Ethanol- and Methanol-Coordinated and Solvent-Free Dodecahydro closo -Dodecaborates of 3 d Transition Metals and of Magnesium

: Magnesium and 3 d transition metals closo -borates were prepared by mechanosynthesis (ball milling) of the mixtures Na 2 B 12 H 12 + MCl 2 (M = Mn, Fe, Co, Ni, Mg), followed by addition of ethanol or methanol and drying under dynamic vacuum. The dead mass of NaCl is partly removed by ﬁltration. The crystal structures of solvent-coordinated and solvent-free closo -borates have been characterized by temperature dependent synchrotron radiation X-ray powder di ﬀ raction, ab initio calculations, thermal analysis and infrared spectroscopy. Various solvated complexes containing six, four, three, two or one solvent molecules were obtained by successive removal of the solvent until in most case the solvent-free metal closo -borates were obtained with the exception of Mg whose hypothetical crystal structure, however, could have its prototype in MnB 12 H 12 . The 3 d transition metal closo -borates were studied in the view of their potential use as Na- or Li-ion battery electrodes in combination with Na or Li closo- borate solid electrolytes. The metal oxidation state (II) obtained in compounds presented here does not allow such application.

Our interest in TM closo-borates started with studies of closo-borates as solid electrolytes [13][14][15][16][17][18][19]. The TM closo-borates were considered as materials for electrodes in Li-and Na-ion solid-state batteries thus providing chemical compatibility of electrode/electrolyteinterface. We have recently reported a method of nickel closo-borate synthesis by mechanochemistry from precursors Na 2 B 12 H 12 and NiCl 2 followed by hydration, crystallization and drying [20]. The precursor Na 2 B 12 H 12 is usually prepared by synthetic routes involving toxic B 2 H 6 , or expensive B 10 H 14 [9]. As an alternative, there exists a method of Na 2 B 12 H 12 synthesis from non-toxic and cheap NaBH 4 [21]. We have then applied the synthetic method to closo-borates of 3d transition metals and of magnesium using water as a solvent [22]. Only a low oxidation state (II) of the TMs has been stabilized in these compounds, which are therefore not applicable as insertion electrodes. In this manuscript, we report on closo-borates of 3d transition metals and of magnesium synthesized using ethanol (EtOH) or methanol (MeOH) as a solvent. We show that the oxidation state of TMs is (II), and that a removal of coordinating solvent molecules by drying is an easy way of preparing the solvent-free closo-borates for most 3d TMs.

Synchrotron Radiation X-Ray Powder Diffraction (SR-XPD)
The data used for crystal structure and refinements were collected at the Swiss Norwegian Beamlines of ESRF (European Synchrotron Radiation Facility, Grenoble, France) between 25 and 400 • C (T-ramps). SR-XPD were recorded on Dectris Pilatus M2 detector at wavelengths of 0.69425, 0.7225, 0.77936, 0.7849 or 0.8187 Å calibrated with NIST SRM640c Si standard. The 2D images were integrated and treated with the locally written program Bubble.
For all measurements, the samples were sealed into borosilicate capillaries of diameter 0.5 mm (under argon atmosphere), which were spun during data acquisition, and the temperature was controlled with a hot air blower calibrated with silver thermal dilatation. The dynamical vacuum was used to eliminate the solvent molecules in the capillaries during the measurement contrary to the measurements labelled as "closed system" where the capillaries were only heated. Crystal structures were solved ab-initio using the software FOX [23] and refined with the Rietveld method using TOPAS [24]. Closo-borate anion B 12 H 12 2was modelled as rigid body with ideal icosahedral shape and with corresponding B-H and B-B distances. The interatomic distances and coordination polyhedra were analyzed using programs DIAMOND [25] and PLATON [26].

Fourier Transform Infrared Spectrosopy (FTIR)
FTIR experiments were performed using a Bio-Rad Excalibur Instrument between 600 and 4000 cm −1 with a spectral resolution set to 1 cm −1 .

