- © 2012 The Mineralogical Society
The effect of pressure on the naturally occurring hydroxide-perovskite stottite, FeGe(OH)6, has been studied in situ by micro-Raman spectroscopy to 21 GPa at 300 K. The ambient spectrum contains six OH-stretching bands in the range 3064–3352 cm−1. The presence of six non-equivalent OH groups is inconsistent with space group P42/n. In view of this inconsistency a new ambient structure determination of stottite from Tsumeb was carried out, but this did not allow the clear rejection of P42/n symmetry. However, a successful refinement was also carried out in space group P2/n, a subgroup of P42/n, which allows for six non-equivalent O atoms. The two refinements are of comparable quality and do not allow a choice to be made based purely on the X-ray data. However, taken with the ambient and 150 K Raman spectra, a good case can be made for stottite having P2/n symmetry at ambient conditions. On this basis, the pressure induced spectroscopic changes are interpreted in terms of a reversible phase transition P2/n ↔ P42/n.
Hydroxide-perovskites are A-site empty, perovskite-structured, charge-neutral frameworks of corner-linked B(OH)6 and B′(OH)6 cation octahedra in which all of the oxygen atoms form OH groups, giving a general stoichiometry BB′(OH)6. The H atoms form hydrogen bonds across the faces of the dodecahedral A-site cavity. These structures provide a rare opportunity to study the deformation of a perovskite-type framework in the absence of interactions with an A cation.
The framework can accommodate homovalent and heterovalent cations. All heterovalent hydroxide-perovskites, BB′(OH)6, have ordered frameworks in which different cations alternate on crystallographically non-equivalent sites, examples include schoenfliesite, MgSn(OH)6; burtite, CaSn(OH)6; stottite, FeGe(OH)6; and mopungite, NaSb(OH)6. The ordering of the heterovalent cations is dictated by bond-valence constraints on bridging O atoms, each of which is bonded to an H atom as OH. Disorder in the heterovalent structures is prohibited by under-bonding or over-bonding of these O atoms: e.g. Mg–O–Mg + O–H = 1.67 valence units (vu); Sn–O(H)–Sn + O–H = 2.33 vu; compared with Mg–O–Sn + O–H = 0.33 + 0.67 + 1 = 2 vu. The structures of burtite and stottite are shown in Fig. 1. Burtite and schoenfliesite are the only members of the group for which H atoms have been located (by neutron powder diffraction).
With one exception, mushistonite CuSn(OH)6 (tetragonal, P42/n), all known stannate hydroxide-perovskites are cubic (space group Pn3İ or Pn3İm) with two non-equivalent H atoms. Stottite, FeGe(OH)6 (Strunz et al., 1958; Strunz and Giglio, 1961; Ross et al., 1988) and mopungite, NaSb(OH)6 (Williams, 1985) are reported to be tetragonal in space group P42/n with three non-equivalent H atoms. Three natural homovalent hydroxide-perovskites have been described: bernalite, Fe(OH)3 (Birch et al., 1993; Welch et al. 2005); dzahlindite, In(OH)3; and söhngeite, Ga(OH)3. The last two have Im3İ symmetry, whereas bernalite is orthorhombic. The space group of bernalite remains uncertain, although it is primitive (e.g. Pmmn; Welch et al., 2005) and not Immm (Birch et al., 1993). Synthetic δ-Al(OH)3 (Matsui et al., 2011) is orthorhombic with space group P212121, and has a single Al site. Wunder et al. (2011) reported that the synthetic high-pressure “3.65 Å-phase”, MgSi(OH)6, is also an hydroxide-perovskite. Originally, the space group was determined as orthorhombic, Pnam, but it has been shown subsequently to be monoclinic P21 by density-functional theory (Wunder et al. 2012), or monoclinic P21/n by single-crystal XRD (Welch and Wunder, in press).
As expected, hydroxide-perovskites are much more compressible than perovskites, due to the absence of an A cation which is the primary stiffening component of the perovskite structure (Zhao et al., 2004). For example, bernalite Fe(OH)3 has K0 = 75 GPa (Welch et al., 2005), compared with GdFeO3 perovskite which has K0 = 182 GPa (Ross et al., 2004). Ross et al. (2002) determined the bulk and axial moduli of stottite (P42/n) to 8 GPa at 300 K by single-crystal XRD as follows: K0 = 78.4 GPa, (K0′ = 6.2); K0(a) = 81.3 GPa, (K0′ = 6.4); K0(c) = 73.3 GPa, (K0′ = 5.7). No evidence for transitional behaviour was observed in this pressure range. The bulk modulus of stottite is very similar to the bulk moduli of MgSn(OH)6 (72 GPa), ZnSn(OH)6 (73 GPa), Fe(OH)3 (78 GPa) and MgSi(OH)6 (76 GPa).
