Thermal conversion of the hydrous aluminosilicate LiAlSiO₃(OH)₂ into γ-eucryptite

2.1 Thermal behavior of LiAlSiO₃(OH)₂

As described in the Experimental Section, LiAlSiO₃(OH)₂ is obtained as fine aggregated needles with lengths up to several μm. Earlier thermogravimetric (TG) investigations showed that the dehydration of LiAlSiO₃(OH)₂ takes place in the temperature interval 350–500 °C, with the associated weight loss in very good agreement with the stoichiometric amount of water (12.5%). Above 550 °C there was no further weight change. During scanning electron microscopy studies of this material it was noticed that dehydration can also be induced in the electron beam and that the anhydrous product retained the needle-like crystal shape of LiAlSiO₃(OH)₂. We then studied the evolution of the dehydration on individual LiAlSiO₃(OH)₂ crystals in the transmission electron microscope (TEM, Figure 2d–f). Upon beam exposure the dimension of the needle-shaped crystals is reduced by about 10% and at the same time the edges become rounded. The process can also be followed with energy-dispersive X-ray (EDX) analysis and seems to be completed after 25–30 min. After beam irradiation the material appears to be amorphous.

Figure 2:

TEM bright field images of LiAlSiO₃(OH)₂ sample before (a) and after heat treatment at 500 °C (b) and 800 °C (c). Shrinkage of a LiAlSiO₃(OH)₂ crystal upon dehydration (e–f) as observed in TEM investigation after respectively 0, 275 and 1434 s of continuous beam exposure. The inset shows a selected area diffraction pattern of the amorphized particle after the investigation.

Figure 3 shows the dehydration of LiAlSiO₃(OH)₂ as observed in a multi-temperature powder X-ray diffraction (PXRD) experiment. In good agreement with the TG analysis, the diffraction pattern of LiAlSi₃(OH)₂ is maintained up to 300 °C. At 400 °C dehydration is apparent by a changed pattern (which still contains reflections from a minor fraction of LiAlSiO₃(OH)₂). At 500 °C a pure pattern of converted phase is seen, which is distinguished from the pattern of the precursor by the presence of a low angle reflection at 2θ ≈ 12° (d ≈ 7.7 Å) and two peaks in the 2θ range 27–29°. This new pattern then appears unchanged up to 700 °C. In the pattern at 800 °C clear changes can be noticed (e.g. the Bragg peak at high d is absent and in the 2θ range 27–29° there is a single peak), which manifest an additional structural transition after dehydration. This pattern coincides with that of γ-eucryptite. Upon cooling there was no further structural change.

Figure 3:

Dehydration of LiAlSiO₃(OH)₂ as observed in PXRD investigations (Cu Kα radiation).

Attempts to index the PXRD pattern resulted in possible monoclinic and triclinic unit cells. However, attempts to solve the structure from the PXRD data or from continuous rotational electron diffraction (cRED) analysis were unsuccessful. Figure 2a–c compares TEM images of the original LiAlSiO₃(OH)₂ sample with that of the sample treated at 500 and 800 °C. While all of the samples exhibit needle-like morphology and a lack of uniform particle size distribution, the sharpness of the edges changes upon heating. Well-defined edges of the initial hydrous aluminosilicate crystals clearly become more rounded.