Differential Scanning Calorimetry (DSC) and Thermogravimetry (TG)
DSC/TG data were measured using a NETZCH STA449 F3 Jupiter instrument from 30 • C to 350 • C. The measurements were performed under inert atmosphere of nitrogen with a purge rate of 20 mL/min. The sample were contained in sealed Al crucibles which were drilled before closing the DSC to allow solvent molecules to escape during the heating. The heating rate was 10 • C/min. Approximately 2 mg of sample was used for each run.

Ab Initio Calculations
In order to gain some insight into charge distribution within EtOH or MeOH coordinated and solvent free samples; also to determine hydrogen atoms positions and confirm the crystal structures density functional theory (DFT) calculations were performed within the spin polarized plane wave DFT method implemented in Vienna ab initio Simulation Package (VASP) [27]. The atoms were replaced with projected augmented wave potentials [28] with the valence electron configurations (1s 1 ) for H, (2s 2 2p 1 ) for B, (2s 2 2p 4 ) for O, (3d 8 4s 2 ) for Ni, (3d 8 4s 1 ) for Co, (3d 7 4s 1 ) for Fe, (3s 2 ) for Mg and (3d 6 4s 1 ) Crystals 2019, 9, 372 3 of 15 for Mn. The gradient-corrected (GGA) exchange-correlation functional and the non-local correction accounting for a weak dispersive interactions were applied [29,30]. In order to account for strong correlation effects in the valence orbitals of transition metals the GGA+U method was used with Hubbard U = 5.0 eV [31]. We have assumed antiferromagnetic ordering for all crystalline structures. Detailed analysis of the magnetic properties are beyond the scope of the present study.
For each crystal structure atomic positions were relaxed according to conjugated gradient (CG) algorithm. The lattice parameters were not optimized, and they were frozen at experimental values. For configurations with partial occupancies we have constructed at least two independent configurations with all sites fully occupied. Each of these structures was independently optimized and the one with lower ground state energy was used for structural analysis and charge density analysis. The later was performed with Bader method [32]. In order to improve the accuracy the charge density sampling grid was doubled for the charge analysis.

Phase Analysis by SR-XPD and DSC/TG
The thermal stability and quantitative phase composition were investigated by SR-XPD and DSC/TG. Selected DSC/TG curves are shown in Figures 1 and 2. As already discussed in [20,22] the temperature-dependent phase evolution is influenced by the gas pressure in the sample container. While the calorimetry was performed in an open crucible under nitrogen flow, the SR-XPD data were collected in a closed system or under dynamic vacuum using glass capillaries. The difference in experimental conditions lead to discrepancies observed between DSC/TG and SR-XPD experiments.
All crystal structures were solved ab initio from SR-XPD data, followed by the assignment of structure types. Selected representatives for each structure type were validated by DFT calculations. Space groups (corroborated by DFT calculations) and experimental lattice parameters are summarized in Table 1. Diffraction data at variable temperature are shown for the cobalt (Figures 3 and 4)  strong correlation effects in the valence orbitals of transition metals the GGA+U method was used with Hubbard U = 5.0 eV [31]. We have assumed antiferromagnetic ordering for all crystalline structures. Detailed analysis of the magnetic properties are beyond the scope of the present study. For each crystal structure atomic positions were relaxed according to conjugated gradient (CG) algorithm. The lattice parameters were not optimized, and they were frozen at experimental values. For configurations with partial occupancies we have constructed at least two independent configurations with all sites fully occupied. Each of these structures was independently optimized and the one with lower ground state energy was used for structural analysis and charge density analysis. The later was performed with Bader method [32]. In order to improve the accuracy the charge density sampling grid was doubled for the charge analysis.

Phase Analysis by SR-XPD and DSC/TG
The thermal stability and quantitative phase composition were investigated by SR-XPD and DSC/TG. Selected DSC/TG curves are shown in Figures 1 and 2. As already discussed in [20,22] the temperature-dependent phase evolution is influenced by the gas pressure in the sample container. While the calorimetry was performed in an open crucible under nitrogen flow, the SR-XPD data were collected in a closed system or under dynamic vacuum using glass capillaries. The difference in experimental conditions lead to discrepancies observed between DSC/TG and SR-XPD experiments.
