Introduction

The P21/cC2/c phase transformation in natural and synthetic pyroxenes occurs both at high temperature and high pressure for several compositions (e.g. Angel et al. 1992; Arlt and Angel 2000a; Tribaudino et al. 2001, 2002; Nestola et al. 2004; Pommier et al. 2005). The high temperature and high pressure C2/c phases, however, are characterized by a different structural asset of the tetrahedral chains running along c. Based upon this difference, we call HTC2/c those clinopyroxenes showing extended tetrahedral chains (e.g. spodumene LiAlSi2O6, O3–O3–O3 = 170°, Arlt and Angel 2000a) and HPC2/c those having kinked tetrahedral chains (e.g. iron-free pigeonite Ca0.15Mg1.85Si2O6, O3–O3–O3 = 137°, Nestola et al. 2004). The complete sequence of phase transformations from HTC2/c to P21/c to HPC2/c has been found for ZnSiO3 clinopyroxene (Arlt and Angel 2000a) at room temperature through a pressure range included between room pressure and 7.4 GPa. For Li-bearing pyroxenes instead only the HTC2/c to P21/c phase transition has been observed so far (Arlt and Angel 2000a; Pommier et al. 2005), whereas P21/c Li-bearing pyroxenes appear to be very stable with increasing pressure (Gatta et al. 2005). The pyroxene composition plays a substantial role in determining the transition pressure or temperature as well as the compressibility; high pressure and high temperature studies of pyroxene solid solution are, therefore, crucial to better constrain the structure—physical properties’ relationships in these minerals.

Here, we present high-pressure data obtained for two synthetic samples belonging to the LiAlSi2O6–LiGaSi2O6 solid solution using single-crystal X-ray diffraction, in order to define a possible trend between the size of the M1 crystallographic site and the HTC2/cP21/c transition pressure, as well as to improve the systematic compressibility of Li-bearing pyroxenes. The samples studied can be defined as HTC2/c as they show at room conditions a strongly extended tetrahedral chain (e.g. Ohashi et al. 2003).

Experimental

Two synthetic single-crystals with compositions Li(Al0.53Ga0.47)Si2O6 and LiGaSi2O6 (both showing space group HTC2/c at room pressure) and crystal sizes 175 × 120 × 50 and 100 × 70 × 55 μm3, respectively, were selected for the high-pressure X-ray diffraction on the basis of their sharp optical extinction, absence of twinning and sharp diffraction profiles (about 0.06° in ω). The samples are selected from the same synthesis run of Ohashi (2003) done in a belt-type apparatus using mixtures of crystalline Li2Si2O5, Al2O3, Ga2O3 and SiO2 sealed in platinum capsules and maintained at 1,770 K and 6 GPa for 20 h. The crystals were loaded in a BGI-type and an ETH diamond anvil cells using T301 steel gaskets preindented to 80 and 85 microns for Li(Al0.53Ga0.47)Si2O6 and LiGaSi2O6, respectively, and with holes of 250 μm in diameter. A methanol:ethanol = 4:1 mixture was used as hydrostatic medium and two single-crystals of quartz (Angel et al. 1997) were used as internal pressure standards. Unit-cell parameters were determined using a HUBER four-circle diffractometer. Full details of the instrument and the peak-centring algorithms are provided by Angel et al. (2000). During the centring procedure, the effects of crystal offsets and diffractometer aberrations were eliminated from refined peak positions by the eight-position centring method of King and Finger (1979). Unconstrained unit-cell parameters were found to be within one estimated standard deviation of the symmetry constrained ones, calculated using SINGLE04 (Angel 2004) and reported in Table 1. In order to monitor the possible space group change of the two samples investigated, scans of three intense b-type reflections, i.e. (102), (031), and \( {\left( {\ifmmode\expandafter\bar\else\expandafter\=\fi{2}31} \right)} \) (selected among those relative to spodumene, Arlt and Angel 2000a), typical of the primitive P21/c lattice and forbidden in the C2/c space group, were performed at all measured pressures.