Zhao et al. (2004) developed a novel approach to predicting the high-P behaviour of perovskites that is based upon bond-valence analysis correlated with the volume behaviour (compressibility) of the dodecahedrally coordinated A site and the octahedrally coordinated B site. Briefly, their analysis of the published structures and bulk moduli of diverse 3:3 perovskites in the literature indicated that those with βA/βB >1 become more distorted with increasing P, whereas those with βA/βB <1 become less distorted and can undergo P-induced phase transitions. On this basis, as hydroxide-perovskites have vacant and A sites which are very compressible, we would expect them to distort primarily by A-site compression. However, it is conceivable that a limit to compression by rigid rotation of the octahedrally coordinated B site will be reached at some pressure because the non-bonded B–B′ distances shorten considerably in highly rotated structures. If there is such a limit to rigid rotation, we might expect a switch to B–O and B′–O bond shortening as the dominant compression mechanism. According to Zhao et al. (2004) such behaviour may lead to phase transitions at high pressure in low-symmetry structures. Hydroxide-perovskites provide a means of testing such ideas, with the additional interest provided by H-mediated interactions.
In this context, our motivation here is to recognize changes in the compression mechanism of hydroxide-perovskites, to understand their compositional dependencies and, ultimately, to identify their structural origins. We have chosen to study stottite, FeGe(OH)6, as it is tetragonal and has the potential to transform on compression and also because high quality natural crystals are available. No synthetic stottite analogues are available and no synthesis recipes have been reported.
The structure of stottite was first determined by Strunz and Giglio (1961) who reported it to be tetragonal in space group P42/n. This structure has three non-equivalent O atoms, which implies three non-equivalent H atoms, but the H atom positions have yet to be located. Ross et al. (1988), also reported P42/n symmetry, but expressed significant reservations about assigning tetragonal symmetry to stottite. However, they felt that the single-crystal XRD data did not fully justify assigning lower symmetry. Ross et al. (1988) noted that Strunz et al. (1958) reported a biaxial optical character for stottite. There is, consequently, some doubt about the true symmetry of stottite that X-ray diffraction may have difficulty in resolving.
The stottite crystal used in this experiment was taken from the same locality as that studied by Ross et al. (2002), from the Tsumeb mine in Namibia. This sample was generously donated for study by Mr William Pinch of Rochester, New York, USA. The crystal used in this study is a dark brown single-crystal fragment, 42 × 50 μm in size, which is free from inclusions and wedge-shaped with a thickness varying from 5–25 μm (Fig. 2). Unpolarized Raman spectra were collected at <300 K at atmospheric pressure, and at increasing and decreasing pressures at 300 K.
For the high-pressure experiment the crystal was mounted in a diamond-anvil cell (DAC) with a ruby sphere for pressure calibration. Synthetic diamonds from Sumitomo Europe with 0.6 mm culets and a stainless steel gasket pre-indented to 50 μm with a 150 μm hole drilled in it were used. The ruby sphere was 8.5 μm in diameter. Fluid helium was used as a pressure-transmitting medium at 0.2 GPa using the gas-loading technique described by Jephcoat et al. (1987).
All of the Raman spectra were recorded in 135° scattering geometry using a SPEX Triplemate equipped with a back-illuminated, liquid-N2-cooled CCD detector in the range of 100 to 4000 cm−1. The intrinsic resolution of the spectrometer is 1.5 cm−1 and calibrations are accurate to ±1 cm−1. For some of the bands the fitted peak position varied outside the intrinsic limits of accuracy due to the large full width at half maximum (FWHM) and/or strong band overlap. In particular, the OH-stretching bands at <3200 cm−1 at atmospheric pressure overlap and broaden significantly under pressure: their FWHM increases from 80–90 cm−1 to 130 cm−1 at 5.06 GPa limiting the accuracy in band position to ±15–20 cm−1.
The Raman spectra were excited by the 514.5 nm line of an argon ion laser, focussed to a 10 μm diameter spot on the sample. The laser power (<80 mW) was low enough to avoid heating the sample. Ambient Raman spectra were collected in air from different parts of the crystal fragment at varying beam orientations. Raman spectra at pressures of up to 21.1 GPa were obtained from the area marked by a red oval in Fig. 2. This area of the crystal surface appeared optically flat and, for the 135° scattering geometry of the DAC experiment, gave the best intensity and resolution for the OH-stretching modes. The frequency of each Raman band was obtained by fitting Voigtian line profiles.