All crystal structures were solved ab initio from SR-XPD data, followed by the assignment of structure types. Selected representatives for each structure type were validated by DFT calculations. Space groups (corroborated by DFT calculations) and experimental lattice parameters are summarized in Table 1 Figures S1-S37).
The phase evolution of as prepared samples with temperature as observed by SR-XPD under dynamic vacuum and in closed system is schematically shown in Figure 5. In the following, we discuss the crystalline phase evolution in dried samples, the corresponding crystal structures are presented in Section 3.2.   The phase evolution of as prepared samples with temperature as observed by SR-XPD under dynamic vacuum and in closed system is schematically shown in Figure 5. In the following, we discuss the crystalline phase evolution in dried samples, the corresponding crystal structures are presented in Section 3.2.       In all studied systems, i.e., EtOH and MeOH syntheses with all metals, the phase observed in as prepared samples is the complex with six ligands. The presence of this complex in the systems Fe-EtOH and Ni-EtOH is not proved by structural analysis, but is very probable as the phases formed by thermolysis of as prepared samples contain less than six ligands. The phase obtained by the thermolysis of hexa-complex in EtOH systems is usually a complex with four ligands. The Mn-EtOH system is an exception where the tetra-complex is not observed, and in some systems it may be present as an unknown phase. The next phase obtained by the thermolysis is the three ligands complex, which is formed in both, EtOH and MeOH syntheses. In some systems, the synthesized sample contains two or three phases, i.e., the complexes with six, four and three ligands. The coexistence of the complexes in as prepared samples is explained by different solvation rates, as the amount of added solvent is not precisely measured. In the systems with Mn and Fe the successive removal of ligands by thermolysis proceed further, and the complexes with two and one ligands are formed. The mono-complex in Mn-EtOH system contains also one molecule of EtOH that does not coordinate the metal.
The unsolvated metal closo-borates have been obtained only for Mn-EtOH, Fe-MeOH, Ni-MeOH and both Co-systems by further heating under dynamic vacuum. The desolvation was, however, complicated by the reaction with NaCl if the latter was not removed by filtration. It is especially the case of EtOH systems where the formation of Na6MnCl8, Na6FeCl8 and Na2Fe3Cl8 was observed. The absence of unsolvated MgB12H12, not formed also in the water assisted synthesis [22], is understood as its instability in a broad interval of temperatures.
The evolution of the phases with temperature differs slightly from the situation in water assisted systems [20,22]. While the first complex formed in all syntheses is always the octahedral hexacomplex, the next complex obtained by the thermolysis is the tetra-complex in water and some EtOH systems, while it is absent in all MeOH assisted systems. The tri-complex is then absent in water assisted syntheses. The crystal structure of the complexes varies with the ligand (see below).
The thermal stability of the complexes depends rather on the ligand than on the metal. While all EtOH complexes are the lest stable, the Mn and Fe complexes with MeOH are more stable than those with water, and for Co, Ni and Mg complexes the relation is inversed. The lower stability of EtOH complexes compared with MeOH is surprising comparing the boiling point of EtOH and MeOH. The Mg-complexes are the most stable among all studied metals. One can also note the less rich polymorphism of hexa-complexes with EtOH and MeOH compared to the water. It can be explained by the ratio between ligand size and the size of closo-borate cage as analyzed in [22]. In all studied systems, i.e., EtOH and MeOH syntheses with all metals, the phase observed in as prepared samples is the complex with six ligands. The presence of this complex in the systems Fe-EtOH and Ni-EtOH is not proved by structural analysis, but is very probable as the phases formed by thermolysis of as prepared samples contain less than six ligands. The phase obtained by the thermolysis of hexa-complex in EtOH systems is usually a complex with four ligands. The Mn-EtOH system is an exception where the tetra-complex is not observed, and in some systems it may be present as an unknown phase. The next phase obtained by the thermolysis is the three ligands complex, which is formed in both, EtOH and MeOH syntheses. In some systems, the synthesized sample contains two or three phases, i.e., the complexes with six, four and three ligands. The co-existence of the complexes in as prepared samples is explained by different solvation rates, as the amount of added solvent is not precisely measured. In the systems with Mn and Fe the successive removal of ligands by thermolysis proceed further, and the complexes with two and one ligands are formed. The mono-complex in Mn-EtOH system contains also one molecule of EtOH that does not coordinate the metal.