Table 1 Unit-cell parameters at different pressures for Li(Al0.53Ga0.47)Si2O6 and LiGaSi2O6 samples

Results

Phase transitions

The unit-cell parameters for the samples studied in this work were determined at 12 and 10 different pressures and room temperature reaching 8.85 and 7.32 GPa for Li(Al0.53Ga0.47)Si2O6 and LiGaSi2O6, respectively. Their evolution with pressure together with the evolution of the unit-cell volumes are reported in Figs. 1 and 2. For the Li(Al0.53Ga0.47)Si2O6 sample the unit-cell parameters decrease up to about 1.710 GPa, and between this pressure and 2.732 GPa they show a strong discontinuity (1.7% of volume change), which indicates a first order phase transition from C2/c to P21/c as indicated by the appearance of the b-type reflections, forbidden in C2/c. Then, to better bracket the transition, we decreased the pressure and at 2.156 GPa the sample symmetry was still P21/c as confirmed by the b-type reflections. At a polarizing microscope, during a further decrease of pressure from 2.156 to 1.814 GPa we observed the coexistence of the P21/c and C2/c domains in the pyroxene crystal separated by a sharp Becke line (Fig. 3). This phenomenon was already reported in a previous work on spodumene LiAlSi2O6 by Arlt and Angel (2000b), and was related to the occurrence of pressure buffering in the DAC during a first-order phase-transition. In that work, the authors proposed that the phase transition boundary is parallel to the (100) plane. Since we used the same crystal orientation in the DAC used by Arlt and Angel (2000b), it is likely that also in our experiment the (100) plane represents the transition boundary, with the crystal oriented with the (001) plane parallel to the diamond culet. Once the pressure reached 1.814 GPa the P21/c domain disappeared as also confirmed by the zero intensity of the (102) reflection and the evident change in volume (Table 1; Fig. 2). During the second compression we observed that the P21/c symmetry is stable to 8.85 GPa (the maximum pressure reached in this experiment).

Fig. 1
figure 1

Evolution of the unit-cell parameters as a function of pressure for Li(Al0.53Ga0.47)Si2O6 and LiGaSi2O6 samples

Fig. 2
figure 2

Evolution of the unit-cell volumes with pressure for Li(Al0.53Ga0.47)Si2O6 and LiGaSi2O6 samples

Fig. 3
figure 3

Coexistence of the two C2/c and P21/c phases in Li(Al0.53Ga0.47)Si2O6 sample (larger crystal in the picture) viewed under a polarizing microscope decreasing the pressure from 2.156 GPa to slightly higher pressure than 1.814 GPa: from the top left the single domain is the P21/c transforming to C2/c starting to grow in the top middle image and evolving through the other images. The smaller crystal is the quartz used as an internal standard pressure. The phenomenon is perfectly reversible from one phase to the other across this two-phase coexistence field. The scale of the images is given by the size of the crystal

At the transition the a and c axes as well as the β angle decrease discontinuously, whereas the b axis slightly increases. This behaviour was already observed for spodumene, Sc-spodumene (Arlt and Angel 2000a), clinoenstatite and pigeonite (Angel and Hugh-Jones 1994; Nestola et al. 2004).

The high-pressure behaviour of LiGaSi2O6 is different with respect to that observed for Li(Al0.53Ga0.47)Si2O6. Going from room pressure (C2/c symmetry) to 0.39 GPa (Table 1) the sample already showed a phase transformation to the P21/c symmetry as confirmed by the appearance of the b-type reflections, obviously absent at room pressure. For this composition, we were not able to observe optically the coexistence of the P and C domains as described above for Li(Al0.53Ga0.47)Si2O6. By extrapolating the high-pressure P21/c data down to room pressure we can estimate that the change in volume due to the transition is smaller than 0.2%, much smaller than that showed by Li(Al0.53Ga0.47)Si2O6 and LiAlSi2O6; however, since we do not have any data between room pressure and 0.39 GPa we cannot speculate on the thermodynamic character of the transition.

