Reaction of MnCl₂ and MnBr₂ with ammonia gas at room temperature led to white air-sensitive powders of the ammine complexes, 1 and 2 respectively. Raman spectra for 1 and 2 are shown in Figure 2 By comparison with published data for transition metal hexa-ammine complexes, the bands for 1 and 2 could be fully assigned to N–H and Mn–N vibrations (Table 3) [29,30,31,32,33]. Previous IR studies on the M(NH₃)₆I₂ (M = Mn, Fe) complexes have suggested that the rotation of ammonia molecules about the M-N axis is possible [34]. From the results presented here we can establish that this is also the case at least for Mn(NH₃)₆Br₂ owing to the degenerate (δ_d) band observed in the Raman spectra at 1590 cm⁻¹ and also the large H-H bond lengths suggested from XRD refinements (Table 3, Table 4, Table 5 and Table 6). Despite the apparent absence of the equivalent δ_d band in the Raman spectrum of 1, there are sizeable intermolecular distances between neighbouring H atoms in the ammonia molecules (indeed the partially occupied H positions in the model assume dynamic disorder) and no obvious structural reason why rotation should not occur, suggesting that the band is simply too weak in the spectrum of 1 to be observed. This is consistent with the study by Eβmann et al. who infer that the rotation of the ammonia molecules about the M-N axis in a number of transition metal ammine chlorides and bromides is free at room temperature [25]. Further, the rotational effects observed for the ammonia molecules in single crystals of Ni(NY₃)₆X₂ (Y = H, D and X = Br, I, NO₃, PF₆) as investigated by neutron diffraction are described in more detail by Schiebel et al. [35,36].

Initial search-matching procedures for the PXRD powder pattern of 1 revealed a match with that of the corresponding Mg analogue (PDF card No: 01-075-6756) and the pattern could be indexed to a cubic cell with a = 10.442 Å [9]. Although there is no refined crystal structure for Mn(NH₃)₆Br₂, Watt and Manhas reported XRD data (and elemental analysis) for the hexa-ammine formed by the reaction of MnBr₂ with liquid ammonia [37]. The PXRD data obtained for 2 matched well with the pattern of Fe(NH₃)₆Br₂ (PDF card No: 01-085-2095), which was generated using the Powdercell software package from single crystal structural data, and could thus be indexed to a cubic unit cell with a = 10.249 Å [25,38]. From the observed reflections, it was thus deduced that Mn(NH₃)₆Br₂ and Mn(NH₃)₆Cl₂ were isostructural. Subsequent Rietveld refinements for 1 and 2 were performed taking the previous single crystal X-ray structure of Mn(NH₃)₆Cl₂ and isostructural Fe(NH₃)₆Br₂ as the respective starting models [25]. The refinements of 1 and 2 were performed following an identical strategy in which the background coefficients (GSAS Function 8, a reciprocal interpolation function), scale factor and the unit cell parameters were varied in the initial cycles. The atomic parameters, peak profile parameters and isotropic displacement parameters (for non-hydrogen atoms) were varied subsequently. Modeling of the peak shapes used function 2, the Thompson-Cox-Hastings pseudo-Voigt function. The final crystallographic results for 1 and 2 are presented in Table 4, atomic parameters in Table 5 and interatomic distances angles in Table 6. Profile plots for the refinements are shown in Figure 3 and Figure 4.