The unsolvated metal closo-borates have been obtained only for Mn-EtOH, Fe-MeOH, Ni-MeOH and both Co-systems by further heating under dynamic vacuum. The desolvation was, however, complicated by the reaction with NaCl if the latter was not removed by filtration. It is especially the case of EtOH systems where the formation of Na 6 MnCl 8 , Na 6 FeCl 8 and Na 2 Fe 3 Cl 8 was observed. The absence of unsolvated MgB 12 H 12 , not formed also in the water assisted synthesis [22], is understood as its instability in a broad interval of temperatures.
The evolution of the phases with temperature differs slightly from the situation in water assisted systems [20,22]. While the first complex formed in all syntheses is always the octahedral hexa-complex, the next complex obtained by the thermolysis is the tetra-complex in water and some EtOH systems, while it is absent in all MeOH assisted systems. The tri-complex is then absent in water assisted syntheses. The crystal structure of the complexes varies with the ligand (see below).
The thermal stability of the complexes depends rather on the ligand than on the metal. While all EtOH complexes are the lest stable, the Mn and Fe complexes with MeOH are more stable than those with water, and for Co, Ni and Mg complexes the relation is inversed. The lower stability of EtOH complexes compared with MeOH is surprising comparing the boiling point of EtOH and MeOH. The Mg-complexes are the most stable among all studied metals. One can also note the less rich polymorphism of hexa-complexes with EtOH and MeOH compared to the water. It can be explained by the ratio between ligand size and the size of closo-borate cage as analyzed in [22].
tg-Mn(EtOH) 6    The formation of hexa-EtOH closo-borates was observed for Mn, Co and Mg. The complex forms probably also for Fe, but its structure is not yet solved. Their crystal structures all contain the octahedral complex M(EtOH) 6 2+ centered inside a closo-borate triangular bipyramid (B 12 H 12 ) 5 of hcp anion packing (Figure 6a). The arrangement of the ions has the topology of the hexagonal BN-b, i.e., the cations are displaced from the tetrahedral site of the ZnS-wurtzite type on the triangular face shared between two tetrahedra. The formation of hexa-EtOH closo-borates was observed for Mn, Co and Mg. The complex forms probably also for Fe, but its structure is not yet solved. Their crystal structures all contain the octahedral complex M(EtOH)6 2+ centered inside a closo-borate triangular bipyramid (B12H12)5 of hcp anion packing (Figure 6a). The arrangement of the ions has the topology of the hexagonal BN-b, i.e., the cations are displaced from the tetrahedral site of the ZnS-wurtzite type on the triangular face shared between two tetrahedra.

m-M(EtOH) 4 (B 12 H 12 ) (M = Co, Ni)
Heating under dynamic vacuum leads in two cases (Co and Ni) to the formation of closo-borates m-M(EtOH) 4 (B 12 H 12 ) containing square pyramidal complexes with four ethanol and one hydrogen ligands (Figure 7). The complex is located on a triangular face shared between two closo-borate tetrahedra (B 12 H 12 ) 4 of a close to hcp anion packing along the b-axis.       In the case of iron, the closo-borate m-Fe(EtOH) 3 (B 12 H 12 ) releases two ethanol molecules, and the coordination polyhedron changes from the octahedron Fe(EtOH) 3 H 3 to a triangular bipyramid Fe(EtOH)H 4 in o-Fe(EtOH)(B 12 H 12 ). The complex is also bridging two closo-borate anions and forming chains along the c-axis (Figure 9b). The closo-borate anions are packed in nearly regular square layers perpendicular to b-axis forming cubes (B 12 H 12 ) 8 containing ethanol molecules. Please, note that the experimental structure has contained the EtOH molecule as non-coordinating one, but the DFT optimization of the experimental structure moved the EtOH molecule in the coordination sphere of the iron.