In the low-temperature Raman experiment a single-crystal fragment of stottite was mounted in a liquid He cooled optical microscope cryostat (Microstat® He, Oxford Instruments) equipped with a Spectrosil water-free window. The crystal was pressed in indium to ensure good thermal contact with the cold finger and the temperature was controlled within ±0.5 K.
Single-crystal X-ray diffraction
A tabular crystal of stottite (0.086 × 0.189 × 0.202 mm) of good optical and diffraction quality from the same locality as that of the crystal studied by Ross et al. (2002) was mounted on an XcaliburE kappa-diffractometer operating at 50 kV/45 mA using MoKα radiation (λ = 0.71073 Å). A CCD-based EOS detector was used for data collection. Unit-cell parameters were refined from 2408 reflections having I > 7σ(I). More than a sphere of reflection intensities was collected from 3.8–30.1° θ to a resolution of 0.77 Å. A Lorentz-polarization correction was applied. A Gaussian absorption correction based upon face-indexing was used (μ = 10.39 mm−1) and the reflection intensities converted to structure factors using CrysalisPro software (Agilent Technologies). Structure solution and refinement were carried out using the SHELX suite of programs (Sheldrick, 2008).
Ambient Raman spectrum
The ambient Raman spectrum of stottite is shown in Fig. 3. The relative band intensities depend upon the orientation of the crystal relative to the incident laser beam. The minimum number of peaks needed to fit the spectra in the OH-stretching region is five: three broad bands centred at 3064, 3159, 3240 cm−1; and two narrow bands centred at 3320 and 3352 cm−1. However, the shape of the 3064 cm−1 band suggests two components. To clarify this, spectra were collected at temperatures <300 K. The spectrum at 150 K is shown in Fig. 4. The two-component structure of the 3064 cm−1 feature is evident (Fig. 4, bands 5 and 6). The remaining four bands are well resolved in the 150 K spectrum and correlate with their ambient temperature counterparts.
High-pressure Raman spectra
In fitting the high-pressure spectra in the OH-stretching frequency range of stottite we recognize that those spectra that are readily identifiable as related to the ambient spectrum are adequately fitted with five bands, although the broad low-wavenumber band actually consists of two unresolved bands (Fig. 4). Representative Raman spectra of stottite as a function of pressure are shown in Fig. 5. The variation with pressure of OH-stretching bands is shown in Fig. 6 and summarized in Table 1.
The most striking feature on compression is the appearance of three new OH-stretching bands above 9.5 GPa. Furthermore, the 9.5 GPa spectrum appears to contain the three new bands (indicated by blue arrows in Fig. 5b) as part of a mixed spectrum that also contains a major contribution from the original five-band spectrum. The 11.3 GPa spectrum contains only the three new bands, with no vestiges of the original stottite spectrum. On compression to 21 GPa the three bands shift to lower wavenumber. On decompression, the three-band spectra shift to higher wavenumber. Below 10 GPa all of the spectra are mixtures of the three band and the original low pressure stottite spectra. The final ambient spectrum collected with an open DAC, is almost identical to the initial ambient spectrum. Thus, despite the hysteresis, as evidenced by mixed spectra, the spectral changes, including the FWHM line widths, are fully reversible.
Unlike the dramatic spectral changes in the OH-stretching region, the framework modes of stottite show no major changes over the full compression cycle, which shows that the main structural framework is preserved up to 21 GPa. In the range 100–900 cm−1 all of the observed Raman modes have a positive pressure dependence (Fig. 5a; Table 1) with the strongest modes being tracked reliably up to the maximum pressure.
Single-crystal X-ray diffraction
Crystallographic data and information about the X-ray data collection and structure determination are summarized in Table 2. Systematic absences indicate P42/n symmetry. Reflection merging in Laue groups 1, 2/m, mmm, 4/m and 4/mmm gave Rint values of 0.035, 0.037, 0.093, 0.041 and 0.095, respectively. The unconstrained unit cell has a = 7.5454(5), b = 7.5553(4), c = 7.4728(6) Å, α = 89.983(5), β = 89.957(6), γ = 89.982(5)° and V = 426.00(5) Å3. The corresponding constrained tetragonal cell has a = 7.5520(1), c = 7.4694(2) Å and V = 426.01(2) Å3. Reconstructed pseudo-precession photos of hk0, h0l and 0kl sections of the reciprocal lattice of stottite were obtained using Unwarp in CrysalisPro (Fig. 7). The hk0 section has a strong a/2 subcell. No indications of twinning were evident in diffraction patterns.
