] and was observed not during thermal composition, but obtained as an adduct by ball-milling NH3BH3 with Ca(BH4)2. The structure therein was described as containing layers of AB alternating with layers (square lattice) of borohydride stacked along the long axis, b.
A close inspection of the structural model of Ca(BH4)2·AB ( Figure 6) reveals that the borohydride layers formed by a square lattice are conserved, however buckled by 90°. A further relationship is found in α-Ca(BH4)2-like ribbons, running parallel to the a-axis in Ca(BH4)2·AB. These Ca(BH4)2-ribbons are highlighted in the grey box in Figure 6. In this description the NH3BH3 molecules form a further one-dimensional structural element (zig-zag chain), which alternates with Ca(BH4)2 ribbons In this context, the relationship between the AB-complex Ca(BH4)2·AB and the reported Ca(BH4)2·2AB may therefore be described best as a reduction in dimensionality of the structural elements of the precursors in the former. In α-Ca(BH4)2 the square lattice (layers) of Ca-BH4 conserved in both AB-complexes is oriented perpendicular to the b-axis, while the ribbons conserved only in Ca(BH4)2·AB connect these layers and run along the a-axis. If the buckled Ca-BH4 square lattice (blue and grey sheets in Figure 6) is flattened out the α-Ca(BH4)2 structure is recovered in Ca(BH4)2·AB, however with 50% of the ribbons replaced by zig-zag chains of NH3BH3. There is therefore no major reconstruction of the α-Ca(BH4)2 structure in this AB-complex.
The B-N lengths in Ca(BH4)2·AB at 414 K amount to 1.613(5) Å, significantly longer than those in the reported Ca(BH4)2·2AB (1.546 Å at 298 K) and very close to those in solid NH3BH3 (1.592 Å) [13,14]. The intermolecular distances between hydride and proton ions BH···HN pertaining to adjacent AB molecules amount to 2.05 Å, as opposed to 1.736 Å in the reported Ca(BH4)2·2AB, and are slightly larger to those in AB (2.02 Å). This attests to a larger degree of interaction to BH4 groups as opposed to Ca(BH4)2·2AB, which could lead to a slightly modified decomposition mechanism. The closest refined BH···HN distance between NH3BH3 and adjacent hydridic H − of BH4 groups is 2.02 Å, which is in quite good agreement with the values of 1.986 and 2.037 Å reported for Ca(BH4)2·2AB. The main difference hence lies in intermolecular AB contacts, which are significantly larger in Ca(BH4)2·AB. Nevertheless, we need to recall that the determination of interatomic hydrogen distances from X-ray powder diffraction data are severely biased by the application of restraints during refinement, and the determined values need not be very accurate.

Synthesis
Ca(BH4)2, LiBH4 and NH4Cl were used as purchased from Sigma-Aldrich (Buchs, Switzerland). Chloride based samples were prepared by high-energy ball milling (mechanochemistry) in a Pulverisette P7 planetary ball mill (Fritsch, Idar-Oberstein, Germany). The ball-to-powder mass ratio was approximately 50, 60 milling cycles at 600 rpm of 2 min each were interrupted by 5 min cooling breaks to prevent heating and powder agglomeration on the walls of the milling jar. For the second approach based on NH4BH4, the latter was prepared according to ref. [20] with some minor modifications. Milling was performed in the Fritsch Pulverisette P7 planetary ball mill in a stainless steel vial cooled down to approximatively 240 K prior to milling. The milling speed was set to 500 rpm and the ball to powder mass ratio was approximately 50 for both dry and liquid assisted (500 μL (350 mg) of Et2O at 500 mg of sample mass) milling. The temperature of the milling jar was not kept low during milling. In the case of dry milling, the milling time was 30 min and the temperature of the jar was around 35 °C (308 K) at the end of the milling. In the case of the ether assisted milling, the milling time was of 5 min, resulting in a much smaller temperature increase during milling. The ether was removed by evacuating the sample directly after the milling.

Synchrotron Radiation X-ray Powder Diffraction (SR-XPD)
Synchrotron radiation powder X-ray diffraction (SR-PXD) data were collected at the Swiss Norwegian Beamlines of the European Synchrotron Radiation Facility (ESRF, Grenoble, France) and at the Materials Science Beamline of the Swiss Light Source (Paul Scherrer Institute, Villigen, Switzerland) 0.6888 and 0.8271 Å, respectively, as calibrated by a Si standard. Samples were loaded into 0.5 mm borosilicate capillaries. Structures were solved in direct space using the software FOX [21] and treating BH4 as a rigid body. After structure solution the respective models were refined with Fullprof [22] or TOPAS [23].

Conclusions
NH4Ca(BH4)3 is to our knowledge the first perovskite-type inorganic compound that contains two inherently dynamic hydrogen-rich molecules, both on the anion and cation sites. The usage of ammonium NH4 + as A-site cation allows stabilizing ammonium borohydride for the first time at room temperature and pressure. This describes an approach that is not restricted to perovskite-type borohydrides and the incorporation of NH4 + may be generalized to search for new hydrogen-rich compounds. Besides the chemical benefit to hydrogen storage applications, the ammonium cation also enriches the spectrum of structural tools by a further parameter, providing, next to the non-spherical borohydride anion, a cation with tetrahedral charge distribution. The combination of non-spherical ions with chemically distinct hydrogen species may be exploited in complex hydrides to design polar structures.
NH4Ca(BH4)3 decomposes at 380 K to new ammonia-borane complex Ca(BH4)2·NH3BH3 and releases one hydrogen molecule in the process. A major drawback of boron and nitrogen-based hydrogen storage systems is the possible evolution of ammonia NH3 and diborane B2H6, which both lead to the failure of fuel cells. Neither of them are observed during the hydrogen release of NH4Ca(BH4)3, while minor NH3 evolution occurs during decomposition of Ca(BH4)2·NH3BH3, similar to previously reported AB-complexes of metal-borohydrides. NH4Ca(BH4)3 is the first compound of a series of novel metal borohydrides which is expected to grow quickly thanks to a concept that is easily applied to a large and growing number of mixed-metal borohydride salts that contain cations similar in size and thus readily replaced by NH4 + . For instance, replacing NH4 + for K in K3Mg(BH4)5 [24] or KAl(BH4)4 [25,26] would result in hydrogen densities of 21.2 and 19 wt %, which is already very close to NH4BH4 itself. Generally, the decomposition of such proposed compounds is assumed to be dominated by the transformation of NH4 + and BH4 − to AB-complexes. While this mechanism is bound to release pure hydrogen, the subsequent decomposition of AB-complexes poses the known problems of rehydrogenation and fuel cell contamination.