In Figure 4 the attenuated total reflection Fourier-transform infrared (ATR-FTIR, or simply IR hereafter) spectra of original LiAlSiO₃(OH)₂ are compared with that of the samples treated at 500 and 800 °C. Generally, IR spectra of lithium aluminosilicates are characterized by the bands of the framework of tetrahedra. There are two distinct regions where T(Si, Al)–O stretches and Si–O–Al inter-tetrahedral bending modes appear (i.e. bending of T–O–T angles), respectively. For LiAlSiO₄ glass these regions have broad bands with absorption maxima at 936 and 695 cm⁻¹. In the crystalline materials α- and β-eucryptite these modes are resolved into more or less narrow bands. Clearly, the IR spectrum of the 800 °C product (i.e. γ-eucryptite) follows this pattern. LiAlSiO₃(OH)₂ is a structurally different material and its IR spectrum was described earlier [10]. There are distinct O–H stretches from structural hydroxy groups in the spectral region 2900–3800 cm⁻¹. Further, the T–O–T bending bands at around 700 cm⁻¹ are very weak because of the small proportion of corner-condensed TO₄ tetrahedra in the structure. The IR spectrum of the product immediately after dehydration (i.e. the 500 °C treated sample) shows sharp bands in the 400–1250 cm⁻¹ region, characteristic for a well crystallized product. Their rather large number indicates a larger unit cell and/or low space group symmetry of the structure. Bands at around 700 cm⁻¹ manifest the presence of a framework with corner-connected tetrahedra (and thus the absence of AlO₆ octahedra). There are also weak but significant bands in the range 2800–3000 cm⁻¹ (which are at wavenumbers too low for O–H stretches unless they reflect a very strong hydrogen bond environment [11]) and in the range 1200–1600 cm⁻¹ (that could relate to O–H bending). At this point, there is no further information about the structure of the initial dehydration product. As mentioned above, all attempts to solve the structure – either from PXRD or electron diffraction data – failed.

Figure 4:

ATR-FTIR spectra of LiAlSiO₃(OH)₂ before and after heat treatment at 500 and 800 °C.

We note a striking (but coincidental) relationship with the thermal behavior of Li-ABW, which is summarized in Scheme 1. The Li-ABW dehydration has been described in detail by Norby [8]. Complete water loss occurs at around 300 °C upon which the zeolite structure collapses. Rehydration is possible to a limited extent. The powder pattern of anhydrous Li-ABW was indexed with an orthorhombic cell, and it was assumed that the structure has narrowed elongated, 8-ring channels in which 4-coordinated Li ions are situated. Based on this model, a transformation path was then proposed for the transition to γ-eucryptite. Yet, the structure model of anhydrous Li-ABW did not conform to the intensities of the PXRD data. Although the dehydration temperature and occurrence of an intermediate phase is strikingly similar for LiAlSiO₃(OH)₂, it can be safely assumed that its transition follows a different route, through nucleation and growth of the new phase, since the coordination of Al and Li changes from octahedral to tetrahedral. As a matter of fact, we initially envisioned that the octahedral coordination for Al may be maintained in a metastable intermediate phase. However, this is at least not realized in the product obtained at 500 °C because the IR spectrum clearly shows the presence of a framework of tetrahedral units.

Scheme 1:

Comparison of the thermal behavior of LiAlSiO₃(OH)₂ and Li-ABW.

2.2 Analysis of γ-eucryptite

The γ-eucryptite structure has rarely been observed. Its presence was first recognized in rapidly cooled melts of LiAl_0.5Ga_0.5SiO₄ [12] and then seen in the crystallization of glasses in the systems LiAlSiO₄-LiGaGeO₄, LiAlSiO₄-LiGaGeO₄, LiAlGeO₄ [13], and LiAlSiO₄-NaAlSiO₄ [14]. It was further observed as a byproduct in the thermal conversion of Li-LTA (i.e. Linde type A zeolite in which Na ions were exchanged by Li) [15]. The thermal decomposition of Li-ABW established γ-eucryptite unambiguously as a polymorph of LiAlSiO₄. Its metastable nature with respect to the ground state, rhombohedral α-LiAlSiO₄, has been shown from DFT calculations of the equation of states [16]. The dehydration of LiAlSiO₃(OH)₂ provides another way to access pure γ-eucryptite.