Attempts to solve the structure by direct methods or Patterson synthesis were unsuccessful. The E-statistics suggested non-centrosymmetry (E = 0.741). However, attempts to determine the structure in the non-centrosymmetric space groups P42 and P4İ, both subgroups of P42/n, were unsuccessful. Consequently, as an initial way forward we used the atom coordinates of the P42/n structure reported by Ross et al. (1988) as a starting model. Full anisotropic refinement of the stottite structure in space group P42/n proceeded smoothly to give final agreement indices of R1 = 0.038 and wR2 = 0.086. No H atoms were evident in difference-Fourier syntheses. Atom coordinates, and Uij and Ueq values are listed in Table 3 and the structure is shown in Fig. 8. Bond lengths and polyhedral distortion indices (Robinson et al. 1971) are given in Table 4.
As indicated above, reflection merging in 4/m (Rint = 0.041) is clearly superior to merging in 4/mmm (Rint = 0.095). Consequently, merohedral twinning is unlikely to be present or significant. Although not evident from diffraction patterns, we nonetheless checked for twinning using PLATON (Spek, 2005) which indicated possible twins related by a 180° rotation about . For this twin law PLATON calculated a batch scale factor of only 0.08 (8%) for the minor component. We attempted a twin refinement in P42/n, starting with a batch scale factor of 0.2 and using as starting atom coordinates those of the refined untwinned structure. Refinement proceeded satisfactorily and used reflection weighting and an extinction correction in the final stages. The minor twin refined to a proportion of 9% (which is close to that indicated by PLATON). However, slightly poorer agreement indices were obtained (R1 = 0.040 and wR2 = 0.134) compared with the refinement with no twinning. No clear improvement was achieved by taking into account possible twinning. We therefore consider the case for twinning to be inconclusive, and it does not appear to affect the refined structure significantly.
As reflection merging in 2/m also gave a good agreement index (Rint = 0.037) we attempted a refinement in space group P2/n (Z = 4), a non-standard setting of P2/c, which is a subgroup of P42/n. We used the non-standard setting as a comparison with the P42/n structure. Data reduction to the (pseudo-tetragonal) monoclinic unit cell gives a = 7.5509(1), b = 7.4695(2), c = 7.5531(2) Å, β = 90.005(2)° and V = 426.01(2) Å3. As structure solution was unsuccessful, we used the Powdercell program to derive an initial unique model P2/n structure (Z = 4) for refinement, that is derived from a stottite-like pseudo-tetragonal structure. The refinement proceeded smoothly to give final values R1 = 0.044 and wR2 = 0.096 and a reasonable structure. No H atoms were found in the difference-Fourier maps. Experimental details of the refinement in P2/n are given in Table 2 and atom coordinates and displacement parameters in Table 3. The refined P2/n structure is shown in Fig. 8. It has a +−+ tilt system which is analogous to ++− of stottite. Compared with P42/n, the P2/n structure has all its atom sites doubled, with six non-equivalent oxygen atoms: Ge(1,2), Fe(1,2) and O(1–6). Thus, it is consistent with the six OH bands of the 150 and 300 K Raman spectra. Finally, it should be noted that refinement in the monoclinic space group P21/n failed. Thus, P2/n is the only plausible alternative to P42/n as the ambient structure of stottite; P2/n is preferred over P42/n on the grounds that it allows for six OH bands in the Raman spectrum.
The ambient and 150 K Raman spectra of stottite in the OH-stretching region are inconsistent with the P42/n structure determined by Strunz and Giglio (1961), Ross et al. (1988) and ourselves. In this respect the reservations expressed by Ross et al. (1988) are relevant: “… the unit-cell symmetry appeared lower than tetragonal – either orthorhombic or, more likely, monoclinic (based on subgroup relations) with β ~90°. For neither crystal was the evidence so convincing as to force refinement in a lower symmetry than P42/n, but the discrepancies were a trifle disturbing, particularly considering the anomalous optical properties previously mentioned.” Thus, there are grounds for uncertainty concerning the space group of stottite as determined by XRD.
Our refinement in P2/n converged to give a reasonable result, and the structure is topologically similar to that of P42/n stottite. The non-H P2/n structure has a clear tetragonal pseudosymmetry that may not be possible to resolve conclusively using X-ray diffraction. However, the P2/n result indicates that such a structure, which has six non-equivalent oxygen atoms, is consistent with the diffraction data. Thus, a critical link with the Raman spectra has been identified and this allows us to propose a phase transition from P2/n → P42/n on compression. Taking the XRD and Raman results together, we propose that stottite has the P2/n structure at ambient conditions.