Both structures, m-Mn(EtOH) 2 (B 12 H 12 )•EtOH and o-Fe(EtOH)(B 12 H 12 ) have the pcu topology of anion packing, and CsCl as a structural prototype.   The solvent-free closo-borate forms by further heating under dynamic vacuum only for manganese and cobalt and ethanol synthesis, and for nickel and iron in methanol synthesis. Its formation was hindered in the ethanol synthesis for iron and nickel samples by formation of ternary chlorides, and in the case of magnesium in both syntheses by melting of the sample. The Fe 2+ , Co 2+ and Ni 2+ borates are isostructural, and crystallize with bcc packing of anions as well as the disordered Cu + borate [22]. On the contrary, the disordered tg-Mn(B 12 H 12 ) crystallizes in a trigonal structure with close to hcp anion packing (Figure 6b). The octahedral complex MnH 6 is located inside a triangular bipyramid (B 12 H 12 ) 6 in a similar way as in tg-Mn(EtOH) 6 (B 12 H 12 ) with the topology of hexagonal BN-b (Figure 6a). In the experimental structure, the complex is disordered around the trigonal axis. The DFT modelling with one from the three equivalent positions led to the stable octahedral configuration (insert in Figure 6b). The Mn 2+ cation is now bridging two closo-borate anions, which explains its disordering from the center of the triangular face on three equivalent positions.

Discussion
The preparation of solvent free 3d transition metal closo-borates was the main target of this project in a view of their possible application as insertion electrodes. Solvent free magnesium closo-borate is, on the other side, interesting as a possible solid electrolyte conducting Mg 2+ ions. The former has been achieved by a post-milling solvent addition followed by annealing using any polar protic solvent such as water, ethanol and methanol. Drying under a dynamic vacuum results in a removal of solvent molecules for low boiling point solvents and in a crystallization of solvent free metal closo-borates. The NaCl dead mass may be partly removed by filtration using ethanol. All the prepared metal closo-borates contain the metal in the lowest oxidation state of (II), which unfortunately prohibits their use as insertion electrodes, as already discussed previously [22]. The attempts of syntheses using halides with a metal in higher oxidation state such as FeCl 3 were unsuccessful due to high reducing power of hydroborate. The simple and cheap synthetic procedure unfortunately has not produced magnesium solvent free closo-borate.
As already discussed, most of the solvent free 3d TM closo-borates crystallize with the bcc anionic sublattice. The only exception is tg-Mn(B 12 H 12 ) crystallizing with hcp anionic sublattice. It may be directly compared to the closo-borates of bigger alkali earths such as SrB 12 H 12 and BaB 12 H 12 where the cation is located in the tetrahedral site [33], and to CaB 12 H 12 where smaller cation is located on the triangular face shared by two tetrahedra [34]. Both coordination polyhedra are in agreement with Pauling limits of stability, which are based on the ratio of cation and closo-borate anion radii [35]. For even smaller Mn 2+ this simple calculation predict linear coordination, but is partly hindered by the coordination ability of the transition metal resulting in the coordination octahedron H 6 and Mn 2+ bridging two closo-borates. This leads to a bigger volume per formula unit compared to Ca 2+ and Sr 2+ contrary to their ionic radii that are bigger than Mn 2+ . For Mg 2+ the ionic radii ratio predict also the linear coordination by two closo-borates [33]. The tg-Mn(B 12 H 12 ) is therefore a probable model for the long time searched structure of MgB 12 H 12 . Recent calculations predict MgB 12 H 12 .structure that is stabilized by the van der Waals attraction between aromatic closo-borane anions rather than covalent bond between the metal and B 12 H 12 [36].