Structure refinement against PXRD data confirmed that Mn(NH₃)₆Cl₂ (1) forms with the cubic K₂[PtCl₆]-type structure as previously reported from single crystal X-ray analysis [24]. Rietveld refinement has also established for the first time that Mn(NH₃)₆Br₂ also crystallizes with the K₂[PtCl₆]-type structure and is thus isostructural with the chloride and Fe(NH₃)₆Br₂ (Figure 4) [25]. The hexa-ammine structures can be considered to be composed of a FCC lattice of isolated octahedral complex ammoniate cations, i.e., [Mn(NH₃)₆]²⁺, with halide anions, X (X = Cl⁻, Br⁻), occupying the tetrahedral interstices of the unit cell in a distorted close-packed cubic configuration (Figure 1) [39]. Although these [Mn(NH₃)₆]²⁺ octahedra are effectively isolated, Sørensen et al. haveevoked the concept of weakly bound-[Mn(NH₃)₆]–X—“chains” in rationalizing the ammonia sorption mechanism [23]. These chains are thus composed of metal octahedra connected along the chain length via the interceding halide anions. Alternatively, the cubic hexa-ammine structure could be viewed as a simple cubic lattice of halide anions with nitrogen occupying the faces of each halide cube and manganese located in the body centers of alternate halide cubes.

We observe that the cubic structure expands as a result of the replacement of Cl⁻ by Br⁻ as would be anticipated from ionic radii arguments [40,41]. Previously reported structural models of the hexa-ammines have defined partially occupied hydrogen positions (96 k sites that are 75% occupied) reflecting the dynamic disorder of the ammonia ligands in the octahedral cations. Although it would not be expected that powder X-ray diffraction methods could resolve the positions and disorder of the hydrogen with any great accuracy, we were tentatively able to vary the hydrogen coordinates and achieve stable, convergent refinements with reasonable N–H bond distances. The Mn–N distances in the complex cations are similar to those observed for other transition metals in equivalent ammoniates and the bond length increases slightly from chloride to bromide [25].

The TGA results collected to 773 K for 1 and 2 (Figure 5, Figure 6 and Table 7) show a 3-stage mass loss in each case which, by simultaneous MS analysis of the gases released, is associated solely with the loss of ammonia. The DTA data indicate four distinct endothermic events with minima at 379.3 K, 490.6 K, 556.7 K and 574.5 K respectively for 1, and three distinct endothermic events at 407.5 K, 516.6 K and 565.2 K for 2 that correlate to each mass loss for both complexes. Careful isolation of the di- and mono-ammoniated compounds using TG-DTA has been possible by ceasing the heating process at the appropriate stage of decomposition. For the chloride complex, TGA analysis to 398 K and 498 K allowed isolation of the di- and mono-ammoniated decomposition products respectively. Slightly higher temperatures were used to isolate the di- and mono-ammoniates for the bromide; 423 K and 523 K respectively. The increase in deammoniation temperatures from the 1 to 2 is unsurprising upon consideration of the respective enthalpies for the ammines in Table 2.

Reaction Scheme 1 hence describes the anticipated route by which the chloride and bromide ammoniates release ammonia in accordance with the TPD data observed in this study. These are comparable with data observed in literature for various dichloride ammines [9]. The TGA profile in Figure 6 suggests that weight loss for 1 occurs almost immediately (marginally above room temperature) and in contrast to previous work, ammonia desorption is complete below 600 K. The bromide, 2, begins to desorb ammonia at a higher temperature than the chloride commensurate with its higher desorption enthalpy (Table 2), but this onset still lies below 373 K.