Using PXRD and powder neutron diffraction data of pure γ-eucryptite samples Norby could determine its structure as a “stuffed” cristobalite [8]. We performed a Rietveld refinement of the PXRD pattern of our product obtained at 800 °C (Figure 5) and could confirm Norby’s model (cf. Tables 1 –3). The structure is shown in Figure 6, together with (tetragonal) cristobalite. The topology of the tetrahedra (i.e. the arrangement of tetrahedral nodes) with the six-ring motif is identical for cristobalite and γ-eucryptite. The rings are largely skewed in γ-eucryptite in order to establish the coordination of the Li ions. When adding tetrahedra, further deviations are seen regarding their orientation.

Figure 5:

(a) Overnight PXRD pattern and Rietveld refinement of γ-eucryptite with the 40–90° 2θ range being 12.5 times magnified. (b) Zoom in on the 34–38° 2θ range. Arrows denote the reflection at 35.5°, which could not be reasonably described with the existing model, and an unindexed low-intensity peak at 37.8°.

Figure 6:

Comparison of the crystal structure of tetragonal cristobalite (a) and monoclinic γ-eucryptite (b). The upper row depicts the framework of T atoms (which has been artificially decorated for cristobalite), the lower row depicts the polyhedral representation. (c) Is the same as (b) but with Li atoms included (grey spheres/tetrahedra). The structure of γ-eucryptite is based on the DFT-optimized model (cf. Tables 1 and 2).

In the monoclinic γ-eucryptite structure (space group Pc) the Li ions are distributed over two positions, Li1 and Li2. Li1 attains a rather regular tetrahedral coordination by O atoms belonging to four different Al/SiO₄ tetrahedra. In contrast, Li2 is coordinated to an edge of a SiO₄ tetrahedron and its O environment appears rather distorted. Norby could establish the Li positions from neutron diffraction data. For us it was not possible to perform a reliable refinement of the Li atom positions from in-house PXRD data, and eventually these were constraint to correspond to those of the DFT optimized structure (which is discussed subsequently). Also, we note that the Rietveld refinement was not perfect (cf. Figure 5). While the overall fit produced low R values, the peak at 35.5° could not be reasonably described with the existing model. Further, a number of low-intensity peaks were not indexed, which hints to the possibility of either triclinic symmetry or the presence of an unidentified impurity. In order to explore a possible lower (triclinic) symmetry for γ-LiAlSiO₄ we subjected model structures with P1 symmetry to DFT optimization. In the optimized structure the monoclinic symmetry was restored with unit cell parameters and Al, Si, and O atom position parameters being in good agreement with the results of the Rietveld refinement (Tables 1 –3). The Li–O distances in the DFT optimized structure are in a narrow range around 2.0 Å.

Lastly, we evaluated the temperature dependence of the volume and lattice parameters for LiAlSiO₃(OH)₂ and γ-eucryptite, obtained respectively upon heating and cooling in multi-temperature PXRD measurements (Figure 7). The volumetric thermal expansion coefficients are 36 × 10⁻⁶ and 47 × 10⁻⁶ K⁻¹, respectively, which is similar to Li-ABW in the same temperature range (43 × 10⁻⁶ K⁻¹) [9], and thus in sharp contrast to the zero thermal expansion behavior of β-eucryptite. The expansion of both LiAlSiO₃(OH)₂ and γ-eucryptite is anisotropic, affecting primarily the a, b axes and a, c axes, respectively, whereas the axis parallel to the framework channels stays virtually unchanged (c axis for LiAlSiO₃(OH)₂ and b axis for γ-eucryptite).

Figure 7:

Volume (a) and lattice parameters of LiAlSiO₃(OH)₂ (b) and γ-eucryptite (c) as a function of temperature. The lattice parameters were refined from the multi-temperature PXRD data obtained on heating and cooling, respectively. Both phases show little to none thermal expansion in the axis parallel to the channels, while in-plane dimensions change significantly.