The coordination of the metal in solvated closo-borates has been verified by comparing their volume per formula unit (V/Z). The dependence of the experimental values as a function of number of solvent molecules is presented in the Figure 12. It shows linear behavior similar to that of water solvated closo-borates [22]. From the slope of the linear relation, we can obtain the volume per solvent molecule in the crystal. The values of 83(4) Å 3 for ethanol and 50(1) Å 3 for methanol are slightly lower than estimated molecules volume in liquid ethanol and methanol at RT of 97 and 67 Å 3 , respectively [37]. To shed more light on the bonding in the solvated metal closo-borates, we have calculated Bader charges on complex cations and on metallic centers using the DFT optimized structural models. The charges for octahedral hexa-complexes with water, MeOH and EtOH ligands are presented in the Figure 13a. The charges for other selected closo-borates are given in the Figure 13b.
Within the accuracy of the calculations, the positive charge on the whole hexa-complex does not change with the metal, i.e., with the number of 3d electrons (±0.02e for MeOH and EtOH ligands; ±0.06e for H2O); with increasing number of metal 3d-electrons charge of metal decreases. For Mg the charge on the metal is larger than on the whole complex, this means that the charge on octahedral set of ligands is negative. This gives ionic attraction between Mg and ligands, as Mg cannot make orbital hybridization easily and explains higher thermal stability of Mg complexes. Ligands are thus polarized, i.e., have dipole moment that is largest for the case of Mg. An important feature in the optimized structural models is an apparent H + … H − interaction between the protic hydrogen from the ligand and hydridic hydrogen from the closo-borate anion. However, as a main interaction stabilizing the crystal one should consider the ionic interaction between complex cation and complex anion. The H + … H − interaction then dictates the fine details of the structure.  To shed more light on the bonding in the solvated metal closo-borates, we have calculated Bader charges on complex cations and on metallic centers using the DFT optimized structural models. The charges for octahedral hexa-complexes with water, MeOH and EtOH ligands are presented in the Figure 13a. The charges for other selected closo-borates are given in the Figure 13b.
Within the accuracy of the calculations, the positive charge on the whole hexa-complex does not change with the metal, i.e., with the number of 3d electrons (±0.02e for MeOH and EtOH ligands; ±0.06e for H2O); with increasing number of metal 3d-electrons charge of metal decreases. For Mg the charge on the metal is larger than on the whole complex, this means that the charge on octahedral set of ligands is negative. This gives ionic attraction between Mg and ligands, as Mg cannot make orbital hybridization easily and explains higher thermal stability of Mg complexes. Ligands are thus polarized, i.e., have dipole moment that is largest for the case of Mg. An important feature in the optimized structural models is an apparent H + . . . H − interaction between the protic hydrogen from the ligand and hydridic hydrogen from the closo-borate anion. However, as a main interaction stabilizing the crystal one should consider the ionic interaction between complex cation and complex anion. The H + . . . H − interaction then dictates the fine details of the structure. To shed more light on the bonding in the solvated metal closo-borates, we have calculated Bader charges on complex cations and on metallic centers using the DFT optimized structural models. The charges for octahedral hexa-complexes with water, MeOH and EtOH ligands are presented in the Figure 13a. The charges for other selected closo-borates are given in the Figure 13b.
Within the accuracy of the calculations, the positive charge on the whole hexa-complex does not change with the metal, i.e., with the number of 3d electrons (±0.02e for MeOH and EtOH ligands; ±0.06e for H2O); with increasing number of metal 3d-electrons charge of metal decreases. For Mg the charge on the metal is larger than on the whole complex, this means that the charge on octahedral set of ligands is negative. This gives ionic attraction between Mg and ligands, as Mg cannot make orbital hybridization easily and explains higher thermal stability of Mg complexes. Ligands are thus polarized, i.e., have dipole moment that is largest for the case of Mg. An important feature in the optimized structural models is an apparent H + … H − interaction between the protic hydrogen from the ligand and hydridic hydrogen from the closo-borate anion. However, as a main interaction stabilizing the crystal one should consider the ionic interaction between complex cation and complex anion. The H + … H − interaction then dictates the fine details of the structure.