Although ostensibly both halides exhibit a three-step loss of ammonia with increasing temperature, it is evident from close examination of the chloride DTA profile (Figure 5) that the transformation from the presumed mono-ammine to the deammoniated chloride, MnCl₂, appears to progress through two closely spaced (almost overlapping) thermal events. Indeed, a change in the slope of the TGA profile at the same temperature as the double minimum in the DTA trace (556.7 and 574.5 K) would also suggest that the loss of the final ammonia molecule is not a stoichiometric process from mono-ammine to MnCl₂. This TG-DTA profile is in agreement with two (or more) gas releases from nominal Mn(NH₃)Cl₂ measured by Sørensen et al., although the equivalent desorption in their study appeared to occur at a temperature ca. 80 K above that seen herein [9]. By contrast, the temperature range over which the dibromide deammoniates is narrower, which results in the successive deammoniation steps being less widely spaced (in temperature and time) and, for example a stability domain for the nominal mono-ammine of only several K. Although the final desorption step has not yet been resolved into more than one event in either the DTA or TGA traces, the asymmetry of the DTA peak and the fluctuations in slope of the TGA trace would suggest an analogous non-stoichiometric decomposition from mono-ammine to dihalide to that observed for 1. A mechanism for ammonia absorption/desorption for M(NH₃)_nCl₂ (M = Mg, Mn, Ni) ammines was proposed by Sørensen et al. using DFT calculations. This model indicated that ammonia molecules could enter and leave the bulk interface via the zipping/unzipping of the loosely-bound “chains” of octahedral ML₆ groups (where L could be either NH₃ and/or Cl depending on n) to allow conservation of the metal coordination geometry independent of n. This model necessarily also involves structural rearrangement in the bulk after each successive sorption step. In the later deammoniation steps, for example, it would be necessary not only for Cl to substitute for NH₃ in each ML₆ octahedron, but also for the resulting octahedra to condense and form edge sharing chains. The structures of the lower Mn ammines are discussed in more detail below, but the non-stoichiometric decomposition from mono-ammine to dihalide observed above could be rationalized in the intermediate loss of one NH₃ unit for every two Mn centers in the edge-sharing octahedral clusters in Mn(NH₃)X₂ and a process as described in Scheme 2:

Decomposition of 1 and 2 to 398 K and 423 K respectively in the TG-DTA experiments led to small samples (ca. 50 mg) of air-sensitive, light pink and light orange/brown powders (3 and 4) respectively. Decomposition of 1 and 2 conducted to higher temperatures, 498 and 523 K respectively, led to air-sensitive, pink and light orange/brown powders to give the nominal mono-ammoniated complexes, 5 and 6 respectively. After weight losses were complete for both systems a blackened solid mass (7) and an orange/brown powder (8) respectively were the result. Raman spectra for 3 and 4 are shown in Figure 7a,b respectively, and assignments are given in Table 8. The bands for 3 and 4 have been assigned to N–H and Mn–N vibrations based on the premise that similar bonding modes exist in both the hexa- and di-ammines. The presence of the weak degenerate δ_d band at 1599 cm⁻¹ in 4 would suggest that there is similar dynamic disorder (rotation) of NH₃ groups in the diammines to that observed in the hexammines (and again this band is observable in the bromide but not the chloride).

Initial search-matching procedures for the PXRD powder pattern of 3 revealed a match with the Mg analogue (PDF card No: 01-073-4070) and the pattern could be indexed to an orthorhombic unit cell with a = 8.180 Å, b = 8.200 Å, c = 3.750 Å [41]. From the observed reflections, it was deduced that 3 was isostructural with Fe(NH₃)₂Cl₂ and Mg(NH₃)₂Cl₂ and matched well to a powder pattern generated using Powdercell in which Mn was placed on the Fe positions of the orthorhombic Cmmm Fe(NH₃)₂Cl₂ structure [15]. Inspection of patterns for 4 revealed a match with that of the corresponding Mg analogue (PDF card No: 01-089-6789) [38,41]. The pattern for 4 could be indexed to an orthorhombic unit cell with a = 5.944 Å, b = 11.876 Å, c = 3.983 Å and it was apparent that 4 was isostructural with Fe(NH₃)₂Br₂ and Mg(NH₃)₂Br₂ (orthorhombic Pbam) [16,38,41].

Subsequent Rietveld refinements for 3 and 4 were performed taking the previous single crystal X-ray structures of the relevant isostructural iron ammines as the respective starting models [16]. The refinements were conducted using PXRD data collected from the di-ammine products from the TG-DTA experiments,i.e., 3 and 4, which were isolated by heating 1 and 2 to 398 K and 423 K respectively. The refinements of 3 and 4 were performed following a similar sequence in which the background coefficients (GSAS Function 8, a reciprocal interpolation function), scale factor and the unit cell parameters were varied in the initial cycles. The peak profile parameters, atomic parameters including isotropic displacement parameters (for non-hydrogen atoms) were varied subsequently. Modeling of the peak shapes was performed using the Thompson-Cox-Hastings pseudo-Voigt function (function 2 in GSAS). It was not possible in the refinements of either 3 or 4 to refine the positions or displacement parameters of the hydrogen atoms, so these were fixed at values from the literature models of the equivalent iron compounds [16]. The final crystallographic results for 3 and 4 are presented in Table 9, atomic parameters in Table 10 and interatomic distances and angles in Table 11. (STA data for the isolation of the di-ammines and profile plots for the refinements can be found in the supplementary material).