Equations of state

In this study we could determine the equation of sate for the HTC2/c and P21/c phases of Li(Al0.53Ga0.47)Si2O6 and for the P21/c phase of LiGaSi2O6. The limited pressure range in which the HTC2/c of LiGaSi2O6 is stable did not allow evaluating any compressibility data for such symmetry. For the P21/c symmetry of LiGaSi2O6, we fitted the pressure–volume data (including those measured during decompression) using a third-order Birch–Murnagahan equation of state (Birch 1947, BM3-EoS) with the EoS-FIT5.2 program (Angel 2002) refining simultaneously the unit-cell volume, V 0, the bulk modulus, K T0 and its first pressure derivative, K′. The following EoS parameter values were obtained: V 0 = 404.32(15) Å3, K T0 = 86(2) GPa, K′ = 10.4(9) (Table 2). These values are in excellent agreement with those obtained from the “normalized stress–finite strain” (Ff) plot calculated using as V 0 the BM3-EoS value (Fig. 4). The values obtained with F Ef E plot gave K T0 = 86 GPa and K′ = 10.4. A second-order BM-EoS (BM2-EoS) was used to fit the pressure–volume data up to 1.814 GPa for the C2/c symmetry of Li(Al0.53Ga0.47)Si2O6 giving the following coefficients: V 0 = 396.89(2) Å3, K T0 = 135.2(9) GPa, K′ was fixed to 4 during the refinement. Also for the C2/c phase the F Ef E plot gave a very good agreement with the refined K T0, being the obtained value of 137 GPa. Moreover, the plot (Fig. 4) clearly indicates that the BM2-EoS well describes the PV trend for C2/c symmetry of Li(Al0.53Ga0.47)Si2O6, as the slope of the linear regression is practically zero within one standard deviation. For the high pressure P21/c symmetry of Li(Al0.53Ga0.47)Si2O6 the PV data were fitted using a BM3-EoS from 2.156 GPa up to the maximum pressure reached (8.849 GPa). In a first cycle of refinement V 0, K T0 and K′ were refined simultaneously and gave the following coefficients: V 0 = 396.3(3) Å3, K T0 = 81(4) GPa, K′ = 12(1). The relatively large error in K T0 is due to the limited number of measured points and to the fact that an experimental V 0 is not available. Since the “normalized stress–finite strain” (Ff) plot calculated using as V 0 the volume obtained from the BM3-EoS fit, indicates a K′ = 12 (Fig. 4), we decided to perform a further refinement cycle with fixed K′ = 12 obtaining the following results: V 0 = 396.63(7) Å3, K T0 = 80.5(4) GPa (Table 2). Note that large values of K′ are quite common for Li-clinopyroxenes (see for example Arlt and Angel 2000a for spodumene: K′ = 8.9(6) and Gatta et al. 2005 for Fe–Mg spodumene: K′ = 9.6).

Table 2 Equation of state coefficients (see the text) for Li(Al0.53Ga0.47)Si2O6 and LiGaSi2O6 samples (this study) and LiAlSi2O6 (Arlt and Angel 2000a)
Fig. 4
figure 4

F Ef F plot (see the text) for LiGaSi2O6 (P21/c symmetry) and LiAl0.53Ga0.47Si2O6 (P21/c and C2/c symmetry) studied in this work

The volume compressibilities, calculated as β V  = −1/K T0, are the following: β V  = 0.00740(2) GPa−1 and β V  = 0.01242(3) GPa−1 for the C2/c and P21/c Li(Al0.53Ga0.47)Si2O6, respectively, and β V  = 0.01163(3) GPa−1 for the P21/c LiGaSi2O6. For comparison the volume compressibilities of the C2/c and P21/c phases of spodumene LiAlSi2O6 (Arlt and Angel 2000a) are β V  = 0.00694(3) GPa−1 and β V  = 0.01099(3) GPa−1, respectively.

Axial compressibility

In order to determine the axial moduli for a, b, and c for the C2/c and P21/c phases of the Li(Al0.53Ga0.47)Si2O6 and for the P21/c symmetry of the LiGaSi2O6 we used a parameterised form of the BM-EoS in which the individual axes are cubed, following the procedure implemented in the EoS-FIT5.2 program (Angel 2002). The results obtained are reported in Table 2. For Li(Al0.53Ga0.47)Si2O6 the axial bulk modulus ratios are 1.00:1.05:1.40 and 2.60:2.67:1.00 for the C2/c and the P21/c phase, respectively, indicating both a change in the compressibility scheme and a more pronounced anisotropy associated with the transformation. For the P21/c LiGaSi2O6 end-member the axial bulk modulus ratios is 1.23:2.23:1.00, which is similar to that of the P21/c Li(Al0.53Ga0.47)Si2O6 sample, the major difference being the compressibility along the a axis which is softer for LiGaSi2O6 than for Li(Al0.53Ga0.47)Si2O6. Concerning the LiAlSi2O6 (Arlt and Angel 2000a) the axial bulk modulus ratios are 1.00:1.05:1.29 and 1.54:2.22:1.00 for the C2/c and the P21/c phases, respectively, with the Ga causing an increase in anisotropy for both the C and P symmetries.