4.1 Synthesis

Phase-pure LiAlSiO₃(OH)₂ was obtained when treating mixtures of LiAlSiO₄ glass and water with a molar ratio 1:1–1:2 at 5 GPa and 400–600 °C for 8 h in a multi-anvil high pressure device. Mixtures were confined in Au–Pd alloy capsules which were sealed by laser welding to prevent water loss. The recovered capsule was pinched open, verified to still contain free H₂O, and the water then removed by evaporation at ∼50 °C in a drying oven. For details on the preparation of the LiAlSiO₄ glass precursor and its crystallization in hydrothermal environments at gigapascal pressures see Ref. [10].

4.2 Powder X-ray diffraction (PXRD) analysis

Multi-temperature PXRD studies were performed on a Panalytical X′Pert PRO instrument operating with CuKα radiation and in θ-2θ diffraction geometry using the XRK 900 chamber from Anton Paar. The powder sample was mounted on a gold sample holder. Measurements were performed in air atmosphere from room temperature up to 500 or 800 °C and back. The data was collected in the 2θ range of 10–50° with steps of 100 °C between individual measurements. Each measurement had 5 min equilibration time, 45 min acquisition time and 5 K min⁻¹ heating or cooling rate between the steps. Measurements for Rietveld refinement were recorded on a Panalytical X’Pert Alpha1 diffractometer operated with CuKα₁ radiation. The sample was mounted on a zero-background Si wafer and measured at room temperature in Bragg-Brentano geometry in a 2θ range of 10–110° with 0.013° step size. Powder patterns were indexed using the programs Dicvol06 [17] and McMaille [18]. Structure solution and Rietveld refinement were performed using the FOX [19] and Fullprof [20] software packages, respectively. The following parameters were refined: background, zero shift or sample displacement, unit cell parameters, peak profile and asymmetry, and anisotropic peak broadening using spherical harmonics whenever necessary [21]. Pseudo-Voight or Thompson-Cox-Hastings Pseudo-Voight functions were used for the peak shape refinement.

CCDC 2104036 contains the supplementary crystallographic data of monoclinic γ-eucryptite, which can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif.

4.3 Transmission electron microscopy (TEM) investigations

Transmission electron microscopy (TEM) morphological observations and in-situ decomposition studies were performed on a JEOL 2100F instrument operating at 200 kV accelerating voltage. A 100 µm condenser aperture and spot size three were used in order to slow down the beam induced decomposition. Morphological observations were made through conventional bright-field (BF) imaging. The in-situ study comprised time series of a combination of BF imaging and energy dispersive X-ray (EDX) measurements. All measurements were performed on powder samples deposited onto a copper micro-grid coated with holey carbon. Electron diffraction investigations intended for obtaining structural information by applying the continuous rotation electron diffraction (cRED) technique [22], [23], [24] were performed on a JEOL 2100 LaB₆ microscope operating at 200 kV. In order to reduce the electron beam damage, a 70 µm condenser aperture was used and spot size 3. To further reduce the potential beam damage during data acquisition, the sample was cooled to approximately −170 °C in a cryo TEM sample holder. The cRED data were collected using the Timepix Quad, a high-speed hybrid detection camera.

4.4 Attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy

Attenuated total reflection Fourier transform infrared (ATR-FTIR) spectra were acquired on a Varian 610-IR FTIR spectrometer in the 400–4000 cm⁻¹ wavenumber range (32 scans, resolution 4 cm⁻¹) using a Specac Goldengate micro-ATR accessory equipped with KRS-5 lenses and a diamond ATR element. Spectra were normalized in a range from 0 to 1.

4.5 DFT calculations

DFT total energy calculations were performed using the Vienna Ab Initio Simulation Package (VASP) [25, 26] in the framework of the projector augmented wave method (PAW) [27] within generalized gradient approximation (GGA), and employing the Perdew-Burke-Ernzerhof (PBE) parametrization of the exchange-correlation functional [28, 29]. The cutoff energy for the plane wave basis set was 600 eV. Structural relaxations employed a 2 × 2 × 2 Γ-centered k-point grid, and Brillouin zone integration was done with the tetrahedron method [30]. Forces on all atoms were converged to at least 0.01 eV Å⁻¹.