The unit cell parameters of 3 are slightly larger than those in both the isostructural magnesium and iron chloride ammines as would be expected by comparison of the respective ionic radii of high spin Mn²⁺vs. Mg²⁺ and high spin Fe²⁺ (and as is also the case for the respective hexa-ammines) [37,38,40,41]. A representation of the structure of 3 is given in Figure 8. The structure is composed of edge sharing Mn(NH₃)₂Cl₄ octahedra (which can be termed as chains in which the trans-NH₃ groups are thus terminal and the halide groups are bridging) which propagate parallel to the [001] direction. The chains are effectively aligned parallel in the (110) plane throughout the structure with each chain translated by half the [001] face diagonal with respect to its nearest neighbour. Leineweber et al have also rationalised the orthorhombic structure in terms of CsCl-type motifs and described isostructural Mg(NH₃)₂Cl₂ as a brick-like arrangement of double CsCl units in the ab plane [38,41]. The Mn–N bond length in 3 is broadly similar to those in the hexa-ammmines, 1 and 2, and close to those observed in the equivalent magnesium and iron compounds [16,38,41].

4 is isostructural with Fe(NH₃)₂Br₂ and Mg(NH₃)₂Br₂ (Figure 9) [16,38,41]. As with 3, the unit cell of 4 is larger than the Fe and Mg analogues. Similar to 3, 4 is composed of chains (where Br replaces Cl as X). The key difference in the structures of 3 and 4 lies in the relative arrangement of these edge sharing octahedral chains. As Bremm and Meyer point out, the packing of the chains in the Pbam structures is hexagonal as opposed to the tetragonal sequence in the Cmmm compounds [16]. Hence, successive rows of edge-sharing octahedra running parallel to the [110] direction are rotated by 90°. Following the double CsCl unit notation used by Leineweber et al, the units are arranged in a herringbone pattern in the ab plane in the bromide (as opposed to the brick-like arrangement in the chloride). Mn–X distances are longer in 4 than 3 as would be expected on coordination of the metal to the larger halides [37,40]. The Mn–N distance in 4 is shorter however and comparable to that observed in Fe(NH₃)₂Cl₂. Inevitably from the mismatch in Mn–N and Mn–X bond lengths in both 3 and 4 the MnX₄(NH₃)₂ octahedra are axially compressed in both compounds. We are unable to draw any firm conclusions regarding the positions or dynamics of the hydrogens from our relatively limited models from PXRD. The Mg(ND₃)₂X₂ compoundshave been studied by neutron diffraction to resolve such issues and a similar approach would be required in the Mn–NH₃–X systems to understand the relationships of NH₃ rotational disorder with halide and temperature [42].

Further de-ammoniation of 1 and 2 to yield the anticipated mono-ammoniated complexes 5 and 6 was performed although isolation of these phases is yet to be achieved owing to the apparent narrow stability range (prior to further desorption). In the available literature, among the transition metal ammines only the Fe(NH₃)Cl₂ and Ni(NH₃)X₂ (X = Cl, Br) have been crystallographically characterized giving hexagonal P6₃/mmc and monoclinic I2/m structures for the Fe and Ni compounds respectively [16,43]. Further heating of 1 and 2 gave rise to 7 and 8 which were amorphous by PXRD. Raman spectra for 7 and 8 showed no notable features over the measurement range and none of the bands that were present in 1–4 from which one could infer that deammoniation was complete as suggested by the thermogravimetric data.