Discussion and conclusions

Phase transition and elasticity: comparison with LiAlSi2O6, LiFeSi2O6 and LiScSi2O6

Beside LiAlSi2O6 (Arlt and Angel 2000a) and the samples studied in this work, high-pressure data are also available for LiFeSi2O6 and LiScSi2O6 (Pommier et al. 2005; Arlt and Angel 2000a). These two end-members, HTC2/c at ambient conditions, show a phase transition to P21/c between 0.7 and 1 GPa and between 0.3 and 0.66 GPa, respectively. For purpose of comparison, we can assume as their transition pressure the average values of 0.85 and 0.48 GPa for LiFeSi2O6 and LiScSi2O6, respectively. For these Li-clinopyroxenes, changes in composition are limited to cation substitution at the M1 site; therefore we can expect some relationship between the M1 cation size and the transition pressure. In Fig. 5, we report the M1 aggregate cation radius (Shannon 1976) versus the transition pressure for all samples. It appears that for the LiAlSi2O6–LiGaSi2O6 join and LiFeSi2O6 the pressure decreases almost linearly with decreasing the cation radius, whereas the LiScSi2O6 datum lies out of the trend. This behaviour is similar to that observed as a function of temperature for the HTC2/cP21/c transition in LiM3+Si2O6 clinopyroxenes (Redhammer et al. 2001; Redhammer and Roth 2004). The following transition temperatures, taken from Redhammer and Roth (2004) and Redhammer et al. (2001): T c  = 330(1) K for LiCrSi2O6; T c  = 286(1) K for LiGaSi2O6; T c  = 205(1) K for LiVSi2O6; T c  = 234(1) K for LiScSi2O6; T c  = 90(1) K for LiSc0.72In0.28Si2O6; and T c  = 229(1) K for LiFeSi2O6, are plotted versus the M1 cation radius for the LiM3+Si2O6 in Fig. 6. Also in this case the transition temperature decreases linearly with decreasing the cation radius with exception of the Sc-bearing pyroxenes. A major difference between the high-pressure and the high-temperature behaviour is given by spodumene LiAlSi2O6 (Arlt and Angel 2000a; Redhammer and Roth 2004), since this C2/c pyroxene transforms to the P21/c symmetry with increasing pressure, but it is C centred at all temperatures. It appears, therefore, that the variation of the M1 cation size is not the only factor to determine the transition pressure or temperature, but that also the different electronic configuration of the M1 cations (as found for the P21/c to HPC2/c phase transition in pyroxenes Arlt et al. 1998), as well as differences in covalent bonds between the M1 cations and the oxygens should be taken into account.

Fig. 5
figure 5

Cation radius at M1 site versus the pressure of C2/cP21/c transition for different Li bearing clinopyroxenes. The solid line is a weighted linear fit

Fig. 6
figure 6

Cation radius at M1 site versus the temperature of C2/cP21/c transition for different Li bearing clinopyroxenes. The solid line is a weighted linear fit

The substitution of Ga for Al (i.e. increase of the cation radius at the M1 site) gives rise to a decrease of the bulk modulus for both C2/c and P21/c phases. In particular, for the P21/c phases, for which we have the bulk moduli of both end-members, it appears that the K T0 variation across the solid solution is not linear (Fig. 7), having the sample with mixed composition a bulk modulus smaller than that expected from an ideal behaviour of the solid solution.

Fig. 7
figure 7

Bulk modulus K T0 versus the Ga content at M1 site for the P21/c phases of LiGaSi2O6, LiAl0.53Ga0.47Si2O6 (this study) and spodumene LiAlSi2O6 (Arlt and Angel 2000a)

The P21/c phases of spodumene and Li(Al0.53Ga0.47)Si2O6 are more compressible than the corresponding C2/c phases, this is due mostly to the softening of the c direction (which is the most rigid in the C2/c phases) as a result of the phase transformation.