Some insight into the structural subtleties of stottite and the potential difficulties associated with strong pseudosymmetries in hydroxide-perovskites, can be gained from a recent study of the synthetic hydroxide-perovskite MgSi(OH)6, also known as the “3.65 Å-phase”, by Wunder et al. (2011). Its structure was inferred by comparison with the X-ray powder diffraction pattern of δ-Al(OH)3 (Matsui et al. 2011), and using Rietveld refinement to obtain structural data in orthorhombic space group Pnam (Z = 2) as follows: a = 5.1900(3), b = 5.1132(3), c = 7.3304(4) Å and V = 194.53(2) Å3. However, two low-angle diffraction peaks of significant intensity were not fitted by the Pnam model. Furthermore, the infrared spectrum in the OH-stretching region of MgSi(OH)6 contains six absorption bands, and is inconsistent with Pnam symmetry (three non-equivalent OH groups).
Wunder et al. (2012) reported the results of a computational study of MgSi(OH)6 using density-functional theory (DFT) in which they proposed that the most likely space group was P21 (monoclinic), with six non-equivalent OH groups, this being consistent with the infrared OH-stretching spectra. This structure is metrically pseudo-orthorhombic with β = 90.03(1)°. Crucially, the P21 structure has an ordered alternation of Mg(OH)6 and Si(OH)6 octahedra, whereas the Pnam structure has a single mixed cation site, implying Mg–Si disorder.
A structure determination of MgSi(OH)6 using single-crystal XRD on coarse, well crystallized material synthesized at 10 GPa and 450°C for 205 h has been reported by Welch and Wunder (in press). The structure is monoclinic P21/n (Z = 2), but metrically pseudo-orthorhombic, with a = 5.11606(9), b = 5.19282(9), c = 7.33434(14) Å, β = 90.021(2)° and V = 194.850(6) Å3. However, a few weak violators of the n-glide were observed, consistent with a P21 component. Refinements of the structure in P21 were problematic and unstable (non-convergence; non-positive-definite values for several oxygen Uijs; and very large standard errors on the Flack parameter). Consequently, the X-ray-based structure has space group P21/n, which has three non-equivalent H atoms. The very weak X-ray signature of the P21 structure may reflect very similar non-H structures for the two phases, and also the fact that the primary distinction is due to H location, which is registered by Raman and infrared spectroscopy, but not by XRD.
A comparison of the Raman spectra of stottite and the infrared spectrum of MgSi(OH)6 (Wunder et al., 2011, fig. 2a) is instructive. The 150 K and ambient spectra of stottite have six OH-absorption bands and are qualitatively very similar to the ambient infrared spectrum of MgSi(OH)6. The wavenumber shifts between the two studies reflect the different nearest-neighbour cations bonded to OH groups: Fe–O(H)–Ge vs. Mg–O(H)–Si. The close correspondence of the stottite and MgSi(OH)6 spectra suggests that the OH groups of P21 (Z = 2) MgSi(OH)6 and P2/n (Z = 4) stottite occupy analogous local environments. Hence, by analogy with the high-pressure behaviour of stottite, a transition P21 → P21/n in MgSi(OH)6 on compression seems likely. Such a transition could be identified as a change from six to three OH bands in Raman or infrared spectra.
A further interesting subtlety concerning the transitional behaviour of stottite is the apparent hysteresis, particularly on decompression. We have no evidence for significant pressure variation within the DAC and the irradiated sample area was 0.01 mm in diameter, and so we infer that the observed mixed spectra, implying two coexisting phases, reflect the nature of the phase transition. Thermal hysteresis has been observed in first-order phase transitions in numerous materials, including perovskites. In the present case, the hysteresis involves pressure rather than temperature. The origin of hysteresis in stottite, and its possible existence in related hydroxide-perovskites, merits further study.
Conclusive structural data for stottite, at ambient conditions and at high pressure, are highly desirable, but obtaining these data from X-ray diffraction is not a trivial task. It appears that pseudosymmetry in some hydroxide-perovskites presents problems for definitive structure determination, even though high-quality structure refinements from XRD can be obtained. In H-rich structures, such as hydroxide-perovskites, with their extensive H-bonded connectivity throughout the octahedral framework, H behaviour may have a significant role in phase transitions. In this regard, spectroscopic studies are likely to provide key information on symmetry breaking in hydroxide-perovskites, as registered by the OH-stretching region.
This work was supported by Natural Environment Research Council fellowship NER/I/S/2001/00723 and Natural Environment Research Council grant NER/B/S/2003/00258 to AKK, and Natural Environment Research Council grants GT59801ES and GR3/10912 to APJ; MDW and WAC thank the ESRF for beam time.
- Manuscript received 19 January 2012.
- Manuscript Accepted for publication 8 March 2012.